JP2007249555A - Image processor and approximate straight line calculation apparatus - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve a highly accurate approximate straight line by efficiently removing an outlier from coordinates included in a coordinate group. <P>SOLUTION: A CPU acquires the coordinate group composed of the coordinates indicating an edge position (step S2) and selects three or more coordinates from the coordinate group according to a prescribed rule (step S4). Then, the CPU calculates an evaluation value indicating the degree of being arranged on the same straight line for the selected coordinates (step S6), and cumulatively adds the evaluation value in association with each of the selected coordinates (step S8). Further, the CPU repeatedly executes the steps S4-S10 so that the respective coordinates constituting the coordinate group are selected for the same number of times. Thereafter, the CPU extracts the plurality of coordinates with the relatively high degree of being arranged on the same straight line from the coordinate group on the basis of the cumulatively added evaluation values (cumulative values) of the respective coordinates (step S12) and calculates the approximate straight line (step S14). <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing device and an approximate line calculation device that calculate an approximate line from a group of coordinates including a plurality of coordinates, and in particular, efficiently removes an outlier that causes an error and calculates the approximate line with high accuracy. It is about technology.

  In the manufacturing field, automation is promoted from the viewpoint of labor saving and high efficiency. In order to realize such automation, many sensors are used. In particular, an image processing apparatus that can photograph a product, a semi-finished product, and the like and process the photographed image to perform dimension measurement inspection and positioning of the product is effective.

  In such an image processing apparatus, an edge (contour) of an object (product, semi-finished product, etc.) is detected based on information (lightness, brightness, color difference, etc.) in the captured image. In particular, edge detection is often performed on a linear portion of an object. This is because most of the objects include a linear portion, and by using the linear portion as a reference, highly accurate dimensional measurement and positioning can be easily realized.

  However, an erroneous edge position is often detected due to dirt in the background of the object or a defect (molding burr or chipping) of the object. In view of this, it is common practice to calculate an approximate straight line indicating a linear shape portion of an object based on a plurality of edge positions detected within a predetermined region. For example, Japanese Patent No. 2850608 (Patent Document 1) discloses that the coordinates of a plurality of edge points are measured and linear approximation is performed.

Furthermore, in order to further improve the accuracy of linear approximation, for example, an LTS estimation method (Least Trimmed Squares Estimate Method) as disclosed in Non-Patent Document 1 has been proposed. According to the LTS estimation method, a provisional approximate line is calculated from a plurality of edge positions by the least square method, and a plurality of edge positions having a small distance (residual) from each edge position to the provisional approximate line are extracted. . Then, an approximate straight line is calculated again from the extracted edge position by the least square method. According to this LTS estimation method, it is possible to reduce the influence of outliers that deviate significantly from the original values.
Japanese Patent No. 2850608 Aya Sakuma, "Research on Robustness Theory of Regression Analysis and its Application", Summary of 2004 Graduation Thesis, March 7, 2005, Nanzan University, Department of Mathematical Sciences, Internet <URL: http: // www .seto.nanzan-u.ac.jp / msie / gr-thesis / ms / 2004 / kimura / 01mm075.pdf>

  However, even if the above-described LTS estimation method is used, if a large number of outliers are included in a plurality of detected edge positions, the provisional approximate line itself is greatly affected by errors. For this reason, even if the edge position is normally normal, the distance from the provisional approximate straight line may be large. In such a case, the normal normal edge position is removed, and an error is caused based on the outlier. In some cases, a large approximate straight line is calculated.

  Accordingly, the present invention has been made to solve such a problem, and an object of the present invention is to perform image processing capable of calculating an approximate straight line with high accuracy by efficiently removing outliers from coordinates included in a coordinate group. An apparatus and an approximate line calculation device are provided.

  According to the first invention, a coordinate group acquisition unit that acquires a coordinate group including four or more coordinates indicating a specific position of an image, and a coordinate selection unit that selects three or more coordinates from the coordinate group according to a predetermined rule And, with respect to three or more coordinates selected by the coordinate selection means, an evaluation value calculation means for calculating an evaluation value indicating the degree of arrangement on the same straight line, and each of the three or more coordinates, In the coordinate value selecting means, the evaluation value calculating means, and the evaluation value adding means, the evaluation value adding means for accumulating the evaluation values calculated by the evaluation value calculating means and the coordinates of the coordinate group are selected the same number of times. Repetitive means for repeatedly executing the processing, and extraction means for extracting a plurality of coordinates having a relatively high degree of arrangement on the same straight line from the coordinate group based on the evaluation values accumulated and added by the evaluation value addition means , Extraction An image processing apparatus and a approximate line calculating means for calculating an approximate line based on a plurality of coordinates extracted by stage.

  According to the first invention, an evaluation value indicating the degree of arrangement on the same straight line is calculated for each combination of three or more coordinates selected from the coordinate group. As described above, the evaluation value is not calculated based on the provisional approximate straight line calculated from the entire coordinate group, but is calculated microscopically between three or more selected coordinates. Coordinates that are outliers are less likely to be placed on the same straight line in any combination, while normal coordinates are more likely to be placed on the same straight line in a combination of normal coordinates. As a result, since the evaluation values are accumulated in association with each of the coordinates, the difference between the normal coordinates and the coordinates of the outlier appears significantly, so that the outlier can be efficiently removed.

  Preferably, the coordinate group acquisition unit includes an image acquisition unit that acquires an image, an edge detection unit that acquires a coordinate group by detecting an edge position that causes a change in color information in the image acquired by the image acquisition unit, and including.

