JPS6324473A - Graphic recognizing device - Google Patents

Abstract

PURPOSE:To detect a corner desired to be detected by separating from a noise component highly accurately to a graphic in which different corners are mixed by obtaining the difference between adjacent contour point angles to an entire contour point angle string and obtaining the sum of contour point angle different strings existing between a zero string and a next zero string of the contour point angle difference strings. CONSTITUTION:By making a point S0 on the contour a start point, an angle theta0 formed by a segment V0 directed to a point E0 separated by (m) in the number of the contour points from the S0 and a horizontal line is obtained by the calculation from the coordinate values of the two points of the S0, and the E0. Then, an angle theta1 formed by a segment V1 directed to an E1 (m) separated in the number of the contour points from an S1 separated by (n) from S0 and the horizontal line is calculated. Thereafter, similarly, the calculation of the contour point angle string thetai is carried out throughout all the contour points. Then, by using the contour point angle string, the difference thetai between the adjacent contour point angles thetai and thetai+1 is calculated on all the contour point angle strings. Finally, by the use of the contour point angle difference string, the average angle difference of an L point in the vicinity of the contour point angle difference thetai is calculated on all the contour point angle difference strings and this contour point angle difference average string is defined to be a contour point angle difference function.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a figure recognition device that recognizes the shape of a figure by extracting contour features (corner parts) of the figure captured in a silhouette image. .

In recent years, figure recognition devices have been used as inspection devices in the FA field.
Although it has been introduced into component mounting machines, etc., there is a need for a shape recognition device that can perform higher precision and faster processing. One method that can satisfy such requirements is the present invention, which performs shape recognition by focusing on contour information of a figure.

First, this conventional recognition method will be briefly explained with reference to the drawings.

FIG. 8 is a block diagram of a recognition system including a figure recognition device. In the figure, 21 is an image signal input unit such as a camera, and 22
is a figure recognition device which processes an image signal outputted from an image signal input section 21 and recognizes a figure. Furthermore, the figure recognition device 22 includes a binarization circuit 23 that binarizes the image signal using an appropriate threshold value, and a contour extraction circuit that extracts outline information of the figure from the binary image created by the binarization circuit 23. Part 24
, a corner detection section 25 that detects corner information of a figure from the contour information extracted by the contour extraction section 24, and a recognition section 26 that recognizes the figure based on the corner information 25. Next, the operation of the contour extraction section 24 will be explained with reference to FIG. 9. Normal contour extraction involves detecting one point on the outline of a figure, such as 8 points 31 in Figure 9, and performing a 3x3 mask operation as shown in 32 in Figure 9, centering on this 8 points 31, to detect the object. A connection direction indicating the direction to the next contour point on the contour of No. 33 is calculated and stored. Next, the position of the next contour point is calculated based on the connection direction, and the 3×3 mask operation described above is performed centering on that position to calculate the connection direction and the position of the next contour point. Thereafter, similar processing is repeated to detect the connecting directions and positions of all contour points.

Next, the operation of the corner information detection section 26 will be explained.

Based on the connection direction of each contour point detected by the contour extraction section, the difference between adjacent connection directions (angles) is calculated, and the tenth
As shown in the figure, a contour point 44 for which an angular difference function 43 in which the horizontal axis 41 is the contour point number and the vertical axis 42 is the difference in connection direction takes an extreme value is taken as a point representing a corner. In other words, points where the connection direction changes significantly indicate that the contour line is not straight but curved, and among these points, where the thick square of the difference in connection direction takes an extreme value, the contour line is the most Indicates a large bend and is generally considered the apex of a corner.

Problems to be Solved by the Invention However, as described above, in the method of calculating the angular difference function of contour points from the connecting direction determined by mask calculation, since the connecting direction changes only in 45° increments, the angular difference function contains errors, making it difficult to detect the exact location. Furthermore, since it calculates the difference in the connection direction of adjacent contour points, it is suitable for detecting very small corners, but on the other hand, even small irregularities such as noise on the contour can be detected as corners. On the other hand, there is a problem in that it is difficult to detect a corner that is a gentle curve with a large bend.

