JP2006331322A - Search tag and information search system using the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a search tag and an information search system capable of easily extracting a search tag with a fixed focus imaging device and easily searching related information based on the extracted search tag. To do.
SOLUTION: A search tag 3 in which five identification points P0 to P4 representing a specific geometric invariant are arranged on each side of a quadrangular region 3a is formed in a search information holding object 2. An imaging unit 11 that captures an image of the search information holding object 2 on which a search tag is formed, and a tag identification unit 12 that identifies each identification point from the image data captured by the imaging unit 11 and calculates an identification point position. A geometric invariant calculating means 13 for calculating a geometric invariant based on the identification point position calculated by the tag identifying means 12, and a geometric invariant calculated by the geometric invariant calculating means 13. Search information extraction means 14 for extracting search information on a plane including the search tag based on a variable is provided.
[Selection] Figure 1

Description

  The present invention relates to a search tag that can easily specify search information that is a key when searching for information, and an information search system using the search tag.

In order to perform various processes with a processing device such as a computer using search tags that specify various search information existing in space, the position (spatial coordinates), distance and orientation (direction vector) of the search tag ), It is necessary to recognize the type.
In order to detect the position of the search tag, the search tag is imaged by an imaging means such as a digital camera without using artificial landmarks, and feature points are extracted from the imaged data by image processing to obtain the position and orientation. However, in this case, since advanced image processing is required, there are unsolved problems such as a long processing time and a recognition result that is not stable.

In order to solve the above unsolved problems, for example, a specific pattern is attached as a mark at the position of the search information, and the search information holding object is imaged by an imaging means such as a digital camera. A method of detecting a position of search information by searching for a specific pattern is employed.
Furthermore, as a search information, for example, a bar code or a non-contact type radio tag (RFID) is attached to a desired search information holding object, and a method in which an identification code is read by a reading device is adopted. Alternatively, a technique is adopted in which character information printed on a retrieval information holding object is imaged by an imaging unit such as a digital camera, and the captured data is recognized by a character recognition unit to recognize search information. However, in any case, in order to recognize the search information, it is necessary to read the identification code or image the character information by bringing the reading device or imaging means close to the position of the search information. There is an unsolved problem that it cannot be applied when identifying search information held in objects scattered over a wide range.

In order to solve such an unsolved problem, a tag in which five black dots are arranged on an octagonal plate is provided, the tag is photographed with a digital camera, and the type, position and orientation of the plate are determined from the image data. Research on worker information support using fixed signs based on geometric invariants that are simultaneously recognized has been conducted (for example, see Non-Patent Document 1).
Here, the geometric invariant is, for example, when five points (P0 to 04) on the plane are not on the same straight line, the following two geometric invariants may exist. It is known (see, for example, Non-Patent Document 2).

When a plane on which these five points are photographed is photographed with a camera, geometric distortion generally occurs due to projective transformation. For example, image data captured by the camera changes when the orientation of the camera rotates or the distance from the plane changes. Further, depending on the tilt of the camera, a rectangle may appear as a parallelogram (a right angle is not preserved), or a rectangle appears as a trapezoid (a parallelism is not preserved). However, the two values I ₁ and I ₂ calculated by the above equation (1) from the coordinates of the five points in the camera image are unchanged even if there is a geometric distortion due to the projective transformation described above. A geometric invariant having such a property may be used as a feature amount for image recognition.

  That is, an ID marker generating means for generating an ID marker serving as a solid marker for recognizing the ID code of an object based on a geometric invariant, and the ID marker obtained by photographing from an arbitrary viewpoint are attached. An ID marker area extracting means for extracting an ID marker area by taking an image of an object with an image input device and performing image processing on the image; and a feature for extracting positions of a plurality of feature points from the image of the ID marker area A point position extracting unit, a geometric invariant calculating unit for calculating at least one geometric invariant from the positions of a plurality of feature points, and an ID code of an object associated with the geometric invariant in advance An object ID code recognizing device having ID code recognizing means that is selected from a database that stores a physical invariant and an object ID code in association with each other is known (for example, , See Patent Document 1).

Toshiki Ito, “Study on worker support using solid labels based on geometric invariants”, Graduate School of Engineering, Osaka University, Department of Electronic Control Mechanical Engineering, 2000 Master thesis, http: //www-cape.mech. eng.osaka-u.ac.jp/ccm06adm/oldboys/2000/ito.pdf Akihiro Sugimoto “Extracting Information Regardless of how Objects are Visible”, Journal of Information Processing Society of Japan, December 1996, Vol. 37 No. 12 Pages 1125 to 1131 JP 2004-192342 A (first page, FIG. 1)

  However, in the conventional example described in Non-Patent Document 1, the tag has an excellent function of simultaneously recognizing the type, distance, and posture of an object. In order to find the tag, it is necessary to adjust the color for each use environment because the tag is given a rare color in that environment. When using the tag in various environments, the tag is not necessarily stable. There is an unsolved problem that it cannot be detected.

In addition, since the tag is as small as about 5 cm in diameter, if the tag is found from the camera image data of the entire space, it is necessary to zoom the camera to obtain a clear image of five points. Naturally, prior to zooming, There is an unsolved problem that the camera direction (pan, tilt) must be controlled, which increases the cost of the camera and its control system and increases the processing time.
Furthermore, in calculating the geometric invariant, it is necessary to accurately measure the coordinates of the black spot. In the camera image data, a portion having a luminance below a certain threshold is recognized as a black spot, and the center of gravity is determined. Since the coordinates are adopted, as is generally known, the brightness of the camera image data varies greatly depending on the illumination conditions, so detection of black spots is not always stable, and the coordinate measurement accuracy is adversely affected. There are unresolved issues.

In addition, in the conventional example described in Patent Document 1, an ID marker is formed in which two sets of feature points of five color feature points and five black feature points are formed in a 7 × 7 square cell. The ID marker is affixed to the object, the object image captured by the object ID code recognition device is captured by the camera or the like, the image information is subjected to Fourier transform and converted to a spatial frequency, and then the object ID marker area Since the feature point positions are extracted after extracting the similar or coincident shape with the spatial frequency data of the background image by pattern matching, there is an unsolved problem that the feature point extraction process becomes complicated . In addition, there is an unsolved problem that search information cannot be arranged inside the ID marker.
Therefore, the present invention has been made paying attention to the above-mentioned unsolved problems of the conventional example, and the search tag can be easily extracted by a fixed-focus imaging device, and the extracted search tag can be used. It is an object of the present invention to provide a search tag and an information search system that can easily search related information based on the search information.

In order to achieve the above object, a search tag according to claim 1 includes five identification points representing specific geometric invariants on each side of a rectangular region formed on a plane. It is characterized by being arranged so as to specify a position in a plane including
Further, in the search tag according to claim 2, in the invention according to claim 1, an equally divided scale is formed on each side of the quadrangular region, and the identification point is arranged on the formed scale. It is characterized by that.

Further, the search tag according to claim 3 is characterized in that, in the invention according to claim 1 or 2, the identification point is arranged to be movable on each side of the quadrangular region.
Furthermore, the search tag according to claim 4 is the search tag according to any one of claims 1 to 3, wherein each of the identification points has a luminance change pattern along a radial direction starting from a circular center point. It is the same in the radial direction.
Still further, in the search tag according to claim 5, in the invention according to claim 4, the brightness change pattern includes a decrease pattern in which the brightness gradually decreases and an increase pattern in which the brightness gradually increases as the distance from the center point increases. It is characterized by being either one of these.

