JP2004272885A - Device and method for edge extraction - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device and a method for edge extraction capable of stably extracting edges of an object even if an image for edge extraction contains noise components in the device and method for extracting the edges of the object in an image search area. <P>SOLUTION: The method for extracting the edges of the object in an image search area comprises a step of calculating difference values between contiguous inquiry figures about an integrated value obtained through integration of brightness projection value along perimeters of a plurality of inquiry figures similar to the object with reference to a reference point placed at an arbitrary position within the image search area, extracting the difference values to be maximum, and holding the extracted values as the candidate maximum difference values. The method also comprises a step of determining as the edge of the object the inquiry figure showing the greatest candidate maximum difference value among a plurality of candidate maximum difference values obtained by shifting the reference point. <P>COPYRIGHT: (C)2004,JPO&NCIPI

Description

  The present invention relates to an edge extracting device and an edge extracting method for extracting an edge of an object in an inspection image.

  When specifying the position of a product or a component of a product being manufactured, there is a method of extracting the edge of the component based on the luminance distribution of the inspection area. This method utilizes the property that "the difference in brightness is very large near the edge on the figure", that is, the brightness gradient near the boundary. In this method, the brightness of the brightness projection value in the inspection area of the product is used. An edge is a set of points at which the change is approximately maximum (for example, see Patent Document 1).

  FIG. 6 is a diagram illustrating an edge extraction method according to a conventional example.

  Here, an example of extracting an edge L of a target as shown in FIG. 6A will be described. The xy coordinate axes are taken as shown.

  First, for an object Q in an inspection image, an edge sampling region R having a certain size that includes the edge L is set at an approximate position where the edge L is considered to exist.

  Next, in the edge sampling region R, a luminance distribution in a direction perpendicular to the approximate direction of the edge L (x-axis direction in FIG. 6) (that is, y-axis direction) is measured. By this measurement, a bright and dark luminance distribution as shown in FIG. 6B is obtained.

  Then, the luminance distribution shown in FIG. 6B is differentiated along the y-axis direction. By this differentiation processing, a differential value is obtained as shown in FIG. At this time, the position where the differential value is maximum, that is, the position where the luminance change is maximum means one point on the edge L of the object Q.

  A plurality of the above processes are performed in the x-axis direction in the edge sampling region R to obtain a set of points having the maximum differential value. Then, an edge L is extracted from the obtained set of points using the least squares method.

JP 2001-4326 A

  The above-described method is effective when the luminance distribution of the inspection image is relatively stable and there is no noise component, or when the edge does not include a curved portion.

  However, for example, when the product to be inspected is damaged, the contrast between light and dark is weak, or the unevenness of the reflected luminance is large, the luminance change may be abnormally large at a position other than the original edge. At this time, the maximum value of the differential value that appears when the luminance distribution is differentiated appears as an abnormal point at a position that is extremely different from the original tendency. In this case, if the least squares method is applied as described above, an edge containing many errors will be extracted. At this time, even if robust estimation is used, it is difficult to remove the error when the influence of noise is large.

  When the object has a shape including a curved portion such as a circle or an ellipse, the width of the edge sampling area in the direction in which an edge can exist (for example, the x-axis direction in FIG. 6) is reduced. There is a need. That is, when the target has a shape including a curved portion, the curved portion needs to be linearly approximated. Therefore, since the edge sampling area is made finer, the error due to the subsequent least squares process is likely to be larger especially when the product to be inspected is damaged, when the contrast between light and dark is weak, or when the unevenness of the reflected luminance is large. .

  Therefore, an object of the present invention is to provide an edge extraction apparatus and method for extracting an edge of an object in an image search area, in which an edge extraction target image containing a large amount of noise components is stably An object of the present invention is to provide an edge extracting device and an edge extracting method capable of extracting an edge of an object.

  In order to achieve the above object, according to the present invention, when an object included in an image search area to be extracted has a circular shape, an edge of the object is extracted by the following processing.

  First, a reference point is set at an arbitrary position in the image search area, and an integrated value obtained by integrating luminance projection values along the circumference of a plurality of concentric exploration figures centered on the reference point, A difference value between adjacent search patterns is calculated, and the maximum difference value is extracted and held as a maximum difference candidate value at the reference point.

  Then, among the plurality of maximum difference candidate values obtained as described above by shifting the position of the reference point, the search pattern showing the largest maximum difference candidate value is determined as the edge of the object.

  The fact that the difference value of the integrated luminance projection values between adjacent search patterns is large means that the luminance on the circumference between adjacent search patterns has changed greatly, that is, the difference in brightness is large. means. In other words, since the edge of the object is the boundary line between the brightness and darkness, the maximum difference candidate value that maximizes the difference between the integrated brightness projection values between adjacent search patterns is obtained only in the circle of the search pattern. This is when the circumference is closest to the edge of the object. The greater the circumference of the search pattern close to the edge of the object, the greater the maximum difference candidate value. In other words, it can be said that the vicinity of the search pattern when the maximum difference value that is the largest difference among the plurality of maximum difference candidate values is present is the portion where the probability that the edge of the target object exists is the highest. Therefore, in the edge extraction device and the method thereof according to the present invention, the search pattern showing the maximum difference value is determined to be the edge of the edge extraction target. That is, in the present invention, the brightness gradient near the edge of the object is used.

FIG. 1 is a basic block diagram of an edge extraction device according to the present invention.
According to the present invention, the edge extraction device 1 that extracts an edge of a circular object included in an image search area,
First means 11 for setting a reference point at an arbitrary position in the image search area;
A second means 12 for setting a plurality of exploration figures so as to form a concentric circle centering on the reference point;
Third means 13 for integrating the luminance projection values along the circumference for each of the exploration figures to obtain a set of integrated values;
Fourth means 14 for calculating a difference value of an integrated value between adjacent search patterns,
A fifth means 15 for extracting a maximum difference value from a set of difference values calculated by the fourth means 14 and holding the same as a maximum difference candidate value at the reference point;
A plurality of maximum difference candidate values obtained by the second, third, fourth and fifth means 12, 13, 14 and 15 by switching the position of the reference point set by the first means 11 within the image search area. A sixth means 16 for extracting a maximum maximum difference candidate value among the maximum values and holding this as a maximum difference value;
Seventh means 17 for determining the search pattern at the time of the maximum difference value as an edge of the object.

