JP2004102603A - Data characteristics extracting device and data collating device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image collating device which determines that two data, which are obtained by imaging or expressing one object, are related to the same object. <P>SOLUTION: In the image collating device 100, a binary image (A) of a line drawing is obtained; a multilevel image, which highly expresses the luminance of a point that density of the line is low, is created by a luminance level determination part 102; a characteristic data creating part 103 finds a value, which indicates gradient to a position of the luminance of a local area, only for the local area where distribution of the luminance related to each pixel of the multilevel image is enough plane; and data for collation is created based on the gradient; then another multilevel image (B) is obtained; a multiple tone image, which is converted to the binary image in a line drawing converting part 107 and highly expresses the luminance of a point that density of the line is low, is recreated; and a value, which indicates a gradient in the luminance of the local area where distribution of the luminance of the multilevel image is plane, is found in a characteristic data creating part 109; and a similarity tendency between the binary image (A) and the multilevel image (B) is determined by the value and the data for collation. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to a data collation technique for collating data such as image data.
[0002]
[Prior art]
In recent years, with the remarkable development of hardware technology and software technology, image data obtained by imaging has been widely used as an information processing target.
For example, an image database in which a considerable amount of image data is collected and managed collectively is used for various purposes such as medical, office business, academic research, surveillance, etc. Research and development of image recognition technology for determining whether or not a registered object or person is also actively pursued.
[0003]
By the way, when searching for image data in which a specific object or the like is captured from an image database in which image data obtained by imaging is integrated, it is necessary to determine whether or not a specific object or the like is included in certain image data. In some cases, it is possible to use a method of comparing the certain image data with data representing the specific object or the like as a line drawing or a multi-tone image.
[0004]
As a conventional image data comparison technique for comparing these two data, for example, there is a method using a generalized Hough transform or a high-speed generalized Hough transform in which the conversion is accelerated.
Hereinafter, an outline of this method will be described.
The person performing the search draws a specific object or the like desired to be searched on a paper as a line drawing, and inputs the line drawing to the computer as binary image data through an image scanner or the like. The computer generates high-speed generalized Hough transform of the image data of the line drawing to generate template data, and extracts contours by spatial differentiation of the image based on multi-tone image data to be searched in the image database. Generates binary image data to be searched, performs a process of voting template data for binary image data to be searched using high-speed generalized Hough transform, and determines that the number of votes to a certain point in the image is equal to or greater than a threshold. If there is a certain point, it is determined that an image corresponding to a specific object or the like is captured at that point. The generalized Hough transform and the high-speed generalized Hough transform are described in Non-Patent Document 1, for example.
[0005]
As another image data matching technique, there is an image data matching technique according to the prior invention by the present inventor, in which each image data is wavelet-decomposed to create specific matching data and then both image data are matched. (See Patent Document 1).
Each of these various image data matching technologies has its own advantages and disadvantages.
[0006]
[Non-patent document 1]
Akio Kimura et al., Fast Generalized Hough Transform-Similar Transformation Invariant Arbitrary Figure Detection Method, "Transactions of the Institute of Electronics, Information and Communication Engineers (D-2)", April 1998, J81-D-2, No. 4, p. . 726-734
[0007]
[Patent Document 1]
JP 2001-283221 A
[0008]
[Problems to be solved by the invention]
According to the present invention, even when two data obtained by imaging or expressing an object such as the same object, person, and scenery include the object relatively rotated, the two data are the same object. It is an object of the present invention to provide a new data collating technology which can be determined to be related to the above, and which uses a method different from the conventional one.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, a data feature extraction device according to the present invention is a data feature extraction device that generates feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space, Three-dimensional data acquisition means for acquiring three-dimensional data, scanning means for sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the scanning means, the vicinity of the focused position An xyz three-dimensional coordinate space of a predetermined number of four or more local areas on the xy two-dimensional plane, of a three-dimensional position having an x coordinate and a y coordinate in the local area among the three-dimensional positions indicated by the three-dimensional data. Are obtained, and each of the average positions obtained for all the local regions is substantially on the same plane in the xyz three-dimensional coordinate space. It is determined whether or not there is, and only when it is present on substantially the same plane, the information indicating the gradient of the plane with respect to the xy two-dimensional plane is associated with the information indicating the focused position, And a feature information generating means for generating as one element of the feature information.
[0010]
A data feature extraction method according to the present invention is a data feature extraction method for generating feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space, and acquiring three-dimensional data. A three-dimensional data acquisition step, a scanning step of sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the scanning step, four or more predetermined numbers located around the focused position For each of the local regions on the xy two-dimensional plane, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region among the three-dimensional positions indicated by the three-dimensional data is obtained, Whether each of the average positions obtained for all the local regions are substantially on the same plane in the xyz three-dimensional coordinate space Is determined, and only when the plane is substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane and information indicating the focused position are associated with one of the characteristic information. And generating a feature information generated as an element.
[0011]
Here, being substantially on the same plane means being present at a distance of a predetermined value or less from a certain plane.
This makes it possible to obtain feature information on three-dimensional data that can be used for matching between three-dimensional data. When the three-dimensional data are collated with each other based on the obtained feature information, the three-dimensional data representing the same object has a similarity to some extent in a rotating relationship about a straight line perpendicular to the xy two-dimensional plane. It can be determined to be high.
[0012]
Further, a data collation device according to the present invention is a data collation device for collating three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space, and is a first device for acquiring first three-dimensional data. Three-dimensional data acquisition means, first scanning means for sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the first scanning means, it is located around the focused position. Xyz three-dimensional coordinates of a predetermined number of four or more local regions on the xy two-dimensional plane, among three-dimensional positions indicated by the first three-dimensional data, three-dimensional positions having x and y coordinates in the local region An average position in the space is determined, and whether or not each of the average positions determined for all the local regions is substantially on the same plane in the xyz three-dimensional coordinate space; The determination is made, and the vector V on the xy two-dimensional plane indicating the gradient of the plane with respect to the xy two-dimensional plane is specified only when the plane V exists on the substantially same plane, and the vector V and the information indicating the focused position are determined. A first feature information generating unit to be associated, and a unit vector S starting from one position Q in the xyz three-dimensional coordinate space and heading in one direction are determined and associated with each vector V specified by the first feature information generating unit. For each piece of information obtained, a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space is determined based on a focused position related to the information, and a plurality of sets of the vector V and the vector PQ and a unit vector S are shown. Matching data generating means for generating matching data; second three-dimensional data obtaining means for obtaining second three-dimensional data; A second scanning means for sequentially focusing on each of the above positions, and a predetermined number of four or more xy two-dimensional planes located around the focused position each time one position is focused on by the second scanning means For each of the local regions in, among the three-dimensional positions indicated by the second three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region is determined. It is determined whether or not each of the average positions obtained for the local area is substantially on the same plane in the xyz three-dimensional coordinate space. Only when the average positions are on substantially the same plane, the xy2 dimension of the plane is determined. A second feature information generating unit that specifies a vector V ′ on an xy two-dimensional plane indicating a gradient with respect to the plane and associates the vector V ′ with information indicating the focused position; By sequentially applying each vector V in the matching data to each vector V ′ specified by the second feature information generating means, the azimuth relationship between the vector V ′ and the new unit vector S ′ becomes a vector. Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V applied to V ′ and the unit vector S, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, It is characterized by comprising a matching means for determining a similar tendency between the first three-dimensional data and the second three-dimensional data.
[0013]
The data collation method according to the present invention is a data collation method for collating three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space, wherein a first three-dimensional data is acquired. A three-dimensional data acquisition step, a first scanning step that sequentially focuses on each position on the xy two-dimensional plane, and each time one position is focused on by the first scanning step, it is located around the focused position. Xyz three-dimensional coordinates of a predetermined number of four or more local regions on the xy two-dimensional plane, among three-dimensional positions indicated by the first three-dimensional data, three-dimensional positions having x and y coordinates in the local region The average position in the space is determined, and all the average positions determined for all the local regions are substantially on the same plane in the xyz three-dimensional coordinate space. Is determined, and only when the plane is substantially on the same plane, the vector V on the xy2D plane indicating the gradient of the plane with respect to the xy2D plane is specified, and the vector V and the focused position are determined. A first feature information generating step of associating the information with the information indicating the unit vector S, and a unit vector S starting from one position Q in the xyz three-dimensional coordinate space and heading in one direction is determined. For each piece of information associated with the vector V, a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space is determined based on a focused position related to the information, and a plurality of sets of the vector V and the vector PQ and a unit are defined. A collation data generating step of generating collation data indicating the vector S, and second three-dimensional data for acquiring second three-dimensional data An obtaining step, a second scanning step of sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the second scanning step, four or more positions located around the focused position For each of a predetermined number of local regions on the xy two-dimensional plane, of three-dimensional positions indicated by the second three-dimensional data, an average of three-dimensional positions having x and y coordinates in the local region in the xyz three-dimensional coordinate space It is determined whether or not each of the average positions obtained for all of the local regions is substantially on the same plane in the xyz three-dimensional coordinate space. As long as the vector V ′ on the xy2D plane indicating the gradient of the plane with respect to the xy2D plane is specified, and the vector V ′ and the focused position are indicated. A second feature information generating step of associating each vector V ′ in the matching data with the vector V ′ specified in the second feature information generating step, so that the vector V ′ is newly added. Each unit vector S ′ is determined so that the azimuth relationship with the unit vector S ′ is the same as the azimuth relationship between the vector V applied to the vector V ′ and the unit vector S, and xyz3 of all the determined unit vectors S ′ A matching step of determining a similar tendency between the first three-dimensional data and the second three-dimensional data according to a distribution state in a three-dimensional coordinate space.
[0014]
This makes it possible to appropriately collate the three-dimensional data having a rotational relationship.
[0015]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, an image matching device according to an embodiment of the present invention will be described with reference to FIGS.
<1. Configuration>
FIG. 1 is a functional block diagram of the image matching device according to the embodiment of the present invention.
[0016]
The image matching device 100 is used in an image search system or the like that searches for specific image data from an image database, and has a function of matching two pieces of image data. Here, the image retrieval system uses an image database, which is a set of multi-gradation two-dimensional image data (hereinafter, referred to as “retrieved data”) formed by imaging, to write objects such as objects, people, and landscapes by hand. For searching for data to be searched, which is an image of the same object as the two-dimensional image data (hereinafter referred to as “search reference data”) of the line drawing obtained by digitizing the line drawing drawn by using a scanner or the like. It is a system having.