  Preferably, the image acquisition unit acquires an image obtained by photographing an object having a linear part.

  Preferably, the evaluation value calculation means defines a straight line including two coordinates among the three or more coordinates, and calculates an evaluation value based on a distance between each of the remaining coordinates and the defined straight line.

  Preferably, the evaluation value calculation means calculates an evaluation value based on a correlation coefficient for three or more coordinates.

  Preferably, the repeating unit repeatedly executes the processes in the coordinate selecting unit, the evaluation value calculating unit, and the evaluation value adding unit until all the combinations for selecting three or more coordinates from the coordinate group are selected by the coordinate selecting unit. .

Preferably, the coordinate group includes two-dimensional or three-dimensional coordinates.
According to the second invention, the coordinate group acquisition means for acquiring a coordinate group including four or more coordinates, the coordinate selection means for selecting three or more coordinates from the coordinate group according to a predetermined rule, and the coordinate selection means For the three or more selected coordinates, the evaluation value calculating means for calculating the evaluation value indicating the degree of arrangement on the same straight line and the evaluation value calculating means in association with each of the three or more coordinates. An evaluation value adding means for accumulating the evaluation values to be accumulated, and a process for repeatedly executing the processes in the coordinate selecting means, the evaluation value calculating means and the evaluation value adding means so that each coordinate of the coordinate group is selected the same number of times. Based on the evaluation value cumulatively added by the evaluation value addition means, an extraction means for extracting a plurality of coordinates arranged on the same straight line from the coordinate group and having a relatively high degree, and extracted by the extraction means The An approximate straight line calculation device and a approximate line calculating means for calculating an approximate line based on the number of coordinates.

  According to the second invention, for each combination of three or more coordinates selected from the coordinate group, an evaluation value indicating the degree of arrangement on the same straight line is calculated. As described above, the evaluation value is not calculated based on the provisional approximate straight line calculated from the entire coordinate group, but is calculated microscopically between three or more selected coordinates. Coordinates that are outliers are less likely to be placed on the same straight line in any combination, while normal coordinates are more likely to be placed on the same straight line in a combination of normal coordinates. As a result, since the evaluation values are accumulated in association with each of the coordinates, the difference between the normal coordinates and the coordinates of the outlier appears significantly, so that the outlier can be efficiently removed.

  According to the present invention, it is possible to realize an image processing device and an approximate line calculation device that can efficiently remove outliers from coordinates included in a coordinate group and calculate an approximate line with high accuracy.

  Embodiments of the present invention will be described in detail with reference to the drawings. Note that the same or corresponding parts in the drawings are denoted by the same reference numerals and description thereof will not be repeated.

[Embodiment 1]
FIG. 1 is a schematic configuration diagram of an image processing apparatus 100 including an approximate straight line calculation apparatus according to Embodiment 1 of the present invention.

  With reference to FIG. 1, the image processing apparatus 100 acquires an image photographed by the camera 2, calculates an approximate straight line, and then outputs the calculation result to the display device 4 and the external device 6.

  As an example, the camera 2 includes an imaging element such as a CCD (Coupled Charged Device) or a CMOS (Complementary Metal Oxide Semiconductor) sensor and a lens. The captured image is output to the image processing apparatus 100. The image captured by the camera 2 may be either a grayscale image (black and white image) or a color image, and may be either a still image or a moving image. In the first embodiment of the present invention, a case where a color image is used will be described.

  The display device 4 displays the approximate straight line calculated by the image processing device 100, the detected edge position, and the like together with the image taken by the camera 2. As an example, the display device 4 includes a liquid crystal display (LCD), a plasma display, an organic EL display (Electro Luminescence display), and the like.

  The external device 6 includes, for example, a programmable controller that monitors and controls a production line and the like, and executes various types of control and measurement according to the approximate straight line calculated in the image processing apparatus 100.

  The image processing apparatus 100 includes a CPU 10, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 14, an HDD (Hard Disk Drive) 16, an input / output unit 18, and an image input unit 20. , A filter 22, an image memory 24, an edge detection unit 26, and an image output unit 28. The units other than the image input unit 20 are connected to each other via the bus 30. The image processing apparatus 100 is realized by a personal computer or the like as an example.

  The image input unit 20 is electrically connected to the camera 2, receives a video signal captured by the camera 2, performs predetermined signal conversion processing, acquires color information of each pixel, and then filters the acquired color information 22 to output. Specifically, the image input unit 20 performs frame synchronization on the video signal received from the camera 2, demodulates the color information of each pixel developed and transmitted on the time axis, and determines the color for each pixel. Get information.

  The filter 22 receives images sequentially output from the image input unit 20 and performs filter processing such as color information conversion on each pixel of the image. Then, the filter 22 outputs the filtered image to the image memory 24 and the edge detection unit 26. As an example, the filter 22 is a LUT (Look-Up Table) type filter that performs processing in accordance with a predetermined rule (table) set by the CPU 10. For this reason, the filter 22 is configured to be able to change the content of the filter process in accordance with the object, shooting conditions, and the like.

  The image memory 24 temporarily stores an image output from the filter 22 at predetermined intervals or at a timing when a command is received from the CPU 10. The image memory 24 also temporarily stores the edge position received from the edge detection unit 26. Then, the image memory 24 outputs the stored image to the image output unit 28, the CPU 10, or the like.