  Further, even when the shape of the object to be edge-extracted is other than the above-described circular shape or when the shape is not known in advance, the edge of the object in the image search area is extracted using the present invention. It is possible to do. In this case, information on the approximate contour of the object to be edge-extracted is obtained. A reference search pattern is generated based on this information, and a search pattern is set using the reference search pattern as a similar reference.

  Naturally, even when the shape of the object to be edge extracted is known in advance, it is possible to extract the edge of the object in the image search area using the present invention. For example, this corresponds to a case where the shape of the object can be defined by an equation representing an ellipse, a polygon, or the like. Based on this equation, a search pattern similar to the target object is set.

  Using the search pattern, the edge of the target object is extracted by the following processing.

  First, an integrated value obtained by taking a reference point at an arbitrary position in the image search area and integrating luminance projection values along the outer perimeter of a plurality of exploration figures similar to an object based on this reference point, A difference value between adjacent search patterns is calculated, a maximum difference value is extracted, and this is held as a maximum difference candidate value at the reference point.

  Then, of the plurality of maximum difference candidate values obtained by shifting the position of the reference point, the search pattern showing the largest maximum difference candidate value is determined as the edge of the object.

  The system configuration in this case is almost the same as that of FIG. 1 described above, but the functions of the following units are particularly expanded.

  That is, first, the first means 11 sets a reference point at an arbitrary position in the image search area, and the second means 12 sets a plurality of search patterns based on the reference point and similar to the object. Set. Then, the third means integrates the luminance projection values along the outer periphery of the search pattern with respect to each search pattern to obtain a set of integrated values. Then, the seventh means 17 determines the search pattern at the time of the maximum difference value obtained through the above-described fourth means 14, fifth means 15, and sixth means 16 as an edge of the object. .

  Thus, the above system includes the system already described with reference to FIG. Therefore, the present invention can be applied to a case where the target object has a circular shape.

  According to the above-described method, a plurality of maximum difference candidate values are obtained by shifting the position of the search pattern, and the search pattern indicating the maximum maximum difference candidate value is determined as the edge of the object. When accuracy is required for the edge extraction result, it is preferable to set the position of the reference point of the search pattern finely (for example, so as to be shifted by one pixel). As a modification of this method, in order to reduce the amount of calculation to be performed by the system and to speed up the processing, as a pre-processing of the above processing, a processing of calculating a candidate point to be a reference point in advance (hereinafter referred to as “rough search Process).

  The rough search processing when the shape of the object to be edge-extracted is circular is as follows. First, a coarse search reference point is set at an arbitrary position in the image search area, and two concentric coarse search figures having a predetermined radius difference centered on the coarse search reference point are set. Next, the luminance projection values along the outer periphery and the inner periphery of the rough search graphic are integrated to obtain the outer periphery integrated value and the inner periphery integrated value. Next, an outer-internal difference value, which is a difference between the outer peripheral integrated value and the inner peripheral integrated value, is calculated. Then, with reference to the outer-internal difference value obtained by shifting the position of the coarse search reference point by a distance equal to or less than a predetermined radius difference in the image search area, the reference point is determined from the coarse search reference points of the coarse search graphic. Determine candidate points to be formed.

  Further, when the shape of the object to be edge-extracted is other than the circular shape as described above, or when the shape is not known in advance, the coarse search processing is performed by using the coarse search graphic described above in the image search area. Based on a coarse search reference point set at an arbitrary position, the rough search reference point is set to be composed of two outer peripheral figures and inner peripheral figures having different sizes similar to the object.

  According to the present invention, in an edge extraction apparatus and method for extracting an edge of an object in an image search area, an edge of the object is stably extracted even for an edge extraction target image containing a large amount of noise components. be able to.

  That is, since the direction of the edge is specified without using the least square method as in the conventional example, even if there is a large noise in the image search area, it is not affected. For example, even when a product to be edge-extracted has a flaw, a light-dark contrast is weak, or unevenness of reflection luminance is large, an edge of the object can be stably extracted.

  In particular, according to the present invention, even when the object has a shape including a curved portion such as a circle or an ellipse, it is possible to stably extract the edge of the object without using the least squares method. Can be.

  In addition, it is possible to easily realize an inspection as to how accurate the shape of the object from which an edge is extracted according to the present invention is, that is, whether the shape is normal or defective as a circle. it can.

  The present invention is also used when the shape of the object to be edge-extracted is other than a circle, when the shape is not known in advance, or when the geometrical expression that defines the shape of the object is known in advance. Thus, the edge of the object in the image search area can be extracted, and the same effect as when the shape of the object to be edge-extracted is circular can be obtained.

  Further, if a rough search process for calculating in advance a candidate point to be a reference point is provided as a pre-process of the above process, the amount of calculation to be performed by the system can be reduced and the speed of the process can be increased.

  Here, an embodiment will be described in which the edge extraction target in the image search area is circular. Note that each process described below can be realized on software on a computer, for example.

  FIG. 2 is a flowchart of an edge extraction method according to an embodiment of the present invention.

  First, in step S101, a reference point is set at an arbitrary position in the image search area. Although the details will be described later, a plurality of reference points are ultimately set. The method of setting the reference point in the image search area, for example, the setting place and the setting interval are not particularly limited to the present invention, and the application place, application, cost, processing time, and other various points of the edge extracting device according to the present invention are various. What is necessary is just to determine in consideration of a factor. For example, when accuracy is required for the edge extraction result, the reference point is set finely (for example, shifted by one pixel), or when cost and processing time are taken into consideration, the reference point installation interval is increased to make the reference point coarse. Just set it. Alternatively, the reference points may be intensively set at a specific place in the image search area.

  After setting the reference point in step S101, the process proceeds by setting a plurality of exploration figures so as to form a concentric circle centered on the reference point. In this embodiment, since the object to be extracted has a circular shape, the search pattern is circular, and a plurality of circular search patterns having different radii are set concentrically around the reference point. For each of the search patterns set concentrically in this way, the brightness projection values along the respective circumferences are integrated to generate a set of integrated values. The details are as follows.

  First, in step S102, the radius of the search pattern is set.

  Next, in step S103, a luminance integrated value along the circumference of the search pattern is calculated.

  Then, in step S104, it is determined whether or not the above-described processing has been completed for all the search patterns for which luminance projection values are to be calculated for the concentric search patterns centered on the reference point set in step S101.

