JPH07220089A - Position detection method - Google Patents

Abstract

(57) [Abstract] [Purpose] To provide a position detection method by image recognition that can be applied to an object of any shape, has general versatility, can be processed at high speed, and has high noise resistance. [Structure] Whether an image contour point detected by scanning is already detected or not is determined by an overlay memory which stores the already detected image contour point G, and the contour line tracking H is performed with the undetected image contour point G as a starting point. Start, and access the binarized image based on the search direction order of the image contour point search direction number table according to the arrival direction number to the reference point to detect the next image contour point, and the overlay memory traces the contour line. Detect the end point,
The current contour line tracking is ended, and the scanning is continued from the point next to the start point, and the outer image contour line I, the inner contour lines D and E,
F is detected, the characteristic points of the detected contour line are detected, shape matching is performed using the cumulative calculation result of the distance angle between these characteristic points, and the position is detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting the position of an object in a visual system, and more particularly to detecting the contour line of an image of the object on a binarized image and detecting the position of the object based on the contour line. To a position detection method for detecting

[0002]

2. Description of the Related Art Recently, in the United States and Japan, a large number of general-purpose visual systems (vision systems) used for identifying objects, recognizing positions, etc. have been developed. The conventional general methods used for object identification and position recognition in these systems are mainly of the following two types.

The first method is a method of statistically processing an image area of an object on image data. For example, it is a method of detecting the shape, position, and direction of an object with reference to statistics such as area, major axis, minor axis, and moment of inertia.

The second method uses a template (reference image) according to the characteristic shape of the object, scans the image obtained by picking up the object with this template, and scans the image and the template. Correlation calculation between
Based on the correlation coefficient, the position where the characteristic shape exists in the image is identified, and the position of the object is detected by the position.

However, in the first method, since the contour shape is not directly used, there is a problem that the accuracy of contour detection is insufficient.

In addition, the second method has a problem that it is difficult to generalize because the limiting conditions such as the size and direction of the template to be used are derived.

As a measure against the above-mentioned problems, as shown in FIG. 16, the contour line of the image B of the object in the image window frame A is traced in program mode by using a computer, and the detected contour line is used. There is a position detection method for detecting the position of the predetermined position C of the object.

The outline tracking procedure in this position detecting method is as follows.

(1) The inside of the image window frame A is scanned to detect the first point satisfying the outline detection standard as an image outline point. The first detected image contour point becomes the start point of contour line tracking, that is, the first reference point. In this case, the outline detection reference is made programmatically.

(2) Based on the contour point tracking reference, the first
The eight neighboring points of the reference point are accessed to detect the next image contour point. If the detected next image contour point does not satisfy the contour end condition, this process is repeated. In this case, the contour point tracking reference and the contour end condition are made programmatically.

(3) When the detected image contour point satisfies the above-mentioned contour line end condition, at this point, this algorithm ends.

[0012] In this procedure, various contour line tracking algorithms can be constructed by changing various programs of the contour line detection standard, the contour point tracking standard, and the contour line end condition, and the contour line tracking can be performed even for an object having a complicated shape. It is possible.

(4) Regarding the contour line obtained above,
A characteristic point is detected, and the detected characteristic point and the reference data of the characteristic point relating to the shape of the object registered in advance are collated by a programmatic collation method to detect the position.

[0014]

However, in the conventional position detecting method, the contour line detection reference for detecting the contour line, the contour point tracking reference, the contour line end condition, and the like are executed, and the feature points of the detected contour line are detected. Since the position and the reference data of the feature points registered in advance are collated with each other programmatically,
When this is applied to a large amount of image data, there is a problem in that the amount of processing increases, it takes too much time, and practicality is lost.

In particular, in the step of obtaining the starting point of contour line tracing by scanning the image, if the image contour point reached by scanning is set as the starting point without distinguishing between detected and undetected points,
The detection of the starting point is limited to the outer contour of the image of the object, and since the scanning does not reach the inside of the image, there is a problem that the contour line inside the image cannot be detected. There is a problem that it takes too much time to distinguish detections by a program and it is not practical.