  The edge detection unit 26 receives an instruction from the CPU 10 and detects an edge position (contour) that causes a change in the color information in the image based on the color information of each pixel constituting the image output from the filter 22. . Then, the edge detection unit 26 outputs the detected plurality of edge positions to the CPU 10 and the image memory 24 as a coordinate group. Furthermore, the edge detection unit 26 is configured to be able to detect an edge position with respect to a specific region of the image output from the filter 22 in accordance with a mask region given via the input / output unit 18 or the like. When a grayscale image (black and white image) is input, the color information is a grayscale value. When a color image is input, the color information is an RGB value (red, green, and blue grayscale values). , CMY values (cyan values of cyan, magenta, and yellow) or color variables (hue, value, and saturation).

  The image output unit 28 receives an image from the image memory 24 and various display information from the CPU 10, etc., generates a video signal, and outputs it to the display device 4. As an example, the image output unit 28 outputs an analog signal (composite signal or component signal) or a digital signal corresponding to the display device 4.

  The ROM 12 stores programs executed when the CPU 10 is activated and various settings.

  The RAM 14 stores programs executed by the CPU 10, work data related to approximate straight line calculation described later, and the like. The RAM 14 includes, for example, a semiconductor storage element such as a DRAM (Dynamic Random Access Memory) and an SRAM (Static Random Access Memory).

  The HDD 16 stores programs executed by the CPU 10, calculation results in the CPU 10, and the like, and reads and writes in response to requests from the CPU 10. Instead of the HDD 16, a semiconductor memory (flash memory card, SD memory card, IC memory card, etc.), an optical disk (CD (Compact Disk), a DVD (Digital Versatile Disc) -ROM / RAM / R / RW, MO ( A device capable of supporting data in a nonvolatile manner such as Magnetic Optical Disc) or MD (Mini Disc) may be used.

  The input / output unit 18 outputs a calculation result in the CPU 10 to the external device 6. As an example, the input / output unit 18 includes a contact output (DO) composed of a photodiode, a transistor, or a relay, a field bus, a USB (Universal Serial Bus), an RS-232C (Recommended Standard 232 version C), an IEEE ( It consists of communication means according to Institute of Electrical and Electronic Engineers (1394), SCSI (Small Computer System Interface), Ethernet (registered trademark), and the like.

  The CPU 10 acquires a coordinate group indicating a plurality of edge positions (two-dimensional coordinates) detected by the edge detection unit 26 with respect to the image captured by the camera 2. In particular, in Embodiment 1 of the present invention, the CPU 10 acquires a coordinate group including four or more edge positions. Then, the CPU 10 selects three or more coordinates from the acquired coordinate group according to a predetermined rule, and calculates an evaluation value indicating the degree of arrangement on the same straight line for the selected three or more coordinates. Further, the CPU 10 accumulates and adds the calculated evaluation values in association with each of the selected three or more evaluation values.

  Similarly, the CPU 10 sequentially selects the three or more coordinates and repeatedly executes the above-described process so that the coordinates of the coordinate group are selected the same number of times.

  Then, the CPU 10 extracts a plurality of coordinates having a relatively high degree of arrangement on the same straight line from the coordinate group based on the accumulated and added evaluation value, and calculates an approximate straight line based on the extracted coordinates.

  Further, the CPU 10 outputs the calculated approximate straight line to the display device 4 and the external device 6 through the image output unit 28 and the input / output unit 18, respectively.

  In the first embodiment of the present invention, the image input unit 20 corresponds to an “image acquisition unit”, and the image input unit 20 and the edge detection unit 26 implement a “coordinate group acquisition unit”. Further, the CPU 10 implements “coordinate selection means”, “evaluation value calculation means”, “evaluation value addition means”, “repetition means”, “extraction means”, and “approximate straight line calculation means”.

(Edge position detection process)
FIG. 2 is a diagram for explaining the outline of the edge position detection processing in the edge detection unit 26.

  With reference to FIG. 2, as an example, the edge detection unit 26 scans an image IMG taken by the camera 2 in the x direction, and sequentially detects edge positions.

  Specifically, the edge detection unit 26 defines a region 42.1 having a predetermined width around a predetermined scanning position 40.1 in the x direction in the image IMG. The edge detection unit 26 then associates the color information of the pixels included in the defined region 42.1 with the value in the y direction, and calculates a color information histogram 46.1 that is cumulatively added in the x direction. In the case of FIG. 2, the color information histogram 46.1 is calculated from the one-dimensional color information. Therefore, in the case of a color image, the one-dimensional lightness (red, blue, and green gray values are used as the color information. For example) can be used. Then, the edge detection unit 26 detects a change position 48.1 in the y direction that causes a sudden change in the calculated color information histogram 46.1.

By such a procedure, the edge detection unit 26 detects the edge position 44.1 which is the intersection of the scanning position 40.1 in the x direction and the change position 48.1 in the y direction, and this edge position 44.1. The coordinates (x ₁ , y ₁ ) are acquired.

Subsequently, the edge detection unit 26 defines a region 42.2 having a predetermined width around the next scanning position 40.2 in the x direction. Then, the edge detection unit 26 calculates the color information histogram 46.2 as described above. Further, the edge detection unit 26 detects a change position 48.2 in the y direction that causes a sudden change in the calculated color information histogram 46.2. The edge detection unit 26 detects the edge position 44.2 from the scan position 40.2 in the x direction and the change position 48.2 in the y direction, and acquires the coordinates (x ₂ , y ₂ ).

  Similarly, the edge detector 26 sequentially scans in the x direction, detects an edge position group 44 composed of a plurality of edge positions, and acquires a coordinate group composed of coordinates indicating each edge position.