  If the processing has been completed for all the search patterns, the process proceeds to step S105.

  If the process still remains, the process proceeds to step S102, the radius of the search pattern is set to another value, and the process proceeds to step S103. If accuracy is required for the edge extraction result, the radius of the circular search pattern is relatively frequently set so that the space between the concentrically adjacent search patterns becomes narrower (for example, one pixel interval). Just set it. Conversely, if cost and processing time are taken into consideration, it is only necessary to set a relatively coarse interval so that the interval between adjacent search patterns becomes wider.

  With respect to the reference point set in step S101, the process is completed for all the search figures centered on this reference point, whereby a set of the integrated values corresponding to each concentric search figure is obtained. That is, by going through steps S101 to S104, a set of integrated values corresponding to the set reference points is obtained.

  In step S105, a difference value between adjacent search patterns is calculated for each integrated value obtained by integrating the luminance projection values along the circumference of the plurality of search patterns as described above. As a result of this calculation, difference values having various magnitudes are obtained.

  Then, in step S106, a maximum difference value is extracted from the plurality of difference values obtained in step S105, and this value is held. In this specification, this value is referred to as a “maximum difference candidate value”. Through steps S101 to S106 as described above, the maximum difference candidate value corresponding to the set reference point on a one-to-one basis is obtained.

  In step S107, it is determined whether or not the above processing has been completed for all reference points that can be set in the image search area.

  If the processing has been completed for all the reference points, the process proceeds to step S108. If the process still remains, the process proceeds to step S101, the position of the reference point is switched, and the process proceeds to step S102.

  By completing the processing for all assumed reference points in the image search area, that is, through steps S101 to S107, the maximum difference candidate value corresponding to each reference point in the image search area is obtained. Since the edge of the object is a boundary line between brightness and darkness, the maximum difference candidate value that maximizes the difference value of the integrated brightness projection values between adjacent search patterns is obtained only when the edge of the object is searched. This is when the circumference of the figure is closest to the most.

  In step S108, the largest maximum difference candidate value among the plurality of maximum difference candidate values obtained as described above is extracted and held. In this specification, this value is referred to as a “maximum difference value”.

  The fact that the difference value of the integrated luminance projection values between adjacent search patterns is large means that the luminance on the circumference between adjacent search patterns changes greatly, that is, the difference in brightness is large. means. Since the edge of the object is a boundary line between brightness and darkness, the maximum difference candidate value that maximizes the difference value of the integrated brightness projection values between adjacent search patterns is obtained only when the edge of the object is searched. This is when the circumference of the figure is closest to the most. The greater the circumference of the search pattern close to the edge of the object, the greater the maximum difference candidate value. From this, it can be said that the vicinity of the search pattern when the maximum difference value which is the largest difference among the plurality of maximum difference candidate values is indicated is the portion where the probability that the edge of the object exists is the highest.

  Therefore, in step S109, the position and radius of the center of the search pattern at the time of the maximum difference value are specified, and the circular figure defined by the position and radius of the center is determined as the edge of the object.

  FIG. 3 is an explanatory diagram specifically illustrating edge extraction according to the embodiment of the present invention.

  Here, as an example, reference points P1, P2, and P3 are set at positions as illustrated in FIGS. 3A, 3B, and 3C with respect to an object Q existing in the image search area, A case will be described where concentric search figures Sa-i, Sbi and Sc-i (i is a natural number) centered on this reference point are set. In this example, in the object Q, the luminance inside the circle is assumed to be brighter than the luminance outside the circle. In FIG. 3, the intervals between the radii of the circles forming the concentric circles are particularly schematically illustrated.

  Since the edge of the object Q becomes the boundary line of brightness, the maximum difference candidate value that maximizes the difference value of the integrated brightness projection values between adjacent search patterns is obtained only in the circumference of the search pattern. However, this is the case where the portion closest to the edge of the object is the most. In each case of FIGS. 3A to 3C, the maximum difference candidate value is obtained as follows.

  First, a case where the reference point P1 is set so as to have a positional relationship as shown in FIG. An exploration figure indicating the maximum difference candidate value is obtained through the processing of steps S101 to S106 in FIG. 2, but in this example, it is set to Sa-7.

  Next, a reference point P2 is set at a position different from the reference point P1. This state is shown in FIG. The search figure indicating the maximum difference candidate value is, for example, Sb-6.

  Further, a reference point P3 is set at a position different from the reference point P1 or 2. This state is shown in FIG. The search pattern indicating the maximum difference candidate value is, for example, Sc-2.

  Here, the above three reference points are illustrated, but as described above, the position of the reference point Pi (i is a natural number) is relatively shifted to obtain many maximum difference candidate values. Then, the largest one is searched for from the obtained maximum differential candidate values, and this is held as the maximum difference value.

  As described above, the maximum difference value means that the luminance on the circumference between the adjacent search patterns changes most, that is, the difference between light and dark is the largest. In other words, the greater the circumferential portion of the search pattern closer to the edge of the object, the greater the maximum difference candidate value. In the example shown in FIG. 3, a search pattern Sa-7 based on the reference point P1 as shown in FIG. 3A, and a search pattern based on the reference point P2 as shown in FIG. Of Sb-6 and the search figure Sc-2 based on the reference point P3 as shown in FIG. 3C, the search figure Sc-2 based on the reference point P3 is close to the edge of the target object. Since the circumferential part of the search pattern is the largest and approaches almost the entire circumference, the maximum difference candidate value at this time becomes the maximum difference value.

  Therefore, the position and radius of the center of the search pattern Sc-2 at the time of the maximum difference value shown in FIG. 3C are specified, and the circular figure defined by the position and radius of the center is determined. Determined as an edge.

  As described above, according to the present embodiment, in the edge extraction apparatus and method for extracting the edge of the target object in the inspection image, even if the edge extraction target image contains many noise components, The edge of an object can be extracted.

  As described above, according to this embodiment, the direction of the edge is specified without using the least squares method as in the conventional example. Therefore, even if there is a large noise in the image search area, it is not affected.

  In the present embodiment, the difference value is calculated immediately after calculating the brightness integrated value. However, a plurality of brightness integrated values are stored in a memory or the like, and the respective difference values are collectively calculated later to extract the edge. May be used.