[0017]
The image matching device 100 is realized as one function of a computer, and as shown in FIG. 1, a search reference data acquiring unit 101, a brightness level determining unit 102, a feature data generating unit 103, a range limiting unit 104, a matching data generating unit The storage unit 105 includes a search data acquisition unit 106, a line drawing unit 107, a luminance level determination unit 108, a feature data generation unit 109, a collation unit 110, and an evaluation value recording unit 111. The function of each unit constituting the image matching apparatus 100 is realized by the CPU executing a control program stored in the memory of the computer.
[0018]
Hereinafter, each unit constituting the image matching apparatus 100 will be described.
<1-1. Search criteria data acquisition unit 101>
The search criterion data acquisition unit 101 has a function of specifying search criterion data, which is one line drawing data, as a processing target and transmitting the same to the luminance level determination unit 102. This search criterion data is 1 if a line is drawn by a scanner or the like for each pixel position after a person performing a search draws a line drawing on paper by hand, and if there is no line (dot). It is obtained by giving a data value of 0 and converting it into binary two-dimensional image data, for example, data of 512 pixels in the horizontal direction and 512 pixels in the vertical direction.
[0019]
<1-2. Brightness level determination unit 102>
The brightness level determination unit 102 performs a predetermined brightness determination process based on the transmitted search reference data, determines a multi-level brightness value for each pixel of the search reference data, and sets, for example, a 256-level multi-level brightness value. It has a function of generating image data.
The brightness determination processing is performed by sequentially setting each pixel position on the xy two-dimensional plane related to the search reference data as a first scanning point, and centering on the first scanning point, all pixel positions within a predetermined circular area with a constant radius r1. A luminance value that is approximately inversely proportional to the number of all pixel positions whose binary data value is 1 (that is, the number of all pixel positions corresponding to a line or a point drawn in the circular area) is assigned to the first value. This is a process for determining a luminance value for a scanning point.
[0020]
That is, in the brightness determination process, while moving the first scanning point on the search reference data, the number of all pixel positions having a data value of 1 in a circular area of a constant radius r1 centered on the first scanning point is k. When the luminance value is determined assuming that the arithmetic expression for calculating the luminance value Z at the first scanning point is, for example, the following equation 1, the multiplicity of 256 gradations from 0 to 255 is obtained. This is a process for generating gradation image data.
[0021]
[Equation 1] Z = 255 / (k + 1)
In this case, for example, the radius r1 of the circular region is set to 10 pixels or the like, and the size of the circular region is determined so that k can be changed from 0 to a value of 255 or more. Is an integer value obtained by rounding the result of 255 / (k + 1).
[0022]
In practice, when the first scanning point is located at the end of the search reference data, among the positions in the circular area centered on the first scanning point, those that protrude from the area of the search reference data appear. When the radius r1 of the circular area is set to 10 pixels, horizontal coordinates 10 to 501 and vertical coordinates 10 to 501 in the search reference data in which a pixel exists at each of the horizontal coordinates 0 to 511 and the vertical coordinates 0 to 511. With each coordinate position as a first scanning point, a luminance value at each first scanning point is determined, and no luminance value is determined for pixel positions within a range of 10 pixels from each end.
[0023]
By scanning, the coordinates of the first scanning point sequentially change in the horizontal direction as (10, 10), (11, 10), (12, 10), ..., (500, 10), (501, 10). Then, the next row changes in the same manner as (10, 11), (11, 11), (12, 11),..., (500, 11), (501, 11). An xy two-dimensional plane relating to image data such as search reference data is a two-dimensional plane represented by an x-axis and a y-axis, and all pixel positions constituting the image data are integer values in the x-coordinate and y-coordinate. Shall be represented. It is assumed that each pixel has a position as a point, and the distance between two pixels that are adjacent in the horizontal or vertical direction on the xy two-dimensional plane is one.
[0024]
FIG. 2 is a diagram illustrating a relationship between a line drawing portion and first scanning points in the search reference data.
The figure shows a circular area 203 centered on a first scanning point 202 at a certain moment during scanning, as an example of search reference data 200 representing a line drawing 201 in which two triangles are connected.
[0025]
In this example, since the number k of the pixel positions on the line drawing in the circular area is 16, the brightness value Z of the first scanning point 202 is Z = 255 / (16 + 1) = 15 from Equation 1.
For example, when a simple straight line is represented in the search reference data, if the first scanning point is on the straight line by the above scanning, the most pixel positions are included in the circular area 203. The luminance of the first scanning point becomes the lowest, and the luminance of the first scanning point decreases to some extent as the first scanning point moves away from the straight line.
[0026]
<1-3. Feature data generation unit 103>
The feature data generation unit 103 has a function of performing a predetermined feature extraction process based on the multi-tone image data generated by the luminance level determination unit 102 and storing feature data that is information indicating the features of the image.
In the feature extraction processing, each position in the two-dimensional image space related to the multi-tone image data generated by the luminance level determination unit 102 is sequentially set as a second scanning point, and a predetermined constant radius is set around the second scanning point. Only when the average brightness values of the first to fourth quadrants obtained by dividing the circular area of r2 into horizontal and vertical crosses have a planar relationship, the area around the second scanning point A vector V, which is a spatial gradient of each average luminance value, and a three-dimensional coordinate P obtained by adding a luminance value of the second scanning point to another two-dimensional pixel position of the second scanning point as data of another dimension. This is the process of saving the pair. Hereinafter, V (1), V (2),... When discriminating individual vectors V, and P (1), P (2),. ··· etc., and typically V (i), P (i).
[0027]
FIG. 3 is a diagram showing an example of the relationship between the second scanning point used in the feature extraction process and each of the first to fourth quadrants. In the figure, the radius r2 is set to 10 pixels.
Here, the feature extraction processing will be described in more detail with reference to FIG.
The multi-gradation image data generated by the luminance level determining unit 102 is data in which a luminance value is determined for each pixel position in the xy two-dimensional image space. A portion between a negative direction of x and a negative direction of y with respect to the second scanning point within a circular area having a radius r2 around the second scanning point while moving the second scanning point on the two-dimensional image space. The average value z1 of each luminance value corresponding to (region R1 in FIG. 3) and the average value of each luminance value corresponding to a portion (region R2) between the positive direction of x and the negative direction of y. z2, an average value z3 of each luminance value corresponding to a portion (region R3) between the negative direction of x and the positive direction of y, and a portion (region R4) between the positive direction of x and the positive direction of y. ), And an average value z4 of the respective luminance values is obtained. Each average luminance value of the peripheral scanning point to calculate the vector V (i) is the spatial gradient of luminance values at the periphery of determining whether or not and the second scanning point in the plane relationship.
[0028]
Note that the second scanning point is set when the multi-tone image data is composed of 502 pixels in the horizontal direction (x coordinate is 10 to 501) and 502 pixels in the vertical direction (y coordinate is 10 to 501) on the xy two-dimensional plane. , X coordinate and y coordinate move in the range of 20 to 491 respectively. That is, the coordinates of the second scanning point sequentially change in the horizontal direction as (20, 20), (21, 20), (22, 20), ..., (490, 20), (491, 20). Then, (20, 21), (21, 21), (22, 21),..., (490, 21), (491, 21) also change in the same manner for the next row.
[0029]
The determination as to whether or not the respective average luminance values of the portions (regions R1 to R4) around the second scanning point have a planar relationship is based on whether or not the relationship represented by the following inequality expression 2 is satisfied. Done.
[Equation 2] | z1-z2-z3 + z4 | ≦ C1
In Equation 2, C1 is a constant, and a relatively small value is determined in advance so that when z1−z2−z3 + z4 is close to 0, the relationship of Equation 2 is satisfied. When the maximum value of the luminance value is 255, C1 may be set to, for example, one value of about 0 to 5.
[0030]
If the relationship represented by the inequality expression 2 is satisfied, the above-described regions R1 to R4 are further rotated by 30 degrees in the circumferential direction around the second scanning point, and the region R1 ′ in the new range is obtained. Similarly, the average values z1 to z4 of the luminance values in the respective regions are determined for R4 ′ to determine whether or not the relationship represented by the inequality expression 2 is satisfied. If the relationship is still satisfied, the same is performed again. By rotating the regions R1 ′ to R4 ′ in the directions, z1 to z4 are similarly obtained for the regions R1 ″ to R4 ″ in the new range, and it is determined whether or not the relationship represented by the inequality of Expression 2 is satisfied, and the relationship is determined. Is satisfied, it is determined that the average luminance values of the respective areas around the second scanning point have a planar relationship.
[0031]
Hereinafter, the vector V (i), which is a spatial gradient of the luminance value around the second scanning point, which is performed only when the average luminance values of the respective regions around the second scanning point have a planar relationship. The calculation procedure will be described. When z1-z2-z3 + z4 is not 0, z4 is corrected so that z1-z2-z3 + z4 becomes 0, and the vector V (i) is calculated by the following procedure after the correction.
[0032]
First, for each of the regions R1 to R4, an average position (x-coordinate and y-coordinate of the center of gravity) of the region on the xy two-dimensional plane is determined as another-dimensional coordinate (z-coordinate), and an average luminance value in the region is determined. As a result, four three-dimensional coordinates are obtained.
The x coordinate of the average position of the region R1 is x1, the y coordinate is y1, the x coordinate of the average position of the region R2 is x2, the y coordinate is x2, and the x coordinate of the average position of the region R3 is x3, y coordinate. Let y3 be the x coordinate of the average position of the region R4, and y4 be the y coordinate. The four three-dimensional coordinates described above are A1 (x1, y1, z1), A2 (x2, y2, z2), A3 (x3, y3, z3) and A4 (x4, y4, z4).
[0033]
By substituting each of the three-dimensional coordinate values of A1 to A4 into Equation 3 which is a general formula of a plane, the value of each coefficient of the general equation of the plane is calculated, and the vector V (i) is calculated by Equations 4 and 5. Of the components Vx and Vy are obtained.
[Equation 3] a · x + by · y + c · z = d
In Equation 3, x, y, and z are variables, a, b, c, and d are coefficients, and “·” is a multiplication operator.
[0034]
[Equation 4] Vx = −a / c
[Equation 5] Vy = −b / c
Equation 4 represents the inclination of a straight line intersecting the xz plane with the plane in the three-dimensional coordinate space according to Equation 3 in which the coefficient value is determined, that is, the displacement Δz / Δx in the z direction with respect to the displacement in the x direction. The plane in the three-dimensional coordinate space according to the equation (3) having the determined coefficient value represents the inclination of a straight line intersecting with the yz plane, that is, the displacement Δz / Δy in the z direction with respect to the displacement in the y direction.