  The edge detection unit 26 detects an edge position only for a specific region, not the entire region of the image IMG, according to a mask region (not shown) given via the input / output unit 18 (FIG. 1). It is also possible.

  The coordinate group acquired by the edge detection unit 26 in this way is output to the CPU 10.

(Approximate line calculation process)
FIG. 3 is a diagram for explaining the outline of the approximate straight line calculation process in the CPU 10. FIG. 3 schematically shows the image IMG stored in the image memory 24 (FIG. 1).

  Referring to FIG. 3, an image IMG is an image of an object OBJ having a linear shape part. As an example, it is assumed that the edge detection unit 26 performs edge detection in a preset mask region 50 and acquires coordinates a to e corresponding to the edge positions a to e, respectively. At this time, the edge detection unit 26 detects the edge positions a to d on the linear outline EDG of the object OBJ, while the noise region 52 such as dirt on the background of the object OBJ is the edge position e. Is erroneously detected. The CPU 10 calculates an approximate line according to the least square method based on the coordinates a to e corresponding to the edge positions a to e, respectively, and regards the calculated approximate line as the outline EDG of the object OBJ.

  However, the edge position e is an “outlier” greatly deviating from the contour line EDG, and the approximate straight line calculated including such coordinates essentially includes a large error. Therefore, the CPU 10 calculates an approximate straight line after efficiently removing such outliers.

  FIG. 4 is a flowchart showing a control structure related to approximate straight line calculation in the CPU 10.

  With reference to FIG. 4, the CPU 10 first acquires a coordinate group including coordinates indicating the edge position detected by the edge detection unit 26 (step S <b> 2). Then, the CPU 10 selects three or more coordinates from the coordinate group according to a predetermined rule (step S4). Then, CPU10 calculates the evaluation value which shows the degree arrange | positioned on the same straight line about the selected coordinate (step S6). That is, the CPU 10 calculates an evaluation value for the selected coordinates according to a method described later. Then, the CPU 10 accumulates and adds the calculated evaluation values in association with each of the selected coordinates (step S8).

  Further, the CPU 10 determines whether or not the coordinates constituting the coordinate group have been selected the same number of times (step S10). When the coordinates constituting the coordinate group are not selected the same number of times (NO in step S10), CPU 10 repeatedly executes steps S4 to S10 described above.

  When the coordinates constituting the coordinate group have been selected the same number of times (YES in step S10), the CPU 10 determines the coordinate group based on the evaluation value (cumulative value) of each coordinate that has been cumulatively added. To extract a plurality of coordinates having a relatively high degree of arrangement on the same straight line (step S12). Then, the CPU 10 calculates an approximate straight line according to the least square method based on the extracted plurality of coordinates (step S14). The calculation of the approximate straight line according to the least square method is a well-known method, and thus detailed description thereof is omitted. Further, the CPU 10 outputs the calculated approximate straight line to the display device 4 and the external device 6 (step S16), and ends the process.

  As described above, according to the present invention, since the evaluation values are cumulatively added in association with each coordinate, each coordinate must be selected the same number of times. Therefore, the CPU 10 sequentially selects combinations of three or more coordinates from the coordinate group according to a predetermined rule.

  In general, the combination of selecting different r coordinates from a group of coordinates including n coordinates is nCr = n! / {R! × (n−r)! } There are streets. Looking at all the combinations, each coordinate is selected nCr × r / n times. Therefore, as an example in which the coordinates constituting the coordinate group are selected the same number of times, the CPU 10 executes the above-described steps S6 and S8 for all combinations in which three or more coordinates are selected from the coordinate group.

  In the following, in order to facilitate understanding, a case where the number of coordinates selected for calculating the evaluation value is three will be exemplified.

(Evaluation value calculation processing and cumulative addition processing)
CPU10 calculates the evaluation value which shows the degree arrange | positioned on the same straight line with respect to three coordinates to select.

FIG. 5 is a diagram for explaining an example of processing related to calculation of an evaluation value.
FIG. 5A shows a case where evaluation values are calculated for the coordinates a, coordinates c, and coordinates e shown in FIG.

  FIG. 5B shows a case where evaluation values are calculated for the coordinates a, coordinates c, and coordinates d shown in FIG.

  With reference to FIG. 5A and FIG. 5B, the CPU 10 defines a straight line including two coordinates among the selected three coordinates. Then, the CPU 10 calculates an evaluation value based on the distance between the remaining coordinates and the straight line. That is, it can be said that the smaller the distance between the remaining coordinates and the straight line, the higher the selected three coordinates are arranged on the same straight line.

  Referring to FIG. 5A, when a coordinate e that is an outlier is included, the distance L between the straight line 60ae including the coordinate a and the coordinate e and the coordinate c that is the remaining coordinate is a relatively large value. Become.

  On the other hand, referring to FIG. 5B, in the case where each of the coordinates consists of coordinates a, coordinates c and coordinates d detected along the contour line EDG of the object OBJ, a straight line 60ad including the coordinates a and coordinates d. And the coordinate c which is the remaining coordinates are substantially zero.

Therefore, in Embodiment 1 of the present invention, an evaluation value that takes a higher value as the degree of arrangement on the same straight line is larger is adopted. Specifically, assuming that the three selected coordinates are coordinates i (x _i , y _i ), coordinates j (x _j , y _j ), and coordinates k (x _k , y _k ), coordinates i and coordinates The distance L _ijk between the straight line including k and the coordinate j is normalized by the distance from the coordinate i to the coordinate k, and the evaluation value V _ijk shown in the equation (1) is adopted.