  In the above embodiment, the edge extraction in the case where the edge extraction target in the image search area has a circular shape has been described. As a modified example of the present embodiment, it is further inspected whether the circular object from which the edge is extracted is a circle with a certain degree of accuracy, that is, whether the object is normal or defective as a circle. It may be. Hereinafter, this will be described.

  FIG. 4 is a flowchart of an edge extraction method according to a modification of the embodiment of the present invention. FIG. 5 is an explanatory diagram specifically illustrating edge extraction according to a modification of the embodiment of the present invention.

  Here, as shown in FIG. 5, a case will be described in which the accuracy of a target object Q from which a substantially circular edge centered on a reference point P is extracted is inspected. In the present modified example, as shown in FIG. 5, it is assumed that scratches E1 and E2 are provided.

  First, in step S201 of FIG. 4, at least two concentric luminance test figures centered on the center position specified in advance according to the present invention are set so as to sandwich the edge extracted according to the present invention. In the example shown in FIG. 5, the luminance test graphics T1 and T2 around the reference point P are set so as to sandwich the edge of the object Q. In addition, the accuracy of the inspection can be adjusted by changing the radius of the luminance inspection graphic.

  Next, in step S202, for each luminance test graphic, a luminance projection value is measured and held at predetermined intervals (hereinafter, referred to as "sampling intervals") along the circumference of each luminance test graphic. In the example shown in FIG. 5, for each of the luminance test graphics T1 and T2, a luminance projection value is measured along the circumference of the luminance test graphic in units of a sampling interval, and is held.

  Next, in step S203, it is determined whether or not the luminance projection value measured in step S202 is included in a predetermined range for each luminance test graphic. As described above, the edge is the boundary line between the brightness of the object and the brightness, and if there is no flaw on the circumference, the brightness projection value measured in the unit of the sampling interval is substantially constant. However, if there is a flaw, the measured luminance projection value fluctuates. Therefore, in step S203, the presence or absence of a flaw is checked by determining whether the measured luminance projection value falls within a predetermined range. In the example shown in FIG. 5, it is determined whether or not the luminance projection values measured for each of the luminance test graphics T1 and T2 fall within a predetermined range. Further, by changing the size of the predetermined range, the accuracy of the inspection can be adjusted.

  If the measured luminance projection value falls within the predetermined range, it is determined that there is no unacceptable flaw, and it is determined to be normal. At this point, the inspection processing according to the present modification ends.

  If the measured luminance projection value is not included in the predetermined range, it means that an unacceptable flaw exists, and the process proceeds to step S204.

  In step S204, the number of luminance projection values not included in the predetermined range is counted, and this is held as the number of errors. In the example shown in FIG. 5, the number of errors is counted as “1” for both the luminance test graphics T1 and T2.

  Next, in step S205, it is determined whether or not the number of errors is greater than a predetermined number. If the number of errors is smaller than the predetermined number, the object is determined to be normal. If the number of errors is larger than the predetermined number, the object is determined to be defective. By changing the “predetermined number”, the accuracy of the inspection can be adjusted.

  As described above, according to the present modification, the accuracy of the shape of the target object from which the edge is extracted according to the present invention is a circle, that is, whether the shape is a normal circle or a defect. Can be inspected.

  Note that, in the above-described embodiment and its modifications, the case where the shape of the object to be edge-extracted is a circular shape is described. It is possible to extract the edge of the object. If the shape of the object to be edge-extracted is other than circular and the shape is not known in advance, information about the approximate contour of the object to be edge-extracted is obtained in advance before the edge extraction process. I do. Based on this information, a reference search pattern that is a similar reference for setting a plurality of search patterns is generated.

  In the case where the shape of the object to be edge-extracted is not known in advance, the general outline of the object to be edge-extracted may be obtained by using a generally used contour line extraction method. Here, as one specific example, a tracking start point for contour tracking is set, and in the process of tracking the contour, tracking is performed while attaching a tracked mark to a point at which tracking has been completed. A method for extracting a line will be briefly described.

  7 to 9 are diagrams illustrating an example of a method of extracting a contour line of a figure. A case will be described in which an approximate contour line of the figure A existing in the binarized search area S as shown in FIG. 7 is extracted. Here, in order to impress the explanation, one pixel in which the figure A is present is expressed as “black”, but is indicated by one hatched box in FIGS. One pixel where the figure A does not exist is expressed as “white”.

  It is assumed that the search area S is raster-scanned and a pixel represented by black is first encountered. The pixel at this time is labeled as "X0" as shown in FIG. 7, and this is set as the tracking start point.

  Pixels in the vicinity of 8 viewed from the start point X0 at the center are as shown in FIG. At this time, it is checked whether or not there is a pixel represented in black one by one in the numerical order as shown in FIG. In the case of FIG. 8A, since the second checked pixel is black, this pixel is labeled “X1”.

  Pixels in the vicinity of 8 viewed at the center of X1 are as shown in FIG. At this time, it is checked whether or not there is a pixel represented in black one by one in the numerical order as shown in FIG. In the case of FIG. 8B, since the pixel checked third is black, this pixel is labeled “X2”.

  Pixels in the vicinity of 8 viewed at the center of X2 are as shown in FIG. At this time, it is checked whether or not there is a pixel represented in black one by one in the numerical order as shown in FIG. In the case of FIG. 9A, since the pixel checked sixth is “1”, this pixel is labeled “X3”.

  The process described above is repeated until a pixel already labeled as X0 is detected again as a “black pixel”. When the resulting X0 to X6 follow the labeled pixels, the outline of FIG. A is obtained.

  The outline obtained as described above is used as information about the approximate outline of the object to be edge-extracted, and based on this information, a reference exploration pattern serving as a similar reference for setting a plurality of exploration figures is obtained. Generate.

  As described above, when the shape of the object to be edge-extracted is other than a circle and the shape of the object to be edge-extracted is not known in advance, for example, the shape of the graphic described with reference to FIGS. A reference search pattern is generated based on the contour line extraction method. If the shape of the object to be edge extracted is known in advance, the reference search pattern is generated based on an equation defining the shape.

  Note that, even when the shape of the object to be edge-extracted is not circular and the shape of the object to be edge-extracted is known in advance, the edge of the object in the image search area is extracted using the present invention. It is possible. For example, this corresponds to a case where the shape of the object can be defined by an equation representing an ellipse, a polygon, or the like. Based on this equation, a reference search pattern serving as a similar reference for setting a plurality of search patterns is generated.