[0035]
In other words, Vx represents the gradient of the luminance value in the x direction, and Vy represents the gradient of the luminance value in the y direction. Therefore, it can be said that the vector V (i) represents the spatial gradient of the luminance value.
Assuming that the value of the three-dimensional coordinates P (i) to be stored in pairs with V (i) = (Vx, Vy) thus obtained is (Px, Py, Pz), Px is converted to multi-tone image data. Py is the x coordinate of the second scanning point on the xy two-dimensional plane relating to the multi-tone image data, and Pz is a value obtained by the following equation (6).
[0036]
[Equation 6] Pz = (z1 + z2 + z3 + z4) / 4
The feature data generating unit 103 records and stores a plurality of pairs of V (i) and P (i) obtained by such a procedure in a memory. However, exceptionally, for V (i) in which both Vx and Vy, which are the components of V (i), are approximately 0, that is, V (i) that is less than a predetermined threshold, V (i) And P (i) are not stored. For example, if the average luminance values in the regions R1 to R4 around the second scanning point are all the same, V (i) and P (i) for the second scanning point are not stored.
[0037]
<1-4. Range limiting unit 104>
The range limiting unit 104 has a function of specifying a range where an image to be collated exists in the search reference data. For example, a search criterion having 512 pixels in the horizontal and vertical directions in response to an input from a searcher Identify a rectangular area in the data. The rectangular area is defined as, for example, a range where the x coordinate is 30 to 480 and the y coordinate is 30 to 480.
[0038]
<1-5. Matching Data Generation and Holding Unit 105>
The collation data generation / storage unit 105 determines a point Q and a vector S indicating a certain direction on the xyz three-dimensional space, and stores the vector V (i) and the three-dimensional coordinates P ( i), the one in which the x coordinate and the y coordinate of P (i) are included in the range specified by the range limiting unit 104 is extracted, and the extracted V (i) and P (i) are extracted. For each group, PQ (i) indicating the position of Q in relative coordinates from P (i) is obtained, and V (i), PQ (i) and one S are stored in a memory or the like as collation data. Has functions.
[0039]
Number of pairs of V (i) and P (i) stored by the feature data generation unit 103 that include the x coordinate and y coordinate of P (i) within the range specified by the range limiting unit 104 Let n be the number of vectors V (i), n number of three-dimensional relative coordinates PQ (i), and one vector S stored by the matching data generation and holding unit 105 as matching data. Will be.
[0040]
For example, if P (1) and P (2) are within the range specified by the range limiting unit 104, the relative coordinate PQ (1) is associated with the vector V (1), and similarly, the relative coordinate PQ (2) is stored in association with the vector V (2). Each of the vectors V (i), PQ (i) and S is data composed of a component in the x direction, a component in the y direction, and a component in the z direction in the xyz three-dimensional coordinate space. The vector V (i) having the starting point at the point P and the vector S having the starting point at the point Q are represented by V (i), PQ (i) and S, and the relative positional relationship of the starting point and the relative vector direction , That is, a relative azimuth relationship.
[0041]
<1-6. Searched Data Acquisition Unit 106>
The search target data acquisition unit 106 sequentially retrieves one multi-tone image data from an image database storing a large number of search target data, that is, multi-tone image data, and specifies it as a target to be compared with search reference data. And has a function of transmitting it to the line drawing unit 107.
[0042]
The search target data is multi-tone image data generated by a digital camera or the like, and is, for example, data of 512 pixels in the horizontal direction and 512 pixels in the vertical direction at 256 tones.
<1-7. Line drawing unit 107>
The line drawing unit 107 performs spatial differentiation processing for obtaining a gradient of a luminance value on a two-dimensional image plane with respect to one piece of to-be-searched data, that is, multi-tone image data transmitted from the to-be-searched data acquisition unit 106, and a result thereof. And a function of generating binary image data corresponding to a line image and transmitting the generated binary image data to the luminance level determination unit 108 as a result. The continuous processing from the spatial differentiation processing to the binarization processing is, in other words, processing in which a portion where the spatial gradient of the luminance change of the image is steep, that is, a so-called edge portion is set to 1 and other portions are set to 0.
[0043]
<1-8. Brightness level determination unit 108>
The luminance level determining unit 108 receives the binary image data generated based on the data to be searched from the line drawing unit 107, performs a predetermined luminance determining process, and performs multi-level luminance for each pixel in the binary image data. It has a function of determining values and generating multi-gradation image data of, for example, 256 gradations. The brightness determination processing for generating multi-tone image data based on the binary image data is basically the same as the brightness determination processing performed by the brightness level determination unit 102, that is, the image data to be processed is different. Is similar.
[0044]
<1-9. Feature data generation unit 109>
The feature data generation unit 109 has a function of performing a predetermined feature extraction process based on the multi-tone image data generated by the luminance level determination unit 108 and storing feature data indicating the features of the image. This feature extraction process is basically the same as the feature extraction process performed by the feature data generation unit 103 described above, that is, the same except that the image data to be processed is different. If completely identical image data is input to the feature data generation units 103 and 109, both of the feature data generation units 103 and 109 store feature data having the same contents.
[0045]
The feature data stored by the feature data generation unit 109 is a plurality of sets of vectors V and three-dimensional coordinates P obtained based on the multi-tone image data generated by the luminance level determination unit 108. Hereinafter, the vector V and the three-dimensional coordinates P stored by the feature data generation unit 109 are expressed in the form of a vector V ′ (j) and three-dimensional coordinates P ′ (j).
[0046]
<1-10. Collation unit 110>
The collation unit 110 is generated by the feature data generation unit 109, and includes n sets of vectors V (i), relative coordinates PQ (i), and one S, which are the comparison data generated by the comparison data generation and storage unit 105. Using the m sets of vectors V '(j) and the three-dimensional coordinates P' (j), the image shown in the search reference data in a certain range approximated the image shown in the search target data. It has a function of performing a collation process for determining whether or not it is a device.
[0047]
As the matching process, the matching unit 110 first applies all V (i) in the matching data to each vector V ′ (j) in order, and expresses V (i), PQ (i), and S The size 1 is set so that the azimuth relationship between V (i) and S in the three-dimensional coordinate space and the azimuth relationship between V ′ (j) and S ′ (j, i) in the three-dimensional coordinate space are the same. Is defined as a unit vector S ′ (j, i). Here, S ′ (i, j) is a unit determined by applying vector V (i) to vector V ′ (j) so as to maintain the same azimuthal relationship as vector S with respect to vector V (i). This means a vector S ′, and if there are n V (i) as collation data and m m vectors V ′ (j) stored by the feature data generation unit 109, the unit vector S '(I, j) is determined to be nm.
[0048]
If the angle between the reference and the direction of the vector PQ (i) from the point P to the point Q is α with respect to the direction of the vector V (i), α is V (i) and PQ (i). )), And if the angle between the reference and the direction of the vector S is β based on the direction of the vector V (i), β can be obtained from V (i) and S. The distance γ between the point P and the point Q can be obtained from the vector PQ (i).
[0049]
Therefore, to determine the unit vector S ′ (i, j) by applying the vector V (i) to the vector V ′ (j), the starting point of the unit vector S ′ (hereinafter referred to as “point Q ′”) A vector P′Q ′ that is located at a distance γ from the position of the point P ′ (j) that is the starting point of the vector V ′ (j) and that goes from the point P ′ to the point Q ′ is a vector V ′ (j) Is determined to be a vector directed to the angle of α with reference to, and the unit vector S ′ is determined to be a vector directed to the angle of β with reference to the vector V ′ (j).
[0050]
Next, the xyz three-dimensional coordinate space is divided into a three-dimensional array of unit cubes having a predetermined size, for example, each side having a length of 5, and each unit cube is positioned in the unit cube. A process of obtaining a vector addition result of all the vectors S ′ (i, j), that is, a process of combining vectors, is performed. Finally, the maximum value Smax of the size of the combined vector obtained as a vector addition result and the vector S ′ The number Snum of the vectors S ′ (i, j) in the unit cube containing the most (i, j) is evaluated together with the information for identifying the searched data acquired by the searched data acquiring unit 106. If the maximum value Smax of the magnitude of the vector obtained as a result of the vector addition stored in the value recording unit 111 is equal to or greater than a predetermined threshold, the image in the data to be searched and the search reference data The image performing a process of determining and tend similar.
[0051]
The collation unit 110 displays this determination result via, for example, a display device or the like. The display contents of the determination result are, for example, the search target data itself, similarity or dissimilarity, the value of Smax / n, and the like.
If the image data as a result of line drawing of the search target data and the search reference data completely match, the maximum value Smax of the vector obtained as a result of the vector addition described above is the size of the unit cube. If the size is equal to or larger than 1 pixel × 1 pixel × 1 pixel, it is theoretically at least n. Therefore, in consideration of the influence of discrete pixel positions in the digital image, in practice, it is conceivable to set, for example, 0.6 times or 0.7 times n as the above-mentioned predetermined threshold value.
[0052]
<1-11. Evaluation value recording section 111>
The evaluation value recording unit 111 uses the maximum value Smax of the vector size obtained as a result of the vector addition described above and the vector S ′ (i, This is a storage area such as a memory in which the maximum density of j), that is, the maximum value Snum of the number in the unit cube is stored.
[0053]
Each time one piece of image data is acquired as search data from the image database by the search data acquisition unit 106, the evaluation value recording unit 111 stores a line drawing unit 107, a luminance level determination unit 108, and a feature data generation unit 109. By the operation of the matching unit 110, information for identifying a set of data to be searched and the maximum value Smax are additionally recorded. Therefore, if the contents of the evaluation value recording unit 111 are used, the searched data identified corresponding to the largest one of the recorded maximum values Smax is indicated as the search reference data most in the image database. It is possible to evaluate that the searched data has a high tendency to be similar to the image. Further, for example, it is possible to evaluate that the similarity tendency is higher as the number of Snums is larger.
<2. Operation>
Hereinafter, the operation of the image matching apparatus 100 having the above configuration will be described.
[0054]
When roughly classified, the image collating apparatus 100 acquires search reference data, which is binary image data of an input line drawing, and generates collation data, and sequentially identifies data to be searched from an image database. Then, by using the specified data to be searched and the data for matching, a second process of matching the search reference data with the data to be searched is performed.
[0055]
FIG. 4 is a flowchart illustrating a first process of acquiring search reference data and generating the matching data.
