JP5277155B2 - Verification device, verification method, and computer program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To suppress calculation costs and to improve collation precision. <P>SOLUTION: A collating device includes: an edge extracting unit for extracting an edge of a pattern of an input graphic; a first feature amount extracting unit for extracting a feature amount regarding an intersection point between a searching straight line extending in an angular direction of each rotating angle and the edge as a first feature amount for a plurality of rotating angles around a predetermined center point to the pattern of the input graphic; a second feature amount extracting unit for extracting a feature amount regarding an intersection point between a searching straight line extending in an angular direction of each rotating angle and the edge as a second feature amount having a different standard from the first feature amount extracting unit for the plurality of rotating angles around the predetermined center point to the pattern of the input graphic; and a collating unit for collating the patterns of the input graphic based on the first feature amount and the second feature amount. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a technique for collating graphic patterns.

  2. Description of the Related Art Conventionally, a technique for collating whether or not a graphic pattern included in an image matches a previously registered graphic pattern has been proposed. For example, Non-Patent Document 1 discloses a technique for performing the above collation by using an r-θ curve. Further, as a technique for performing the above collation, various techniques other than the method using the r-θ curve have been proposed.

Masahiko Taniuchi, “Robot Vision”, Shosodo, p. 108-109

  However, the matching technique using the r-θ curve has a problem that it is difficult to accurately match similar figures. For this reason, when the pattern of the graphic included in the image is similar to the pattern of the graphic registered in advance, there is a case where the pattern is erroneously collated if it matches. Another verification technique has a problem in that the calculation cost required for the verification process is high, so that the process takes time.

  In view of the above circumstances, an object of the present invention is to provide a technique that can improve the accuracy of matching while suppressing calculation costs.

  One aspect of the present invention is a collation device (for example, the collation device 10 in the embodiment), and an edge extraction unit (for example, an edge extraction unit that extracts an edge of a pattern of an input graphic (for example, a target graphic in the embodiment) In the embodiment, the edge extraction unit 100) and a search extending in the angular direction of each rotation angle for each of a plurality of rotation angles (for example, each rotation θ in the embodiment) at a predetermined center point with respect to the pattern of the input graphic A first feature amount extraction unit (for example, the first feature amount extraction unit 101 in the embodiment) that extracts, as a first feature amount, a feature amount related to the intersection of a straight line (for example, the search straight line 70 in the embodiment) and the edge; For the pattern of the input figure, for each of a plurality of rotation angles at a predetermined center point, a feature amount related to the intersection of the search straight line extending in the angular direction of each rotation angle and the edge is A second feature quantity extraction unit (for example, the second feature quantity extraction unit 102 in the embodiment) that extracts the second feature quantity based on a different standard from the first feature quantity extraction unit, and the first feature quantity and the second feature quantity And a matching unit (for example, the matching unit 104 in the embodiment) that performs matching on the pattern of the input graphic.

  One aspect of the present invention is the collation device described above, in which the first feature amount extraction unit sets a first reference point and a first search start point on the search line, and the search line, the edge, Among the intersection points, the intersection point located in the predetermined first search direction from the first search start point and closest to the first search start point is the first intersection point, and the distance between the first intersection point and the first reference point The second feature amount extraction unit sets a second reference point and a second search start point on the search line, and sets the intersection of the search line and the edge. Among them, an intersection point located in a predetermined second search direction from the second search start point and closest to the second search start point is defined as a second intersection point, based on a distance between the second intersection point and the second reference point. Extracting the second feature amount, the first reference point and the second reference point, the first search start point and the Two search starting point, the first search direction and the second search direction, out of, and wherein the at least one pair are different or directions, collating apparatus according to claim 1.

  One aspect of the present invention is the collation apparatus described above, wherein the predetermined center point is a center of the inputted figure.

  One aspect of the present invention is a collation method, in which the collation device extracts an edge of an input graphic pattern, and the collation device has a predetermined center point with respect to the input graphic pattern. A first feature amount extraction step for extracting, as a first feature amount, a feature amount related to an intersection of the search line extending in the angular direction of each rotation angle and the edge for each of a plurality of rotation angles, and a collation device is input For a figure pattern, for each of a plurality of rotation angles at a predetermined center point, a feature amount relating to an intersection of a search straight line extending in the angular direction of each rotation angle and the edge is different from the first feature amount extraction unit. A second feature amount extracting step for extracting as a second feature amount, and a collation device for collating the pattern of the input graphic based on the first feature amount and the second feature amount. .