Furthermore, it is difficult to flexibly deal with various objects such as objects with large curves and objects with small curves.

In view of the above-mentioned problems, the present invention is a figure recognition device that detects corners to be detected with high accuracy by separating noise components from noise components in a figure having a mixture of corners of different sizes, and corresponding to many target product stems. It provides:

Means for Solving the Problems In order to solve the above problems, the figure recognition device of the present invention includes:
means for binarizing an image signal; means for detecting a point on the contour of a figure of interest in the binary image; and means for sequentially tracking contour points along the contour starting from one point on the contour. and means for determining a complete sequence of contour points of the figure, means for determining a curvature function of a contour line based on the sequence of contour points, and a straight line extending from the starting point of the contour point tracking to a contour point that is m apart from the contour point. Find the angle θ1 of the straight line from the start point of the contour point tracking to the contour point with the number of contour points m apart from the contour point with the number of contour points of nM. A means for determining a point angle sequence θi and a difference between adjacent contour point angles 01'=θ using the contour point angle sequence θi.
A means for determining i+1-θi for all contour point angle sequences, and a total sum θj of qi′ existing between a sequence where θi′ is zero and the next sequence where θi′ is zero in the contour point angle difference sequence, θj=Σθi′ and the means for determining. In addition, the present invention further includes means for determining the number of contour points forming the corner of θj. Furthermore, the present invention is further provided with means for automatically generating both or one of the parameters m and n, or in addition thereto means for storing the parameters for each type of object.

Effect of the Invention With the above-described configuration, the present invention can obtain the angle of a straight line connecting the m-points and the i-shaped contour points on the contour, which has the effect of smoothing small irregularities on the contour, and smoothing out small irregularities such as noise in the figure. Unevenness can be removed.Furthermore,
The smoothing effect described above increases as the value of m increases.

Similar effects can be obtained even if the value of n is increased. By setting the values of m and n described above to values suitable for the shape of the corner of interest, it becomes possible to detect only the corner of interest and remove unnecessary corners or noise. . Furthermore, as the detected corner information, the point where the average contour point angle difference changes from zero to the next zero, that is, the non-straight portion (corner portion) sandwiched between the straight line portions, is detected, and the Since the sum of the average angle differences of the contour points in the section, that is, the angle of the corner, is calculated, by referring to this value, not only the position of the corner but also the angle of the corner can be calculated. Since the method has a means of calculating the number of contour points to be processed, information regarding the shape of the corner, such as the sharpness of the corner, can also be used, allowing for more accurate corner extraction.

In addition, since appropriate parameters are selected according to the object and stored for each product type, the figure recognition device can be used in actual production sites.

EXAMPLE A figure recognition bag # according to an example of the present invention will be described below with reference to the drawings.

FIG. 1 shows the overall configuration of a figure recognition device according to a first embodiment of the present invention. In FIG. 1, 11 is a recognition target, 12 is a lighting device, 13 is a camera, 14 is a figure recognition device, 15 is a binarization circuit, 16 is a binary image frame memory, 17 is a work memo IJ-118 is a calculation circuit, 1
9 is a contour point detection table. FIG. 2 is a flow diagram of corner detection processing. In the figure, 61 is a contour point extraction process, 52 is a contour point angle calculation process, 53 is a contour point angle difference calculation process, 54 is a contour point angle difference average calculation process, 55 is a corner detection process, and 56 is a corner precision search. It is processing.

The operation of the figure recognition device configured as described above will be explained with reference to the drawings.

In FIG. 1, an object 11 to be recognized is illuminated by an illumination device 12, and its state is converted into a video signal by a camera 13. The output video signal of the camera 13 is input to the figure recognition device 14. The video signal input to the graphic recognition device 14 is converted into a binary video signal by the binarization circuit 15 and then recorded in the frame memory 16. next,
Based on the binary image data recorded in the frame memory 16, the arithmetic circuit 18 stores the contour point data of the figure in the form of a table in the work memory 17 while referring to the contour point detection table 19 (contour point detection table 19). Extraction process 51). Next, based on the contour point data table on the work memory 17, the arithmetic circuit 18 performs corner extraction processing according to a procedure described later, and further performs figure recognition based on the extracted corner information.

