JPH0664610B2 - Measurement method of posture angle and position of figure pattern by affine inverse transformation - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial application] At present, automation in all processing industries extends to not only processing machines but also member management and supply of members to processing machines. In such automation, delivery of members, so-called material handling, is a key point, and it is necessary to measure the position and orientation of the members by an appropriate means.

The present invention utilizes such an affine inverse transformation to
That is, the present invention relates to a method for measuring the posture angle and position of a graphic pattern.

[Prior Art] An object is imaged as a two-dimensional graphic pattern by a television camera or the like, and compared with a reference pattern when the graphic pattern is at a reference position, to measure an attitude angle and a positional deviation with respect to the reference pattern. Can use the conventionally proposed pattern recognition technology or other various methods. However, those methods have their own recognition and measurement targets, and for example, not only posture angle and position but also shape recognition are possible, and posture angle and position can be detected only in the same figure pattern. Since it has peculiarities such as, it is necessary to select an appropriate method according to the purpose. Moreover, it is desirable that the device be relatively inexpensive, depending on its capabilities.

[Problems to be Solved by the Invention] An object of the present invention is to enable easy discrimination of a large number of graphic patterns, and at the same time, a device for easily adjusting the posture angle / positional deviation of a graphic pattern to be measured from a reference pattern. The purpose is to provide a measurement method that enables measurement.

[Means for Solving Problems] In the measuring method of the present invention for achieving the above object, the detected figure pattern is binarized, the vertices in the contour portion are extracted, and the reference is based on the data. Performing pattern matching by comparing with a pattern, detecting at least two specific vertices having a large mutual interval required for affine inverse transformation as feature point corners, and further comparing by sequentially shifting the vertex positions of both patterns. By detecting the corresponding vertices of both patterns, the posture angle and displacement of the figure pattern with respect to the reference pattern are measured by affine inverse transformation from the coordinate relationship of the corresponding feature point corners of the reference pattern and the figure pattern. It is characterized by.

Embodiments Embodiments of the present invention will be described in detail below with reference to the drawings.

When measuring the posture angle and position of an object placed in the measurement area, first, the target object is captured as a two-dimensional figure pattern by a pattern input device such as a television camera, and output for each sample point on the scanning line. Is converted into a binarized signal according to the lightness and darkness of the sample points by AD conversion, and thereby the figure pattern is binarized to form a matrix pattern.

In the image processing device to which the matrix pattern is input, attention is paid to arbitrary adjacent 2 × 2 mesh points in the matrix pattern, and any data at these four points is converted into binary 0 Or, depending on whether it is 1, the contour of the figure pattern, that is, the data at the boundary line between the real part and the empty part is classified into a required number, and further traced along the contour line of the figure pattern, Detect coordinates.

Since the present inventor has already proposed such a method of detecting vertices in Japanese Patent Publication No. 56-47581, such a method may be used.

On the other hand, the reference pattern when the graphic pattern is at the reference position is appropriately stored in advance in the storage device. It goes without saying that this storage device can store reference patterns for various object shapes that are predicted to be sent into the measurement area of the graphic pattern.

FIG. 1 schematically shows the relationship between the reference pattern 1 and the graphic pattern 2 described above. The posture angle θ and the position t (X, Y) of the graphic pattern 2 with respect to the reference pattern 1 are as shown in FIG. Is defined in.

As for the graphic pattern 2 for which the vertex detection is performed as described above, the posture angle / positional deviation with respect to the reference pattern 1 is measured by affine inverse transformation as described later.
It is necessary to determine which of the data groups of the vertices to the reference pattern 1 stored in the storage device corresponds to'or ', ....

The setting of such a correspondence is called "numbering" here, and the operation will be described in detail below with reference to FIG.

In this numbering, first, as compared with the above-mentioned graphic pattern 2, whether or not there is a reference pattern that matches in the number of vertices, the type of corners and the number of masks in the reference pattern group in the storage device. Make a decision.

Here, with the type of corner, seeing from the inside of the figure,
A vertex with an apex angle of less than 180 ° is called a convex corner, and a vertex with a vertical angle of 180 ° or more is called a concave corner. When the former is indicated by a symbol of 0 and the latter by a symbol of 1, a figure consisting of convex corners is type 0 Type 1 is a figure in which one concave corner is present, and the arrangement of corners is described as 0-1-0, and is a figure in which two concave corners are present between convex corners. Is a type 2, and a graphic in which both convex corners and concave corners include those belonging to type 1 and type 2 is type 3.

Regarding the number of masks, the arrangement of the convex corners and the concave corners is an array (mask 1) described by 1-0-0-1, and this arrangement is included in the arrangement of the corners forming the figure. The number of masks 1 is represented by the number of masks.
Similarly, regarding the number of masks 2, the arrangement of the convex corners and the concave corners is described as 0-1-1-0 (mask 2).
Considering the above, the number of this array is included in the array of corners.

For example, in the case of the reference pattern 1 and the figure pattern 2 shown in FIG. 1, the data are as follows, and as can be seen from these data, there is a matching reference pattern.