As shown in the figure, first, the search criterion data acquisition unit 101 acquires the search criterion data and transmits it to the brightness level determination unit 102 (step S11). The luminance level determination unit 102 generates multi-tone image data by moving the first scanning point on the search reference data and determining the luminance value for each pixel position (step S12). As a result, multi-tone image data having a high luminance value is generated at a pixel position around a low density of lines or points indicated by the search reference data.
[0056]
Subsequently, the feature data generation unit 103 moves the second scanning point on the xy two-dimensional plane related to the multi-tone image data generated by the luminance level determining unit 102, and moves the second scanning point to the periphery. An average luminance value of each of the four divided areas is obtained, and a vector V (i) indicating a spatial gradient of the average luminance value is calculated as xy of the second scanning point only when each average luminance value of each area has a planar relationship. The coordinates and the average luminance value of the entire four surrounding areas are stored in a memory or the like together with the xyz three-dimensional coordinates P (i) obtained as z-coordinates (step S13).
[0057]
Subsequently, in response to the specification of the range by the range limiting unit 104, the matching data generation and holding unit 105 determines a point Q at an arbitrary three-dimensional position, determines a vector S in an arbitrary direction, and sets the vector S within the specified range. A relative vector P (i) including the three-dimensional coordinates P (i) and a relative coordinate PQ indicating a vector from the point P (i) to the point Q for all points P (i) located within the specified range. (I) and one vector S are stored in a memory or the like as collation data (step S14).
[0058]
FIG. 5 is a diagram exemplifying V (i) and P (i) generated by the feature data generating unit 103, and a point Q and a vector S determined by the matching data generation and holding unit 105.
In the figure, for convenience of explanation, the number of V (i) and P (i) is only three, and the figure shows a mapping on an xy two-dimensional plane for each element.
[0059]
FIG. 6 is a flowchart showing a second process of identifying the data to be searched and comparing the search reference data with the data to be searched using the data for matching.
As shown in the figure, first, the search target data acquisition unit 106 obtains search target data, which is one piece of multi-tone image data, from the image database and transmits it to the line drawing unit 107 (step S21). In step S22, binary image data is generated by binarization for extracting a so-called edge portion of the image represented by the data to be searched and transmitted to the luminance level determination unit 108 (step S22).
[0060]
The luminance level determination unit 108 to which the binary image data has been transmitted determines the luminance of each pixel position of the binary image data and newly generates multi-tone image data in the same manner as step S12 described above (step S12). S23).
Subsequently, the feature data generation unit 109 moves the second scanning point on the xy two-dimensional plane related to the multi-tone image data generated by the luminance level determining unit 108, and moves the periphery around the second scanning point. An average luminance value of each of the four divided areas is obtained, and a vector V ′ (j) indicating a spatial gradient of the average luminance value is calculated for the second scanning point only when each average luminance value of each area has a planar relationship. The xy coordinates and the average luminance value of the entire four surrounding areas are stored in a memory or the like together with the xyz three-dimensional coordinates P ′ (j) obtained as z coordinates (step S24).
[0061]
Subsequently, the matching unit 110 compares the P ′ (j) and V ′ (i) indicating the features of the searched image stored by the feature data generating unit 109 with the matching data stored in the first processing. For reference, for each P ′ (j), when V (i), which is data for verification, is placed at the position of P ′ (j) in the xyz three-dimensional space so as to match V ′ (j), The unit vector S ′ is arranged while maintaining the same azimuthal relationship from the position of V (i) to S (step S25).
[0062]
FIG. 7 shows V (i), PQ (i), and V (i) and PQ (i) held by the matching data generation and holding unit 105 for each set of V ′ (j) and P ′ (j) generated by the feature data generating unit 109. It is a figure which illustrates the mode which determined the unit vector S '(i, j) by applying the relationship of S.
In the figure, for convenience of explanation, the number of V '(j), P' (j) and V (i) is only three, and the mapping of each element on an xy two-dimensional plane is shown. ing. The notation “S (V ′ (1) ← V (2))” in the figure indicates that it is a unit vector S ′ arranged as a result of applying V (2) to V ′ (1). I have.
[0063]
For example, when V (2) is applied to V ′ (1) located at the position indicated by P ′ (1) extracted based on the searched data, the vector PQ (2) is obtained from P ′ (1). ), And the angle formed by the direction from P ′ (1) to Q ′ with respect to the vector V ′ (1) with respect to the vector V ′ (1) with respect to V (2) A point Q ′ is determined so that the angle of the vector of PQ (2) is the same as the angle formed by the vector V ′ (1) and the unit vector S ′ at the position of the point Q ′. The unit vector S ′ is arranged so that the angle formed by the direction of the vector S is the same as the reference.
[0064]
After arranging the unit vector S ′, the matching unit 110 calculates the above-described Smax and Snum, determines whether or not the similarity is based on Smax as described above, and displays the determination result on a display device. Smax and Snum are recorded in the evaluation value recording unit 111 (step S26).
The image matching apparatus 100 sequentially determines the data to be searched from the image database, and repeats the steps S21 to S26 until it determines that the data is similar (step S26).
[0065]
By the first processing and the second processing described above, among the image data in the image database, the image data including the one that matches or is similar to the object drawn as a line drawing by the searcher is detected. Will be.
In the feature data generation units 103 and 109, the relation is such that the quadrants R1 to R4 around the second scanning point on the xy two-dimensional plane relating to the image data, that is, the center of gravity of each area is positioned at the four vertices of the square. , The coordinates of the centroids in an xyz three-dimensional space in which the average luminance value in the area is obtained as the z-coordinate of the centroid of the area are substantially on the same plane. Only when it exists, the change of the peripheral luminance value in the xy coordinate space at the second scanning point is stored as a vector V (i) or a vector V ′ (j), and is treated as indicating the characteristic of the image data. ing. Moreover, the regions R1 to R4 are rotated by 30 degrees and 60 degrees around the second scanning point, and only when the coordinates of the barycenter of each region similarly obtained on the xyz three-dimensional space are approximately on the same plane, The vector V (i) or the vector V '(j) is stored.
[0066]
In other words, even if the image is rotated, the feature data generation unit sets the feature data generation unit so as to make the relationship among the feature data in the feature data group collected as a result of scanning by the second scanning point invariable. Only when the spatial gradient of the luminance in the peripheral area of the two scanning points is uniform, is the spatial gradient of the luminance at the second scanning point obtained using the luminance of each position in the peripheral area of the second scanning point Are stored as feature data that can be used as a basis for collation.
[0067]
Therefore, according to the image collation device 100, the objects represented by the two image data to be collated are rotated even if they have a relationship in which they are rotated about a straight line perpendicular to the image plane. Just as in the case where there is no data, it is possible to determine whether or not they are similar.
<3. Supplement>
As described above, the image collation device as one embodiment of the data collation device and the data feature extraction device according to the present invention has been described. However, the image collation device can be partially modified. It is a matter of course that the present invention is not limited to the embodiment of the image collating device. That is,
(1) The image collating device can perform not only collation between line image data and multi-tone image data but also collation between two multi-tone image data and collation between two line image data. You may do so. For example, if the image data to be collated is all line drawing image data, the line drawing by the line drawing unit 107 is omitted, and if all the image data to be collated is multi-tone image data. May perform line drawing similar to that of the line drawing unit 107 between the search reference data acquisition unit 101 and the luminance level determination unit 102.
[0068]
When two multi-tone image data are collated with each other, the processing by the line drawing unit and the luminance level determination unit is omitted, and the two multi-tone image data are directly processed by the feature data generation units. The image matching device may be configured to be targeted.
(2) The position on the xy two-dimensional plane where the first scanning point or the second scanning point is located in the scanning by the luminance level determination unit and the feature data generation unit of the image matching device is not necessarily the point when the pixel is represented by a point. It doesn't have to be on top. For example, the first scanning point or the second scanning point may transition at an intermediate point between pixel points. In addition, in order to determine the luminance of the first scanning point, binary data for pixels in a range of a circular area centered on the first scanning point is used, and a spatial gradient of luminance for the second scanning point is obtained. For example, the brightness values of pixels within the range of the circular area centered on the second scanning point are used. However, each circular area may be changed to an area of another shape. The size may be different.
(3) In the brightness determination processing by the two brightness level determination units, Equation 1 is used to determine the brightness value Z of the first scanning point from the number k of the pixel positions in the circular area centered on the first scanning point. However, the calculation is such that the larger the number of lines or points around the first scanning point, that is, the number of pixel positions where the value of the binary data is 1, the lower the luminance value Z or the number of such pixel positions is large. If the calculation is such that the luminance value Z increases as the number increases, the luminance value Z may be determined by a calculation other than Equation 1. That is, the brightness determination processing is performed based on whether or not the density of pixel points having binary data such as a line drawing value of 1 in each local area on the xy two-dimensional plane is high or low. This is a process of determining a dimension attribute value (luminance value).
(4) In the feature data generation unit, the regions R1 to R4 each have a range including a plurality of pixel points, but may have a range including only one pixel point. Note that the feature data generation unit has calculated the average position of all points in each of the regions R1 to R4 and the average value of the luminance of all the points. However, each of the regions R1 to R4 has one pixel point. If only the region R1 is included, the average position of each of the regions R1 to R4 is the position of one pixel included in each region, and the average value of the luminance of each of the regions R1 to R4 is It is obtained by the luminance itself of the one pixel included in the area.
(5) The feature data generation unit determines whether or not the luminance distribution around the second scanning point is planar and calculates the spatial gradient of the luminance around the second scanning point. Are divided into four around the circular region, and the regions R1 to R4 are rotated twice by 30 degrees in order to determine whether or not the region is two-dimensional. However, based on the average position and the average luminance value in each of the four or more local regions around the second scanning point, it is determined whether or not the spatial distribution of luminance around the second scanning point is planar. If the method is such that the angle between the plane and the xy two-dimensional plane is calculated as the spatial gradient of the luminance at the second scanning point, the number of local areas is necessarily four. The position of the local region is not limited to the region R1. It is not limited to R4. Further, the rotation by 30 degrees twice is only one method for ensuring that the luminance distribution around the second scanning point is sufficiently planar, and for example, the second rotation on the xy two-dimensional plane If the brightness of each pixel located at each vertex of a minute regular polygon centered on the scanning point is the z coordinate, the position of each pixel in the xyz three-dimensional coordinate space is substantially on the same plane. Whether or not the distribution of luminance around the second scanning point is sufficiently planar may be determined based on whether or not the determination is negative. Only at points where the distribution of the peripheral luminance is sufficiently planar, the spatial gradient of the luminance added to the point is treated as information indicating the characteristics of the image. Is less susceptible to a shift in the sampling direction and a process of rotating the image in digitizing the image.