Next, the contents of the corner extraction process will be explained in detail.

The corner extraction process can be roughly divided into two parts: a process for determining a contour point angle difference function expressing the curve of the contour line from contour data, and a process for determining a corner portion from the contour point angle difference function. First, in the former process of calculating the angular difference between contour points, as shown in FIG. 3, one point S on the contour is determined. With S as the starting point.

Point E whose contour points are m (m=5 in the figure) away from .

A line segment V heading toward . is the angle θ that it forms with the horizontal. S. .

It is determined by calculation from the coordinate values of the two points Eo. Next, the above S. Contour point S1 located n (n = 10 in the figure) away from
From this, the angle θ1 formed by the horizontal line and the line segment v1 heading toward the point E1, which is m away from Sl by the number of contour points, is calculated. Thereafter, a contour point angle sequence θi is calculated for all contour points in the same manner (contour point angle calculation process 52). Next, using the contour point angle series, the difference (contour point angle difference) 01' between adjacent contour point angles θ1 and θi+1 is calculated for all the contour point angle series (contour point angle difference calculation process 53). Finally, in this embodiment, using the contour point angle difference sequence, L points before and after the contour point angle difference θ1'(θi'-L/2.θi'-L/2+1...
・01'+1...-θi'+L/2-1'01
The average angle difference θi' of ' and /2) is calculated in all contour point angle difference sequences (54), and this contour point angle difference average sequence is used as a contour point angle difference function.

Next, a process for extracting corners using the contour point angle difference average sequence θi' calculated above will be described. FIG. 7 shows a graph of the contour point angle difference average sequence θi' obtained for the object, with the contour point number PN on the horizontal axis.

As can be seen from the name of this function, a place where the average contour point angle difference is zero represents a straight line, and a place where it is not zero represents a curved contour, that is, a corner. Therefore, the area where the contour point angle difference average sequence is from zero to the next zero is a corner portion. In addition, the value Σθ1', which is the sum of the contour point angle difference average θi' between the contour point angle difference average of zero and the next zero, indicates how many times the contour line is beveled at the corner part, so Σθ1 A corner portion can be extracted by detecting a portion where ′ exceeds a certain threshold. At the same time, by calculating the number of contour points that make up the corner, if there are many contour points, it can be determined that the corner curves slowly with a large radius of curvature, which will be described later. (corner detection process 65).

Corners A to F on the object in FIG. 4 correspond to a to f on the graph of the contour point angle difference average sequence 01'. In addition, in this example, in order to easily separate the corner portion and the straight portion, a value P determined from the three values of parameters m, n, and L used to obtain the contour point angle difference averaging sequence described above is calculated. Among the values of contour point angle difference averaging column 01',
If the absolute value is smaller than P, it is approximated to zero, and straight lines with gentle curves or small irregularities are removed as noise.

Furthermore, in this embodiment, a two-stage corner extraction process is performed in order to improve the accuracy of corner extraction. In other words, several sets of parameters m, n, and L are set for calculating the above-mentioned contour point angle difference averaging sequence, and first,
The aforementioned corner extraction process is performed on all the obtained contours using a set of parameters suitable for standard corners. In this process, standard corner parts are extracted correctly, but parts where two corners are close together or corners that curve slowly with a large radius of curvature are not extracted correctly because the parameters are inappropriate. . Figure 5 shows an example of this, with standard corners A and C having a large radius of curvature.
, F, and G are correctly extracted as 90° corners, but due to the large curvature, the E corner is extracted smaller than the actual angle, and the D corner is 90° because the two corners are close.
The two corners are extracted as one 18o0 corner.

Therefore, when there are large curved corners like B and E, or when two corners are close together like D,
Another parameter is set and only that portion is subjected to corner extraction processing (corner detailed search processing 56) again to perform correct corner extraction.