  A plurality of search patterns similar to the reference search pattern are obtained based on the above-described reference search pattern, and the subsequent processing follows the flowchart shown in FIG. That is, first, in step S101, a reference point is set at an arbitrary position in the image search area. Next, in step S102, a plurality of search patterns similar to the above-described reference search pattern are set based on the reference point set in step S101. Then, steps S102 to S104 are repeated, and for each search pattern, the luminance projection values along the outer periphery of the search pattern are integrated to obtain a set of integrated values. Next, in step S105, a difference value of the integrated value between the adjacent search figures is calculated. Next, in step S106, the maximum difference value is extracted from the set of difference values calculated in step S105, and this is stored as the maximum difference candidate value at the reference point. Then, steps S101 to S107 are repeated, and in step S108, the largest maximum difference candidate value is extracted from the maximum difference candidate values obtained by switching the position of the reference point set in step S101 within the image search area. Then, this is held as the maximum difference value. Then, in step S109, the search figure at the time of the maximum difference value is determined as the edge of the object.

  When the shape of the object whose edge is to be extracted as described above is other than circular, it is necessary to correct the inclination between the object whose edge is to be extracted and the search pattern. For example, at least two alignment marks are provided on a product in which an object to be edge-extracted is present, and based on the alignment marks, the deviation of the angle from the search pattern on the virtual plane is measured, and the inclination may be corrected. .

  Note that the system block configuration corresponding to the above-described flowchart is as described with reference to FIG. As a result, it is possible to obtain the same effect as when the shape of the object to be edge-extracted is circular.

  In addition, similarly to the case where the object described above is circular, the degree of accuracy of the object whose edge is extracted as described above, that is, whether the object is normal or defective as a product May be inspected. This process follows the flowchart shown in FIG. That is, first, in step S201, at least two luminance test figures similar to the object based on the position of the reference point specified in step S109 in FIG. 2 are sandwiched by the edge determined in step S109 in FIG. Set. Next, in step S202, a luminance projection value is measured and held at predetermined intervals along the outer periphery of each luminance test graphic for each luminance test graphic. Next, in step S203, it is determined whether or not the luminance projection value measured in step S202 is included in a predetermined range for each luminance test graphic. In step S204, the number of luminance projection values not included in the predetermined range is counted, and this is held as the number of errors. In step S205, it is determined whether the number of errors is greater than a predetermined number. If the number of errors is larger than a predetermined number, the object is determined to be defective.

  As described above, according to the above-described methods, a plurality of maximum difference candidate values are obtained by shifting the position of the search pattern, and the search pattern indicating the maximum maximum difference candidate value is determined as the edge of the target object. . When accuracy is required for the edge extraction result, it is preferable to set the position of the reference point of the search pattern finely (for example, so as to be shifted by one pixel). Next, as a modified example of this method, a rough search process of calculating a candidate point to be a reference point in advance will be described. This rough search process is executed as a pre-process of the above process in order to reduce the amount of calculation to be performed by the system and to speed up the process.

  In this modified example, if the shape of the object to be edge-extracted is circular, two concentric coarse search graphics having a predetermined radius difference centered on the coarse search reference point, If it is not known, the rough search reference point set at an arbitrary position in the image search area is used as a reference, and two different sizes of the outer peripheral figure and the inner peripheral figure similar to the object are used. A candidate point to be a reference point is determined using the rough search graphic. That is, in the present modification, the approximate position of the target object from which the edge is to be extracted is grasped in advance in such a manner that a candidate point to be a reference is determined. Since the already described edge extraction processing of the present invention is executed only for a plurality of determined candidate points, the amount of computation to be processed is reduced, and the processing speed can be increased.

  FIG. 10 is an explanatory diagram exemplifying a coarse search graphic in the case where the shape of the object to be edge-extracted is circular according to a modification of the embodiment of the present invention.

  When the shape of the target object whose edge is to be extracted is circular, the coarse search graphic has two concentric circles having a predetermined radius difference centered on the coarse search reference point R. In FIG. 10, the outer circumference circle is Uo, the inner circumference circle is Ui, and the respective radii are ro and ri. Each of the outer circumference Uo and the inner circumference Ui has a width d (= ro-ri, hereinafter referred to as a “radius difference”) between the circumferences.

  FIG. 11 is an explanatory diagram (part 1) illustrating the principle of the coarse search process in the case where the shape of the object to be edge-extracted is circular according to a modification of the embodiment of the present invention. In FIG. 11, the luminance projection value on the circumference of the outer and inner circles of the coarse search graphic when the coarse search reference point of the coarse search graphic is moved (shifted) from R1 to R2 by the distance k is considered. .

  When d <k as shown in FIG. 11A, the luminance projection value cannot be detected for the hatched area in the figure. That is, if an object to be edge extracted exists in the hatched area, the object cannot be detected, and normal edge extraction cannot be performed.

  On the other hand, when d = k as shown in FIG. 11B and when d> k as shown in FIG. 11B, it is possible to detect the luminance projection value over the entire area. it can. Therefore, in the present modification, the coarse search processing is executed by shifting the position of the coarse search reference point by a distance k that satisfies the condition d ≧ k in the image search area.

  FIG. 12 is an explanatory diagram (part 2) illustrating the principle of the coarse search process in the case where the shape of the target object whose edge is to be extracted is circular, according to a modification of the embodiment of the present invention. In FIG. 12, a case is considered where the coarse search reference point of the coarse search graphic is shifted by a distance k that satisfies the condition of d ≧ k, and the approximate position of the object Q is searched.

  There is a brightness difference between the area inside the object Q and the other area with the edge of the object Q as a boundary. When the brightness projection values along the outer circumference and inner circumference of the coarse search figure are integrated to obtain the outer circumference integration value and the inner circumference integration value, and the outer-internal difference value that is the difference between the outer circumference integration value and the inner circumference integration value is calculated. When a rough search graphic exists at the position shown in FIG. 12C, the outer / internal difference value becomes maximum, and the outer / internal difference value becomes larger than in the case of FIGS. 12A and 12B. That is, as shown in FIG. 12C, when the edge of the object Q exists between the outer circumference and the inner circumference of the rough search graphic, the outer-internal difference value becomes maximum. This means that when the external / internal difference value is the maximum, the coarse search reference point of the coarse search graphic exists near the center point of the object Q. In other words, the external / internal difference value is the maximum. If the coarse search reference point of the coarse search graphic is found, it means that the approximate position of the object to be edge-extracted can be grasped in advance. Thereafter, if the above-described edge extraction processing of the present invention is performed only on the determined plurality of candidate points, the amount of computation to be processed is reduced, so that the processing can be sped up.