  One aspect of the present invention is a computer program, an edge extraction step for extracting an edge of an input graphic pattern, and for each of a plurality of rotation angles at a predetermined center point with respect to the input graphic pattern, A first feature amount extracting step for extracting, as a first feature amount, a feature amount related to an intersection of the search straight line extending in the direction of each rotation angle and the edge, and a plurality of predetermined center points with respect to the pattern of the input figure A second feature amount that extracts, as a second feature amount, a feature amount related to the intersection of the search straight line extending in the angular direction of each rotation angle and the edge for each rotation angle of the first feature amount extraction unit A computer program for causing a computer to execute an extraction step and a matching step for matching the pattern of the input graphic based on the first feature value and the second feature value. A gram.

In the present invention, collation is performed based on two feature amounts: a first feature amount and a second feature amount extracted based on a reference different from the first feature amount. Further, processing such as extraction of an edge and detection of an intersection between a search straight line extending in an angular direction of each rotation angle and the edge is processing with low calculation cost in generally performed collation processing. Therefore, according to the present invention, it is possible to improve the accuracy of matching while suppressing the calculation cost.
Further, according to the present invention configured such that the predetermined center point is the center of the inputted figure, it is possible to reduce the influence of the rotation of the inputted figure and improve the accuracy of collation.

It is a schematic block diagram showing the function structure of a collation apparatus. It is a figure showing the specific example of the image data input into an edge extraction part. It is a figure showing the specific example of the edge data produced | generated by the edge extraction part. It is a figure showing the 1st specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 2nd specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 3rd specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 4th specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 5th specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 6th specific example of a 1st feature-value and a 2nd feature-value. It is a figure showing the 7th specific example of a 1st feature-value and a 2nd feature-value. It is a flowchart showing the flow of operation | movement of a collation apparatus. It is a graph showing the specific example of the 1st feature-value extracted about the figure of FIG. 2A and FIG. 2B, and a 2nd feature-value.

  FIG. 1 is a schematic block diagram illustrating a functional configuration of the collation device 10. The collation device 10 includes a CPU (Central Processing Unit) connected via a bus, a memory, an auxiliary storage device, and the like, and executes a collation program to thereby obtain an edge extraction unit 100, a first feature amount extraction unit 101, and a second feature. It functions as an apparatus including the quantity extraction unit 102, the reference feature quantity storage unit 103, and the collation unit 104. All or some of the functions of the verification device 10 may be realized using hardware such as an application specific integrated circuit (ASIC), a programmable logic device (PLD), or a field programmable gate array (FPGA). The collation apparatus 10 has a feature quantity (hereinafter referred to as “reference feature quantity”) extracted in advance from a predetermined figure (hereinafter referred to as “reference figure”), and is included in the input image data. Is checked to see if it matches the reference figure.

  The edge extraction unit 100 accepts input of image data. The image data input to the edge extraction unit 100 is an image of a graphic (hereinafter referred to as “target graphic”) that is a target of verification processing by the verification device 10. The edge extraction unit 100 extracts edges from the input image data by applying an edge extraction filter to the input image data. Specific examples of the edge extraction filter include a Sobel filter, a Prewitt filter, a Kirsch filter, and a Laplacian filter. The edge extraction unit 100 generates data (edge data) representing an edge extracted from the input image, and transmits the data to the first feature amount extraction unit 101 and the second feature amount extraction unit 102.

  The first feature quantity extraction unit 101 and the second feature quantity extraction unit 102 extract the first feature quantity and the second feature quantity from the same edge data, respectively. The first feature value and the second feature value are extracted based on different criteria.

  Processing of the first feature amount extraction unit 101 will be described. First, the first feature quantity extraction unit 101 sets a center point in the edge data. The center point is a point that becomes the center of the rotation angle θ used in the subsequent processing. The center point may be the center of gravity of the target graphic, or may be a point determined by other predetermined criteria. The center of gravity of the target graphic here is, for example, the center of the target graphic if it is a circle, and the intersection of diagonal lines if the target graphic is a regular polygon. In the following description, the center of a circle is used as a specific example of the center point.