As a condition for performing the fine search process, in FIG. 6, if the contour point angle difference average sequence θi' is the number of contour points from zero to the next zero (ST to ED in the figure), one condition is as follows. It corresponds to the corner of D mentioned above, and A u g
When the value of is between 90° and 180°, the second condition corresponds to the corners of B and E mentioned above, and A n g
/ Fine search processing is performed when CN is smaller than r times the value P determined from the parameters m, n, and L (r is the value determined experimentally), but the parameters used in the above two cases are Since the corners to be used are of different types, different sets are used for each. Furthermore, the contour area to be subjected to the fine search process is between the above-mentioned 5T-ED.

As described above, according to this embodiment, a series of processes can be realized with software, so it can be realized with a relatively simple hardware configuration, and by taking advantage of the flexibility of software,
By adding processing such as averaging processing of the contour point angle difference sequence, noise removal processing, and fine search processing, we were able to improve the corner extraction accuracy.

A second embodiment of the present invention will be described below with reference to the drawings.

FIG. 7 is an overall configuration diagram of a figure recognition device showing a second embodiment of the present invention. In the figure, 11 is the recognition target, 1
2 is a lighting device, 13 is a camera, 14 is a figure recognition device, 1
5 is a binarization circuit, 16 is a binary image frame memory, 17
The work memo IJ-11B is an arithmetic circuit, 19 is a contour point detection table, and the above structure is the same as that shown in FIG. Furthermore, 21 is a parameter set holding circuit, and 20 is a product type table.

The operation of the graphic recognition device configured as described above will be described below.

In FIG. 7, an object 11 to be recognized is illuminated by an illumination device 12, and its state is converted into a video signal by a camera 13. The output video signal of the camera 13 is input to the figure recognition device 14. The video signal input to the graphic recognition device 14 is converted into a binary video signal by the binarization circuit 15 and then recorded in the frame memory 16. Next, based on the binary image data recorded in the frame memory 16,
The arithmetic circuit 18 stores the contour point data of the figure in the form of a table on the work memory 17 while referring to the contour point detection table 19. Next, based on the contour point data table on the work memory 17, corners are determined and the figure is recognized as described in detail in the first embodiment, and the parameter m used at that time.

The values of n and L are values registered in advance in the product type table 20 in a form corresponding to the recognition target object 11.

Next, a method for registering parameters m, n, and L in the product type table 20 will be explained.

In the case of a figure like Figure 5, as already mentioned, B, E
The corner of , the corner of D, and other corners need to be detected by changing the parameters m, n, and L.
A set of these parameters is determined by the following method and registered in the product type table 20. First, corner detection is performed using a standard set of first parameters m, n, and L, and the sum of the average contour point angle differences (Ang) is 9, as in the corner D.
When a corner between 0° and 180° exists, a set of second parameters m, n, and L is registered as the detection parameters of the corresponding corner. Further, when there is a corner such as corners B and E where A n g/CN (CN is the number of contour points) is smaller than a predetermined value, the third parameter m is used as a detection parameter for the corresponding corner.

Register the pair n and L.

As described above, by automatically setting necessary parameters according to the figure and registering them in the product type table 20, corners can be detected using parameters suitable for the product type of the object to be recognized, and the recognition rate can be improved.

Effects of the Invention As described above, the present invention provides means for binarizing an image signal;
means for detecting a point on the contour of a figure of interest in the binary image; and means for sequentially tracing contour points along the contour using the single point on the contour as a starting point to obtain a sequence of all contour points of the figure. means for determining a curvature function of a contour line based on the contour point sequence; and a means for determining an angle θ1 of a straight line toward a contour point whose number of contour points is m apart from the starting point of the contour point tracing; Means for determining an angle θ2 of a straight line from a contour point that is n contour points away from the starting point toward a contour point having m contour points, and similarly determining a contour point angle sequence θi for all contour point sequences; means for calculating the difference θi'-θi+1-θ1 between adjacent contour point angles for all contour point angle sequences using the point angle sequence θi;
The sum θi′ of θi′ that exists between the sequence of zeros and the next sequence of zeros
j=Σθi′, and based on this corner information, the parameter m and n are re-set to be suitable for the corner shape, for corners where the parameters m and n are inappropriate for the corner shape. , the corner extraction process can be performed only on the peripheral portion of the corner, and the corner to be detected can be separated from the noise component and detected with high precision for a figure in which corners of different sizes coexist.

