JPH0562925B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a method for automatically measuring the side length of a rectangular object, and more specifically to bonding of electronic components such as IC chips and assembly of mechanical components. The present invention relates to a side length measuring method for use in identifying the shape and orientation of a rectangular object, identifying the type of parts, etc.

[Prior Art] In production lines that bond electronic components such as IC chips, assemble mechanical parts, etc., there is often a need to automatically recognize the shape, position, orientation, etc. of objects. The shape of the object to be recognized is very often rectangular. and,
If the shape of the object to be recognized is limited to a rectangle like this, the shape of the object can be recognized simply by measuring the length of its long side and short side, or the ratio of the lengths of both sides. This makes it possible to measure accurately and in real time.

However, in order to perform sophisticated recognition, conventionally known object shape recognition devices have rather complicated configurations and require time for processing, impairing their practicality.

[Problems to be Solved by the Invention] The present invention limits the shape of the object to be recognized to the above-mentioned rectangle. We provide a method for shape recognition that allows the shape of an object to be recognized simply by measuring the ratio of the objects, thereby making it possible to classify irregularly shaped objects easily, accurately, and in real time. This is what I am trying to do.

[Means for Solving the Problems] In order to achieve the above object, the side length measuring method of the present invention converts a two-dimensional figure pattern obtained by a pattern input device into a binary matrix pattern. , pattern conversion is performed to convert adjacent n×n mesh points on the above matrix pattern into different direction codes depending on which direction the boundary line between the real part and the empty part of the figure is facing. Among the direction codes that have been converted and converted, we focus on the data structure of the outline of the figure pattern, count the frequency of appearance of each direction code obtained according to the posture of the figure, and based on these counts, determine the recognition target. This method is characterized by measuring the side length or side length ratio of a rectangular object.

[Example] The present invention is based on the assumption that the object to be recognized is rectangular.

FIG. 1 shows a flowchart of the image processing process in the present invention. To explain the method of the present invention with reference to the figure, when measuring the side length or side length ratio of a rectangular object to be recognized,
First, the image data to be recognized is captured as a two-dimensional graphic pattern using a pattern input device such as an ITV camera, and the output for each sample point on the scanning line is converted into a binary value corresponding to the brightness or darkness of the sample point. A matrix pattern is obtained by converting the graphic pattern into a binary signal.

In the image processing apparatus to which the matrix pattern is input, first, the area S of the graphic pattern is measured, which is necessary when measuring the side length, which will be described later. To measure this area S, various known means that have been conventionally used in graphic processing can be employed.

Next, the image processing device performs pattern conversion as described below. That is, if we focus on the adjacent 2×2 mesh points on the matrix pattern, the data as shown in FIG. Classified into 16 types of data. Then, by focusing on the data structure of the contour part of the figure pattern and extracting those related to the direction in which the boundary line between the real part and the empty part of the figure is facing, we can find the above. The 16 types can be classified into 8 types as shown in Figure 3.

Note that, in general, by focusing on n×n mesh points and extracting direction-related data, they are classified into the required number corresponding to the above eight types. However, if the value of n is increased, measurement becomes more accurate, but the amount of signal processing increases at an accelerated rate, making it impossible to perform simple processing. Therefore, it is desirable to set the value to a small value within a possible range.

In addition, when the value of n is set to 3 or more, n
From the binarized data structure at mesh points of
Ignoring them along with the real and empty data,
It is sufficient to pay attention only to the data structure of the contour part of the graphic pattern from among the converted direction codes and perform the following arithmetic processing, and even with this, large errors will not occur for each individual.

In the following description, the direction codes attached to the numerical values .about. attached to the eight types of patterns shown in FIG. 3 will be explained.

After sequentially performing such pattern conversion on the above matrix pattern while shifting the mesh points one by one, the image processing device counts the number of each of the above eight types of direction codes, and based on the counted value. Measure the side length, etc.

Therefore, as an example, considering the posture of a rectangle as shown in Figure 4 and the appearance frequency of each direction code, when the rectangle takes the postures A to C in the figure, each side forming the outline of the rectangle The direction code obtained in each section is as added in the figure.

Furthermore, in order to find the appearance frequency of each direction code by calculation, if we consider the configuration of the rectangular outline (solid line part) inclined by the angle θ as shown in Fig. 5 in vector terms, A diagonal component () and a horizontal component () as shown in Figure 4 will appear, and when calculated using this, Figure 4 A to C
Each direction code for a rectangle with the attitude is
They will appear in the following proportions:

●When θ=0°, ;W, ;W ;H, ;H ●When θ>0°, ;W・cosθ−W・sinθ ;W・cosθ−W・sinθ ;H・cosθ−H・sinθ ;H・cosθ−H・sinθ ;W・sinθ ;H・sinθ ;W・sinθ ;H・sinθ ●When θ<0°, ;W・cos|θ|−W・sin|θ| ;W・cos |θ|−W・sin|θ| ;H・cos|θ|−H・sin|θ| ;H・cos|θ|−H・sin|θ| ;H・sin|θ| ;W・sin| θ｜ ;H・sin｜θ｜ ;W・sin｜θ｜ Now, when measuring the above side length, the ratio of the count values of the direction codes whose real parts are orthogonal to each other, for example, the ratio of the count values of the direction codes of the + horizontal component. Count value and +
If the ratio k ₁ of the count value of the direction code of the vertical component of is expressed as k ₁ = +/+ ...(1), then the value of k ₁ is k ₁ = 2H (cos θ - sin θ) / 2W (cosθ−sinθ)=
H/W... is expressed by (2), and from this equation (2), H=k ₁ ·W is obtained.