This evaluation value V _ijk takes a maximum value of “1” when the coordinates i, j, and k are arranged on the same straight line. Note that the arrangement order of the coordinates i, the coordinates j, and the coordinates k may be any. For example, the coordinates j do not have to be arranged between the coordinates i and k.

Then, the CPU 10 accumulates and adds the evaluation values V _ijk to the accumulated values P (i), P (j), and P (k) in association with the selected coordinates i, j, and k, respectively. When the calculation of the evaluation values V _ijk for all the combinations and the addition of the evaluation values are completed, the CPU 10 extracts coordinates having a relatively high cumulative value from the coordinate group. As an example, out of the coordinates included in the coordinate group, coordinates having a predetermined number of cumulative values from the top may be extracted.

FIG. 6 is a flowchart for more specifically realizing the processing in steps S4 to S10 shown in FIG. It is assumed that the coordinate group includes n coordinates from coordinate 1, coordinate 2,..., Coordinate n, and each coordinate is (x _i , y _i ; 1 ≦ i ≦ n) Let it be represented.

  Referring to FIG. 6, CPU 10 initializes the accumulated value by substituting zero into accumulated value P (i) (where 1 ≦ i ≦ n) (step S100). Subsequently, the CPU 10 initializes the loop counter i by substituting 1 for the loop counter i (step S102), substitutes a value obtained by adding 1 to the value of the loop counter i for the loop counter j (step S104), and loops to the loop counter k. A value obtained by adding 1 to the value of the counter j is substituted (step S106).

The CPU 10 acquires the coordinates (x _i , y _i ), (x _j , y _j ), (x _k , y _k ) from the coordinate group detected by the edge detection unit 26, and calculates the evaluation value V _ijk . (Step S108). Further, the CPU 10 cumulatively adds the calculated evaluation value V _ijk to the cumulative values P (i), P (j), and P (k), respectively (step S110).

  Then, the CPU 10 adds 1 to the loop counter k (step S112), and determines whether or not the loop counter k has exceeded the coordinate number n (step S114). When the loop counter k does not exceed the coordinate number n (NO in step S114), the CPU 10 repeatedly executes steps S108 to S114.

  On the other hand, when the loop counter k exceeds the coordinate number n (YES in step S114), the CPU 10 adds 1 to the loop counter j (step S116), and the loop counter j (the coordinate number n). -1) is determined whether it has not been exceeded (step S118). If loop counter j does not exceed (number of coordinates n−1) (NO in step S118), CPU 10 repeatedly executes steps S106 to S118.

  On the other hand, when the loop counter j exceeds (the number of coordinates n−1) (YES in step S118), the CPU 10 adds 1 to the loop counter i (step S120), and the loop counter i is It is determined whether or not (the number of coordinates n−2) has been exceeded (step S122). When loop counter i does not exceed (number of coordinates n−2) (NO in step S122), CPU 10 repeatedly executes steps S104 to S122.

  On the other hand, when the loop counter i exceeds (the number of coordinates n−2) (YES in step S122), the CPU 10 executes the processing after step S12 shown in FIG.

Table 1 shows the processing results when the present invention is applied to the edge positions a to e shown in FIG. The coordinates of the edge positions a to e are, for example, (x _a , y _a ) = (1, 0), (x _b , y _b ) = (2, 0), (x _c , y _c ), respectively. = (3, 0), (x _d , y _d ) = (4, 0), (x _e , y _e ) = (5, −1).

Referring to Table 1, there are ₅ C ₃ = 10 combinations for selecting three from the five edge positions. Then, the CPU 10 calculates an evaluation value V for each combination. In the table, “◯” indicates that the corresponding coordinate was used for calculation of the evaluation value V, and the value on the right side indicates the evaluation value V at that time.

  Each of the coordinates a to e is selected six times, and the accumulated value P of the evaluation values V calculated at the time of each selection is the value shown in the bottom column of Table 1. In particular, for the coordinate e which is an outlier, a relatively low evaluation value V is calculated in any combination, while in a combination consisting only of the other coordinates a to d, a high evaluation value V is calculated. Yes.

  For this reason, the cumulative value P at the coordinate e is significantly lower than the cumulative value P at any of the coordinates a to d. Then, when extracting, for example, the top three of the accumulated values P among the coordinates a to e, the coordinates a to c are extracted. As a result, it is understood that the coordinate e which is an outlier is removed and an approximate straight line with high accuracy is calculated.

(Comparative example)
FIG. 7 is a comparative example comparing the effect of the present invention with the effect of the conventional LTS estimation method.

FIG. 7A shows approximate line calculation by a conventional LTS estimation method.
FIG. 7B shows approximate straight line calculation according to the present invention.

  7A and 7B, straight line calculation processing is performed on the same image, and the 32 edge positions (+ marks) detected in the mask area 50 are: Are identical to each other. In the conventional LTS estimation method (FIG. 7A), 50% (16 pieces) of the whole are extracted in order from the smallest distance (residual) from each edge position to the provisional approximate line to calculate the approximate line. did. On the other hand, in the case of the present invention (FIG. 7B), approximate straight lines were calculated by extracting 50% (16 pieces) of the whole in descending order of the cumulative value. In FIG. 7A and FIG. 7B, the edge position used for calculating the approximate straight line is further added with a circle mark at the center of the + mark.