  FIG. 13 is an explanatory diagram illustrating the principle of the coarse search processing when the shape of the object to be edge-extracted is other than a circle according to a modification of the embodiment of the present invention. This figure shows a case of a triangle as an example of a shape other than a circle, but is similar in principle to the case of FIG.

  FIG. 14 is a flowchart of a rough search process according to a modification of the embodiment of the present invention.

  First, in step S301, a coarse search reference point that is the center of two concentric coarse search figures having a predetermined radius difference is set at an arbitrary position in the image search area. Next, in step S302, with respect to the outer circumference and the inner circumference of the rough search graphic, the luminance projection values along the circumference are integrated to obtain an outer circumference integrated value and an inner circumference integrated value. Next, in step S303, an outer / internal difference value which is a difference between the outer peripheral integrated value and the inner peripheral integrated value is calculated. Steps S301 to S304 are repeated, and a plurality of outer / inner difference values obtained in step S303 by shifting the position of the coarse search reference point set in step S301 by a distance k (d ≧ k) less than or equal to the radius difference d in the image search area. Is extracted. With reference to a plurality of outer / inner difference values obtained by repeating steps S301 to S304, the coarse search reference point of the rough search graphic from which the outer / internal difference value is obtained is set to the reference point used in step S101 in FIG. Determine as candidate points to be made.

  For the coarse search graphic having the coarse search reference point in step S301, the outer radius of the coarse search graphic is set so that the radius of the object becomes a value within the range between the outer radius and the inner radius of the coarse search graphic. Set the radius and inner radius.

  When the shape of the object to be edge-extracted is other than the above-described circular shape or when the shape is not known in advance, the coarse search process including the coarse search reference point in step S301 is performed. May be set with reference to a coarse search reference point set at an arbitrary position in the image search area, and to be composed of two outer peripheral figures and inner peripheral figures different in size similar to the object. That is, the coarse search criteria for the outer and inner peripheral figures of the coarse search graphic are set such that the size of the object is larger than the inner peripheral graphic and smaller than the outer peripheral graphic of the coarse search graphic similar to the target object. What is necessary is just to set the distance from the point. The subsequent processing is the same as the operation principle in the case of the circular shape described with reference to FIG.

  As an edge extraction apparatus and method for extracting an edge of an object in an image search area, an edge of an object can be stably extracted even in an edge extraction target image containing a large amount of noise components.

  For example, even when a product to be edge-extracted has a flaw, a light-dark contrast is weak, or unevenness of reflection luminance is large, an edge of the object can be stably extracted.

  Further, it is also possible to easily realize an inspection as to how accurate the shape of the object from which the edge is extracted is, that is, whether the shape is normal or defective as a circle.

FIG. 2 is a basic block diagram of an edge extraction device according to the present invention. 5 is a flowchart of an edge extraction method according to an embodiment of the present invention. FIG. 4 is an explanatory diagram specifically illustrating edge extraction according to an embodiment of the present invention. 9 is a flowchart of an edge extraction method according to a modification of the embodiment of the present invention. It is explanatory drawing which illustrates in detail the edge extraction by the modification of the Example of this invention. FIG. 9 is a diagram illustrating an edge extraction method according to a conventional example. FIG. 9 is a diagram (part 1) for explaining an example of a contour line extraction method of a figure; FIG. 11 is a diagram (part 2) illustrating an example of a contour line extraction method for a figure. FIG. 11 is a diagram (part 3) for explaining an example of a contour line extraction method for a figure; FIG. 13 is an explanatory diagram illustrating a rough search figure in a case where the shape of an object to be edge-extracted is circular according to a modification of the embodiment of the present invention. FIG. 11 is an explanatory diagram (part 1) illustrating the principle of a coarse search process when the shape of an object to be edge-extracted is circular according to a modification of the embodiment of the present invention. FIG. 14 is an explanatory view (part 2) illustrating the principle of the coarse search processing when the shape of the object to be edge-extracted is a circle according to a modification of the embodiment of the present invention. FIG. 13 is an explanatory diagram illustrating the principle of a coarse search process in a case where the shape of an object to be edge-extracted is other than a circle according to a modification of the embodiment of the present invention. It is a flowchart of a rough search process in a modification of the embodiment of the present invention.

Explanation of reference numerals

DESCRIPTION OF SYMBOLS 1 ... Edge extraction device 11 ... 1st means 12 ... 2nd means 13 ... 3rd means 14 ... 4th means 15 ... 5th means 16 ... 6th means 17 ... 7th means P, P1 , P2, P3: Reference point Q: Object

Claims (26)