  Next, the first feature quantity extraction unit 101 extracts an individual first feature quantity r1 (θ) for each of a plurality of predetermined rotation angles θ about the center point. The individual first feature amount r1 (θ) is a value representing the distance between the first intersection detected for each rotation angle θ and the first reference point provided for each rotation angle θ. A set of individual first feature amounts r1 (θ) for all predetermined rotation angles θ is the first feature amount. For example, the first feature quantity extraction unit 101 may be configured to extract 360 individual first feature quantities r1 (θ) from 1 degree to 360 degrees for each rotation angle of 1 degree.

  Next, a method for calculating the individual first feature amount r1 (θ) for each rotation angle θ will be described. First, the first feature quantity extraction unit 101 provides a first search start point on a straight line (hereinafter referred to as “search straight line”) extending from the center point in the angular direction of the rotational angle θ at a certain rotational angle θ. Next, the first feature amount extraction unit 101 starts searching for the first intersection point in a predetermined direction (hereinafter referred to as “first search direction”) along the search line from the first search start point. The first intersection is an intersection between the edge included in the edge data and the search line, and is an intersection closest to the first search start point among the intersections existing in the first search direction from the first search start point. The first feature amount extraction unit 101 starts a search in the first search direction from the first search start point, and determines the first detected intersection as the first intersection. Then, the first feature amount extraction unit 101 calculates the distance between the first intersection and the first reference point. The calculated value becomes the individual first feature amount r1 (θ) at the rotation angle θ. The first reference point is a point provided according to a predetermined reference on the search line. The first feature quantity extraction unit 101 performs the above processing for every predetermined rotation angle θ, and calculates the individual first feature quantity r1 (θ) for each predetermined rotation angle θ.

  Next, the process of the second feature amount extraction unit 102 will be described. First, the second feature amount extraction unit 102 sets a center point in the edge data. This center point is the same as the center point set by the first feature quantity extraction unit 101. The setting of the center point may be performed by the second feature amount extraction unit 102 based on the same standard as the first feature amount extraction unit 101, or a notification may be received from the first feature amount extraction unit 101 regarding one of the center points. .

  Next, the second feature quantity extraction unit 102 extracts an individual second feature quantity r2 (θ) for each of a plurality of predetermined rotation angles θ about the center point. A plurality of predetermined rotation angles θ in the second feature quantity extraction unit 102 are the same as the rotation angles θ in the first feature quantity extraction unit 101. The individual second feature amount r2 (θ) is a value representing the distance between the second intersection detected for each rotation angle θ and the second reference point provided for each rotation angle θ. A set of individual second feature values r2 (θ) for all predetermined rotation angles θ is the second feature value.

  Next, a method for calculating the individual second feature amount r2 (θ) for each rotation angle θ will be described. First, the second feature amount extraction unit 102 provides a second search start point on a search straight line extending from the center point in the angular direction of the rotation angle θ at a certain rotation angle θ. Next, the second feature quantity extraction unit 102 starts searching for a second intersection point in a predetermined direction (hereinafter referred to as “second search direction”) along the search line from the second search start point. The second search direction in the second feature quantity extraction unit 102 and the first search direction in the first feature quantity extraction unit 101 may be the same direction or different directions. The second intersection point is an intersection point between the edge included in the edge data and the search line, and is an intersection point closest to the second search start point among the intersection points existing in the second search direction from the second search start point. The second feature amount extraction unit 102 starts the search in the second search direction from the second search start point, and determines the first detected intersection as the second intersection. Then, the second feature amount extraction unit 102 calculates the distance between the second intersection and the second reference point. The calculated value is the individual second feature amount r2 (θ) at the rotation angle θ. The second reference point is a point provided according to a predetermined reference on the search line. The second feature quantity extraction unit 102 performs the above processing for every predetermined rotation angle θ, and calculates the individual second feature quantity r2 (θ) for each predetermined rotation angle θ. A set of individual second feature values r2 (θ) for all predetermined rotation angles θ is the second feature value.