  With reference to FIG. 7A, in the upper right of the mask area 50 in the drawing, positions other than the contour line of the object OBJ are intensively erroneously detected as edge positions. As described above, the edge positions detected outside the contour line of the object OBJ are inherently outliers. However, since the number of the edge positions is large, the outliers are dominant as a result. Therefore, it can be seen that the edge position where the contour line of the object OBJ is normally detected is removed, and an incorrect approximate line 60a is calculated.

  On the other hand, referring to FIG. 7B, when the present invention is applied, all the outlier edge positions that have been erroneously detected in the upper right of the mask area 50 are removed. Recognize. As a result, the calculated approximate straight line 60b is an accurate one that substantially matches the contour line of the object OBJ.

  According to Embodiment 1 of the present invention, an evaluation value indicating the degree of arrangement on the same straight line is calculated for each combination of three coordinates selected from the coordinate group. Thus, the evaluation value is not calculated based on the provisional approximate straight line calculated from the entire coordinate group, but is calculated microscopically among the three selected coordinates. Coordinates that are outliers are less likely to be placed on the same straight line in any combination, while normal coordinates are more likely to be placed on the same straight line in a combination of normal coordinates. As a result, since the evaluation values are accumulated in association with each of the coordinates, the difference between the normal coordinates and the coordinates of the outlier appears significantly, so that the outlier can be efficiently removed. Therefore, it is possible to realize an image processing apparatus that can efficiently remove outliers from coordinates included in a coordinate group and calculate an approximate straight line with high accuracy.

(Modification 1)
In the first embodiment of the present invention described above, the case has been described in which the evaluation value is calculated by selecting three different coordinates from the coordinate group, but the evaluation value may be calculated by selecting four or more coordinates. . In this case, a straight line including two coordinates among the four or more selected coordinates is defined, and an evaluation value is calculated based on an integrated value of distances from the remaining coordinates with respect to the straight line.

That is, r (r ≧ 4) selected coordinates are represented by coordinates i (x _i , y _i ), coordinates l (x _l , y _l ),..., Coordinates m (x _m , y _m ), If the coordinates are k (x _k , y _k ), the evaluation values V _i to _k like the equation (2) can be calculated by extending the above equation (1).

The evaluation values V _{i to k} also take higher values as the degree of arrangement on the same straight line increases. Except for the method of calculating evaluation values V _{i to k} , it is the same as that of the first embodiment of the present invention described above, and therefore detailed description will not be repeated.

  According to the first modification of the first embodiment of the present invention, the number of coordinates selected from the coordinate group for calculating the evaluation value can be expanded to four or more. Therefore, by appropriately selecting the number of selected coordinates, the number of combinations can be reduced as compared with the case of selecting three different coordinates from the coordinate group. As a result, in addition to the effects of the first embodiment of the present invention described above, the number of processing for calculating the evaluation value can be reduced. Therefore, when the present invention is applied to a coordinate group consisting of a large number of coordinates However, the processing speed can be maintained.

(Modification 2)
In Embodiment 1 of the present invention described above and Modification 1 thereof, a straight line including two coordinates among the selected coordinates is defined, and based on the distance between each of the remaining coordinates and the defined straight line. The configuration for calculating the evaluation value has been described. On the other hand, in Modification 2 of Embodiment 1 of the present invention, a configuration for calculating an evaluation value based on a correlation coefficient for a selected coordinate will be described.

If a combination of r (r ≧ 3) or more coordinates selected for calculating the evaluation value is a coordinate i (x _i , y _i ; 1 ≦ i ≦ r), the coordinate i (1 ≦ i ≦ r) The evaluation value V coinciding with the correlation coefficient can be calculated.

  Since this evaluation value V coincides with the correlation coefficient for the coordinate i, it changes in the range of −1 ≦ V ≦ 1 depending on the degree of arrangement on the same straight line of coordinates, and all the coordinates are identical. When arranged on a line, the maximum value “1” is taken. Therefore, similar to Embodiment 1 of the present invention described above and Modification 1 thereof, an approximate straight line can be calculated by extracting a predetermined number of coordinates in descending order of the cumulative value.

  Since the method other than the method of calculating the evaluation value is the same as that of the first embodiment of the present invention described above, detailed description will not be repeated.

  According to the second modification of the first embodiment of the present invention, the same arithmetic expression can be used regardless of the number of coordinates selected from the coordinate group in order to calculate the evaluation value. As a result, in addition to the effects of the first embodiment of the present invention described above, the processing related to the calculation of the evaluation value can be changed relatively freely. Process optimization.

[Embodiment 2]
In the above-described first embodiment of the present invention and its modifications, the configuration for calculating an approximate line for a two-dimensional image has been described. On the other hand, in Embodiment 2 of the present invention, a configuration for calculating an approximate straight line for a three-dimensional image will be described. As an example, the three-dimensional image includes a so-called stereo image, and each coordinate is defined by a three-dimensional value.

  FIG. 8 is a schematic configuration diagram of an image processing device 100 # provided with the approximate straight line calculation device according to the second embodiment of the present invention.

  Referring to FIG. 8, image processing apparatus 100 # obtains two-dimensional images taken by two cameras 2.1 and 2.2 to generate a three-dimensional image, and at the same time, an edge on the three-dimensional image. An approximate straight line is calculated based on the position. Then, in image processing apparatus 100 according to the first embodiment of the present invention shown in FIG. 1, image processing apparatus 100 # replaces image input unit 20, filter 22, and image memory 24 with image input unit 20.1, 20.2, filters 22.1 and 22.2 and image memories 24.1 and 24.2 are arranged, and instead of the edge detection unit 26 and the image output unit 28, an edge detection unit 26 and an image output unit 28 are respectively provided. It is arranged.