Further, when the contour line tracking start point is stored as the contour line tracking end point and the ending process is performed, the contour line of the image B of the object intersects the four edges of the image window frame A. In this case, a program that can be processed within a practical time cannot finish at this intersection and cannot trace the image contour point at the edge, so there is a problem of straying,
There is a problem that it can be applied only to the outline of a closed outline. Further, there is a problem in that it takes too much time to solve this problem programmatically and it is not practical.

The present invention solves the above problems and is simple, quick, and versatile, detects open contour lines and contour lines inside an image, and based on the characteristic points of these contour lines. An object is to provide a position detection method for detecting a position.

[0018]

The positioning method of the present invention is
In order to solve the above problems, in the first step,
The binarized image data stored in the image storage means is line-sequentially scanned, and the scanning is temporarily stopped at the position where the image contour point is reached,
In the second step, the reached image contour point is judged as already detected or undetected by the overlay memory storing the already detected image contour points, and if already detected, the scanning is restarted, If not detected, the reached image contour point is stored in the overlay memory as a start point,
In the third step, contour tracing is started for the binarized image data using the starting point as a first reference point, the arrival direction number is 7 at the first reference point, and the second reference point and subsequent ones are currently referred to from the previous reference point. An image contour point search in which an image contour point search direction number table of arrival direction numbers that are directions toward a point is determined by the arrival direction number to the reference point and the search direction order at the reference point is determined by eight arrival direction numbers. Selecting from the direction number table, accessing the binarized image based on the search direction order of the selected image contour point search direction number table to detect the next image contour point and store it in the overlay memory; In, the data of the starting point and the already detected image contour point stored in the overlay memory is accessed for eight neighboring points around the current reference point, and the third step is used as a result of this access. It is determined whether the detected image contour point is undetected or already detected, and if it is not detected in the fifth step, the third to fourth steps are repeated, and if it is already detected, it is detected in the third step. Since the image contour point is the end point of the contour line tracking, the current contour line tracking is ended and the procedure returns to the first step, the scanning is continued from the point next to the starting point, and the line sequential scanning is performed in the sixth step. When the lower end of the binarized image is reached, all contour line tracking is terminated, and in the seventh step, the curvature of each contour line detected so far is calculated to detect the feature point based on the curvature, and the eighth step At
The contour line length and the angle between the feature points are sequentially cumulatively calculated for each contour line to store the distance angle cumulative calculation result, and in a ninth step, the distance angle cumulative calculation relating to the shape of the object registered in advance is stored. It is characterized in that the reference data and the distance / angle cumulative calculation result are shifted while matching for each feature point to detect the optimum matching position.

[0019]

In the position detecting method of detecting the position of the object by tracing the contour line of the object image by the program processing, if the required time is practically shortened, the function becomes insufficient, and if the function is satisfied, it takes too much time. In order to solve the problem of the conventional example that is not practical, the contour line detection criteria of the program,
It is necessary to shorten the processing time by simplifying the contour point tracking reference, the contour end condition, and the position matching condition.

Therefore, the present invention (1) uses the image contour point search direction number table as a part of the program, instead of sequentially performing the contour point tracking reference processing by the program, and simply and quickly. It has become. And
This image contour point search direction number table uses the fact that the order of accessing 8 neighboring points at a reference point is determined by the arrival direction number to the reference point. In which direction the next image contour point is located. However, by referring to the table, it is possible to detect in the same process, and it is possible to simplify and speed up. And
In addition to tracking the closed image contour line, the tracking can be ended by detecting the end points of the contour line having the end points as described later. Therefore, not only the processing is simplified and the required time is shortened, but also the function can be expanded in the same short time.

(2) Instead of sequentially executing the processing of the contour line detection reference and the contour line end condition by the program, the overlay memory storing the detected image contour points is used to judge whether it has been detected or not. is doing. Since all the detected image contour points can be stored in this overlay memory, not only the processing is simplified and the required time is shortened, but also repeated contour lines are stored until there are no undetected image contour points. Continuing the tracking, the contours that are inside the outer contour can also be traced through the outer contour.