(6) The range limiting unit does not necessarily need to receive the input of the person performing the search, and may fixedly store the rectangular area in advance. Further, the range limiting unit displays the image captured by the search criterion data obtaining unit via a display device or the like, and specifies a rectangle by a mouse drag operation or the like by a person performing a search on the image. An image range to be collated may be specified based on the collation.
(7) In the feature data generation unit, each component of the vector V (i) indicating the spatial gradient of the luminance is calculated by Expressions 4 and 5, but the change in the luminance and the position of the position on the xy two-dimensional plane are calculated. Data calculated by another operation capable of expressing the relationship with the change may be handled in the same manner as the vector V (i). For example, when the average luminance values of the regions R1, R2, R3, and R4 are z1, z2, z3, and z4 in this order, the x component of the vector V (i) is calculated as Vx and V The y component of (i) may be determined to be Vy obtained by the following Expression 8.
[0069]
[Equation 7] Vx = z1-z2 + z3-z4
[Equation 8] Vy = z1 + z2-z3-z4
Further, a vector obtained by multiplying both of the components Vx and Vy by a constant so that the magnitude of the vector V (i) becomes 1 is determined as a new vector V (i), that is, the vector V (I) may be normalized.
(8) The collation data generation and holding unit generates and stores collation data that is a plurality of sets of the vector V (i) and the vector PQ (i) and the vector S. Is sufficient if the information indicates an azimuthal relationship between the direction and the position between the vector V (i) and the vector S, including, for example, the collation data. It may be composed of an angle in the direction of i), the magnitude of the vector PQ (i), and an angle from the direction of the vector V (i) to the direction of the vector S.
[0070]
It should be noted that the application of the vector V (i) to the vector V ′ (j) may be performed only when the magnitudes of the two vectors substantially match. In this case, the collation data needs to include information indicating the magnitude of the vector V (i), and further, the luminance from the maximum to the minimum luminance of the two image data to be collated. A method such as finding the vector V (i) by an operation as shown in Equations 4 and 5 so that the influence of the luminance range does not appear on the vector V (i) so as not to be affected by the difference in the range. It is necessary to adjust the luminance range and use a method of obtaining the vector V (i) by using the equations 7 and 8 in the feature data generation unit.
(9) Regarding the PQ (i) stored by the collation data generation / holding unit, all PQ (i) are set so that the maximum size of PQ (i) is 1 among all PQ (i). Of the components (x component, y component, and z component) of PQ (i) are uniformly multiplied by a constant and newly defined as PQ (i), and stored. It is good also as making it save. Note that among all PQ (i), the size of PQ (i) having the largest size is set to 1, and the average value of the sizes of all PQ (i) may be set to 1. By these normalizations, when the two image data to be compared indicate the same object and are in a scaled relationship with each other, it can be determined that they are the same or similar.
(10) The collation unit of the image collation device performs collation processing such as fitting by using all the data used for collation as xy two-dimensional data, that is, data consisting of the x component and the y component ignoring the z component in the xyz three dimensional data. Is also good. In this case, a plurality of unit vectors S ′ are arranged on the xy two-dimensional plane, and the two matching target vectors are determined in accordance with the distribution state of the starting points of the plurality of unit vectors S ′ that are substantially in the same direction on the xy two-dimensional plane. It becomes possible to evaluate the similarity tendency of the images. It should be noted that the larger the number of starting points in the local range where the starting points of the unit vectors S ′ that go in approximately the same direction on the xy two-dimensional plane are most concentrated, the more similar the two images can be determined.
(11) The collation unit calculates the size of the angle formed by the direction of the unit vector S ′ in the partial area where the addition result of the unit vector S ′ is the largest when viewed from the direction of the vector S, and Alternatively, the degree to which the two images to be compared are angularly rotated may be output to a display device or the like. In addition, the collation unit records the maximum value Smax of the magnitude of the combined vector and the maximum number Snum of the vectors S ′ in the unit cube, which are the result of the addition of the vector S ′, in the evaluation value recording unit. May be output.
(12) The collating unit divides the xyz three-dimensional coordinate space into a three-dimensional array of unit cubes of a predetermined size, for example, each side has a length of 5, and in each unit cube, the unit cube Is performed to obtain a vector addition result of all the vectors S ′ (i, j) positioned at the position, i.e., a process of combining vectors is performed. Finally, the maximum value Smax of the magnitude of the combined vector obtained as the vector addition result is obtained. , The number Snum of the vectors S ′ (i, j) in the unit cube containing the largest number of the vectors S ′ (i, j), in order to identify the search target data acquired by the search target data acquisition unit 106. However, the unit cube here is not necessarily limited to a cube, and may be a predetermined size instead of a cube. May be the use of a sphere having a radius of no problem even be arranged such that each sphere overlap partially when arranged as a three-dimensional array in the same manner that each sphere and cube.
(13) In addition to the above-described image matching device, the data matching technology according to the present invention can be applied to matching between sets of three-dimensional data, and can be applied to three-dimensional map data and three-dimensionally measured object shapes. The present invention can also be applied to an application in which such data is compared with other similar data to determine similarity.
[0071]
If two three-dimensional data to be collated are in a rotational relationship with each other about a certain straight line, if the straight line is unpredictable and a plane perpendicular to the straight line can be assumed in advance. By treating a plane perpendicular to the straight line as an xy two-dimensional image plane in the case of image matching described in the embodiment, it is possible to appropriately determine similarity. For example, three-dimensional data A indicating the shape of a certain building alone and three-dimensional data B indicating the shape and positional relationship of a plurality of buildings in a section where a plurality of buildings including the building are built are described. If it is necessary to collate and determine whether or not the same or similar building as the three-dimensional data A is represented in the three-dimensional data B, the land surface is treated as an xy two-dimensional plane, By performing the same processing as the processing performed by the feature data generation unit of the image matching device by setting the height, which is the attribute value of the position, as the z coordinate, the determination can be performed in the same manner as the image matching device. become.
(14) A program for causing a computer to execute each process (see FIGS. 4 and 6) in the image matching device can be recorded on a recording medium or distributed and distributed via various communication paths. Such recording media include an IC card, an optical disk, a flexible disk, a ROM, and the like. The distributed and distributed programs are provided for use by being stored in a memory or the like that can be read by a computer, and the computer executes the programs to realize the functions of the image processing apparatus described in the present embodiment.
[0072]
【The invention's effect】
As is apparent from the above description, the data feature extraction device according to the present invention is a data feature extraction device that generates feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space. Three-dimensional data acquisition means for acquiring three-dimensional data, scanning means for sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the scanning means, the vicinity of the focused position Xyz three-dimensional coordinates of a three-dimensional position having x-coordinates and y-coordinates in the local region among the three-dimensional positions indicated by the three-dimensional data, for each of four or more predetermined numbers of local regions on the xy two-dimensional plane located in An average position in the space is obtained, and each of the average positions obtained for all the local regions is substantially the same in the xyz three-dimensional coordinate space. It is determined whether or not the information is present on the same plane, and only when the information is present on the substantially same plane, the information indicating the gradient of the plane with respect to the xy two-dimensional plane is associated with the information indicating the focused position. And feature information generating means for generating the feature information as one element of the feature information.
[0073]
This makes it possible to obtain feature information on three-dimensional data that can be used for matching between three-dimensional data. When the three-dimensional data are collated with each other based on the obtained feature information, the three-dimensional data representing the same object has a similarity to some extent in a rotating relationship about a straight line perpendicular to the xy two-dimensional plane. It can be determined to be high.
[0074]
Here, the three-dimensional data is image data indicating a position on the xy two-dimensional plane and a luminance value of each pixel arranged on the xy two-dimensional plane, and the three-dimensional position is a luminance value of the pixel. And the x-coordinate and y-coordinate of the position of the pixel on the xy two-dimensional plane, and the scanning unit is provided with a pixel corresponding to the image data on the xy two-dimensional plane. Focusing on each position in the region in which the scanning unit is located, the feature information generating unit determines, for each of a plurality of local regions on the xy two-dimensional plane of at least four located around the position noted by the scanning unit, The average position on the xy two-dimensional plane and the average value of the luminance values of all the pixels arranged at positions on the xy two-dimensional plane having the x-coordinate and the y-coordinate within the above-mentioned xyz3 It is determined as an average position in the original coordinate space, and it is determined whether or not all the average positions obtained for all the local regions are present on substantially the same plane in the xyz three-dimensional coordinate space. Only when the information exists, information associating information indicating the gradient of the plane with respect to the xy two-dimensional plane and information indicating the focused position may be generated as one element of the feature information.
[0075]
As a result, since the feature information is generated by extracting the features of only the local region having a flat luminance distribution, the influence of the positional reference of the sampling, the rotation processing of the image data, and the like in the digital image data generation stage. Since it does not appear in the feature information, the use of this feature information makes it possible to appropriately perform the matching of the two image data in a rotational relationship to some extent.
[0076]
In addition, the data feature extraction device further includes a line drawing data acquisition unit that acquires binary image data indicating a position on the xy two-dimensional plane and binary data for each pixel arranged on the xy two-dimensional plane; Binary image scanning means for sequentially focusing on each position on an xy two-dimensional plane in which pixels relating to the value image data are arranged; and a binary image scanning means which is included in an area of a predetermined size centered on the position focused by the binary image scanning means. Among the pixels related to the binary image data, the attribute value is determined such that the attribute value decreases as the number of pixels for which the binary data indicates the first value increases as the number increases. Attribute value determining means for associating a position with the attribute value, and each set of the position on the xy two-dimensional plane and the attribute value associated by the attribute value determining means as the three-dimensional data. A transmission unit for transmitting the data to the original data acquisition unit, wherein the three-dimensional position is indicated by an x-coordinate and a y-coordinate of each position constituting the three-dimensional data, and a z-coordinate which is an attribute value corresponding to the position. It is good also as being something.
[0077]
Thus, since the line density is an attribute value for each position on the xy two-dimensional plane, the obtained feature information is used to compare the binary image data representing the line image with other image data. Even when the line is partially interrupted to form a broken line, the similarity of the image data can be determined appropriately.