Reference pattern Graphic pattern Number of vertices 6 6 Arrangement of vertices 0-0-1-0-0-0 0-1-0-0-0-0-0 Number of mask 1 0 0 Number of mask 2 0 0 In this pattern matching If there is a matching reference pattern, as shown in FIG. 2, it is determined whether or not the rotation operation of the graphic pattern data described later has completed one round. It is determined whether or not the arrangement of vertices and the distance between vertices coincide with each other. If there is no matching pattern, the arrangement of vertices 0-1-0-0-0-0-0 in the figure pattern is sequentially set one The rotation operation is performed by shifting the values to 0-0-1-0-0-0, and the above determination is repeated again.

In this example, since the arrangement of the vertices in the data of both patterns is matched by this rotation operation, if the distances between the vertices match, the vertices of ′ in the figure pattern 2 in FIG. 1 correspond to the vertices in the reference pattern. End the numbering.

In the comparison of the distances between the feature point corners, similar patterns cannot be distinguished from each other when performing pattern matching. Therefore, the pattern matching capability is used as a condition for discrimination.

It should be noted that most of the various types of figures can be classified and identified from the above-described three types of characteristic parameters of the number of vertices, the type of corners, and the number of masks.

Pattern matching is performed using the above-mentioned characteristic parameters, and at least two specific vertices with a large mutual interval required for affine inverse transformation are detected as feature point corners. Should be taken, and the vertex at the farthest position is optimal. In the example of FIG. 1, the characteristic point corners in the reference pattern 1 are indicated by A and B, and the characteristic point corners in the graphic pattern 2 are indicated by A ′ and B ′. The distance between the vertices can be calculated extremely easily based on the coordinates of each vertex.

By using the mask pattern, it is possible to specify, to some extent, the data group of the feature point corners of the reference pattern that is close to the feature point corners of the graphic pattern.

In this way, when the feature point corners required for the affine inverse transformation, that is, the two feature point corners on the reference pattern and the two feature point corners on the graphic pattern corresponding to them are obtained, their coordinate relationships are obtained. From the above, the attitude angle θ and the position ^t (X, Y) of the figure pattern 2 with respect to the reference pattern 1, which are expected measurement values, can be obtained from the affine inverse transformation.

In FIG. 1, assuming that the feature point corners can be associated between the reference pattern 1 and the target pattern 2, that is, in the reference pattern, A = ^t (x ₁ , y ₁ ), B = ^t (x ₂ ,
y ₂ ), and the corresponding feature point corners on the graphic pattern are A ′ = ^t (x ′ ₁ , y ′ ₁ ), B = ^t (x ′
When _{2 y '2)} to the respective amount to be determined is given by the following equation.

Next, the result of the error calculation (computer simulation) when the posture angle and the position are measured by the above method will be described with reference to FIGS. 3 to 5. Mathematically, no error should occur even if the posture angle θ and the position ^t (X, Y) are calculated using the above formula. However, in reality, when processed as a digital image, a quantization error occurs in the variable for calculating the amount described above.

When measuring the posture angle and the positional deviation of the two-dimensional pattern using the affine inverse transformation, the maximum value of the predicted measurement error can be obtained by partially differentiating both sides of the above equation.

FIG. 3 shows that each side V and H of the rectangle is V = H = 20 mm.
The figure shows the posture angle measurement error when the square is set as a figure pattern and it is placed in the measurement area with different angles. Although the maximum value of the predicted measurement error is shown in the same figure, it can be seen that the performance becomes much better as the target object becomes larger.

Figures 4 and 5 show the size of the figure pattern is 20mm x 20.
x when fixing to a square of mm and changing the position to put it
The measurement error in the direction and the y direction is shown. × is the origin, +
It is understood that the measurement accuracy is better when the object to be measured is placed at the origin, where is 5 mm in the x and y directions and o is 10 mm in the x and y directions.

[Effect of the Invention] According to the method of the present invention described in detail above, the posture angle / positional deviation of the graphic pattern to be measured with respect to the reference pattern can be measured by a simple device.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a measuring method according to the present invention, FIG. 2 is a flowchart for explaining numbering in the measuring method, and FIGS. 3 to 5 are graphs for explaining measuring errors. . 1 ... Reference pattern, 2 ... Graphic pattern.

Claims (1)

[Claims] 1. A detected figure pattern is binarized to extract vertices in its contour portion, pattern matching is performed by comparison with a reference pattern based on the data, and at least 2 required for affine inverse transformation. Detecting specific vertices with a large mutual spacing as feature point corners, and comparing the vertices of the two patterns while shifting them sequentially to detect the corresponding vertices of both patterns, and then detect the feature points in the reference pattern. The coordinates (x ₁ , y ₁ ), (x ₂ , y ₂ ) of the corner and the coordinates (x ′ ₁ , y ′ ₁ ) of the corresponding feature point corner in the figure pattern,
From the relationship with (x ′ ₂ , y ′ ₂ ), ^{_{However, L = (x '1 -x}} ' 2) 2 + (y '1 -y' 2)
^2. A method for measuring the posture angle and position of a figure pattern by affine inverse transformation, which measures the posture angle θ and the position X, Y of the figure pattern with respect to the reference pattern by the affine inverse transformation.