JPH07200828A - Reference position deciding method for model image of position measuring device applying image processing - Google Patents

Abstract

PURPOSE:To measure accurate position of a subject of measurement. CONSTITUTION:The gray level image data obtained by photographing a model are stored in an image memory 13, and a model image shown by the gray level image data is displayed on a monitor 16. The gray level image data included in a window designated by an operator are registered as the model image data. The reference position of the registered model image data is calculated in a unit lower than the rank of a pixel. The gray level image data acquired by photographing a subject of measurement are stored in the memory 13, the measuring subject image data are scanned by a window including the model image data, and the normalized correlation value is calculated. Based on this correlation value, a window position of the largest degree of coincidence is calculated in a unit lower than the rank of a pixel. Then, the position of the subject of measurement is calculated based on the reference position of the window, the reference position of the model image and the window position, respectively.

Description

Detailed Description of the Invention

[0001]

TECHNICAL FIELD The present invention relates to a method for determining a reference position of a model image in a position measuring device by image processing.

[0002]

BACKGROUND ART The following is known as a position measuring device by image processing.

Model image data obtained by picking up an image of a model and a reference position of an image representing the model are registered in advance. The target image data obtained by picking up the image of the measurement target is scanned by the window containing the model image data, and the calculation regarding the degree of coincidence between the model image data and the target image data is performed at each position of the window. The window position on the target image data with the highest degree of matching is found. A measurement position of the measurement target corresponding to the reference position of the model image data is calculated based on the reference position of the model image, the reference position of the window including the model image data, and the window position on the found target image data.

However, in such a conventional position measuring apparatus, since the reference position of the model image data is calculated in pixel units, the position of the target object cannot be calculated with a precision smaller than the pixel size.

[0005]

DISCLOSURE OF THE INVENTION An object of the present invention is to enable a position measuring device by image processing to accurately measure the position of a measuring object by accurately calculating a reference position of model image data.

A method for determining a reference position of a model image in a position measuring device by image processing according to the first aspect of the present invention registers model image data and its reference position in advance,
The target image data obtained by picking up the image of the measurement target is scanned by the window containing the model image data, and the calculation of the degree of matching between the model image data and the target image data is performed at each position of the window to obtain the best matching degree. Find the window position on the high target image data,
The measurement position of the measurement target corresponding to the reference position of the model image data is calculated based on the reference position of the model image data, the reference position of the window including the model image data, and the found window position on the target image data. The position measuring device is characterized in that the barycentric position of the model image data obtained by imaging the model is calculated in units lower than the pixel unit, and this barycentric position is used as the reference position of the model image data.

According to the first invention, the barycentric position of the model image data is calculated in units lower than the pixel unit, and this barycentric position becomes the reference position of the model image data. Since the position of the measurement target is measured based on the reference position of the model image calculated in units lower than the pixels, the position can be measured more accurately than in the case of calculating in units of pixels.

In a preferred embodiment of the first aspect of the invention, the model image data obtained by imaging the model is binarized, and the position of the center of gravity of the binary model image data is calculated in units lower than the pixel unit. It is characterized in that the position of the center of gravity is used as the reference position of the model image data.

Therefore, the model image data is binarized, and the center of gravity position of the binary model image data is calculated in a unit lower than the pixel and used as the reference position of the model image data. When the binary model image data described above is obtained, the accurate reference position of the model image data can be calculated.

In another preferred embodiment of the first aspect of the invention, the contour line of the image representing the model is extracted from the model image data, the center of gravity position of the contour line is calculated in units lower than the pixel unit, and the center of gravity is calculated. The position is set as a reference position of the model image data.

Therefore, the contour line of the image representing the model is extracted from the model image data, and the position of the center of gravity of the contour line of the model image is calculated in units lower than the pixels and used as the reference position of the model image data. ， When stable binary image data cannot be obtained inside the model,
For example, when the model is a character, the accurate reference position of the model image data can be calculated.

In a further preferred embodiment of the first invention, the gray
It is characterized in that the barycentric position of the model image data expressed by level is calculated in units lower than the pixel unit, and this barycentric position is used as the reference position of the model image data.

Therefore, the barycentric position of the model image data represented by the gray level is calculated in a unit lower than the pixel and used as the reference position of the model image data, so that the model image data is binarized. Even more accurately, the reference position of the model image data can be calculated.