(3) In addition, the characteristic points of the detected contour line are detected by curvature calculation, and the distance angle cumulative calculation result between the detected characteristic points is shape-matched with the distance angle cumulative calculation reference data to detect the position. Has versatility,
Since the position can be accurately detected even if there is a shadow or noise in other portions, it is possible to perform position detection with high noise resistance.

[0023]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A first embodiment of the position detecting method of the present invention will be described with reference to FIGS.
It demonstrates based on FIG. 15, Table 1-Table 3, and a formula (1) -formula (4).

(1) In the first and second steps, the image window frame A of the binarized image is line-sequentially scanned in the lateral direction from the upper end,
At the reached image contour point G, by the overlay memory,
It is determined whether or not the reached image contour point G is undetected, and if it is not detected, the reached image contour point G is set as the start point of contour line tracking, and the reached image contour point G is detected image. The contour points are stored in the overlay memory.

(2) In the third step, the contour line tracing H is started for the binarized image data with the starting point G as the first reference point, and at the first reference point G, the direction code 7 shown in FIG. Is set as the arrival direction number, and after the second reference point, the direction of the direction number shown in FIG. 2 from the previous reference point to the current reference point is set as the arrival direction number. A different image contour point search direction number table is selected, and the binarized image is accessed based on the search direction order of the selected image contour point search direction number table.

The image contour point search direction number table shown in Table 1 is
The arrival direction number is 5.

[0027]

[Table 1]

Table 1 will be described. Direction numbers 1 to 8 are shown in FIG.
This corresponds to the 8-direction number shown in (a). P ₀ is the current reference point. In direction ID Nos. 1-8 have the following reference points P ₁ to P _8, accesses the counterclockwise arrow J. A circle mark is a background point, and a circle mark is a bit pattern indicating a region point of the object. The next direction number indicates the direction number to the next reference point to be obtained, and the table is referenced with the bit patterns of the direction numbers 1 to 8 obtained by accessing, and in the column where the direction number and the ● mark match. By finding the next underlying direction number, the direction number to the next reference point is obtained. A cross indicates an error. In Table 1, when the next direction number is 1, the next reference point has moved back to the previous reference point, and the current reference point is the end point of the contour line having the end point. Indicates that. That is, in Table 1, the end points of the contour line having the end points can be detected.

For the reasons described below, generally, if the arrival direction number is i, the direction i-3 shown in FIG. 2B always becomes the background point. If the direction of i-3 is a region point, it is treated as an error. Therefore, in Table 1, the direction of 5-3 = 2 always becomes the background point. Therefore, when the arrival direction number shown in Table 1 is 5, the direction number among the 8 direction numbers 1 to 8 shown in FIG. The direction in which eight neighboring points of the current reference point are accessed in the counterclockwise direction from the direction 2 and the area point is first obtained is the direction of the next reference point.

The explanation that the direction of i-3 is always the background point is as follows.

The contour shape of the image B of the object is shown in FIG.
3 is either the closed image B ₁ of FIG. 3B or the open image B ₂ of FIG.

Assuming that the reference point of the closed figure B ₁ is S ₁ and the reference point of the opened image B ₂ is S ₂ , respectively, as shown in FIG.
It becomes as shown in FIG.

FIGS. 5 and 6 are diagrams showing the positional relationship between the background point ◯ and the area point  when the direction number from the previous reference point to the current reference point is 5-8. In this case, FIG. 5 shows a closed figure B.
FIG. 6 shows the case of _{1 and} the case of the opened figure B ₂ . In either case, the i-3 direction is the background point, and i-3
If the search is started counterclockwise from the direction, the next reference point can be detected.

FIG. 7 is a diagram showing the detection on the contour line.
In FIG. 7A, when the current reference point P ₀ is detected in the direction of the direction number 6 from the previous reference point P ₁₀ , for example, the direction 6−3 = 3 from the current reference point P ₀ is always the background point. And there is no next reference point in the clockwise direction. Therefore,
By searching counterclockwise from the direction number 3, the next reference point can be detected. Similarly, in FIG. 7B, when the current reference point P ₀ is detected from the previous reference point P _{10 in} the direction of the direction number 5, for example, the direction 5-3 = 2 from the current reference point P ₀ is It always becomes a background point, and there is no next reference point in the clockwise direction from this. Therefore, by searching counterclockwise from the direction number 2, the next reference point can be detected.