  The information search system according to claim 6 is the search tag described in any one of claims 1 to 5 attached to the search information holding object, and the search information to which the search tag is attached. An imaging unit that captures an image of the holding object, a tag identifying unit that identifies each identification point from the image data captured by the imaging unit and calculates an identification point position, and an identification point position calculated by the tag identifying unit Geometric invariant calculating means for calculating a geometric invariant based on the search information on the plane including the search tag based on the geometric invariant calculated by the geometric invariant calculating means. It is characterized by comprising search information extracting means for extracting.

  Furthermore, an information search system according to claim 7 is the search tag described in any one of claims 1 to 5 attached to the search information holding object, and the search information to which the search tag is attached. An imaging unit that captures an image of the holding object, a tag identifying unit that identifies each identification point from the image data captured by the imaging unit and calculates an identification point position, and an identification point position calculated by the tag identifying unit Geometric invariant calculating means for calculating a geometric invariant based on the search information on the plane including the search tag based on the geometric invariant calculated by the geometric invariant calculating means. Search information extraction means for extracting and image correction means for geometrically correcting the image data of the search information extracted by the search information extraction means based on the search tag are provided.

  Furthermore, an information search system according to claim 8 is a search tag according to any one of claims 1 to 5 attached to a search information holding object, and a search with the search tag attached thereto. An imaging unit that captures an image of an information holding object, a tag identifying unit that identifies each identification point from image data captured by the imaging unit and calculates an identification point position, and an identification point position calculated by the tag identifying unit And a geometric invariant calculating means for calculating a geometric invariant based on the information, and an information searching means for specifying search information based on the geometric invariant calculated by the geometric invariant calculating means. It is characterized by that.

Still further, in the information search system according to claim 9, in the invention according to any one of claims 6 to 8, the tag identification unit stores feature information that can identify the identification point, and the image data The comparison information that can be compared with the feature information is acquired from, the acquired comparison information is compared with the feature information, and the identification point is identified based on the comparison result. .
According to a tenth aspect of the present invention, in the invention according to the ninth aspect, the feature information and the comparison information are directional codes that represent one of a maximum increase direction and a maximum decrease direction of luminance for each pixel. It is characterized by being.

According to the search tag according to the present invention, unlike the conventional tag, five identification points are arranged on each side of the quadrangular region. Therefore, even if the search tag is photographed with a fixed focus camera, the search tag is used. Can be easily identified, and search information is arranged in a rectangular area, and this search information is specified by a geometric invariant calculated based on five identification points. The effect that it can detect is acquired.
Further, according to the information search system of the present invention, the search tag can be easily identified, and image data obtained by capturing the search tag by recognizing a coordinate system based on the identification point of the search tag. From this, it is possible to calculate the geometric invariant, and to extract the search information included in the image data and specify the search information position based on the geometric invariant.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 is a system configuration diagram showing the configuration of the first embodiment in the information search system according to the present invention.
In the figure, reference numeral 1 denotes an information search system, which includes a search tag 3 including search information printed on a search information holding object 2 such as a magazine or a folder advertisement, and reads the search tag 3 to extract search information. A tag reading device 4 and an information search device 5 that performs information search based on the search information extracted by the tag reading device 4 are provided.

As shown in FIG. 3, the search tag 3 has a square area 3a such as a square or rectangle formed on the surface of the search information holding object 2, and a scale 3b divided into, for example, 16 equal parts is formed on each side of the square area 3a. In addition, five circular identification points P0 to P4 having a center point are formed on the scale 3b.
The rectangular area 3a has a height H, a width W, an upper left vertex as the origin (0, 0), a right direction as the x axis, and a lower direction as the y axis. P0 is fixed at the origin (0, 0), and the other four identification points P1 to P4 are arranged at arbitrary positions on each side. In order to avoid the three points being arranged on the same straight line, for example, four identification points other than the identification point P0 placed at the origin are not arranged at the vertices. Here, it is assumed that four identification points other than the identification point P0 are P1, P2, P3, and P4 in order of counterclockwise from the identification point P0 when viewed from the center of gravity of the five identification points.

The discrimination point P0 is set to a large diameter (for example, 20 mm) to distinguish it from the other discrimination points P1 to P4, and the other discrimination points P1 to P4 are set to a small diameter (for example, 15 mm).
Therefore, the coordinates of the identification point P1 are represented by (0, y ₁ ), the coordinates of the identification point P2 are (x ₂ , H), the coordinates of the identification point P3 are (W, y ₃ ), and the coordinates of the identification point P4. Is set to (x ₄ , 0). Here, y ₁ and y ₃ may range from H / 16~15H / 16, x ₂ and x ₄ can range of W / 16~15W / 16. Therefore, there are 15 ⁴ = 50625 arrangement patterns of the five identification points P0 to P4.

Further, as shown in FIG. 4, each identification point P0 to P4 has a circular outer periphery, and the luminance change pattern along the radial direction is the same in all radial directions starting from the center point O. Is set. The luminance change pattern may be any change pattern as long as it is a monotonously decreasing or monotonically increasing pattern in a narrow sense. In the example of FIG. 4 (a), the luminance value decreases linearly as shown in FIG. 5 (a), and in the example of FIG. 4 (b), it is shown in FIG. 5 (b). Thus, the luminance value increases linearly. In addition, the decreasing pattern may be a curve that continuously decreases with a negative tangent slope, and may be an upwardly convex curve as shown in FIG. As shown, it may be a downwardly convex curve, an S-shaped curve as shown in (c), or a curve obtained by combining two S-shaped curves as shown in (d). Similarly, in the case of the increasing pattern, various patterns in which the slope of the tangent is positive and continuously increases can be employed.
For the search tag 3 having the above configuration, by substituting the coordinates of each identification point into the equation (1), the two geometric invariants I ₁ and I ₂ with the identification point P0 as the origin are It is calculated as follows using coordinates normalized by length.

Here, X _i = x _i / W, Y _j = y _j / H, W is the width of the quadrangular area 3a, and H is the height of the quadrangular area 3a. X _i and Y _j are values obtained by dividing the scale numbers (1 to 15) attached to the sides of the width and height by an equal fraction (= 16).
As shown in the above equation (2), the geometric invariants I ₁ and I ₂ do not depend on the length of the sides of the quadrangular region, so that the vertical and horizontal sizes of the search tag 3 can be freely set. An effect is obtained. Further, since it does not depend on the unit of length, (x _i , y _i ) (i = 0 to 4) may be a scale number (0 to 16), or may be an actually measured length (for example, mm unit). It may be a value. Furthermore, by providing a scale, an effect is obtained that the arrangement of the five identification points P0 to P4 can be designated without error.

The search tag 3 having the above configuration is placed at an arbitrary position in the space and photographed by the digital camera 11 to obtain five coordinates (x _i , y _i ) (i = 0 to 4) in the image data. Can do. Then, the identification point having the largest diameter is P0, and the identification points are P1, P2, P3, and P4 in the counterclockwise direction as viewed from the center of gravity of the five identification points. When the geometric invariants I ₁ and I ₂ are calculated by substituting them into the formula, the calculated geometric invariants I ₁ and I ₂ are the same values regardless of the three-dimensional position of the search tag 3. Become.