  The cameras 2.1 and 2.2 are arranged so as to photograph the same object from different directions. Since the other points are the same as those of camera 2, detailed description will not be repeated. Filters 22.1 and 22.2 and image memories 24.1 and 24.2 are similar to filter 22 and image memory 24, respectively, and thus detailed description thereof will not be repeated.

  The image output unit 28 # compares the two images taken by the cameras 2.1 and 2.2 and generates a three-dimensional image in which the object is drawn in the three-dimensional space. Specifically, the image output unit 28 # calculates which of the pixels of one image corresponds to which of the other pixels, generates a parallax image using the pixel shift as pixel information, and generates this A three-dimensional image is generated based on the parallax image.

  The edge detection unit 26 # detects an edge position in each of the images taken by the cameras 2.1 and 2.2, and similarly to the image output unit 28 # described above, the pixel is based on the comparison between the two images. And the coordinates of the detected edge position in the three-dimensional space are calculated. Then, the edge detection unit 26 # outputs the detected plurality of edge positions as a coordinate group to the CPU 10 and the image memories 24.1 and 24.2.

  Others are similar to those of image processing apparatus 100 according to the first embodiment of the present invention, and thus detailed description will not be repeated.

  When the present invention is applied to such a three-dimensional image, the image processing apparatus according to the first embodiment of the present invention described above, except that the mathematical formula for calculating the evaluation value is changed as follows. Since it is the same as the process in 100, detailed description will not be repeated.

(Evaluation value calculation process)
As in the first embodiment of the present invention described above, a straight line including two coordinates among the three or more coordinates selected from the coordinate group is defined, and the remaining coordinates and the defined straight line are defined. An evaluation value is calculated based on the distance.

First, when three coordinates are selected from the coordinate group, the selected coordinates are coordinate i (x _i , y _i , z _i ), coordinate j (x _j , y _j , z _j ), coordinate k. _Assuming that (x _k , y _k , z _k ), it is possible to calculate the evaluation value V _ijk for the three-dimensional coordinates like the equation (4) by expanding the equation (1).

Further, when four or more coordinates are selected from the coordinate group, the selected r (r ≧ 4) coordinates are represented by coordinates i (x _i , y _i , z _i ) and coordinates l (x _l , Y _l , z _l ),..., Coordinates m (x _m , y _m , z _m ), coordinates k (x _k , y _k , z _k ), the above equation (4) is expanded. Evaluation values V _i to _k can be calculated as shown in Equation (5).

The evaluation value V _ijk and the evaluation values V _{i to k} take higher values as the degree of arrangement on the same straight line is larger. The CPU 10 calculates an approximate straight line using the evaluation values for the three-dimensional coordinates calculated in this way, as in the above-described processing.

  In the second embodiment of the present invention, in addition to the effects of the first embodiment of the present invention, an approximate straight line with high accuracy can be calculated for a three-dimensional image that is expected to expand in the future.

(Modification)
Furthermore, as in the second modification of the first embodiment of the present invention described above, the evaluation value can be calculated based on the correlation coefficient for the selected coordinates.

Specifically, when r (r ≧ 3) or more coordinates selected for calculating the evaluation value are coordinates i (x _i , y _i , z _i ; 1 ≦ i ≦ r), Based on the correlation coefficients (V _xy , V _yz , V _zx ) between the components (xy component, yz component, zx component), the evaluation value V can be calculated as in equation (6) or equation (7).

  As can be seen from these equations, since the correlation coefficients for the three components cannot be calculated, after calculating the correlation coefficients for any two of the three-dimensional coordinate values, An evaluation value V is calculated from the number of relations.

  In the evaluation value V shown in the expression (6), when the coordinates are arranged in the range of −3 ≦ V ≦ 3 according to the degree of arrangement on the same straight line of coordinates, any coordinate is arranged on the same straight line. Takes the maximum value of “3”.

  On the other hand, the evaluation value V shown in the equation (7) changes in the range of −1 ≦ V ≦ 1 depending on the degree of arrangement on the same straight line of coordinates, and all the coordinates are arranged on the same straight line. In this case, the maximum value “1” is taken.

  Therefore, as in the second embodiment of the present invention described above, an approximate straight line can be calculated by extracting a predetermined number of coordinates in descending order of the cumulative value after cumulative addition. Except for the method of calculating the evaluation value, the method is the same as that of the second embodiment of the present invention described above, and the detailed description will not be repeated.

[Other embodiments]
In the description of the above-described embodiment, an image processing apparatus including an edge detection unit is shown as an example. However, the edge detection unit is not necessarily included in the image processing apparatus. For example, an approximate straight line calculation device that receives the coordinates of the edge position detected by the external device and calculates an approximate straight line based on the coordinates may be configured.

  In the above description of the embodiment, the configuration has been described in which the edge detection unit accumulates color information for an area having a predetermined width in an image to detect the edge position. However, it is not limited to this, and a well-known edge position detecting means can be applied. Furthermore, the scanning direction for detecting the edge position may also be scanned, for example, in an oblique direction of the image in accordance with the position or direction of the linear shape portion of the object.

  In the description of the above-described embodiment, the configuration in which the evaluation value is calculated for all the combinations of three or more coordinates selected from the coordinate group detected by the edge detection unit has been described. There is no need to calculate an evaluation value.