4 to 7 for each of the direction numbers 1 to 8
If you try, you can get the same result. Table 1 is an image contour point search direction number table for starting the search counterclockwise from the direction i-3, which is created by induction.

Further, in the case of a search from the starting point arrived by scanning, it is easily understood that the arrival direction number should be 7.

As described above, the image contour point search direction number table of Table 1 is accessed to detect the next image contour point, and the detected image contour point is stored in the overlay memory.

(3) In the fourth step, the data of the starting point and the already detected image contour points stored in the overlay memory are accessed for eight neighboring points around the current reference point, and the image detected this time is accessed. The presence / absence of the start point and the detected image contour point which overlap the contour point is determined. From this result, it is possible to determine whether the image contour point detected in the third step is undetected or already detected.

(4) In the fifth step, if not detected, the third to fourth steps are repeated, and if already detected, the image contour point detected in the third step is the contour tracing end point. Therefore, the tracing of the current contour line is terminated, the process returns to the first step, and the scanning is continued from the point next to the start point.

(5) In the next scan, even if an already detected image contour point is detected, the overlay memory judges that it passes without setting it as a starting point, and an undetected image contour point is detected. The image contour line is searched using as a starting point, and the process is repeated until no undetected image contour point is detected. Therefore, referring to FIG. 1, first, after detecting the outer side image contour line I of the image B of the object by the contour line tracking H,
The contour lines D, E, F inside the image B of the object are detected,
The position of the predetermined position C is recognized. In this case, in the scanning of the outer image contour line I of the image B of the object and the scanning of the contour lines D, E, and F of the inside of the image B of the object, "" of the image data at the detection position of the image contour point is detected. "0" → "1", "1" → "0"
Changes in the opposite.

(6) Incidentally, if the points on the four edges of the image window frame of the binarized image data stored in the image storage means are treated as the already detected image contour points, the image shown in FIG. You can search the image contour of the opened image. To do this easily,
The points on the four edges of the image window frame of the binarized image data stored in the image storage means may be stored in the overlay memory as already detected image contour points.

A first embodiment for carrying out the position detecting method of the present invention will be described with reference to FIGS. 8 and 9.

In this embodiment, a part of the contour line tracking is performed by hardware based on random logic rather than program control to speed up the process. In FIG.
Reference numeral 02 is composed of an image memory and an overlay memory in which coordinates are superposed on the image memory, and the start point and the detected image contour point are sequentially written in this overlay memory.
The control unit 107 uses the overlay memory of the image storage unit 102 while scanning the binary image stored in the image memory of the image storage unit 102 using the DMA (direct memory access) unit 103. Then, the already detected points are checked to detect the image contour points. Control unit 1
Reference numeral 07 designates an image contour point which is reached and detected by the above scanning as a starting point of image contour line tracing, and the starting point address and the number of start directions in the search direction order based on the image contour point search direction number table selected by the arrival direction number are set. It is sent to the address calculation unit 105 and the contour point search unit 104 to start the image contour line tracking. When the tracking end information is output from the DMA unit 103 and the image contour line tracking is completed, the scanning is continued. Then, when an undetected image contour point is detected, the above operation is repeated until the image data ends, that is,
The above line-sequential scanning is executed until the lower end of the binarized image is reached.

The DMA unit 103 has a start point or a reference point of 8
The neighboring points are accessed from the image memory of the image storage unit 102. In this case, it is possible to access both the image memory and the overlay memory with one addressing access. The data accessed from the image memory is sent to the contour point searching unit 104, and the overlay data accessed from the overlay memory is temporarily stored. The contour point search unit 104
The image contour points are detected as described below, and the DMA unit 103
Compares the image contour point detected by the contour point searching unit 104 with the start point and the already detected image contour point stored in the overlay memory to detect the tracking end information. When the tracking end information is detected, it is sent to the control unit 107.