In the reference position determining method of the model image in the position measuring device by the image processing according to the second invention, the model image data and its reference position are registered in advance,
The target image data obtained by picking up the image of the measurement target is scanned by the window containing the model image data, and the calculation of the degree of matching between the model image data and the target image data is performed at each position of the window to obtain the best matching degree. Find the window position on the high target image data,
The measurement position of the measurement target corresponding to the reference position of the model image data is calculated based on the reference position of the model image data, the reference position of the window including the model image data, and the found window position on the target image data. In the position measurement device, the model image data obtained by imaging the model is displayed on the display device, and the gray level distribution of the model image data is displayed on the two cutting lines specified on this display screen which are orthogonal to each other. Create each of them, calculate the coordinates of the middle point of the two edges at both ends specified on each gray level distribution in units lower than the pixel unit, and use the calculated coordinates as the reference position of the model image data. It is characterized by

According to the second aspect of the invention, the gray level distribution is created on the specified cutting line on the display screen on which the model image data is displayed, and the two edges specified on the gray level distribution are created. The coordinates of the intermediate point are calculated in units lower than the pixel, and the coordinates of this intermediate point are used as the reference position. Therefore, it is possible to calculate an accurate reference position based on the two edges at both ends of the cutting line designated by the operator.

A reference position determining method of a model image in a position measuring device by image processing according to a third aspect of the present invention registers model image data and its reference position in advance,
The target image data obtained by picking up the image of the measurement target is scanned by the window containing the model image data, and the calculation of the degree of matching between the model image data and the target image data is performed at each position of the window to obtain the best matching degree. Find the window position on the high target image data,
The measurement position of the measurement target corresponding to the reference position of the model image data is calculated based on the reference position of the model image data, the reference position of the window including the model image data, and the found window position on the target image data. In the position measuring device, the model image data obtained by capturing the image of the model is binarized, two straight lines representing the contour lines of the image representing the model are calculated, and the coordinates of the intersection of these two straight lines are calculated in pixel units. It is characterized in that it is calculated in a lower unit and the calculated coordinates are used as the reference position of the model image data.

According to the third invention, the model image data is binarized, two straight lines representing the contour line of the image representing the model are calculated for the binary model image data, and the intersection of the two straight lines is further calculated. Are calculated in units lower than the pixel, and the coordinates of this intersection are used as the reference position. Therefore, an accurate reference position can be determined when the model has, for example, a triangular shape.

[0018]

[Explanation of Examples]

Table of contents 1 Hardware configuration 2 Position measurement of target object 3 Registration of model image and its reference position 3.1 Registration of model image 3.2 Method of using center of gravity of gray level image data of model image as reference position 3.3 Binary image of model image Method of setting the center of gravity of the data as the reference position 3.4 Method of setting the center of gravity of the contour line of the model image as the reference position 3.5 Method of setting the midpoint between the two edges of the model image as the reference position 3.6 Representing the contour line of the model image 2 How to use the intersection of straight lines in a book as the reference position

1 hardware configuration

FIG. 1 is a block diagram showing the electrical construction of a position measuring device using image processing.

A model or a target object is imaged by the camera 11, and an analog image of the model image obtained by the imaging
The video signal is applied to the A / D converter (analog / digital converter) 12. The A / D converter 12 converts the video signal supplied from the camera 11 into gray level digital image data. The gray level image data is represented by, for example, 256 gradations (8 bits / pixel). The gray level image data is temporarily stored in the image memory 13.

The D / A converter 14 converts the gray level digital image data stored in the image memory 13 into an analog video signal. The D / A converted video signal is applied to the video monitor 16 and the image represented by the analog video signal is displayed. The character memory 15 stores font data for displaying characters and the like. The font data is appropriately D / A converted by the D / A converter 14 and displayed on the video monitor 16.

The timing control circuit 10 is responsive to the timing control signal given from the CPU 21 to respond to the A / D converter.
12, image memory 13, D / A converter 14 and character memory 15
, A timing signal is output to each of them to synchronize operations in these circuits, transmission and reception of data between them, and the like.