Since the calculated geometric invariants I ₁ and I ₂ are obtained as real values, the following formulas (3) and (4) are calculated using predetermined unit quantities I ₁ _unit and I ₂ _unit. As a result, it is quantized and converted into integer identification codes I1 and I2.
I1 = INT [I ₁ / I ₁ _unit] (3)
I2 = INT [I ₂ / I ₂ _unit] (4)
Here, INT [] is a function that rounds down the decimal point of the real number in square brackets to make it an integer.

In this way, the identification codes I1 and I2 can be calculated for each arrangement pattern of the five identification points P0 to P4. The geometric invariants I ₁ and I ₂ are expressed by the above formula (3) and ( 4) Since it is converted into an integer in the equation, the arrangement pattern of the identification points representing the same identification codes I1 and I2 is not necessarily one, and a reading error occurs when reading the coordinates of the identification points from the captured image data. Therefore, it is preferable that the arrangement patterns that can be reliably identified in consideration of reading errors are limited to the 50625 portions.

For this reason, in order to determine the arrangement pattern which can be used, the identification point arrangement pattern determination process shown in FIG. 7 is executed using the information retrieval apparatus 5 including a microcomputer in advance. As described above, since the coordinates of the identification point do not depend on the unit of length, here, (x _i , y _i ) will be described as a scale number.
As shown in FIG. 7, in the discrimination point arrangement pattern determination process, first, in step S1, the coordinates x ₀ , y ₀ , x ₁ , y ₄ are all set to “0”, and the coordinate y _{2 is set} to H (= 16). In addition, initialization is performed so that the coordinate x ₃ is W (= 16) and the coordinates y ₁ , x ₂ , y _3, and x ₄ are all “1”.

Next, the process proceeds to step S2, and (x ₀ , y ₀ ) to (x ₄ , y ₄ ) are substituted into the above-described equation (1) to calculate geometric invariants I ₁ and I _2. The calculated geometric invariants I ₁ and I ₂ are converted into integers according to the equations (3) and (4) to calculate the identification codes I1 and I2.
Next, the process proceeds to step S3, where the tolerance ε for each of the _four parameters y ₁ , x ₂ , y ₃ and x ₄ and the fixed coordinates x ₀ , y ₀ , x ₁ , y ₂ , x ₃ , y _4. The identification codes I1 and I2 are calculated by performing the operations of the above formulas (1), (3), and (4) for 2 ¹⁰ combinations of all the allowable values obtained by adding and subtracting the minimum values I1 _MIN , I2 _MIN and maximum values I1 _MAX and I2 _MAX are calculated.

Then, the processing proceeds to step S4, the _{y 1, x 2, y 3} , x 4, I1 MIN, I1, I1 MAX, I2 MIN, I2, I2 step S5 after storing all the storage table T of _MAX Transition.
In this step S5, it is determined whether or not the recording of I1 _MIN , I1, I1 _MAX , I2 _MIN , I2 and I2 _MAX has been completed for all the arrangement patterns {y ₁ , x ₂ , y ₃ , x ₄ }. If there is an arrangement pattern that has not been recorded, the process proceeds to step S6, and any of the coordinates y ₁ , x ₂ , y ₃ , x ₄ is changed (for example, y ₁ = 1 → y ₁ = 2). Etc.) to generate another arrangement pattern {y ₁ , x ₂ , y ₃ , x ₄ }, and then return to the step S2, and all the arrangement patterns {y ₁ = 1 to 15, x ₂ = 1 to 15 , Y ₃ = 1 to 15, x ₄ = 1 to 15}, when the recording to the storage table T is completed, the process proceeds to step S7.

In step S7, the two-dimensional array I for recording the identification code (I1, I2) to be employed is initialized with “0”, and then the process proceeds to step S8, where the fluctuation range setting value V is initialized with “2”. Shift to step S9.
In this step S9, the variation width ΔI of the identification code (I1, I2) with respect to the allowable error ε for all the arrangement patterns {y ₁ , x ₂ , y ₃ , x ₄ } stored in the storage table T is expressed as (5 ) Calculate according to the formula.
ΔI = (I1 _MAX −I1 _MIN +1) + (I2 _MAX −I2 _MIN +1) (5)

Next, the process proceeds to step S10, and whether there is an arrangement pattern {y ₁ , x ₂ , y ₃ , x ₄ } in which the calculated fluctuation width ΔI of each identification code (I1, I2) matches the fluctuation width setting value V. If there is an arrangement pattern {y ₁ , x ₂ , y ₃ , x ₄ } that satisfies ΔI = V, the process proceeds to step S11 and the corresponding arrangement pattern {y ₁ , x ₂ , Y ₃ , x ₄ } using {I1 _MIN , I1 _MAX , I2 _MIN , I2 _MAX }, elements I [i1, i2] of all two-dimensional arrays I (I1 _MIN ≤ i1 ≤ I1 _MAX , I2 _MIN ≦ i2 ≦ I2 _MAX ) is determined to be “0”, and if these are satisfied, the process proceeds to step S12 and the corresponding arrangement pattern {y ₁ , x ₂ , y of the recording table T Record “Recruit” in ₃ , x ₄ }.

In step S13, the value of the element I [i1, i2] (I1 _MIN ≤ i1 ≤ I1 _MAX , I2 _MIN ≤ i2 ≤ I2 _MAX ) of the two-dimensional array I is changed to "1" indicating used. To do.
Next, the process proceeds to step S14, where it is determined whether or not the investigation has been completed for all the arrangement patterns {y ₁ , x ₂ , y ₃ , x ₄ } in the storage table T, and all the arrangement patterns {y ₁ , When the survey for x ₂ , y ₃ , x ₄ } is completed, the discrimination point arrangement pattern determination process is terminated, and when the survey for all arrangement patterns {y ₁ , x ₂ , y ₃ , x ₄ } is not completed. The process returns to step S10.

On the other hand, when the array pattern {y ₁ , x ₂ , y ₃ , x ₄ } with ΔI = V does not exist as a result of determination in step S10, the process proceeds to step S15, and the fluctuation range setting value V is set to “1”. Then, the process returns to step S10.
Furthermore, when the determination result of step S11 is any one of the elements I (i1, i2) (I1 _MIN ≤ i1 ≤ I1 _MAX , I2 _MIN ≤ i2 ≤ I2 _MAX ) of the two-dimensional array I, In step S16, it means that the fluctuation range ΔI of I1 or I2 overlaps the used value, so that the corresponding arrangement pattern {y ₁ , x ₂ , y ₃ , x ₄ } of the recording table T After “not adopted” is recorded in step S14, the process proceeds to step S14.

In this way, the arrangement pattern {y ₁ , x ₂ , y ₃ , x ₄ } of the recording table T that has been determined to be “adopted” in the identification point arrangement pattern determination process is used as the actual arrangement pattern. Then, the arranged arrangement patterns {y ₁ , x ₂ , y ₃ , x ₄ } are sorted, for example, in ascending order of the identification code I1, and are rearranged in this order and stored in a predetermined storage device such as a hard disk.
Then, when creating the search information holding object 2, for example, as shown in FIG. 8, by connecting the scales 3b in the search tag 3, the right direction from the origin (0, 0) in 16 rows × 16 columns Sequentially assigns “1” to “16”, then assigns “17” to “32” sequentially from the left end of the second row to the right, and finally assigns “256” to the right end of the bottom row. Formation regions SA ₁ , SA ₂ ,... SA ₂₅₆ are formed.