Further, the three-dimensional data is image data indicating a position and an attribute value of each pixel arranged on the xy two-dimensional plane on the xy two-dimensional plane, and the three-dimensional position is an attribute value of the pixel. It is indicated by a certain z-coordinate and an x-coordinate and a y-coordinate of a position of the pixel on an xy two-dimensional plane, and the scanning unit is provided with a pixel related to the image data on an xy two-dimensional plane. Focusing sequentially on each position in the area, the feature information generating means uses the position as a reference within a certain range of the area on the xy two-dimensional plane centered on the position noted by the scanning means, Average value z1 of each attribute value corresponding to the first local area between the negative direction of x and the negative direction of y, and corresponding to the second local area between the positive direction of x and the negative direction of y Average of each attribute value z2, the average value z3 of each attribute value corresponding to the third local region between the negative direction of x and the positive direction of y, and the fourth locality between the positive direction of x and the positive direction of y. Each of the average positions is obtained by obtaining an average value z4 of each attribute value corresponding to the region. When the value of the expression | z1-z2-z3 + z4 | is equal to or less than a predetermined value d1, all the local regions are obtained. It is determined that each of the obtained average positions is substantially on the same plane in the xyz three-dimensional coordinate space, and information indicating a gradient of the plane with respect to the xy two-dimensional plane is associated with information indicating the focused position. The attached information may be generated as one element of the feature information.
[0078]
Accordingly, in image data or the like in which pixels are arranged in a two-dimensional array on an xy two-dimensional plane, the inclination of the luminance value with respect to the xy two-dimensional data plane when the luminance value of the pixel is set to the z-coordinate is defined by the x and y directions It is possible to easily determine whether or not the distribution of the luminance values serving as the basis for calculating the gradient of the luminance value when the characteristic information of the image data is used as the characteristic information of the image data is planar. . In the case where the average luminances z1 to z4 of the regions R1 to R4 shown in the embodiment are in a planar relationship, the spatial gradient of the luminance of the scanning point is obtained based on these z1 to z4 as a basis for calculation to indicate the feature of the image. If a method of treating as image information is used, for example, for each of two image data obtained by imaging the same object by rotating a camera so as to have a relationship rotated by 90 degrees with respect to each other, the image data of the two When the information indicating the feature is obtained, the information indicating the feature of each image becomes the same.
[0079]
Further, the area of the certain range according to the feature information generating means is a circular area having a constant radius centered on a position focused by the scanning means, and the feature information generating means is further focused by the scanning means. Are rotated by a predetermined angle around the reference position and correspond to the first to fourth regions after the rotation. An average value z1 ′, z2 ′, z3 ′ and z4 ′ of each attribute value is obtained, and the feature information generating means determines that the value of the mathematical expression | z1-z2-z3 + z4 | is equal to or less than a predetermined value d1, In addition, when the value of the expression | z1'-z2'-z3 '+ z4' | is equal to or smaller than a predetermined value d1, each of the average positions obtained for all the local regions is in the xyz three-dimensional coordinate space. Almost on the same plane It determines that that, those associated with information indicating the information and the focus position indicating a gradient with respect to the xy2 dimensional plane of the plane may be generated as a component of the characteristic information.
[0080]
As a result, the same feature information indicating the features of the two pieces of image data representing the same object in a relationship rotated by an angle other than 90 degrees may be the same.
Further, the feature information generating means further associates information indicating the gradient with information indicating the focused position only when the gradient of the plane with respect to the xy two-dimensional plane is equal to or more than a predetermined amount. , May be generated as one element of the feature information.
[0081]
As a result, since only points having a certain amount of change in luminance are extracted as features of image data, the number of data used for collation can be effectively reduced, and the calculation time required for collation can be reduced.
Further, when all of the average positions obtained for all the local regions are substantially on the same plane, the feature information generation unit converts the mathematical expression representing the plane into a × x + by × y + cz × d ( a, b, c, and d are constant values), a vector V on an xy two-dimensional plane having -a / c and -b / c as x and y components, respectively, is obtained, and the magnitude of the vector V is set to 1 May be specified as information indicating the gradient relating to the plane.
[0082]
By such normalization, information indicating the characteristics of three-dimensional data such as multi-tone image data is not affected by whether the range of attribute values corresponding to the z-coordinate such as luminance is narrow or wide. It can be used for general purposes.
Further, when all of the average positions obtained for all of the local regions are substantially on the same plane, the feature information generating unit generates a vector V on an xy two-dimensional plane indicating the gradient of the plane. Identifying, specifying the vector V as information indicating the gradient related to the plane, and generating, as one element of the feature information, a correspondence between the vector V and the information indicating the focused position; The data feature extraction device further determines a unit vector S starting from one position Q on the xyz three-dimensional coordinate space and heading in one direction, and sets each unit vector S associated with each vector V generated by the feature information generation unit. With respect to the information, a vector PQ indicating the relative positional relationship of the position Q in the xyz three-dimensional coordinate space is determined based on the focused position related to the information, and the vector V and the vector PQ are determined. Matching data for generating information indicating a plurality of sets of torque PQs and a unit vector S as matching data for use in matching the three-dimensional data acquired by the three-dimensional data acquiring means with other three-dimensional data. It may include a generation unit.
[0083]
As a result, data that can be directly used for collation is generated.
Further, the collation data generating means uniformly multiplies the magnitude of each vector PQ by a constant so that the magnitude of the largest vector PQ among the plurality of sets of vectors PQ becomes 1. Information indicating each vector PQ, each vector V, and the unit vector S obtained as a result of multiplication by a constant may be generated as the comparison data.
[0084]
According to the data generated in this way, it is possible to appropriately determine when two data to be collated indicate the same target and are in a relation of enlargement / reduction.
Further, a data collation device according to the present invention is a data collation device for collating three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space, and is a first device for acquiring first three-dimensional data. Three-dimensional data acquisition means, first scanning means for sequentially focusing on each position on the xy two-dimensional plane, and each time one position is focused on by the first scanning means, it is located around the focused position. Xyz three-dimensional coordinates of a predetermined number of four or more local regions on the xy two-dimensional plane, among three-dimensional positions indicated by the first three-dimensional data, three-dimensional positions having x and y coordinates in the local region An average position in the space is determined, and whether or not each of the average positions determined for all the local regions is substantially on the same plane in the xyz three-dimensional coordinate space; The determination is made, and the vector V on the xy two-dimensional plane indicating the gradient of the plane with respect to the xy two-dimensional plane is specified only when the plane V exists on the substantially same plane, and the vector V and the information indicating the focused position are determined. A first feature information generating unit to be associated, and a unit vector S starting from one position Q in the xyz three-dimensional coordinate space and heading in one direction are determined and associated with each vector V specified by the first feature information generating unit. For each piece of information obtained, a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space is determined based on a focused position related to the information, and a plurality of sets of the vector V and the vector PQ and a unit vector S are shown. Matching data generating means for generating matching data; second three-dimensional data obtaining means for obtaining second three-dimensional data; A second scanning means for sequentially focusing on each of the above positions, and a predetermined number of four or more xy two-dimensional planes located around the focused position each time one position is focused on by the second scanning means For each of the local regions in, among the three-dimensional positions indicated by the second three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region is determined. It is determined whether or not each of the average positions obtained for the local area is substantially on the same plane in the xyz three-dimensional coordinate space. Only when the average positions are on substantially the same plane, the xy2 dimension of the plane is determined. A second feature information generating unit that specifies a vector V ′ on an xy two-dimensional plane indicating a gradient with respect to the plane and associates the vector V ′ with information indicating the focused position; By sequentially applying each vector V in the matching data to each vector V ′ specified by the second feature information generating means, the azimuth relationship between the vector V ′ and the new unit vector S ′ becomes a vector. Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V applied to V ′ and the unit vector S, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, It is characterized by comprising a matching means for determining a similar tendency between the first three-dimensional data and the second three-dimensional data.
[0085]
This makes it possible to appropriately collate the three-dimensional data having a rotational relationship.
Here, the matching unit may perform the fitting of each vector V to each vector V ′ only when the difference between the magnitude of the vector V ′ and the magnitude of the vector V is less than a predetermined value. Good.
[0086]
Thereby, the matching can be performed more accurately.
Further, the data collation device further includes a cube region where the starting point of the unit vector S ′ determined by the collation unit is the largest among the cube regions when the xyz three-dimensional coordinate space is divided into cube regions of a predetermined size. In addition, a recording unit that records the number of unit vectors S ′ including the starting point on a recording medium may be provided.
[0087]
As a result, the similarity between the data to be collated can be represented by numerical values. Therefore, it is possible to specify the data having the highest similarity.
Further, when the xyz three-dimensional coordinate space is divided into cubic regions having a predetermined size, the data collating device further synthesizes all unit vectors S ′ including the starting point in the cubic region for each cubic region. It is also possible to provide a recording means for obtaining the magnitude of the obtained combined vector and recording the maximum value of the obtained combined vector magnitudes on a recording medium.
[0088]
As a result, an index that can evaluate the similarity between data with a relatively high degree of accuracy can be obtained.
Further, instead of the collation data generating means, a unit vector S starting from one position Q on the xy two-dimensional plane and heading in one direction is determined, and each vector V specified by the first characteristic information generating means is determined. For each piece of associated information, a vector PQ indicating a relative positional relationship of the position Q on the xy two-dimensional plane is determined based on the x coordinate and the y coordinate of the focused position related to the information, and a plurality of sets of the vector V and the vector PQ are determined. And a two-dimensional collation data generating unit for generating collation data indicating the unit vector S and the unit vector S. In place of the collation unit, the collation is performed on each vector V ′ specified by the second feature information generation unit. Vector V in the application data is sequentially applied, so that the azimuth relationship between the vector V ′ and the new unit vector S ′ is a vector obtained by applying the vector V ′. Each unit vector S ′ is determined so as to have the same azimuthal relationship between V and the unit vector S, and the first three-dimensional data and the first three-dimensional data are determined according to the distribution state of all the determined unit vectors S ′ in the xy two-dimensional plane. The image processing apparatus may further include a two-dimensional matching unit that determines a similar tendency with the second three-dimensional data.
[0089]
This makes it possible to easily perform a rough collation.
Further, the data collation device further includes a line drawing data acquisition unit that acquires binary image data indicating a position on the xy two-dimensional plane and binary data for each pixel arranged on the xy two-dimensional plane; Binary image scanning means for sequentially focusing on each position on an xy two-dimensional plane in which pixels related to image data are arranged, and a binary image scanning means which is included in a region of a predetermined size centered on the position focused by the binary image scanning means. Among the pixels related to the binary image data, the attribute value is determined such that the attribute value decreases as the number of pixels for which the binary data indicates the first value increases, and Attribute value determining means for associating the attribute value with the attribute value, and each set of the position on the xy two-dimensional plane and the attribute value associated by the attribute value determining means as the three-dimensional data, Transmission means for transmitting the three-dimensional data to the three-dimensional data acquisition means, wherein the three-dimensional position is indicated by an x coordinate and a y coordinate of each position constituting the three-dimensional data and a z coordinate which is an attribute value corresponding to the position. It is good also as being something.