The contour point searching unit 104 detects the image contour points of the data of the eight neighboring points from the DMA unit 103, based on the number of starting directions in the search direction order based on the image contour point searching direction number table. The contour point search direction number table is composed of a ROM, and the input data is given as an address by a total of 11 bits of 3 bits of the arrival direction number and 8 bits of the data of 8 neighboring points. Data 3
A total of 4 bits, which are a bit and 1 bit of error information, are read. As the error information, "1" is output when all 8 neighboring points are "1" or all "0", and when the direction number i-3 is "region point of the object", the error information is indicated as an error. To do.
The next direction number read out is sent to the address calculation unit 105.
The error information is sent to the control unit 107.

The address calculation unit 105 includes a contour point search unit 1
X, Y according to the next direction number obtained in 04
Increase or decrease in each direction to make the address of the next reference point,
This address is sent to the DMA unit 103, and also sent to the contour data storage unit 106 together with the detected next direction number, and the contour data storage unit 106 stores this.

FIG. 9 is a block diagram showing an example of details of the DMA unit 103, the contour point searching unit 104 and the address calculating unit 105 of FIG.

In FIGS. 8 and 9, the line-sequential memory access circuit 1008 of the DMA unit 103 is sent a line-sequential scanning start point address from the control unit 107, and the line-sequential memory access circuit 1008 starts line-sequential scanning from the start point address. To do.

Simultaneously with the line-sequential scanning, the line-sequential memory access circuit 1008 reads the image memory data and the overlay data at the same coordinate point from the image storage unit 102. The comparison circuit 1009 compares the read image memory data at the same coordinate point with the overlay data. At this time, when the image memory data is a coordinate changing from “0” to “1” and the overlay data is “0”, the coordinate is the contour line tracing start point, and in other cases, the contour line tracing is performed. Not the starting point.

When the contour tracing start point is detected, the start point information is sent to the control unit 107. When the control unit 107 receives the start point information, it sends a line-sequential memory access stop command to the line-sequential memory access circuit 1008 and stores the address next to the start point as the line-sequential scanning start point address.

Subsequently, the control unit 107 sends the line-sequential scanning start point address to the buffer 1007 and the start direction number to the buffer 1004. The memory access circuit 100 according to the line sequential scanning start point address stored in the buffer 1007.
1 accesses the 8-neighbor point data and the overlay data and sends them to the search table 1003 and the comparison circuit 1002. The overlay data sent to the comparison circuit 1002 is temporarily stored. Search table 1003 as described in Table 1 above.
The next direction number is obtained, and this direction number is stored in buffer 1.
004, calculation table 1005, and contour data storage unit 10
6 and sent to.

The comparison circuit 1002 compares the direction number stored in the buffer 1004 with the overlay data stored previously to obtain tracking end information. In the calculation table 1005, increase / decrease values ΔX, Δ in the X and Y directions corresponding to the respective direction numbers.
Y is written, and this is sent to the arithmetic circuit 1006 to calculate the next reference point address. The reference point address is sent to the contour data storage unit 106 and the buffer 1007.

The feature point detection unit 108 subdivides the detected contour line and calculates the curvature, and detects and stores the feature point position from the result. More specifically, the feature point detection unit 108
As shown in FIG. 10, is composed of a curvature value calculation unit 111, a curvature value comparison unit 112, and a feature point storage unit 113. An example of the actual shape of the contour data stored in the contour data storage unit 106 is shown in FIG. 11, which will be described below with reference to this.

In FIG. 11, the number of each contour point indicates a tracking number. In the curvature value calculation unit 111, each contour point i
For each time, the curvature value is calculated based on the detected direction code i, for example, according to the curvature calculation formula shown in the following formula (1).

[0055]

[Equation 1]

However, the addition / subtraction is modulo 8. The curvature value comparison unit 112 compares the curvature value calculated by the curvature value calculation unit 111 for each contour point with the absolute value of the curvature value and the set threshold value. Then, the characteristic points are determined. For example, when [1] | curvature calculation value | ≦ threshold value 1, straight line [2] threshold value 1 <| curvature calculation value | <threshold value 2 curve [3] threshold value 2 ≦ | curvature calculation value | To determine. Table 2 shows the determination results of the direction code i, the curvature value, and the feature points in the contour line of FIG. In this example, the calculation is performed with n = 2.