Any one of the subdivided areas SA _{1 to} SAT ₂₅₆ is selected, and a print position RP for printing a search code RC including search information such as a two-dimensional barcode is specified in the selected area, and the information search apparatus 5 , The arrangement pattern {y1, x2, y3, x4} having the identification code I1 corresponding to the numerical value i (i = 1 to 256) assigned to the specified printing position RP is selected. The identification points P0 to P4 of the selected arrangement pattern {y1, x2, y3, x4} are determined, the identification points P0 to P4 are printed with the determined arrangement pattern {y1, x2, y3, x4}, and the printing position A search code composed of a two-dimensional code such as an optically readable QR code is printed on the RP.

As shown in FIG. 1, the tag reading device 4 receives a digital camera 11 as an imaging unit that captures the search tag 3 and outputs image data D, and video data D output from the digital camera 11. The tag identifying unit 12 as tag identifying means for identifying the identification points P0 to P4 of the search tag 3 based on the video data D, and the coordinates (x0) of the identification points P0 to P4 identified by the tag identifying unit 12 , Y0) to (x4, y4), geometric invariants I ₁ and I ₂ are calculated as integers, and the identification codes I ₁ and I ₂ are calculated to calculate the identification codes I ₁ and I _2. The search information position is specified based on the statistical invariant calculation unit 13 and the identification codes I1 and I2, the image data at the specified search information position is read, and this is output as search information. A search line information extraction unit 14.

  As shown in FIG. 2, the tag identification unit 12 includes a partial image extraction unit 31 that sequentially extracts partial images PG of a predetermined size from target image data D that is image data for one screen input from the digital camera 11. A direction code calculation unit 32 that calculates an extraction direction code ED for the extracted partial image PG, a direction code storage unit 33 that stores in advance a reference direction code RD as feature information that can identify the identification points P0 to P4, The direction code comparison unit 34 that compares the extracted direction code ED calculated by the direction code calculation unit 32 with the reference direction code RD stored in the direction code storage unit 33 and determines whether or not they match. When the comparison result of the direction code comparison unit 34 moves the partial image of the partial image extraction unit 31 pixel by pixel and the extraction direction code ED matches the reference direction code RD, Is constructed the center point of the partial image at the identification point coordinate calculating unit 35 that calculates the center point of the identification points P0-P4.

  Here, for example, as shown in FIG. 9, when an arbitrary pixel PE0 of one screen data is considered, the direction code is the luminance values of the eight pixels PE1 to PE8 that are adjacent to and surround the pixel PE0. The information is obtained by encoding the direction in which the increase (or decrease) width from the brightness value of the pixel PE0 is the largest by comparing the brightness values of the pixel PE0. In this case, for example, numbers 0 to 7 in the right direction, the diagonally upper right direction, the upward direction, the diagonally upper left direction, the left direction, the diagonally lower left direction, the downward direction, and the diagonally lower right direction toward the page. The information is encoded by appending. Then, the direction code RD is obtained by obtaining the direction codes 0 to 7 with respect to the entire original data of the identification points P0 to P4 of the search tag 3 and storing them as two-dimensional data. Here, the eight directions are focused on the eight pixels adjacent to the pixel PE0. However, the present invention is not limited to this, and is not limited to this. For example, in the middle of each of the directions in which the surrounding eight pixels are arranged. Subdivision including the direction is also possible, and in this case, a number from 0 to 15 may be assigned to each direction to perform more detailed encoding.

  Further, here, the method of generating the reference direction code RD from the original data of the identification points P0 to P4 has been described. However, the present invention is not limited to this, and the reference direction code RD having a predetermined size is directly calculated. You may make it produce | generate. In the reference direction code RD generated in this way, all the direction codes (0 to 7 in the above example) appear uniformly in the circumferential direction, and all have the same direction on the same radial direction. There is a characteristic which becomes a code.

The partial image extraction unit 31 of the tag identification unit 12 extracts a partial image PG having the same size as the image size used when generating the reference direction code RD, and the direction code calculation unit 32. Is configured to calculate the direction code of the partial image PG by performing an operation in the same procedure as that for generating the reference direction code RD for the extracted partial image PG.
Here, for example, as shown in FIG. 10A, the size of the image data D is 480 pixels high and 640 pixels wide, and the reference direction code RD as shown in FIG. 10B. Consider an example in which the image size is 32 pixels in both height and width. That is, in such an example, as shown in FIG. 10A, the partial image extraction unit 31 sequentially extracts partial images PG every time one pixel moves from the uppermost left end of the image data D toward the right end. When the right end of one row is reached, the second row is moved down by one pixel, and the partial image PG is sequentially extracted every time one pixel moves from the left end to the right end, and this is continued until the last row. That is, the partial image PG is extracted.

The direction code comparison unit 34 adds up the difference of the direction codes for all the corresponding pixels of the reference direction code RD and the partial image PG. If the total value is 0, it is determined as “match”, and if it is within a predetermined threshold, it is determined as “approximate”. The direction code difference is calculated as follows. For example, the case where the direction code is divided into eight and the direction codes 0 to 7 are used will be described. The difference d between the two direction codes D1 and D2 can be calculated by the following equation (6).
d = Min {G, 8-G} where G = | D1-D2 | (6)

  That is, when the code of the extraction direction code ED is “0” and the code of the reference direction code RD is “0”, the difference d becomes “0” because G = 0, but the extraction direction code ED When the code of the reference direction code RD is “1” (or “7”), G = 1, 8-G = 7 (or G = 7, 8-G = 1). Therefore, when d = 1, the code of the extraction direction code ED is “0” and the code of the reference direction code RD is “3” (or “5”), G = 3, 8−G = 5 (or Since G = 5, 8-G = 3), d = 3.

For example, if the upper left of the target image data D is the coordinate origin (0, 0) and the comparison result of the direction code comparison unit 34 is “0”, the identification point coordinate calculation unit 35 is the center point of the partial image PG at this time. Is stored as the center point of any of the identification points P0 to P4, and the center point coordinates of the partial image PG at this time are also used as the identification points P0 to P4 even when the comparison result is not “0” but below a predetermined threshold. Is stored as any one of the center point coordinates (x _i , y _i ), and when the center point coordinates of the five identification points are detected, the identification point having the largest contour line diameter is identified. Set as the point P0, and further calculate the centroid points of the five identification points, sequentially set as the identification points P1 to P4 counterclockwise from the identification point P0 as seen from the calculated centroid points, and set the identification points P0 to P4. And its center point coordinates (x ₀ , y ₀ ), (x ₁ , y ₁ ), ( x ₂ , y ₂ ), (x ₃ , y ₃ ), and (x ₄ , y ₄ ) are output to the geometric invariant calculator 13.