  The reference feature amount storage unit 103 is configured using a storage device such as a magnetic storage device or a semiconductor storage device, and stores the reference feature amount in advance. The reference feature amount includes a reference first feature amount and a reference second feature amount extracted in advance from the reference figure. The reference first feature value and the reference second feature value are a set of individual reference first feature values and individual reference second feature values for all predetermined rotation angles θ. The reference feature amount storage unit 103 may store reference feature amounts related to a plurality of reference figures in advance. As will be described later, the first feature value and the second feature value acquisition method (setting method for the first reference point, the second reference point, the first search start point, the second search start point, the search method, etc.) There are a plurality of methods, the method applied when the reference first feature amount and the reference second feature amount of the reference feature amount are extracted from the reference figure, the first feature amount extraction unit 101, and the second feature amount. The method applied when the extraction unit 102 extracts the first feature amount and the second feature amount is the same method.

The matching unit 104 uses the first feature value and the second feature value extracted by the first feature value extraction unit 101 and the second feature value extraction unit 102, and the reference feature value stored in the reference feature value storage unit 103. By comparison, the pattern of the graphic included in the image data input to the edge extraction unit 100 is collated. Specifically, the collation unit 104 calculates a correlation value C _min based on the following Equation 1 using the first feature value, the second feature value, and the reference feature value.

The correlation value C _min includes a first feature amount r1 _n (θ + θ ₀ ) and a second feature amount r2 _n (θ + θ ₀ ) and a reference when the phase difference θ ₀ changes from 0 degrees to 360 degrees with a predetermined width. This is the smallest value among the values representing the correlation between the feature quantity r1 _m (θ) and the reference feature quantity r2 _m (θ).

More specifically, the collation unit 104 operates as follows, for example. First, assuming θ = 0 and θ ₀ = 0, the difference between the individual reference first feature value r1 _m (0) and the individual first feature value r1 _n (0 + 0), and the individual reference second feature value r2 _m (0 ) And the individual second feature amount r2 _n (0 + 0) are calculated and added. Next, the matching unit 104 increases the value of the rotation angle θ up to 360 with a predetermined width while keeping the value of the phase difference θ ₀ as it is. At this time, the collation unit 104 increases the difference between the individual reference first feature value r1 _m (0) and the individual first feature value r1 _n (0 + 0) and the individual reference second value at each rotation angle θ each time it is increased. The difference between the feature amount r2 _m (0) and the individual second feature amount r2 _n (0 + 0) is calculated and added. Then, the matching unit 104 calculates a cumulative value of values calculated at each rotation angle θ. The collation unit 104 stores the calculated value as a correlation value at the phase difference θ ₀ = 0. Next, the collation unit 104 increases the phase difference θ ₀ to 360 with a predetermined width. At this time, the collation unit 104 calculates and stores the correlation value at each phase difference θ ₀ each time it is increased. Then, the collation unit 104 compares the calculated correlation values at each phase difference θ _{0 and} selects the correlation value having the smallest value as the correlation value C _min .

The matching unit 104 compares the correlation value C _min with a preset threshold value. When the correlation value C _min is smaller than the threshold value, the collation unit 104 determines that the target graphic is a graphic having the same pattern as the reference graphic. When a plurality of reference figures and reference feature amounts are stored in the reference feature amount storage unit 103, the matching unit 104 calculates correlation values C _min for all reference feature amounts, and the correlation value is smaller than the threshold value. selecting the smallest correlation value _{C min} in the C _min. The collating unit 104, the subject graphic is determined to be a figure of the same pattern as the reference graphic relating smallest correlation value C _min that is selected.

  FIG. 2 (FIGS. 2A and 2B) is a diagram illustrating a specific example of image data input to the edge extraction unit 100. In FIG. 2A, a cup with a handle and a steam pattern are shown in white inside a uniformly colored circular region. In FIG. 2B, a teacup and steam pattern without handles are shown in white inside a uniformly colored circular region. 2A and 2B are merely examples of graphics included in the image data input to the edge extraction unit 100. Therefore, the figure included in the image data input to the edge extraction unit 100 may be any other figure.