Furthermore, by outputting the contour points of the corners, the sharpness of the corners can be tested.

In addition, by automatically setting the parameter required for corner detection and registering and maintaining it for each object, it is possible to flexibly respond to many types of objects and improve the recognition rate. be able to.

[Brief explanation of the drawing]

゛ Fig. 1 is an overall configuration diagram of a figure recognition device according to a first embodiment of the present invention, Fig. 2 is a flowchart of corner detection processing in the first embodiment of the present invention, and Fig. 3 is a diagram of the first embodiment. Explanatory diagrams of the contour point angle calculation method in Figures 4 and 5.
6 is an explanatory diagram of the corner detection method in the first embodiment, FIG. 7 is an overall configuration diagram of the figure recognition device in the second embodiment of the present invention, and FIG. 8 is an illustration of the conventional figure recognition system. Overall configuration diagram, Figure 9 is an explanatory diagram of the conventional contour point extraction method,
FIG. 1o is an explanatory diagram of a conventional corner detection method. 15...--Binarization circuit, 16---Binary image frame memory, 17...--Work memory, 18-
... Arithmetic circuit, 19... - Contour point detection table, 2o.
...Type table, 51...Contour point extraction processing, 5
2...- Contour point angle calculation process, 53... Contour point angle difference calculation, 54... Contour point angle difference average calculation, 55
. . . corner detection processing, 56 . . . - precise search processing. Name of agent: Patent attorney Toshio Nakao and 1 other person 2nd
Dimensions
...-, 0...al's IQ) Figure 5, Figure 6-11110N-1111

Claims (5)