Here, the method of extracting the partial image PG from the target image data D and obtaining the direction code for the partial image PG has been described. A code may be obtained, and a direction code corresponding to the partial image PG may be extracted therefrom and supplied to the direction code comparison unit 34.
In the geometric invariant calculation unit 13, the center point coordinates (x ₀ , y ₀ ), (x ₁ , y ₁ ), (x ₂ , y ₂ ), (x ₃ ) of the inputted identification points P0 to P4. , Y ₃ ), (x ₄ , y ₄ ) to calculate geometric invariants I ₁ and I ₂ by calculating the following equations (7) and (8), and calculating the calculated geometric invariant The values obtained by converting I ₁ and I ₂ into integers using the equations (3) and (4) are output to the search information extraction unit 14 as identification codes I1 and I2.

In the search information extraction unit 14, the segmentation area SA _k (k) containing the search information based on, for example, the value of the identification code I 1 out of the identification codes I 1 and I 2 input from the geometric invariant calculation unit 22. = 1 to 256), and the image data is extracted and output to the information search device 5.

  The information search device 5 includes a search code decoding program, for example, a communication program for connecting to the Internet, a transmission program for transmitting the decoded code information, and the like, and a search output from the search information extraction unit 14 of the tag reader 4 When the image data of the code is input, the image data is analyzed, the search information stored in the search code is decoded, and the search process is executed based on the decoded search information. As this search process, for example, if the search code includes the Internet connection address (URL) and product code, the connection address and product code are decrypted by the search code decoding program, and the communication program and transmission are performed. By transmitting the product code to the connection destination address decoded by the program, the server of the connection destination address is accessed, and the product information corresponding to the product code stored in the server is received and displayed on the display device 5a.

Next, the operation of the above embodiment will be described.
Now, as shown in FIG. 8, the search code in the subdivision area SA ₂₁₃ in the quadrangular region 3a becomes in the QR code is printed, so as to specify the subdivision area SA _213, an identification code I1 is "213" It is assumed that the search tag 3 on which the identification points P0 to P4 are printed in the arrangement pattern to be printed is printed on the search information holding object 2 such as a magazine or an insertion advertisement.

  In order to read the search tag 3 of the search information holding object 2, first, the digital camera 11 is set so that the search tag 3 falls within the visual field range, and then the shutter button is pressed, thereby the search tag 3 The image data D including the image is formed, and the image data D is output to the tag identification unit 12. In the tag identification unit 12, when image data D is input from the digital camera 11, the image data D is supplied to the partial image extraction unit 31, and the partial image extraction unit 31 temporarily stores the image data. A partial image PG having a predetermined size is sequentially extracted from the origin position of the image data D, and the extracted partial image PG is supplied to the direction code calculation unit 32.

  The direction code calculation unit 32 calculates an extraction direction code ED based on the luminance change between the pixels included in the partial image PG. At this time, when the partial image PG has not reached any position of the identification points P0 to P4 of the search tag 3, there is no direction code calculated based on the luminance change, or a character, a figure, or the like is described. However, there is no direction code pattern in which the luminance decreases or increases sequentially in the radial direction from the center of the identification points P0 to P4 in any case. The extraction direction code ED calculated by the direction code calculation unit 32 and the reference direction code RD calculated from the regular search tag 3 stored in the direction code storage unit 33 in advance do not coincide with each other. The total value of the differences d calculated by the equation (6) is a large value and is not recognized as the identification points P0 to P4.

However, any one of the identification points P0 to P4 is included in the partial image PG. For example, as shown in FIG. 11A, the extracted partial image PG has the central point of any of the identification points P0 to P4. In the crossing state, the direction codes of the identification points included in the partial image PG are sequentially changed as shown in FIGS. 11B to 11E, and the reference direction stored in the direction code storage unit 23 is changed. As shown in FIG. 11 (f), if the code is the direction code at the center of the search tag 3, in FIG. 11 (b), the sum of the direction code differences d is large. ), The sum of the direction code differences d becomes smaller, and in FIG. 11 (d), the sum of the direction code differences d falls within a predetermined threshold, but in FIG. 11 (e), Since it completely matches the reference direction code RD, the total difference d of the direction codes is There "0", the center point coordinates of a partial image PG in this state is stored as one of the center point coordinates of the identification points _{P0~P4 (x i, y i)} .

By repeating the above identification point extraction processing, the center point coordinates (x _i , y _i ) (i = 0 to 4) of the five identification points P0 to P4 of the search tag 3 are extracted.
At this time, even when the posture of the search information holding object 2 with respect to the digital camera 11 changes and the search tag 3 does not face the digital camera 11, the identification points P0 to P4 are the same for all 360 degrees. Thus, the processing of the tag identification unit 12 is not affected.

  Further, when the search tag 3 approaches or moves away from the digital camera 11, even if the circular pattern of the identification points P <b> 0 to P <b> 4 captured thereby is enlarged or reduced, the luminance is all 360 degrees from the center. The characteristics of the change in the same direction with respect to the direction of are not changed, and since the extraction direction code ED is the same as the reference direction code RD, the identification points P0 to P4 can be accurately identified. Will not be affected.

  Further, even if visual distortion occurs in the circular pattern of the imaged identification points P0 to P4 due to the directions of the identification points P0 to P4 and the digital camera 11, the luminance changes in all directions of 360 degrees from the center. The extraction point code ED of the imaged identification points P0 to P4 and most of the reference direction code RD coincide with each other so that the identification points P0 to P4 can be accurately identified. Thus, the processing of the tag identification unit 12 is not affected.

When the center point coordinates of the five identification points of the search tag 3 are extracted as described above, the identification point having the largest diameter of the boundary line of the identification points is set as the identification point P0, and The center of gravity point is obtained from the center point coordinates, and when the identification point P0 is viewed from the obtained center of gravity point, the identification points that appear sequentially in the counterclockwise direction are sequentially set as P1, P2... P4, and the set identification points P0 to P4 Their center point coordinates (x ₀ , y ₀ ), (x ₁ , y ₁ ), (x ₂ , y ₂ ), (x ₃ , y ₃ ), (x ₄ , y ₄ ) are expressed in geometric Output to the variable calculation unit 13.

In this geometric invariant calculation unit 13, the center point coordinates (x ₀ , y ₀ ), (x ₁ , y ₁ ), (x ₂ , y ₂ ), (x ₃ , y) of the respective identification points P0 to P4. ₃ ) Based on (x ₄ , y ₄ ), the operations of the equations (7) and (8) are performed to calculate the geometric invariants I ₁ and I _2. (4) The identification codes I1 and I2 converted to integers using the formula are output to the search information extraction unit 14.
In the search information extracting unit 14, it determines the granularity area SA ₂₁₃ based on the internal identification code of the identification codes I1 and I2 are input I1 (= 213) as the position of the search information, images of the subdivision area SA ₂₁₃ Data, that is, image data of a search code composed of a QR code is extracted, and this image data is output to the information search device 5.

  In the information retrieval device 5, the image data of the input search code is decrypted by the decryption program, the connection destination address and the product code included in the search code are decrypted, the server of the decrypted connection destination address is accessed, and the product code is transmitted. To do. As a result, the product information corresponding to the product code is transmitted from the connection destination server to the information search device 5, and when the product information from the server is received, the information search device 5 displays the received product information on the display device 5a.

Thus, according to the above embodiment, it calculates the geometric invariant I ₁ and I ₂ on the basis of the arrangement pattern of the identification points of the search for the tag 3, calculated geometric invariant I ₁ and I ₂ The identification codes I1 and I2 are calculated by converting the numbers into integers. Since the calculated identification code I1 is information indicating the position of the search information describing the search code, the search code is composed of a two-dimensional code such as a QR code. Even if it is, the position of the search code can be easily detected, the image data of the search code can be easily extracted from this position, and even if the imaging position of the digital camera 11 is set roughly, The search code can be detected accurately.