  FIG. 3 is a diagram illustrating a specific example of edge data generated by the edge extraction unit 100. The edge data in FIG. 3 is edge data generated from the image in FIG. 2A. Hereinafter, using the edge data shown in FIG. 3 as a specific example, using FIGS. 4 to 10, the first reference point, the second reference point, the first search start point, the second search start point, the first search direction, Specific examples of the second search direction will be shown, and each of the seven specific examples of the first feature value and the second feature value will be described.

  FIG. 4 is a diagram illustrating a first specific example of the first feature amount and the second feature amount. In FIG. 4, the first reference point, the second reference point, and the first search start point are all at the center of the circle. Therefore, even if the rotation angle θ changes, the positions of the first reference point, the second reference point, and the first search start point are not changed and are constant. 4 and FIGS. 5 to 10 described below, the rotation angle θ is represented as an angle that is counterclockwise from a reference line 30 that extends in the horizontal right direction from the center point 20 of the circle. The second search start point is the position of the first intersection. The first feature amount extraction unit 101 performs a search along the search straight line 70 from the first search start point toward the outer periphery of the circle, and detects the position of the edge detected first as the first intersection. The second feature quantity extraction unit 102 sets the position of the first intersection detected by the first feature quantity extraction unit 101 as the second search start point, and follows the search straight line 70 from the second search start point to the outer circumference of the circle. A search is performed to detect the second intersection. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 5 is a diagram illustrating a second specific example of the first feature amount and the second feature amount. In FIG. 5, the first reference point, the second reference point, and the first search start point are all the center point 20 of the circle. Therefore, even if the rotation angle θ changes, the positions of the first reference point, the second reference point, and the first search start point are not changed and are constant. The second search start point is an intersection 40 between the search straight line 70 and the outer circle. Then, the first feature quantity extraction unit 101 performs a search along the search straight line 70 from the first search start point toward the outer circumference of the circle, and the second feature quantity extraction unit 102 performs the center of the circle from the second search start point. The search is performed along the search straight line 70 in the direction. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 6 is a diagram illustrating a third specific example of the first feature amount and the second feature amount. In FIG. 6, the first reference point and the first search start point are both the center point 20 of the circle. Therefore, even if the rotation angle θ changes, the positions of the first reference point and the first search start point are not changed and are constant. The second reference point and the second search start point are the intersection 40 of the search straight line 70 and the outer circle. Then, the first feature quantity extraction unit 101 performs a search along the search straight line 70 from the first search start point toward the outer circumference of the circle, and the second feature quantity extraction unit 102 performs the center of the circle from the second search start point. The search is performed along the search straight line 70 in the direction. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 7 is a diagram illustrating a fourth specific example of the first feature amount and the second feature amount. In FIG. 7, the first reference point and the second reference point are both the center point 20 of the circle. Therefore, even if the rotation angle θ changes, the positions of the first reference point and the second reference point do not change and are constant. Further, the first search start point and the second search start point are set based on the inner circle 50. The inner circle 50 is a circle whose radius is a predetermined value R, and is a concentric circle of the circle of the target graphic. Specifically, the first search start point and the second search start point are the intersection 60 of the search straight line 70 and the inner circle 50. Then, the first feature quantity extraction unit 101 performs a search along the search line 70 from the first search start point toward the center of the circle, and the second feature quantity extraction unit 102 searches for a circle from the second search start point. A search is performed along the search line 70 in the outer circumferential direction along the straight line 70. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 8 is a diagram illustrating a fifth specific example of the first feature value and the second feature value. In FIG. 8, the first reference point, the second reference point, the first search start point, and the second search start point are set based on the inner circle 50. The inner circle 50 is a circle whose radius is a predetermined value R, and is a concentric circle of the circle of the target graphic. Specifically, the first reference point, the second reference point, the first search start point, and the second search start point are the intersection 60 of the search straight line 70 and the inner circle 50. Then, the first feature quantity extraction unit 101 performs a search along the search line 70 from the first search start point toward the center of the circle, and the second feature quantity extraction unit 102 performs an outer periphery of the circle from the second search start point. The search is performed along the search straight line 70 in the direction. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 9 is a diagram illustrating a sixth specific example of the first feature amount and the second feature amount. In FIG. 9, the first reference point and the second reference point are both the center point 20 of the circle. Therefore, even if the rotation angle θ changes, the positions of the first reference point and the second reference point do not change and are constant. The first search start point is set based on the first inner circle 51. The second search start point is set based on the second inner circle 52. The first inner circle 51 and the second inner circle 52 are circles having radii of predetermined values R1 and R2, respectively, and are concentric circles of the circle of the target graphic. Specifically, the first search start point and the second search start point are a first intersection 61 and a second intersection 62, which are intersections of the search line 70, the first inner circle 51, and the second inner circle 52, respectively. . Then, the first feature quantity extraction unit 101 performs a search along the search straight line 70 from the first search start point toward the outer circumference of the circle, and the second feature quantity extraction unit 102 performs the center of the circle from the second search start point. The search is performed along the search straight line 70 in the direction. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 10 is a diagram illustrating a seventh specific example of the first feature amount and the second feature amount. In FIG. 10, the first reference point and the first search start point are set based on the first inner circle 51. The second reference point and the second search start point are set based on the second inner circle 52. The first inner circle 51 and the second inner circle 52 are circles having radii of predetermined values R1 and R2, respectively, and are concentric circles of the circle of the target graphic. Specifically, the first reference point and the first search start point are the intersection (first intersection 61) between the search straight line 70 and the first inner circle 51. The second reference point and the second search start point are intersection points (second intersection point 62) between the search straight line 70 and the second inner circle 52. The first feature amount extraction unit 101 performs a search along the search straight line 70 from the first search start point toward the outer periphery of the circle, and the second feature amount extraction unit 102 moves from the second search start point to the center direction of the circle. A search is performed along the search line 70. The first feature quantity extraction unit 101 extracts the distance between the first reference point and the first intersection as an individual first feature quantity r1 (θ), and the second feature quantity extraction unit 102 extracts the second reference point and the second reference point. The distance from the intersection is extracted as the individual second feature amount r2 (θ).