On the other hand, as mentioned above, the area S of the graphic pattern is determined by a well-known technique, so from the above equation and the relationship S=W・H, the side lengths W and H can be calculated as W=√ ₁ H ＝√ ₁ ... can be obtained by (3).

That is, by determining the value of the parameter _K1 from the count value of the direction code, the side length ratio of the rectangular object can be determined, and by further measuring the area S, the side lengths W and H can be determined. .

Also, if we pay attention to the four types of oblique components (~) among the direction codes obtained by pattern conversion, ●When θ>0°, ●When θ<0°, ;W・sinθ;H・sin|θ|; H・sinθ; W・sin|θ|; W・sinθ; H・sin|θ|; H・sinθ; W・sin|θ|.

Here, considering the parameter k ₂ = +/+ ... (4), which is the ratio of the direction codes of oblique components with different inclination directions, the value of K ₂ when the recognition target takes a positive or negative attitude angle is calculated. When calculating, ●When θ>0° ●When θ<0° K ₂ = H/W, K ₂ = W/H ...(5) The value of K ₂ changes depending on whether θ is positive or negative. However, if the sign of the attitude angle θ is known, this parameter K ₂ can be used to measure the side length.

In addition to the side length, the attitude angle θ of the recognition target can also be easily measured from the count values for each direction code.

To measure this attitude angle θ, the ratio of the number of direction codes of the oblique component to the number of overall direction codes, that is, K ₃ = +++/++++++++
Consider the parameter ……(6). Calculating the value of K ₃ for each posture in Figure 4 A to C: ●When θ=0°, K ₃ =0/2W+2H=0 (=tanθ) ……(7) ●｜θ｜＞0° When, K ₁ = 2Wsin|θ|+2Hsin|θ|/2W(cos|θ|−sin|
θ｜)+2H(cos｜θ｜−sin｜θ｜)+2Wsin｜θ｜+2
Hsin | θ | = (W + H) sin | θ | / (W + H) cos | θ | =
sin|θ|/cos|θ|=tan|θ|...(8)

The value of the parameter _K3 in the above equation (6) is obtained by calculation based on the image data taken in from the pattern input device, and the value of the parameter
Since K ₃ has the relationship of equations (7) and (8),
The attitude angle θ can be determined by |θ|=tan ⁻¹ (k ₁ ) (9).

Figures 6 and 7 show the amount of error when measuring the side lengths W and H of objects with various attitude angles, with a rectangle of 40× It shows the maximum error, average error, and minimum error of the results measured using 40 matrix patterns.

Judging from the results shown in the figure, the amount of error is considered to be relatively large, but the figure pattern is
The amount of error increased because the measurement was performed under very poor conditions of fitting within a 30 x 30 matrix pattern.If the amount of information is increased by increasing the number of divisions of the matrix pattern, the error will increase. It can be expected that the amount will decrease significantly.

In the above, we are considering the eight types of direction codes shown in Figure 3. For example,
By making the direction codes of , , and 4 the same, there can be four types.

[Effects of the Invention] As detailed above, according to the side length measurement method of the present invention, since it is assumed that the recognition target is a rectangle, it is possible to accurately measure the length using extremely simple means and in real time. It is possible to measure the shape, etc. of a target object.

[Brief explanation of the drawing]

Fig. 1 is a flowchart of the image processing process in the method of the present invention, Fig. 2 is an explanatory diagram of data classification at 2 × 2 mesh points for pattern conversion, and Fig. 3 is a direction code for pattern conversion. Figures 4A to C are explanatory diagrams of the relationship between the orientation of a rectangular object to be recognized and the direction code. Figure 5 is the appearance of each direction code in the contour of a rectangle tilted by an angle θ. The explanatory diagrams of frequency, FIGS. 6 and 7, are graphs showing the amount of error when measuring the side lengths W and H of a rectangular object at various attitude angles.

Claims (1)

[Claims] 1. A two-dimensional graphic pattern obtained by a pattern input device is converted into a binary matrix pattern, and for adjacent n×n mesh points on the matrix pattern,
A pattern conversion is performed to convert to a different direction code depending on which direction the boundary line between the real part and the empty part of the figure is facing, and from among the converted direction codes, the data structure of the contour part of the figure pattern is converted. It is characterized by counting the appearance frequency of each direction code obtained according to the orientation of the figure, and measuring the side length or side length ratio of the rectangular object to be recognized based on these counted values. How to measure the side length of a rectangular object.