[0057]

[Table 2]

The straight line is not a characteristic point, the curve is a continuous point, but the first point thereof is a curve characteristic point, and the corner is one point or a continuous point. In the case of a continuous point, there is a maximum value. Determines the point having the maximum value as the corner feature point when there is no maximum value. The feature point information (x, y coordinates and feature information) thus obtained is stored in the feature point storage unit 113. The results of FIG. 11 are shown in Table 3.

[0059]

[Table 3]

In this way, by calculating the characteristic points of the contour line by using the curvature calculation, the amount of data for specifying the contour line can be significantly reduced, and high-speed processing becomes possible.

The distance angle cumulative calculation unit 109 sequentially cumulatively calculates, for each contour line, the contour length between feature point positions stored in the feature point storage unit 113 of the feature point detection unit 108 and the angle with respect to the horizontal direction of the image, Memorize the result. That is, with respect to the feature point data stored in the feature point storage unit 113, the distance angle cumulative calculation is performed using the following equations (2) and (3).

[0062]

[Equation 2]

[0063]

[Equation 3]

However, in tracking point number 1, S = 0, Φ = 0
And

FIG. 12 shows the result of cumulative distance angle calculation by the above calculation. Using this distance angle calculation result, even if there are some irregularities or deformed portions on the contour line, only minute and local fluctuations appear on the line, and if the shape is the same as a whole, Since the line patterns are the same, it is possible to detect the contour line shape having high noise resistance.

The shape matching unit 110 matches the distance angle cumulative calculation reference data of the template registered in advance with the calculation result stored in the distance angle cumulative calculation unit 109 while shifting each feature point to obtain the optimum value. Detect the matching position.
The actual shape of the template is shown in FIG.
(B) shows the distance angle cumulative calculation reference data. Then, the point that minimizes the degree of coincidence D shown in the following equation (4) is
The shape matching point is used as the detection result.

[0067]

[Equation 4]

However, Φ is the object, and Φt is the template. Next, the second embodiment using the method of the present invention is shown in FIGS.
It will be described based on 5.

FIG. 14 is a block diagram showing the configuration of the second embodiment, which is an image input section 1101 and an image storage section 1102.
And a CPU 1103 and a temporary storage unit 1104. The operation of the CPU 1103 will be described with reference to the flowchart of FIG.

In step # 1, the stored image is line-sequentially scanned to detect boundary point coordinate data.

In step # 2, an undetected boundary point is detected as a starting point from the detected boundary point coordinate data.
If it is the start point, proceed to step # 3, and if it is not the start point,
Return to step # 1.

In step # 3, the contour line search is started from the obtained starting point, and the scanning point next to the starting point is stored as the next starting point.

In step # 4, the next contour point is detected according to the search table.

In step # 5, it is determined based on the overlay data whether or not the detected contour line satisfies the search end condition. If it is not the search end condition, the process returns to step # 4, and if it is the search end condition, the process proceeds to step # 6.

In step # 6, scanning is continued up to the end of the image contour line, and it is determined whether or not there is another starting point. If there is another starting point, the process returns to step # 3, and if there is no other starting point, the scanning is ended and the process proceeds to step # 7.

In step # 7, the curvature of the detected image contour line is calculated, and the feature point is detected based on the result.

In step # 8, the result of sequentially cumulatively calculating the contour length and the angle between the detected feature points for each contour line is stored as a distance angle cumulative calculation result.

In step # 9, the distance-angle cumulative calculation reference data of the registered template is checked against the above-mentioned distance-angle cumulative calculation result while shifting for each feature point, the optimum matching position is detected, and the process ends.

[0079]

According to the position detecting method of the present invention, instead of sequentially processing the scanning of the starting point of contour line tracing and the detection of the ending point of contour line tracing by a program, a used starting point or an already detected image is detected. Tracking the contour line of an open image, which was not possible with the conventional technique, even though the processing is simplified by using the overlay memory in which the coordinates of the contour point are stored by superimposing the coordinates on the image data, and It is possible to detect a closed contour line inside the image.