An edge extraction method for extracting an edge of a circular object included in an image search area,
A reference point is set at an arbitrary position in the image search area, and adjacent integrated values obtained by integrating luminance projection values along the circumference of a plurality of concentric search patterns around the reference point are adjacent to each other. Calculating a difference value between the search patterns, extracting the maximum difference value and holding this as a maximum difference candidate value at the reference point,
Among the plurality of maximum difference candidate values obtained by shifting the position of the reference point, determining the search pattern when indicating the largest maximum difference candidate value as an edge of the object. An edge extraction method characterized by the following.   2. The edge extraction method according to claim 1, further comprising a step of determining a candidate point to be the reference point. The step of determining a candidate point to be the reference point,
Setting a coarse search reference point at an arbitrary position in the image search area and setting two concentric coarse search graphics having a predetermined radius difference centered on the coarse search reference point;
Obtaining an outer peripheral integrated value and an inner peripheral integrated value by integrating luminance projection values along the outer and inner peripheries of the coarse search graphic;
Calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value,
The position of the coarse search reference point is shifted by a distance equal to or smaller than the predetermined radius difference within the image search area, and the outer search result is referred to an outer-internal difference value. 3. The method according to claim 2, further comprising: determining a candidate point to be a reference point. An edge extraction method for extracting an edge of a circular object included in an image search area,
A first step of setting a reference point at an arbitrary position in the image search area;
A second step of setting a plurality of exploration figures to have a concentric shape centered on the reference point;
A third step of integrating the luminance projection values along the circumference of each of the search patterns to obtain a set of integrated values;
A fourth step of calculating a difference value of the integrated value between adjacent search patterns,
A fifth step of extracting a maximum difference value from the set of difference values calculated in the fourth step, and storing the maximum difference value as a maximum difference candidate value at the reference point;
Of the plurality of maximum difference candidate values obtained in the second, third, fourth and fifth steps by switching the position of the reference point set in the first step within the image search area, A sixth step of extracting the maximum maximum difference candidate value and holding this as the maximum difference value;
A seventh step of specifying the position and radius of the center of the search figure at the time of the maximum difference value, and determining a circular figure defined by the position and radius of the center as the edge of the object. An edge extraction method characterized by the following. An eighth step of setting at least two concentric luminance test graphics centered on the center position specified in the seventh step so as to sandwich the edge determined in the seventh step;
A ninth step of measuring and holding a luminance projection value at predetermined intervals along a circumference of each of the luminance test graphics, for each of the luminance test graphics;
A tenth step of determining whether or not the luminance projection value measured in the ninth step is included in a predetermined range for each of the luminance test graphics;
An eleventh step of counting the number of the luminance projection values that are not included in the predetermined range, and holding this as an error number;
A twelfth step of determining whether the number of errors is greater than a predetermined number;
The edge extracting method according to claim 4, further comprising: if the error number is larger than a predetermined number, determining that the object is defective. A fourteenth step of setting a coarse search reference point at the center of the two concentric coarse search graphics having a predetermined radius difference at an arbitrary position in the image search area;
A fifteenth step of, for the outer periphery and the inner periphery of the coarse search graphic, integrating luminance projection values along the circumference to obtain an outer periphery integrated value and an inner periphery integrated value;
A sixteenth step of calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value;
A plurality of outer and inner difference values obtained by the fifteenth and sixteenth steps by shifting the position of the coarse search reference point set in the fourteenth step by a distance equal to or less than the predetermined radius difference in the image search area; A seventeenth step of extracting
Referring to the outer / inner difference value obtained in the seventeenth step, a coarse search reference point of the coarse search graphic from which the outer / internal difference value is obtained is determined as a candidate point to be the reference point. The edge extraction method according to claim 4, further comprising:   In the fourteenth step, the radius of the outer periphery and the inner radius of the coarse search graphic are set so that the radius of the object becomes a value within a range between the outer radius and the inner radius of the coarse search graphic. The edge extracting method according to claim 6, further comprising a step of setting a radius. An edge extraction method for extracting an edge of an object included in an image search area,
For an integrated value obtained by taking a reference point at an arbitrary position in the image search area and integrating luminance projection values along the outer perimeter of a plurality of search patterns similar to the object based on the reference point, Calculating a difference value between the adjacent search patterns, extracting the maximum difference value, and holding this as a maximum difference candidate value at the reference point;
Among the plurality of maximum difference candidate values obtained by shifting the position of the reference point, determining the search pattern when indicating the largest maximum difference candidate value as an edge of the object. An edge extraction method characterized by the following.   The edge extraction method according to claim 8, further comprising: generating a reference search pattern serving as a similar reference for setting the plurality of similar search patterns.   9. The edge extraction method according to claim 8, further comprising the step of determining a candidate point to be the reference point. The step of determining a candidate point to be the reference point,
A step of setting a coarse search graphic made up of two different-sized outer peripheral figures and inner peripheral figures similar to the object with reference to a coarse search reference point set at an arbitrary position in the image search area; When,
Obtaining the outer peripheral integrated value and the inner peripheral integrated value by integrating the luminance projection values along the outer peripheral graphic and the inner peripheral graphic of the coarse search graphic;
Calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value,
Referring to an outer / inner difference value obtained by shifting the position of the coarse search reference point within the image search area by a distance equal to or less than the difference value of the distance between the outer peripheral figure and the inner peripheral figure from the coarse search reference point. The method according to claim 8, further comprising: determining a candidate point to be the reference point from the coarse search reference points. An edge extraction method for extracting an edge of an object included in an image search area,
A first step of setting a reference point at an arbitrary position in the image search area;
A second step of setting a plurality of exploration figures similar to the object based on the reference point,
A third step of, for each of the search patterns, integrating luminance projection values along the outer periphery of the search pattern to obtain a set of integrated values;
A fourth step of calculating a difference value of the integrated value between adjacent search patterns,
A fifth step of extracting a maximum difference value from the set of difference values calculated in the fourth step, and storing the maximum difference value as a maximum difference candidate value at the reference point;
The position of the reference point set in the first step is switched within the image search area, and the maximum difference candidate value obtained in the second, third, fourth, and fifth steps is selected. A sixth step of extracting the maximum difference candidate value that becomes, and holding this as the maximum difference value;
A seventh step of determining a search pattern at the time of the maximum difference value as an edge of the object. Eighth step of setting at least two luminance test graphics similar to the object based on the position of the reference point specified in the seventh step so as to sandwich the edge determined in the seventh step When,
A ninth step of measuring and holding a luminance projection value at predetermined intervals along an outer periphery of each of the luminance test graphics,
A tenth step of determining whether or not the luminance projection value measured in the ninth step is included in a predetermined range for each of the luminance test graphics;
An eleventh step of counting the number of the luminance projection values that are not included in the predetermined range, and holding this as an error number;
A twelfth step of determining whether the number of errors is greater than a predetermined number;