  FIG. 11 is a flowchart showing an operation flow of the collation device 10. First, the edge extraction unit 100 accepts input of image data (step S101). Next, the edge extraction unit 100 extracts edges from the input image data and generates edge data (step S102). Next, the first feature quantity extraction unit 101 extracts the first feature quantity from the edge data (step S103). Further, the second feature amount extraction unit 102 extracts the second feature amount from the edge data (step S104). Either the process of step S103 or the process of step S104 may be performed first or may be performed in parallel. Next, the matching unit 104 executes a matching process based on the first feature value and the second feature value (step S105). Then, the collation unit 104 outputs the collation process result (collation result) (step s106), and the process of the flowchart shown in FIG.

  FIG. 12 is a graph showing a specific example of the first feature value and the second feature value extracted for the figures in FIGS. 2A and 2B. In FIG. 12, the horizontal axis represents the rotation angle θ (deg), and the vertical axis represents the distance r1 (r2) between the first intersection (second intersection) and the first reference point (second reference point). The first feature value and the second feature value shown in FIG. 12 are feature values extracted according to the second specific example. r1: The polygonal line of the mark 1 represents the first feature amount extracted from FIG. 2A. r1: The broken line of the mark 2 represents the first feature amount extracted from FIG. 2B. r2: The broken line of the mark 1 represents the second feature amount extracted from FIG. 2A. r2: The broken line of the mark 2 represents the second feature amount extracted from FIG. 2B. As is clear from FIG. 12, even if the figure is similar to that shown in FIGS. 2A and 2B, which is likely to be erroneously recognized by the conventional r-θ curve method, the first feature quantity and the second feature quantity are used. It is possible to extract significantly different feature quantities from the respective figures. Further, the calculation cost for extracting the first feature value and the second feature value in the collation device 10 is not changed in order compared with, for example, the conventional r-θ curve method. Therefore, the collation apparatus 10 can increase the accuracy of collation for similar graphic patterns based on the first feature amount and the second feature amount while suppressing an increase in calculation cost.