In addition, since the geometric invariants I ₁ and I ₂ calculated for the search tag 3 regardless of the facing angle with the digital camera 11 are constant values, the relative attitude between the search information holding object 2 and the digital camera 11 is fixed. Even when the angle of the search tag 3 with respect to the digital camera 11 changes, the identification code can be calculated accurately, and erroneous recognition of the search tag 3 can be reliably prevented.

Moreover, the search tag 3 is arranged so that the identification points P0 to P4 are arranged on each side of the quadrangular area 3a, and the area inside each side where the identification points are arranged is searched except for the search information position where the search code is printed. Is not used, it is possible to describe arbitrary characters, figures, etc. in areas other than the search information position in this inner area, and advertisement information, company information, etc. can be described.
Further, the search tag 3 has a scale formed on each side of the quadrangular area 3a, and the identification points P0 to P4 are arranged on the scale, so that only the outer diameter frame and the scale of the quadrangular area 3a are on the label. The search tag 3 can be configured by pasting the identification points P0 to P4 printed in advance on the scale according to the position of the search information. The manufacturing cost can be reduced compared with the case where it is provided in advance.

  In the search tag and information search system according to the present invention, the “direction code method” is used as the template matching process, and the luminance of the circular pattern of the identification points P0 to P4 of the search tag 3 is 360 from the center. It has the characteristic of changing in the same way for all directions. For this reason, for example, when the search tag 3 is imaged by the camera 11, illumination fluctuation occurs and a shadow is generated on the search tag 3, or the image is captured in the dark without obtaining a sufficient amount of light. However, since the extraction direction code ED based on the captured target image data D is the same as the reference direction code RD, the processing of the tag identification unit 12 is not easily affected.

Thus, since the search tag 3 has a circular pattern in which the luminance changes in the same manner in all directions of 360 degrees from the center point O, the search tag 3 is imaged by the camera 11. Even if the situation changes, it will not be affected.
For this reason, at the time of template matching, by preparing a single template (reference direction code RD), it is possible to deal with a wide range of imaging conditions with the camera 11 and to reduce the calculation cost and speed up the image identification process. Is obtained.

In addition, for the reason described above, there is an effect that the search tag 3 can be identified without being restricted by the positional relationship between the search tag 3 and the camera 11 and the imaging situation.
In the above first embodiment, when outputting the image data of the identification code I1, I2 search information position or subdivided areas SA _k set based on either or both directly to the information retrieval device 5 for a description However, the present invention is not limited to this, and the image data of the subdivided area SA _k may be enlarged and output to the information search device 5.

In the first embodiment, the case has been described in which the search tag 3 is printed on the search information holding object 2 such as a magazine or an insertion advertisement and the position of the search information is specified. However, the present invention is not limited to this. The present invention can also be applied to a case where the position of a transmission light source or the like for visible light communication is specified and the direction of the communication light receiver is controlled.
Furthermore, in the first embodiment, the case _where the identification code I1 is assigned to the subdivision area SA _k in order from “1” has been described. However, the present invention is not limited to this, and the identification code I1 is assigned to a specific position ( For example, an upper left corner, a lower right corner, etc.), a predetermined number of subdivided areas, or an arbitrary allocation method can be applied.

Next, a second embodiment of the present invention will be described with reference to FIGS.
In the second embodiment, the image data at the search information position can be geometrically corrected.
That is, in the second embodiment, as shown in FIG. 12, in the configuration of FIG. 1 in the first embodiment described above, a search information extraction unit is provided between the search information extraction unit 14 and the information search device 5. 14 except that an image correction unit 15 is inserted as image correction means for correcting the image data of the search code RC extracted in 14 into image data in a state in which the digital camera 11 is identified as the search information holding object 2. 1 has the same configuration as in FIG. 1, and the same reference numerals are given to the corresponding parts in FIG. 1, and the detailed description thereof will be omitted.

  Here, the image correction processing in the image correcting unit 15 will be described. As shown in FIG. 13, there are two image data A and B, and the same search tag is included in each of them. Assume that the positions of the identification points P0 to P4 of the search tag 3 in the image data A are (xAi, yAi) (i = 0 to 4). Similarly, the positions of the identification points P0 to P4 of the search tag 3 in the image data B are (xBi, yBi) (i = 0 to 4).

Considering the case where the same point P is included in the image data A and B, the coordinates (xA, yA) of the point P in the image data A are a coordinate system defined by the search tag 3 in the image data A. It is based on. Similarly, the coordinates (xB, yB) of the point P in the image data B are based on the coordinate system defined by the search tag 3 in the image data B.
In general, since the two image data A and B have a projective transformation relationship, the following equation (9) is used between the point (xA, yA) in the image data A and the point (xB, yB) in the image data B. There is a relationship expressed.

  When this equation (9) is expanded and the constant λ is eliminated, two simultaneous equations are obtained. Accordingly, by substituting the coordinates of the four identification points P0, P1, P2, and P3 among the five identification points P0 to P4 of the search tag 3, a total of eight constraint equations can be obtained. . Since the matrix M has 8 degrees of freedom, all the elements mij (i = 1 to 3, j = 1 to 3, except m33) can be obtained using these eight constraint equations.

  By using the transformation matrix M thus obtained, the points in the image data A and B can be projectively transformed with each other. That is, the point of coordinates (xA, yA) in the image data A corresponds to the point of coordinates (xB, yB) in the image data B. On the contrary, the point of coordinates (xB, yB) in the image data B corresponds to the point of coordinates (xA, yA) in the image data A. Both coordinates can be calculated by the above equation (9).

The image data A is assumed to be image data taken by the tag reading device 4, while the image data B is virtual image data when the search code is read at a predetermined size from the front position. And At this time, in the image data B, the search tag 3 should also be in an actual form (the five identification points P0 to P4 are in the design layout) or a similar form thereof.
Therefore, the tag reader 4 reads the coordinates of the four points P0, P1, P2, and P3 among the five identification points P0 to P4 of the search tag 3 to be substituted into the constraint equation obtained from the above equation (9). The point (xAi, yAi) (i = 0, 1, 2, 3) of the obtained image data A and the point (xB, yB) (i = 0, 1, 2, 3) in the design arrangement of the search tag 3 Is adopted.

By using the conversion matrix M obtained as a result, an arbitrary point in the image data obtained by the tag reader 4 can be found in the image data when the search code is read at a predetermined size from the front position. Correctly positioned.
That is, even if the tag reader 4 does not face the search code, the image data of the search code can be obtained by using the design arrangement of four of the five identification points P0 to P4 of the search tag 3. It is possible to obtain image data of a search code of a predetermined size as viewed from the front by correcting geometrically.

  As described above, according to the second embodiment, when image data including the search tag 3 is captured by the digital camera 11, even if the digital camera 11 and the search information holding object 2 are not facing each other, the search information The image data corresponding to the position is geometrically corrected to obtain image data of a search code of a predetermined size as if it was photographed from the front, and the image data after this geometric correction is stored in the information search device 5. By outputting, the information search device 5 can accurately analyze the image data, decode the search information stored in the search code, and execute the search process based on the decoded search information.