  For example, three or more coordinates may be selected so that the distance on the image is a predetermined value (predetermined pixel) or less. Since the pixels selected in this manner are close to each other, it is possible to evaluate the degree of arrangement on the same straight line more appropriately.

  Further, a combination of coordinates for calculating an evaluation value may be determined according to the degree of change in color information at the edge position corresponding to each coordinate. Specifically, the “color distance” of the pixels around the edge position is evaluated, and the coordinates corresponding to the edge position having a small color distance (color information change is small) are defined as coordinates corresponding to the edge position having a large color distance. You may make it combine. The “color distance” is defined as a distance between two pixels in a three-dimensional color space composed of a red axis, a green axis, and a blue axis arranged so as to be orthogonal to each other. By adopting such a combination, the difference in evaluation value between coordinates with a small color distance and low reliability (highly likely to be erroneously detected) and coordinates with a small color distance and high reliability is more pronounced. Therefore, the rate at which coordinates with low reliability are removed increases. For this reason, a more accurate approximate straight line can be calculated.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a schematic block diagram of an image processing apparatus provided with the approximate line calculation apparatus according to Embodiment 1 of this invention. It is a figure for demonstrating the outline | summary of the detection process of the edge position in an edge detection part. It is a figure for demonstrating the outline | summary of the calculation process of the approximate line in CPU. It is a flowchart which shows the control structure which concerns on the approximate line calculation in CPU. It is a figure for demonstrating an example of the process which concerns on calculation of an evaluation value. 5 is a flowchart for more specifically realizing the processing in steps S4 to S10 shown in FIG. It is a comparative example which compares the effect of this invention with the effect in the conventional LTS estimation method. It is a schematic block diagram of an image processing apparatus provided with the approximate line calculation apparatus according to Embodiment 2 of this invention.

Explanation of symbols

  2, 2.1, 2.2 Camera, 4 Display device, 6 External device, 18 Input / output unit, 20, 20.1, 20.2 Image input unit, 22, 22.1, 22.2 Filter, 24, 24.1, 24.2 Image memory, 26, 26 # edge detection unit, 28, 28 # Image output unit, 30 bus, 40.1, 40.2 Scan position, 42.1, 42.2 area, 44 edge Position group, 44.1, 44.2 Edge position, 46.1, 46.2 Color information histogram, 48.1, 48.2 Change position, 50 Mask area, 52 Noise area, 60a, 60b Approximate line, 100 images Processing device, EDG contour, IMG image, OBJ object, P cumulative value, V evaluation value.

Claims (8)

Coordinate group acquisition means for acquiring a coordinate group including four or more coordinates indicating a specific position of the image;
Coordinate selecting means for selecting three or more coordinates from the coordinate group according to a predetermined rule;
Evaluation value calculation means for calculating an evaluation value indicating the degree of arrangement on the same straight line for the three or more coordinates selected by the coordinate selection means;
An evaluation value adding means for accumulating the evaluation values calculated by the evaluation value calculating means in association with each of the three or more coordinates;
Repeating means for repeatedly executing the processes in the coordinate selecting means, the evaluation value calculating means, and the evaluation value adding means so that the coordinates of the coordinate group are selected the same number of times.
Extraction means for extracting a plurality of coordinates having a relatively high degree of arrangement on the same straight line from the coordinate group based on the evaluation values cumulatively added by the evaluation value addition means;
An image processing apparatus comprising: an approximate line calculation unit that calculates an approximate line based on a plurality of coordinates extracted by the extraction unit. The coordinate group acquisition means includes
An image acquisition unit for acquiring images;
The image processing apparatus according to claim 1, further comprising: an edge detection unit that detects an edge position that causes a change in color information in the image acquired by the image acquisition unit and acquires the coordinate group.   The image processing apparatus according to claim 2, wherein the image acquisition unit acquires an image obtained by photographing an object having a linear shape portion.   The evaluation value calculating means defines a straight line including two coordinates among the three or more coordinates, and calculates the evaluation value based on a distance between each of the remaining coordinates and the defined straight line; The image processing apparatus according to claim 1.   The image processing apparatus according to claim 1, wherein the evaluation value calculation unit calculates the evaluation value based on a correlation coefficient for the three or more coordinates.   The repeating means performs the processes in the coordinate selecting means, the evaluation value calculating means, and the evaluation value adding means until all the combinations for selecting three or more coordinates from the coordinate group are selected by the coordinate selecting means. The image processing apparatus according to claim 1, which is repeatedly executed.   The image processing apparatus according to claim 1, wherein the coordinate group includes two-dimensional or three-dimensional coordinates. Coordinate group acquisition means for acquiring a coordinate group including four or more coordinates;
Coordinate selecting means for selecting three or more coordinates from the coordinate group according to a predetermined rule;
Evaluation value calculation means for calculating an evaluation value indicating the degree of arrangement on the same straight line for the three or more coordinates selected by the coordinate selection means;
An evaluation value adding means for accumulating the evaluation values calculated by the evaluation value calculating means in association with each of the three or more coordinates;
Repeating means for repeatedly executing the processes in the coordinate selecting means, the evaluation value calculating means, and the evaluation value adding means so that the coordinates of the coordinate group are selected the same number of times.
Extraction means for extracting a plurality of coordinates having a relatively high degree of arrangement on the same straight line from the coordinate group based on the evaluation values cumulatively added by the evaluation value addition means;
An approximate line calculation device comprising: an approximate line calculation means for calculating an approximate line based on a plurality of coordinates extracted by the extraction means.

Images