Further, as described above, by setting the center of the target graphic as the center point, the following effects can be obtained. In this case, even if the appearance of the target graphic in the edge data is rotated, the center point serving as the reference for the rotation angle θ is always set as the same point regardless of the rotation of the appearance of the target graphic. Therefore, the first feature value and the second feature value acquired from the target graphic are affected only as a phase shift in the rotation angle θ-axis direction. Accordingly, by calculating the correlation value for each phase difference θ ₀ that changes with a predetermined width from 0 degrees to 360 degrees as in the above formula 1, and performing the matching using the smallest value C _min among them. Thus, it is possible to improve the robustness against the rotation of the target graphic.

<Modification>
The image data input to the edge extraction unit 100 may be configured to include an image other than the target graphic. In this case, the edge extraction unit 100 detects a target graphic from the input image data, and generates edge data for the detected target graphic. With this configuration, even if the image data input to the edge extraction unit 100 includes various graphics other than the graphics to be collated, only the portion to be collated is used. A first feature value and a second feature value are extracted. Therefore, it is possible to improve the accuracy of the matching process in the matching device 10.

  In the above description, a circular figure including a pattern on the inside is given as a specific example of a figure to be collated. However, the figure to be collated need not be limited to a circle, and may be any geometric figure such as a square, a star, or a waveform. According to each figure, a reference for extracting the first feature value and the second feature value (first reference point, first search start point, second reference point, second search start point, first search direction, Second search direction) is set.

  In the above description, the collation device 10 is configured to extract two feature amounts, but may be configured to extract three or more feature amounts. For example, the collation device 10 may be configured to extract and collate r2 represented in FIG. 5 as the third feature quantity in addition to r1 and r2 represented in FIG. As described above, the collation apparatus 10 performs collation using three or more feature amounts, whereby the accuracy of the collation processing can be further increased.

  The embodiment of the present invention has been described in detail with reference to the drawings. However, the specific configuration is not limited to this embodiment, and includes designs and the like that do not depart from the gist of the present invention.

DESCRIPTION OF SYMBOLS 10 ... Verification apparatus, 20 ... Center point, 30 ... Base line, 40 ... Intersection, 50 ... Inner circle, 51 ... First inner circle, 52 ... Second inner circle, 60 ... Point, 61 ... First intersection, 62 ... Second intersection point, 70 ... search straight line, 101 ... first feature quantity extraction unit, 102 ... second feature quantity extraction unit, 103 ... reference feature quantity storage unit, 104 ... collation unit

Claims (5)