In the first and second embodiments, the case where the QR code is applied as the search code RC has been described. However, the present invention is not limited to this, and a one-dimensional barcode or other two-dimensional code is used. Any search code can be applied.
In the first and second embodiments, the search code is in one subdivision area. However, the search code is not limited to this, and the search code extends over a plurality of subdivision areas. May be present. In this case, for example, when the upper left of the plurality of areas is SA _L and the lower right is SA _R , two identification codes can be designated such as I1 = L and I2 = R.

Next, a third embodiment of the present invention will be described with reference to FIG.
In the third embodiment, the identification codes I1 and I2 calculated by the geometric invariant calculation unit 13 are directly output to the information search device 5 without using the search code, and the information search device 5 performs identification. The related information is searched based on one or both of the codes I1 and I2.

That is, in the third embodiment, as shown in FIG. 14, in the configuration of FIG. 1 in the first embodiment described above, the subdivision areas SA _{1 to} SA ₂₅₆ of the search tag 3 and the subdivision areas SA _{1 to} SA ₂₅₆ are described. The search code is omitted, the search information extraction unit 14 in the tag reader 4 is omitted, and the identification codes I1 and I2 calculated by the geometric invariant calculation unit 14 are directly supplied to the information search device 5. Except for the above configuration, the configuration is the same as in FIG. 1, and the same reference numerals are given to the corresponding parts to FIG. 1, and the detailed description thereof is omitted.

Here, when the identification codes I1 and I2 calculated by the geometric invariant calculation unit 14 are input, the information search device 5 refers to a related information storage table set in advance based on the identification codes I1 and I2. To select related information.
As this related information, a URL (Uniform Resource Locator) of a WWW (World Wide Web) server in which detailed information is stored is adopted, and this URL is stored in the related information storage table in association with the identification codes I1 and I2. ing.
When the information retrieval device 5 obtains the corresponding URL by referring to the related information storage table based on the identification codes I1 and I2 and obtains the detailed information by accessing the home page of the obtained URL. Displayed on the display device 5a.

According to the third embodiment, geometric invariants I ₁ and I ₂ are calculated based on the position coordinates of the respective identification points P0 to P4 by reading the search tag 3 with the tag reader 4. The geometric invariants I ₁ and I ₂ are converted into integers to calculate identification codes I ₁ and I ₂ , and the identification codes I ₁ and I ₂ are output to the information search apparatus 5. By referring to the storage table, the URL represented by the identification codes I1 and I2 is selected, and the homepage of the selected URL is accessed to obtain the detailed information of the advertisement content and other necessary detailed information. By displaying on the display device 5a, detailed information required by the user can be acquired easily and accurately.

In this case, since the search tag 3 can be written in the outer frame of an article on the advertisement side of a magazine, the search tag 3 itself can be displayed in a large size, and a low-resolution mobile phone camera is used. Even when the search tag 3 is photographed from a distance, the tag reading device 4 can accurately detect the search tag 3 to calculate the geometric invariants I ₁ and I ₂ , which are converted into integers. The identification codes I1 and I2 can be accurately calculated, and the detailed information can be obtained reliably without causing erroneous reading of the search code.

  In the third embodiment, the case where the search tag 3 is directly printed on the search information holding object 2 has been described. However, the present invention is not limited to this, and the search tag 3 is printed on a transparent sheet. The tag reading device 4 is obtained by photographing the search tag 3 printed on the transparent sheet with the digital camera 11 by pasting the transparent sheet on which the search tag 3 is printed on the search information holding object 2. By reading the search tag 3, the identification codes I1 and I2 can be accurately calculated. In this way, by using the transparent sheet on which the search tag 3 is printed, the existing advertisement document can be used as it is, and a transparent sheet can be pasted on the advertisement document to create a new advertisement document. The trouble of doing can be omitted.

In the third embodiment, the tag identification unit 12 identifies the point having the largest diameter among the five identification points as P0, and coordinates (x0, y0) to (x4, y4) of the identification points. However, the present invention is not limited to this, and the upper left identification point in the image photographed by the digital camera 11 is P0, and it is counterclockwise when viewed from the center of gravity of the five points. The coordinates may be calculated as P1, P2, P3, and P4 in order. The geometric invariants I ₁ and I ₂ have different values depending on which of the five discrimination points is taken as P0. Therefore, all five geometric invariants I ₁ and I ₂ are calculated in advance for each search tag when the five identification points are P 0, and the related information storage table in the information search device 5 is calculated. Registered with the same related information. By doing so, the tag identifying unit 12 can omit the process of specifying P0 by comparing the sizes of the identification points, and can simplify and speed up the entire process.

  Furthermore, in the first to third embodiments, if the search tag 3 is installed and used so that the identification point P0 is always at a specific position (for example, upper left), the tag identification unit 12 is Since the identification point at the upper left in the image photographed by the digital camera 11 may be set as P0, the process of specifying the identification point by comparing the size of the identification point can be omitted, and the entire process is simplified. The speed can be increased.

  Next, in the first to third embodiments, when extracting the identification points P0 to P4 of the search tag 3, the partial image PG having a predetermined size is applied to the entire area of the image data D having a predetermined size. The case where the identification points P0 to P4 are extracted by sequentially moving each pixel has been described. However, the present invention is not limited to this, and prior to the template matching process by the direction code comparison unit 34, the image data D is adjacent. Investigate whether or not there is continuity in a predetermined range of direction codes (whether or not the difference d in the direction codes of adjacent pixels is within the predetermined range) for each pixel to be grouped. The mask is selected from each group according to a predetermined selection criterion as shown in FIG. 15, and the identification points P0 to P4 are extracted by the partial image PG only for the pixels belonging to the selected mask. The Ukoto, it is possible to shorten the processing time of image identification. Here, since the identification points P0 to P4 have a circular pattern whose luminance changes in the same manner in all directions of 360 degrees from the center point, the direction codes generated from the images are continuous. Since the value is taken, each of the identification points P0 to P4 is extracted as one group G, that is, a mask by the above grouping process. In addition, as selection criteria, the area which the group G occupies, roundness, etc. are mentioned, for example. Further, as shown in FIG. 16, by obtaining the center point of the mask based on the coordinates at the top, bottom, left, and right ends of the mask and performing the identification point extraction process with the partial image PG only in the vicinity of the center point, the image Processing time can be shortened.

As a template, as shown in FIG. 17, there is a template of 32 pixels × 32 pixels. When the direction code is 0 to 15, a direction code 8 (leftward) is placed on the left half of the center point Q. In the right half, direction code 0 (toward the right) is arranged as follows. Here, the direction code of the center point Q is set to 0.
Example: ... 888Q000 ...
The direction code sequence is stored in the direction code storage unit 33 as the above-described reference direction code RD in FIG. 9 as a horizontal partial template (column-shaped feature information), and the extracted direction code for the image data D is long. The partial image extraction unit 31 sequentially extracts the 32 pixel column-shaped partial image PG as comparison information, compares it with the tag direction code 33 by the direction code comparison unit 34, and compares the comparison information that is identical or approximate to each other. By selecting, the identification point coordinate calculation unit 35 may detect the position coordinates of the identification points P0 to P4 from the image data D.