[0090]
Thus, for example, when collating image data generated by scanning a line drawing representing the outline of an object by hand on paper with a scanner, and image data obtained by imaging the same object with a digital camera Even when a line drawn as a solid line becomes a broken line in the process of being scanned by the scanner, it is appropriate that the similarity between the two image data is higher than when the conventional image matching technology is applied. Is more likely to be determined.
[Brief description of the drawings]
FIG. 1 is a functional block diagram of an image matching device according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating a relationship between a line drawing portion and a first scanning point in search reference data.
FIG. 3 is a diagram illustrating an example of a relationship between a second scanning point used in a feature extraction process and each of first to fourth quadrants;
FIG. 4 is a flowchart illustrating a first process of acquiring search reference data and generating the matching data;
FIG. 5 is a diagram exemplifying V (i) and P (i) generated by a feature data generating unit 103, and a point Q and a vector S defined by a matching data generation and holding unit 105;
FIG. 6 is a flowchart showing a second process of identifying the data to be searched and comparing the search reference data with the data to be searched using the data for matching.
FIG. 7 shows V (i), PQ (i) and V (i) held by the matching data generation and holding unit 105 for each set of V ′ (j) and P ′ (j) generated by the feature data generating unit 109; FIG. 11 is a diagram illustrating a state where a unit vector S ′ (i, j) is determined by applying the relationship of S;
[Explanation of symbols]
100 Image collation device
101 Search criteria data acquisition unit
102 luminance level determination unit
103 Feature Data Generation Unit
104 Range Limiter
105 collation data generation / holding unit
106 Searched Data Acquisition Unit
107 Line drawing unit
108 brightness level determination unit
109 Feature Data Generator
110 collation unit
111 Evaluation value recording section

Claims (21)

What is claimed is: 1. A data feature extraction device for generating feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
Three-dimensional data acquisition means for acquiring three-dimensional data;
scanning means for sequentially focusing on each position on an xy two-dimensional plane;
Each time one position is noted by the scanning means, for each of four or more predetermined number of local regions on an xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x-coordinate and the y-coordinate in the local region is determined, and each average determined for all the local regions is obtained. It is determined whether or not all the positions are substantially on the same plane in the xyz three-dimensional coordinate space. Only when the positions are substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane and the attention A feature information generating unit configured to generate, as one element of the feature information, a piece associated with the information indicating the designated position. The three-dimensional data is image data indicating a position on the xy two-dimensional plane and a luminance value for each pixel arranged on the xy two-dimensional plane,
The three-dimensional position is indicated by a z-coordinate which is a luminance value of the pixel, and an x-coordinate and a y-coordinate of a position of the pixel on an xy two-dimensional plane,
The scanning unit sequentially focuses on each position in a region on the xy two-dimensional plane where the pixel related to the image data is arranged,
The feature information generating means includes, for each of four or more predetermined local areas on the xy two-dimensional plane located around the position noted by the scanning means, an xy two-dimensional image having x and y coordinates in the local area. The average position on the xy two-dimensional plane and the average value of the luminance values of all the pixels arranged at the positions on the plane are determined as the average positions in the xyz three-dimensional coordinate space, and the respective averages determined for all the local regions are obtained. It is determined whether or not the target positions are substantially on the same plane in the xyz three-dimensional coordinate space. Only when the target positions are substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane is provided. 2. The data according to claim 1, wherein the information associated with the information indicating the focused position is generated as one element of the characteristic information. Feature extraction apparatus. The data feature extraction device further comprises:
a line drawing data acquisition unit for acquiring binary image data indicating a position on the xy two-dimensional plane and binary data for each pixel arranged on the xy two-dimensional plane;
A binary image scanning means for sequentially focusing on each position on an xy two-dimensional plane where pixels related to the binary image data are arranged;
Among the pixels related to the binary image data included in a region of a predetermined size centered on the position noted by the binary image scanning unit, the pixel whose binary data indicates the first value for the pixel Attribute value determining means for determining an attribute value so that the attribute value becomes lower as the number increases, and associating the noted position with the attribute value;
Transmitting means for transmitting each set of the position on the xy two-dimensional plane and the attribute value associated with the attribute value determining means to the three-dimensional data acquisition means as the three-dimensional data;
The said three-dimensional position is shown by the x coordinate and y coordinate of each position which comprises the said three-dimensional data, and the z coordinate which is an attribute value corresponding to the said position, The characterized by the above-mentioned. Data feature extraction device. The three-dimensional data is image data indicating a position on the xy two-dimensional plane and an attribute value for each pixel arranged on the xy two-dimensional plane,
The three-dimensional position is indicated by a z-coordinate which is an attribute value of the pixel, and an x-coordinate and a y-coordinate of a position of the pixel on an xy two-dimensional plane,
The scanning unit sequentially focuses on each position in a region on the xy two-dimensional plane where the pixel related to the image data is arranged,
The feature information generating means includes:
A first locality between a negative direction of x and a negative direction of y from the reference with respect to the position within a certain range on an xy two-dimensional plane centered on the position focused by the scanning unit. Average value z1 of each attribute value corresponding to the area, average value z2 of each attribute value corresponding to the second local area between the positive direction of x and the negative direction of y, and the negative value of x and y Average value z3 of each attribute value corresponding to the third local region between the positive direction and average value of each attribute value corresponding to the fourth local region between the positive direction of x and the positive direction of y z4 and the respective average positions are determined. When the value of the expression | z1-z2-z3 + z4 | is equal to or less than a predetermined value d1, each of the average positions determined for all of the local regions is determined. It is determined that they are substantially on the same plane in the xyz three-dimensional coordinate space, 2. The data feature extraction according to claim 1, wherein information indicating a gradient of the plane with respect to the xy two-dimensional plane and information indicating a focused position are generated as one element of the feature information. apparatus. The area of the certain range according to the feature information generating means is a circular area of a certain radius centered on the position noted by the scanning means,
The feature information generating unit further rotates each of the first to fourth regions defined based on the position focused by the scanning unit by a predetermined angle around the reference position, and after the rotation, Average values z1 ′, z2 ′, z3 ′ and z4 ′ of the respective attribute values corresponding to the first to fourth regions of
The characteristic information generating means is configured to determine that the value of the expression | z1-z2-z3 + z4 | is equal to or less than a predetermined value d1, and that the value of the expression | z1'-z2'-z3 '+ z4' | When the value is equal to or less than the value d1, it is determined that each of the average positions obtained for all the local regions is substantially on the same plane in the xyz three-dimensional coordinate space, and the gradient of the plane with respect to the xy two-dimensional plane is determined. 5. The data feature extraction device according to claim 4, wherein information that indicates the indicated information and the information indicating the focused position is generated as one element of the feature information. The feature information generating means further associates information indicating the gradient with information indicating the focused position only when the gradient of the plane with respect to the xy two-dimensional plane is equal to or greater than a predetermined amount, The data feature extraction device according to claim 5, wherein the data feature is generated as one element of the feature information. The characteristic information generating means, when all the average positions obtained for all the local regions are substantially on the same plane, calculates a mathematical expression representing the plane as a · x + by · y + cz · d = (a, (b, c, and d are constant values), a vector V on an xy two-dimensional plane having -a / c and -b / c as x and y components, respectively, is obtained. 2. The data feature extraction device according to claim 1, wherein the unit vector specified is specified as information indicating the gradient related to the plane. The characteristic information generation means specifies a vector V on an xy two-dimensional plane indicating the gradient related to the plane when all the average positions obtained for all the local regions are substantially on the same plane. , Specifying the vector V as information indicating the gradient related to the plane, and generating a correspondence between the vector V and the information indicating the focused position as one element of the feature information,
The data feature extraction device further comprises:
A unit vector S heading in one direction starting from one position Q in the xyz three-dimensional coordinate space is determined, and attention is paid to each piece of information associated with each vector V generated by the feature information generating means. A vector PQ indicating the relative positional relationship of the position Q in the xyz three-dimensional coordinate space is determined with reference to the position, and information indicating a plurality of sets of the vector V and the vector PQ and the unit vector S is obtained by the three-dimensional data obtaining means. 2. The data feature extraction device according to claim 1, further comprising a comparison data generation unit that generates as comparison data for use in comparing the obtained three-dimensional data with other three-dimensional data. The collation data generation means,
Of the plurality of sets of vectors PQ, the magnitude of each vector PQ is uniformly multiplied by a constant so that the magnitude of the vector PQ having the largest magnitude becomes 1, and each vector PQ obtained by multiplying the magnitude by a constant is obtained. 9. The data feature extracting apparatus according to claim 8, wherein information indicating each of the vector V and the unit vector S is generated as the comparison data. A data collation device for collating three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
First three-dimensional data acquisition means for acquiring first three-dimensional data;
first scanning means for sequentially focusing on each position on an xy two-dimensional plane,
Each time one position is noted by the first scanning means, for each of four or more predetermined number of local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the first three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. Only when they are on substantially the same plane, xy2 indicating the gradient of the plane with respect to the xy2D plane is determined. First feature information generating means for specifying a vector V on a two-dimensional plane and associating the vector V with information indicating the focused position;
A unit vector S heading in one direction starting from one position Q in the xyz three-dimensional coordinate space is determined, and each piece of information associated with each vector V specified by the first feature information generating means is related to the information. Matching data for determining a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space based on the focused position and generating matching data indicating a plurality of sets of the vector V and the vector PQ and the unit vector S Generating means;
Second three-dimensional data acquisition means for acquiring second three-dimensional data;
second scanning means for sequentially focusing on each position on the xy two-dimensional plane,
Every time one position is noted by the second scanning means, for each of a predetermined number of four or more local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the second three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. Only when they are on substantially the same plane, xy2 indicating the gradient of the plane with respect to the xy2D plane is determined. A second feature information generating unit that specifies a vector V ′ on a two-dimensional plane and associates the vector V ′ with information indicating the focused position;
By sequentially applying each vector V in the matching data to each vector V ′ specified by the second feature information generating means, the azimuth relationship between the vector V ′ and the new unit vector S ′ becomes vector V ′, Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V applied to ′ and the unit vector S, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, A data collation device comprising: a collation unit configured to determine a similarity tendency between the first three-dimensional data and the second three-dimensional data. The said matching means performs the said fitting of each vector V with respect to each vector V 'only when the difference of the magnitude | size of the said vector V' and the magnitude | size of the said vector V is less than a predetermined value. Item 11. The data collation device according to Item 10. In the case where the xyz three-dimensional coordinate space is divided into cubic regions of a predetermined size, the data collating device further includes, in each cubic region, a cubic region that includes the largest number of starting points of the unit vector S ′ determined by the collating unit, 11. The data collating apparatus according to claim 10, further comprising recording means for recording the number of unit vectors S 'including the starting point on a recording medium. In the case where the xyz three-dimensional coordinate space is divided into cubic regions of a predetermined size, the data matching device is obtained as a result of synthesizing all the unit vectors S ′ including the starting point in the cubic region for each cubic region. 11. The data collating apparatus according to claim 10, further comprising a recording unit that obtains the magnitude of the combined vector and records the maximum value of the obtained combined vector magnitudes on a recording medium. The data collation device according to claim 10,