Further, the feature points of the detected contour line are detected by curvature calculation, and the position of the object is detected by performing shape matching using the distance angle cumulative calculation result between these feature points, thereby providing versatility. However, even if there are shadows and noise in other parts,
This has the effect of enabling accurate position detection, which enables position detection with high noise resistance.

[Brief description of drawings]

FIG. 1 is a diagram showing an operation of the present invention.

FIG. 2 is a diagram showing a search direction order of the present invention.

FIG. 3 is a diagram showing a closed image and an opened image.

FIG. 4 is a diagram showing reference points of a closed image and an open image.

FIG. 5 is a diagram showing previous reference points, current reference points, and background points of a closed image.

FIG. 6 is a diagram showing previous reference points, current reference points, and background points of an opened image.

FIG. 7 is a diagram showing a previous reference point, a current reference point, and a search direction according to the present invention.

FIG. 8 is a block diagram showing a configuration of a first embodiment of a position detecting device using the present invention.

9 is a partial detailed view of FIG. 8. FIG.

FIG. 10 is a detailed block diagram of a feature point detection unit of the first embodiment.

FIG. 11 is a plan view of a contour line stored in a contour data storage unit of the first embodiment.

FIG. 12 is a plan view of a distance angle cumulative calculation result in the first embodiment.

FIG. 13 is a plan view of a template used for position matching and its distance angle cumulative calculation reference data in the first embodiment.

FIG. 14 is a block diagram showing the configuration of a second embodiment of the position detecting device using the present invention.

FIG. 15 is a flowchart showing the operation of the second embodiment.

FIG. 16 is a diagram showing an operation of a conventional example.

[Explanation of symbols]

 A image window frame B image of object C predetermined position D, E, F inner contour line G image contour point (starting point) H contour line tracing I outer image contour line 101 image input unit 102 image storage unit 103 DMA unit 104 contour Point search unit 105 Address calculation unit 106 Contour data storage unit 107 Control unit 108 Feature point detection unit 109 Distance angle cumulative calculation unit 110 Shape matching unit 111 Curvature value calculation unit 112 Curvature value comparison unit 113 Feature point storage unit 1101 Image input unit 1102 Image storage unit 1103 CPU 1104 Temporary storage unit

Continuation of front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location G06T 7/00 9/20 7459-5L G06F 15/70 335 Z

Claims (1)

[Claims] 1. In the first step, the binary image data stored in the image storage means is line-sequentially scanned, and the scanning is temporarily stopped at the position where the image contour point is reached. The image contour points are stored in the overlay memory that stores the previously detected image contour points.
Whether it is already detected or not is detected. If it is already detected, scanning is restarted. If it is not detected, the reached image contour point is stored in the overlay memory as a start point, and in the third step, the start point is determined. The contour line tracking is started for the binarized image data as the first reference point, and the arrival direction number is 7 at the first reference point, and the arrival direction is the direction from the previous reference point to the current reference point after the second reference point. Select the image contour point search direction number table of the number from the image contour point search direction number table that determines the search direction order at the reference point that is determined for each of the eight arrival direction numbers Based on the search direction order of the image contour point search direction number table, the binarized image is accessed to detect the next image contour point and stored in the overlay memory. In the fourth step, the overlay memo is stored. The data of the start point and the already detected image contour point stored in the above are accessed for eight neighboring points around the current reference point, and the access result indicates whether the image contour point detected in the third step is undetected or not detected. If it is not detected in the fifth step, it is judged in the third step.
~ Repeating the fourth step, if already detected, the image contour point detected in the third step is the contour line tracking end point, so the current contour line tracking is terminated and the procedure returns to the first step to return to the start point. Scanning is continued from the next point, and in the sixth step, when the line sequential scanning reaches the lower end of the binarized image, all contour line tracing is completed, and in the seventh step, each of the detected contour lines is detected. The curvature of the contour line is calculated, and the characteristic points based on the curvature are detected. In the eighth step, the contour line length and the angle between the characteristic points are sequentially cumulatively calculated for each contour line, and the distance / angle cumulative calculation result is stored. Then, in the ninth step,
Position detection characterized by shifting the distance angle cumulative calculation reference data relating to the shape of an object registered in advance and the distance angle cumulative calculation result while matching each feature point to detect an optimum matching position. Method.