An edge extraction unit for extracting edges of the pattern of the input figure;
First, a feature amount relating to an intersection of a search straight line extending in the angular direction of each rotation angle and the edge is extracted as a first feature amount for each of a plurality of rotation angles at a predetermined center point with respect to the input graphic pattern. A feature extraction unit;
For the input graphic pattern, for each of a plurality of rotation angles at a predetermined center point, a feature amount relating to an intersection of the search straight line extending in the angular direction of each rotation angle and the edge, and the first feature amount extraction unit Is a second feature amount extraction unit that extracts the second feature amount based on different criteria;
Based on the first feature value and the second feature value, a matching unit that performs matching on the pattern of the input graphic;
Equipped with a,
The first feature amount extraction unit sets a reference point and a first search start point on the search line, and a predetermined first search from the first search start point among intersections of the search line and the edge An intersection located in the direction and closest to the first search start point is defined as a first intersection, and the first feature amount is extracted based on a distance between the first intersection and the reference point.
The second feature amount extraction unit sets a second search start point on the search line, and is located in a predetermined second search direction from the second search start point among intersections of the search line and the edge. Then, the intersection closest to the second search start point is set as the second intersection, and the second feature amount is extracted based on the distance between the second intersection and the reference point,
The reference point is the predetermined center point,
At least any one of the first search start point and the second search start point, the first search direction and the second search direction are different points or directions,
Verification device. The first intersection point and the second search start point are the same point,
  The collation apparatus according to claim 1, wherein the first search direction and the second search direction are the same direction. An edge extraction unit for extracting edges of the pattern of the input figure;
First, a feature amount relating to an intersection of a search straight line extending in the angular direction of each rotation angle and the edge is extracted as a first feature amount for each of a plurality of rotation angles at a predetermined center point with respect to the input graphic pattern. A feature extraction unit;
For the input graphic pattern, for each of a plurality of rotation angles at a predetermined center point, a feature amount relating to an intersection of the search straight line extending in the angular direction of each rotation angle and the edge, and the first feature amount extraction unit Is a second feature amount extraction unit that extracts the second feature amount based on different criteria;
Based on the first feature value and the second feature value, a matching unit that performs matching on the pattern of the input graphic;
With
The first feature amount extraction unit sets a first reference point and a first search start point on the search line, and a predetermined number of points from the first search start point among intersections of the search line and the edge. An intersection point that is located in one search direction and is closest to the first search start point is a first intersection point, and the first feature amount is extracted based on a distance between the first intersection point and the first reference point,
The second feature amount extraction unit sets a second reference point and a second search start point on the search line, and among the intersections of the search line and the edge, a predetermined first number from the second search start point. An intersection located in two search directions and closest to the second search start point is set as a second intersection, and the second feature amount is extracted based on a distance between the second intersection and the second reference point,
The first reference point is set to any one of a point set at a predetermined distance from a predetermined center and the predetermined center,
The second reference point is set to any one of a point set at a predetermined distance from a predetermined center and the predetermined center,
At least any one of the first search start point and the second search start point, and the first search direction and the second search direction are different points or directions . An edge extraction step in which the matching device extracts an edge of the pattern of the input graphic;
The matching device uses, as a first feature amount, a feature amount related to an intersection of the search straight line extending in the angular direction of each rotation angle and the edge for each of a plurality of rotation angles at a predetermined center point with respect to the input graphic pattern. A first feature extraction step to extract;
The collation device, for the input graphic pattern, for each of a plurality of rotation angles at a predetermined center point, a feature amount relating to the intersection of the search straight line extending in the angular direction of each rotation angle and the edge, A second feature amount extraction step for extracting as a second feature amount on a basis different from the amount extraction unit;
A collation step in which the collation device collates the pattern of the input graphic based on the first feature amount and the second feature amount; and
Equipped with a,
In the first feature amount extraction step, a reference point and a first search start point are set on the search line, and a predetermined first search from the first search start point among intersections of the search line and the edge. An intersection located in the direction and closest to the first search start point is defined as a first intersection, and the first feature amount is extracted based on a distance between the first intersection and the reference point.
In the second feature amount extraction step, a second search start point is set on the search line, and the intersection point between the search line and the edge is located in a predetermined second search direction from the second search start point. Then, the intersection closest to the second search start point is set as the second intersection, and the second feature amount is extracted based on the distance between the second intersection and the reference point,
The reference point is the predetermined center point,
Wherein the first search starting point and the second search starting point, the first search direction and the second search direction, among the verification how to wherein at least one pair are different or direction. An edge extraction step of extracting edges for the pattern of the input figure;
First, a feature amount relating to an intersection of a search straight line extending in the angular direction of each rotation angle and the edge is extracted as a first feature amount for each of a plurality of rotation angles at a predetermined center point with respect to the input graphic pattern. A feature extraction step;
For the input graphic pattern, for each of a plurality of rotation angles at a predetermined center point, a feature amount relating to an intersection of the search straight line extending in the angular direction of each rotation angle and the edge, and the first feature amount extraction unit Is a second feature amount extraction step for extracting as a second feature amount based on different criteria;
Based on the first feature value and the second feature value, a matching step for matching the pattern of the input graphic;
And a computer program for causing a computer to execute,
In the first feature amount extraction step, a reference point and a first search start point are set on the search line, and a predetermined first search from the first search start point among intersections of the search line and the edge. An intersection located in the direction and closest to the first search start point is defined as a first intersection, and the first feature amount is extracted based on a distance between the first intersection and the reference point.
In the second feature amount extraction step, a second search start point is set on the search line, and the intersection point between the search line and the edge is located in a predetermined second search direction from the second search start point. Then, the intersection closest to the second search start point is set as the second intersection, and the second feature amount is extracted based on the distance between the second intersection and the reference point,
The reference point is the predetermined center point,
A computer program characterized in that at least any one of the first search start point and the second search start point and the first search direction and the second search direction are different points or directions .

Images