[Claims] (1) means for binarizing an image signal; means for detecting a point on the contour of a figure of interest in the binary image; and means for sequentially detecting contour points along the contour starting from the one point on the contour. means for determining a complete contour point sequence of the figure; means for determining a curvature function of the contour line based on the contour point sequence; and means for extracting a corner portion on the contour of the figure based on the curvature function. In a figure recognition device equipped with figure recognition means, the angle θ_1 of a straight line toward a contour point that is m contour points away from the start point of contour point tracking (m is an integer) is determined by the combination of the corners. Then, calculate the angle θ_2 of a straight line from the contour point that is n contour points (n is an integer) away from the starting point of contour point tracking to the contour point that is m contour points apart, and then calculate the entire contour point sequence in the same way. The contour point angle sequence θ_
i and the difference between adjacent contour point angles θ_i′=θ_i_+_1−θ_ using the contour point angle sequence θ_i
A curvature function calculation means which calculates i for all contour point angle sequences and uses the contour point angle difference sequence as a curvature function of the contour points, and in the contour point angle discrimination, Total sum of θ_i′ existing between the next string of zeros θ_j = Σθ_
A figure recognition device having corner extraction means for determining i' and outputting the θ_j as corner information. (2) means for binarizing an image signal; means for detecting a point on the contour of a figure of interest in the binary image; and means for sequentially detecting contour points along the contour starting from the one point on the contour. means for determining a complete contour point sequence of the figure; means for determining a curvature function of the contour line based on the contour point sequence; and means for extracting a corner portion on the contour of the figure based on the curvature function. and, in a figure recognition device equipped with means for recognizing a figure based on a combination of corners, an angle θ_1 of a straight line toward a contour point that is m contour points away from the start point of contour point tracking (m is an integer) is determined. , calculate the angle θ_2 of a straight line from the contour point that is n contour points (n is an integer) away from the starting point of the contour point tracking to the contour point that is m contour points apart, and then similarly calculate the contour in the entire contour point sequence. Using means for determining a point angle sequence θ_i and the contour point angle sequence θ_i, the difference between adjacent contour point angles θ_i=θ_i_+_1−θ_i is calculated for the entire contour point angle sequence, and the contour Using the curvature function calculation means which is a curvature function of a point and the contour point angle discrimination, the number l of contour points existing between a row where θ_i' is zero and the next row where θ_i' is zero is calculated, and the number l is calculated using the corner information. A figure recognition device having a corner extraction means for outputting as follows. (3) means for binarizing an image signal; means for detecting a point on the outline of a figure of interest in the binary image; means for determining a complete contour point sequence of the figure; means for determining a curvature function of the contour line based on the contour point sequence; and means for extracting a corner portion on the contour of the figure based on the curvature function. , in a figure recognition device equipped with a means for recognizing figures based on the combination of the corners, find an angle θ_1 of a straight line toward a contour point that is m contour points away from the start of the contour point tracking (m is an integer); Obtain the angle θ_2 of a straight line from the contour point that is n contour points (n is an integer) away from the start point of the contour point tracking to the contour point that is m contour points apart, and similarly calculate the contour in the entire contour point sequence. Find the point angle sequence θ_i, then use the contour point angle sequence θ_i to find the difference between adjacent contour point angles θ_i'=θ_i_+_1−θ_i for the entire contour point angle sequence, and use the contour point angle difference a curvature function calculating means for calculating a curvature function of the contour point; A figure recognition device comprising corner extraction means for determining the number of contour points 2 constituting a corner of θ_j and outputting the θ_j, l as corner information, and means for making the parameters m and n variable and independently setting them. (4) means for binarizing an image signal; means for detecting a point on the contour of the figure of interest in the binary image; and means for sequentially detecting contour points along the contour starting from the one point on the contour. means for determining a complete contour point sequence of the figure; means for determining a curvature function of the contour line based on the contour point sequence; and means for extracting a curved portion on the contour of the figure based on the curvature function. In a figure recognition device equipped with figure recognition means, the angle θ_1 of a straight line toward a contour point that is m contour points away from the start point of contour point tracking (m is an integer) is determined by the combination of the corners. Then, calculate the angle θ_2 of a straight line from the contour point that is n contour points (n is an integer) away from the starting point of contour point tracking to the contour point that is m contour points apart, and then calculate the entire contour point sequence in the same way. The contour point angle sequence θ_i
and the difference between adjacent contour point angles θ_i'=θ_i_+_1−θ_i using the contour point angle sequence θ_i.
curvature function calculation means that calculates for all contour point angle sequences and uses the contour point angle discrimination as a curvature function of the contour point; The sum of θ_i′ existing between the strings of zeros θ_j=Σθ_
i′ and outputting the θ_j as corner information; and determining the output of the corner extraction means;
A figure recognition device comprising means for changing set values of the parameters m and n. (5) means for binarizing an image signal; means for detecting a point on the outline of a figure of interest in the binary image; and means for sequentially detecting outline points along the outline starting from the one point on the outline. means for determining a complete contour point sequence of the figure; means for determining a curvature function of the contour line based on the contour point sequence; and means for extracting a corner portion on the contour of the figure based on the curvature function. In a figure recognition device equipped with figure recognition means, the angle θ_1 of a straight line toward a contour point that is m contour points away from the start point of contour point tracking (m is an integer) is determined by the combination of the corners. Then, calculate the angle θ_2 of a straight line from the contour point that is n contour points (n is an integer) away from the starting point of contour point tracking to the contour point that is m contour points apart, and then calculate the entire contour point sequence in the same way. The contour point angle sequence θ_
i and the difference between adjacent contour point angles θ_i′=θ_i_+_1−θ_ using the contour point angle sequence θ_i
curvature function calculation means that calculates i for all contour point angle sequences and uses the contour point angle discrimination as a curvature function of the contour point; Total sum of θ_i′ existing between the next string of zeros θ_1 = Σθ
corner extraction means for determining _i' and outputting the θ_j as corner information; means for determining the output of the corner extraction means and changing the set values of the parameters m and n; and a means for determining the set parameters m and n. 1. A figure recognition device comprising: means for storing the type of an object together with the type of the object.