In this case, when the image size of the image data D is L × M (640 × 480), the total number of matching pixels is L × M × N from the template image size of N × N (N = 32). × N (about 3 × 10 ⁸ ), but the template image size is N × 1 (N = 32), and therefore the total number of pixels to be collated when comparison is performed using columnar feature information is L × M × N (about 1 × 10 ⁷ ), which is N (N = 32) times faster. That is, the speed can be increased by reducing the amount of information to be compared at one time for the feature information and the comparison information.

  Further, as shown in FIG. 18 (a), as described above, the horizontal partial template HT is used to extract an image area that matches or approximates the horizontal partial template HT, and the vertical partial template VT is again extracted with respect to the extracted image area. By performing the comparison process using, it is possible to prevent erroneous detection by performing the comparison process twice. In this case, as shown in FIG. 18B, the vertical direction partial template VT may be used first, and then the horizontal direction template HT may be used.

  By detecting the discrimination points P0 to P4 in this way, the total number when compared in the horizontal direction and then in the vertical direction is L × M × N + N × k. Here, k is the number of matches in the horizontal or vertical direction, and further compared in the vertical or horizontal direction, and has a relationship of k << L × M. That is, by performing comparison processing using a horizontal or vertical partial template, the speed can be increased by N times compared to the first embodiment, and further, after performing comparison in the horizontal (or vertical) direction, When comparing in the vertical (or horizontal) direction, it is possible to reduce false detections and increase the speed almost N times.

  Further, as shown in FIG. 18C, after performing the comparison process using the horizontal partial template HT, the planar template (planar feature information) FT consisting of several vertical partial templates and the horizontal partial template HT are compared. The comparison processing with the partial image PG extracted from the image data D having the image area of the same size as the planar template centered on the center point of the matched partial image PG may be sequentially performed. As shown in d), after the comparison process is performed using the small surface template SFT having a smaller number of pixels than the planar template FT, the center point of the small surface template SFT matching the small surface template SFT is set as the center. Partial image PG and planar template F extracted from image data D having an image area of the same size as planar template FT It may be compared processing.

Thus, after comparing one of the templates having a small amount of information with the corresponding partial image PG of the image data D and narrowing down the detection range of the actual discrimination points P0 to P4 based on the comparison result, If another comparison is made with another, for example, a template having the same amount of information or a larger amount of information, both speeding up and prevention of erroneous detection can be achieved.
Further, in each of the embodiments described above, the case where gradation is applied so that the luminance is changed in all radial directions as the identification points P0 to P4 of the search tag 3 is not limited to this. The luminance may be changed in an arbitrary number of directions such as two directions, three directions, and four directions from the center.
Further, as the search tag 3, one velvet fastener is affixed to the boundary position of the quadrangular region 3a, and the other is affixed to the back of the circular button with the identification point printed on the front surface, so that it can be placed at any scale position. By configuring the identification points P1 to P4 so that they can be arranged, a desired geometric invariant can be set.

1 is a system configuration diagram showing a first embodiment of an information search system according to the present invention. It is a block diagram which shows the specific structure of the tag identification part of FIG. It is a front view which shows the tag for search. It is a figure which shows an example of the pattern which a brightness | luminance reduces as it leaves | separates from the center point of the identification point of a search tag, and a pattern which increases. It is a gradation characteristic diagram which shows an example of the relationship between the brightness | luminance of an identification point and the distance from a center point. It is a gradation characteristic diagram which shows an example of the relationship between the brightness | luminance of an identification point and the distance from a center point. It is a flowchart which shows an example of the identification point arrangement | positioning pattern determination processing procedure performed with an information search device. It is a front view which shows the specific example of the tag for a search. It is an example of the schematic for demonstrating the production | generation of a direction code. It is explanatory drawing which shows image data and a partial image. It is explanatory drawing with which it uses for description of an identification point extraction process. It is a system configuration figure showing a 2nd embodiment of the present invention. It is explanatory drawing with which it uses for description of the image correction process of an image correction part. It is a system configuration figure showing a 3rd embodiment of the present invention. It is explanatory drawing in the case of grouping a direction code. It is explanatory drawing at the time of grouping a direction code and setting the center point as a partial image. It is explanatory drawing which shows the direction code | symbol of a horizontal and a vertical direction. It is explanatory drawing which shows the combination of a template.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Information search system, 2 ... Search information holding object, 3 ... Search tag, 4 ... Tag reader, 5 ... Information search device, 11 ... Digital camera, 12 ... Tag identification part, 13 ... Geometric invariant calculation , 14 ... Search information extraction unit, 15 ... Image correction unit, 31 ... Partial image extraction unit, 32 ... Direction code calculation unit, 33 ... Direction code storage unit, 34 ... Direction code comparison unit, 35 ... Identification point coordinate calculation unit , D ... image data, PG ... partial image

Claims (10)

  Five identification points representing specific geometric invariants are arranged on each side of a quadrangular region formed on a plane so as to specify a position in a plane including the five identification points. Search tag.   2. The search tag according to claim 1, wherein equally divided scales are formed on each side of the quadrangular region, and the identification points are arranged on the formed scales.   The search tag according to claim 1 or 2, wherein the identification point is arranged so as to be movable on each side of the quadrangular region.   The search point according to any one of claims 1 to 3, wherein each of the identification points has the same luminance change pattern along the radial direction starting from a circular center point in all radial directions. tag.   The search tag according to claim 4, wherein the luminance change pattern is one of a decreasing pattern in which luminance gradually decreases and an increasing pattern in which luminance gradually increases as the distance from the center point increases.   The search tag according to any one of claims 1 to 5 attached to a search information holding object, an imaging unit that takes an image of the search information holding object to which the search tag is attached, Tag identification means for identifying each identification point from image data captured by the imaging means and calculating the identification point position, and geometry for calculating a geometric invariant based on the identification point position calculated by the tag identification means And a search information extracting means for extracting search information on a plane including the search tag based on the geometric invariant calculated by the geometric invariant calculating means. Characteristic information retrieval system.   The search tag according to any one of claims 1 to 5 attached to a search information holding object, an imaging unit that takes an image of the search information holding object to which the search tag is attached, Tag identification means for identifying each identification point from image data captured by the imaging means and calculating the identification point position, and geometry for calculating a geometric invariant based on the identification point position calculated by the tag identification means Invariant calculating means, search information extracting means for extracting search information on a plane including the search tag based on the geometric invariant calculated by the geometric invariant calculating means, and the search information extracting And an image correcting means for geometrically correcting the image data of the search information extracted by the means based on the search tag.   The search tag according to any one of claims 1 to 5 attached to a search information holding object, an imaging unit that takes an image of the search information holding object to which the search tag is attached, Tag identification means for identifying each identification point from image data captured by the imaging means and calculating the identification point position, and geometry for calculating a geometric invariant based on the identification point position calculated by the tag identification means An information retrieval system comprising: a dynamic invariant calculating means; and an information searching means for specifying search information based on the geometric invariant calculated by the geometric invariant calculating means.   The tag identification means stores feature information that can identify the identification point, acquires comparison information that can be compared with the feature information from the image data, and compares the acquired comparison information with the feature information. 9. The information search system according to claim 6, wherein the identification point is identified based on the comparison result.   The information search system according to claim 9, wherein the feature information and the comparison information are direction codes that represent one of a maximum increase direction and a maximum decrease direction of luminance for each pixel.

Images