13. The method according to claim 12, further comprising: when the number of errors is larger than a predetermined number, determining that the object is defective.   13. The edge extraction method according to claim 12, wherein the second step includes a step of generating the search pattern similar to the object based on a geometrical expression defining the object.   14. The edge extraction method according to claim 12, further comprising: generating a reference search pattern serving as a similar reference for setting the plurality of similar search patterns. A fourteenth step of setting a coarse search reference point, which is a coarse search graphic reference made up of two different-sized outer peripheral graphics and inner peripheral graphics similar to the object, at an arbitrary position in the image search area;
A fifteenth step of integrating luminance projection values along the outer peripheral figure and the inner peripheral figure of the coarse search figure to obtain an outer peripheral integrated value and an inner peripheral integrated value;
A sixteenth step of calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value;
The position of the coarse search reference point set in the fourteenth step is shifted by a distance equal to or less than a difference value of a distance from the coarse search reference point of the outer peripheral graphic and the inner peripheral graphic in the image search area. A seventeenth step of extracting a plurality of outer and inner difference values obtained by the fifteenth and sixteenth steps;
Referring to the outer / inner difference value obtained in the seventeenth step, a coarse search reference point of the coarse search graphic from which the outer / internal difference value is obtained is determined as a candidate point to be the reference point. 18. The edge extraction method according to claim 15, further comprising:   The fourteenth step is such that the size of the rough search graphic is larger than the inner circumferential graphic and smaller than the outer circumferential graphic of the rough search graphic similar to the object. 17. The edge extraction method according to claim 16, further comprising the step of setting distances of the outer peripheral figure and the inner peripheral figure from the coarse search reference point. An edge extraction device for extracting an edge of a circular object included in an image search area,
First means for setting a reference point at an arbitrary position in the image search area;
A second means for setting a plurality of exploration figures so as to have a concentric shape centered on the reference point;
Third means for integrating the luminance projection values along the circumference of each of the exploration figures to obtain a set of integrated values;
Fourth means for calculating a difference value of the integrated value between adjacent search patterns,
A fifth means for extracting a maximum difference value from a set of difference values calculated by the fourth means, and holding the same as a maximum difference candidate value at the reference point;
Among the plurality of maximum difference candidate values obtained by the second, third, fourth and fifth means by switching the position of the reference point set by the first means within the image search area, Sixth means for extracting a maximum maximum difference candidate value and holding this as a maximum difference value;
And a seventh means for determining the search pattern at the time of the maximum difference value as an edge of the object. Eighth means for setting at least two concentric luminance test figures centered on the center position specified by the seventh means so as to sandwich the edge determined by the seventh means,
Ninth means for measuring and holding a luminance projection value at predetermined intervals along a circumference of each of the luminance test graphics, for each of the luminance test graphics;
A tenth means for determining whether or not the brightness projection value measured by the ninth means is included in a predetermined range for each of the brightness test graphics;
An eleventh means for counting the number of the luminance projection values not included in the predetermined range and holding this as an error number;
Twelfth means for determining whether the number of errors is greater than a predetermined number,
19. The edge extracting apparatus according to claim 18, further comprising: thirteenth means for determining that the object is defective when the number of errors is larger than a predetermined number. Fourteenth means for setting a coarse search reference point at the center of the two concentric coarse search graphics having a predetermined radius difference at an arbitrary position in the image search area;
A fifteenth means for obtaining an outer peripheral integrated value and an inner peripheral integrated value by integrating luminance projection values along the circumference with respect to the outer periphery and the inner periphery of the coarse search graphic;
Sixteenth means for calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value,
A plurality of outer / inner difference values obtained by the fifteenth and sixteenth means by shifting the position of the coarse search reference point set by the fourteenth means by a distance equal to or less than the predetermined radius difference within the image search area. A seventeenth means for extracting
Referring to the outer / inner difference value obtained by the seventeenth means, a coarse search reference point of the coarse search graphic from which the outer / internal difference value is obtained is determined as a candidate point to be the reference point. The edge extraction device according to claim 18, further comprising:   The fourteenth means may be configured such that the radius of the object is a value within a range between the outer radius and the inner radius of the coarse search graphic, and the outer radius and the inner radius of the coarse search graphic are 21. The edge extraction device according to claim 20, further comprising: means for setting a radius. An edge extraction device for extracting an edge of an object included in an image search area,
First means for setting a reference point at an arbitrary position in the image search area;
A second means for setting a plurality of exploration figures similar to the target object, based on the reference point,
Third means for each of the search patterns, integrating luminance projection values along the outer periphery of the search pattern to obtain a set of integrated values;
Fourth means for calculating a difference value of the integrated value between adjacent search patterns,
A fifth means for extracting a maximum difference value from a set of difference values calculated by the fourth means, and holding the same as a maximum difference candidate value at the reference point;
The position of the reference point set by the first means is switched in the image search area, and the maximum difference candidate value obtained by the second, third, fourth, and fifth means is selected. A sixth means for extracting a maximum difference candidate value that becomes, and holding this as a maximum difference value;
And a seventh means for determining the search pattern at the time of the maximum difference value as an edge of the object. Eighth means for setting at least two luminance test figures similar to the object based on the position of the reference point specified by the seventh means so as to sandwich the edge determined by the seventh means When,
Ninth means for measuring and holding a luminance projection value at predetermined intervals along an outer periphery of each of the luminance test graphics, for each of the luminance test graphics,
A tenth means for determining whether or not the brightness projection value measured by the ninth means is included in a predetermined range for each of the brightness test graphics;
An eleventh means for counting the number of the luminance projection values not included in the predetermined range and holding this as an error number;
Twelfth means for determining whether the number of errors is greater than a predetermined number,
23. The edge extracting apparatus according to claim 22, further comprising: thirteenth means for determining that the object is defective when the number of errors is larger than a predetermined number.   24. The edge extracting apparatus according to claim 22, further comprising: means for generating a reference search pattern serving as a similar reference for setting the plurality of similar search patterns. Fourteenth means for setting a coarse search reference point, which is a coarse search graphic reference made up of two different-sized outer peripheral figures and inner peripheral figures similar to the object, at an arbitrary position in the image search area,
A fifteenth means for integrating the luminance projection values along the outer peripheral figure and the inner peripheral figure of the rough search figure to obtain an outer peripheral integrated value and an inner peripheral integrated value;
Sixteenth means for calculating an outer-internal difference value that is a difference between the outer peripheral integrated value and the inner peripheral integrated value,
The position of the coarse search reference point set by the fourteenth means is shifted in the image search area by a distance equal to or less than a difference value of a distance from the coarse search reference point of the outer peripheral figure and the inner peripheral figure. A seventeenth means for extracting a plurality of outside and inside difference values obtained by the fifteenth and sixteenth means;
Referring to the outer / inner difference value obtained by the seventeenth means, a coarse search reference point of the coarse search graphic from which the outer / internal difference value is obtained is determined as a candidate point to be the reference point. 23. The edge extraction device according to claim 22, further comprising:   The fourteenth means may be configured so that the size of the target object is larger than the inner peripheral figure and smaller than the outer peripheral figure of the coarse search graphic similar to the target object. 26. The edge extracting method according to claim 25, further comprising: means for setting a distance between the outer peripheral figure and the inner peripheral figure from the coarse search reference point.

Images