Instead of the collation data generating means,
A unit vector S heading in one direction starting from one position Q on the xy two-dimensional plane is determined, and attention is paid to each piece of information associated with each vector V specified by the first feature information generating means. A vector PQ indicating the relative positional relationship of the position Q on the xy two-dimensional plane is determined based on the x-coordinate and the y-coordinate of the determined position, and comparison data indicating a plurality of sets of the vector V and the vector PQ and the unit vector S is generated. Two-dimensional collation data generation means,
Instead of the matching means,
By sequentially applying each vector V in the matching data to each vector V ′ specified by the second feature information generating means, the azimuth relationship between the vector V ′ and the new unit vector S ′ becomes vector V Are determined so as to be the same as the azimuth relationship between the vector V and the unit vector S applied to ′, and the first unit vector S ′ is determined according to the distribution state of all the determined unit vectors S ′ in the xy two-dimensional plane. A two-dimensional matching unit for determining a similarity tendency between the three-dimensional data and the second three-dimensional data. The data collation device further comprises:
a line drawing data acquisition unit for acquiring binary image data indicating a position on the xy two-dimensional plane and binary data for each pixel arranged on the xy two-dimensional plane;
A binary image scanning means for sequentially focusing on each position on an xy two-dimensional plane where pixels related to the binary image data are arranged;
Among the pixels related to the binary image data included in a region of a predetermined size centered on the position noted by the binary image scanning unit, the pixel whose binary data indicates the first value for the pixel Attribute value determining means for determining an attribute value so that the attribute value becomes lower as the number increases, and associating the noted position with the attribute value;
Transmitting means for transmitting each set of the position on the xy two-dimensional plane and the attribute value associated with the attribute value determining means to the first three-dimensional data acquiring means as the three-dimensional data;
The said three-dimensional position is shown by the x coordinate and y coordinate of each position which comprises the said three-dimensional data, and the z coordinate which is an attribute value corresponding to the said position, The characterized by the above-mentioned. Data collation device. A data feature extraction method for generating feature information indicating features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
A three-dimensional data acquisition step of acquiring three-dimensional data;
a scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted by the scanning step, for each of four or more predetermined number of local areas on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x-coordinate and the y-coordinate in the local region is determined, and each average determined for all the local regions is obtained. It is determined whether or not all the positions are substantially on the same plane in the xyz three-dimensional coordinate space. Only when the positions are substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane and the attention A feature information generating step of generating, as an element of the feature information, a piece of information associated with the information indicating the designated position. A data matching method for matching three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
A first three-dimensional data obtaining step of obtaining first three-dimensional data;
a first scanning step of sequentially focusing on each position on an xy two-dimensional plane;
Each time one position is noted in the first scanning step, for each of a predetermined number of four or more local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the first three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. Only when they are on substantially the same plane, xy2 indicating the gradient of the plane with respect to the xy2D plane is determined. A first feature information generating step of identifying a vector V on a two-dimensional plane and associating the vector V with information indicating the focused position;
A unit vector S heading in one direction starting from one position Q in the xyz three-dimensional coordinate space is determined, and each piece of information associated with each vector V specified in the first feature information generation step is related to the information. Matching data for determining a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space based on the focused position and generating matching data indicating a plurality of sets of the vector V and the vector PQ and the unit vector S Generating step;
A second three-dimensional data acquisition step of acquiring second three-dimensional data;
a second scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted in the second scanning step, the second three-dimensional data is obtained for each of four or more predetermined local areas on the xy two-dimensional plane located around the noted position. Among the three-dimensional positions shown, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x-coordinate and the y-coordinate in the local region is obtained, and each of the average positions obtained for all the local regions is: It is determined whether or not the plane is substantially on the same plane in the xyz three-dimensional coordinate space. Only when the plane is substantially on the same plane, the vector V ′ on the xy2 plane showing the gradient of the plane with respect to the xy2 plane is determined. A second feature information generation step of identifying the vector V ′ and the information indicating the focused position,
By sequentially applying each vector V in the comparison data to each vector V ′ specified in the second feature information generation step, the azimuth relationship between the vector V ′ and the new unit vector S ′ is determined by the vector V ′, Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V applied to ′ and the unit vector S, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, A collating step of determining a similar tendency between the first three-dimensional data and the second three-dimensional data. A computer program for causing a computer to perform data feature extraction processing for generating feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
The data feature extraction process includes:
A three-dimensional data acquisition step of acquiring three-dimensional data;
a scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted by the scanning step, for each of four or more predetermined number of local areas on the xy two-dimensional plane located around the noted position,
The average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region among the three-dimensional positions indicated by the three-dimensional data is obtained, and each average obtained for all the local regions is obtained. It is determined whether or not all the positions are substantially on the same plane in the xyz three-dimensional coordinate space. Only when the positions are substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane and the attention A feature information generating step of generating, as one element of the feature information, a piece of information associated with the information indicating the designated position. A computer program for causing a computer to perform data collation processing relating to collation of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
The data collation processing includes:
A first three-dimensional data obtaining step of obtaining first three-dimensional data;
a first scanning step of sequentially focusing on each position on an xy two-dimensional plane;
Each time one position is noted in the first scanning step, for each of a predetermined number of four or more local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the first three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. Only when they are on substantially the same plane, xy2 indicating the gradient of the plane with respect to the xy2D plane is determined. A first feature information generating step of identifying a vector V on a two-dimensional plane and associating the vector V with information indicating the focused position;
A unit vector S heading in one direction starting from one position Q in the xyz three-dimensional coordinate space is determined, and each piece of information associated with each vector V specified in the first feature information generation step is related to the information. Matching data for determining a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space based on the focused position and generating matching data indicating a plurality of sets of the vector V and the vector PQ and the unit vector S Generating step;
A second three-dimensional data acquisition step of acquiring second three-dimensional data;
a second scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted in the second scanning step, for each of four or more predetermined number of local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the second three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. A second feature information generating step of specifying a vector V ′ on a two-dimensional plane and associating the vector V ′ with information indicating the focused position;
By sequentially applying each vector V in the matching data to each vector V ′ specified in the second feature information generation step, the azimuth relationship between the vector V ′ and the new unit vector S ′ is determined by the vector V ′, Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V and the unit vector S applied to ′ ′, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, A computer program, comprising: a collation step of judging a similar tendency between the first three-dimensional data and the second three-dimensional data. A computer-readable storage medium storing a computer program for causing a computer to perform data feature extraction processing for generating feature information representing features of three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space. ,
The data feature extraction process includes:
A three-dimensional data acquisition step of acquiring three-dimensional data;
a scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted by the scanning step, for each of four or more predetermined number of local areas on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x-coordinate and the y-coordinate in the local region is determined, and each average determined for all the local regions is obtained. It is determined whether or not all the positions are substantially on the same plane in the xyz three-dimensional coordinate space. Only when the positions are substantially on the same plane, information indicating the gradient of the plane with respect to the xy two-dimensional plane and the attention A feature information generating step of generating, as one element of the feature information, a piece associated with the information indicating the designated position. A computer-readable recording medium recording a computer program for causing a computer to perform a data matching process for matching three-dimensional data indicating a plurality of three-dimensional positions in an xyz three-dimensional coordinate space,
The data collation processing includes:
A first three-dimensional data obtaining step of obtaining first three-dimensional data;
a first scanning step of sequentially focusing on each position on an xy two-dimensional plane;
Each time one position is noted in the first scanning step, for each of a predetermined number of four or more local regions on the xy two-dimensional plane located around the noted position,
Among the three-dimensional positions indicated by the first three-dimensional data, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x coordinate and the y coordinate in the local region was determined, and the average position was determined for all the local regions. It is determined whether or not each of the average positions is substantially on the same plane in the xyz three-dimensional coordinate space. Only when they are on substantially the same plane, xy2 indicating the gradient of the plane with respect to the xy2D plane is determined. A first feature information generating step of identifying a vector V on a two-dimensional plane and associating the vector V with information indicating the focused position;
A unit vector S heading in one direction starting from one position Q in the xyz three-dimensional coordinate space is determined, and each piece of information associated with each vector V specified in the first feature information generation step is related to the information. Matching data for determining a vector PQ indicating a relative positional relationship of the position Q in the xyz three-dimensional coordinate space based on the focused position and generating matching data indicating a plurality of sets of the vector V and the vector PQ and the unit vector S Generating step;
A second three-dimensional data acquisition step of acquiring second three-dimensional data;
a second scanning step of sequentially focusing on each position on the xy two-dimensional plane;
Each time one position is noted in the second scanning step, the second three-dimensional data is obtained for each of four or more predetermined local areas on the xy two-dimensional plane located around the noted position. Among the three-dimensional positions shown, the average position in the xyz three-dimensional coordinate space of the three-dimensional position having the x-coordinate and the y-coordinate in the local region is obtained, and each of the average positions obtained for all the local regions is: It is determined whether or not the plane is substantially on the same plane in the xyz three-dimensional coordinate space. Only when the plane is substantially on the same plane, the vector V ′ on the xy2 plane showing the gradient of the plane with respect to the xy2 plane is determined. A second feature information generation step of identifying the vector V ′ and the information indicating the focused position,
By sequentially applying each vector V in the matching data to each vector V ′ specified in the second feature information generation step, the azimuth relationship between the vector V ′ and the new unit vector S ′ is determined by the vector V ′, Each unit vector S ′ is determined so as to have the same azimuthal relationship between the vector V and the unit vector S applied to ′ ′, and according to the distribution state of all the determined unit vectors S ′ in the xyz three-dimensional coordinate space, A collating step of judging a similar tendency between the first three-dimensional data and the second three-dimensional data.

Images