The ROM 22 stores a program for realizing the processing described later. The RAM 22 stores intermediate data of processing as appropriate. The CPU 21 executes the processing described later according to the program stored in the ROM 22. CP
U21 also sends the timing control signal to the timing control circuit.
The data is output to 10 and the data transmission and reception of the entire position measuring device by this image processing are synchronized. Characters are input from the keyboard 25 and positions are specified.

The I / O interface 24 sends and receives data to and from other devices.

CPU 21, ROM 22, RAM 23 and I /
The O interface 24 is generally realized by a microcomputer, a so-called personal computer or the like.

2 Position measurement of target object

The position measuring process of the measuring object by the position measuring device by image processing will be briefly described below.

The CPU 21 pre-registers model image data representing the model of the measurement target before measuring the position of the measurement target. FIG. 2 shows an example of a window containing registered model image data and a model image represented by the model image data. The coordinate P0 (x0, y0) of the window shown in FIG. 2 is the position (window reference position) which is the reference of the window. The model image data registration will be described in detail in the following “3 Model Image Registration and Its Reference Position Registration”.

Further, the CPU 21 determines the reference position of the registered model image data by the sub pixel and registers it. In FIG. 2, the point MG (MGx, MGy) is the reference position of the model image data. There are five modes for determining the reference position, which will be described in detail in "Registration of 3 model images and registration of the reference position".

Next, an image of the object to be measured is picked up by the camera 11, and its video signal is grayed out by the A / D converter 12.
Convert to level image data. The converted gray level image data is stored in the image memory 13. The gray level image data is converted into a video signal by the D / A converter 12, and the measurement target image represented by the video signal is displayed on the video monitor 16. FIG. 3 shows an example of the measurement target image displayed on the video monitor 16.

The CPU 21 scans the target image data in the search region of the target image with the window containing the model image data registered previously, and normalizes the model image data and the target image data existing in the window. Perform correlation calculation. The CPU 21 performs the normalized correlation calculation according to the equation (1) to calculate the normalized correlation value F (xi, yi) at each position (xi, yi).

[0033]

[Equation 1]

The CPU 21 obtains the window position where the degree of matching between the target image data and the model image data is the highest, based on the obtained normalized correlation value. The window position is, for example, the barycentric position of the normalized correlation value, the maximum value position by parabolic approximation in the vicinity of the maximum normalized correlation value, or the like. The window position is calculated in units lower than the pixel.

(1) When the normalized correlation centroid position is set to the window position

The CPU 21 uses equation (2) using the normalized correlation value.
, (3), the barycentric position FG (FGx, FGy) of the normalized correlation value F (xi, yi) is calculated.

[0037]

[Equation 2]

The barycentric position FG (FGx, FGy) of the normalized correlation value calculated in this way is set to the window position U (U
x, Uy).

(2) When the maximum value position by parabolic approximation near the maximum value of the normalized correlation value is the window reference position

The CPU 21 uses the X-coordinates X0, X1 and X2 of the three points in the vicinity of the calculated maximum value of the normalized correlation value and the respective normalized correlation values F0, F1 and F2 to calculate the expression.
Solving the three equations (4), (5) and (6), the coefficient ax
, Bx and cx are calculated. FIG. 4 shows an example of a graph near the maximum normalized correlation value.

[0041]

[Equation 3]

When the CPU 21 calculates the coefficients ax, bx and cx, the X coordinate of the window position is calculated by Ux =-(bx / 2ax). The CPU 21 similarly performs the Y coordinate, and calculates the Y coordinate Uy of the window position.

In this way, the window position U (Ux, Uy) having the highest degree of matching between the measurement target image and the model image is obtained based on the normalized correlation value.

The CPU 21 controls the model image window reference position P0 (x0, y0) and the model image reference position MG.
The measurement position S (Sx, Sy) is calculated based on (MGx, MGy)) and the window position U (Ux, Uy) obtained by the normalized correlation calculation. The CPU 21 calculates the measurement position S (Sx, Sy) according to the equations (7) and (8).

[0045]

[Equation 4]

As described above, the measurement position S of the target object
Is calculated. The position of the measurement target can be accurately measured by calculating the reference position of the model image and the window position by the normalized correlation calculation in units lower than the pixel.

3 Registration of model image and its reference position 3.1 Registration of model image

The model is imaged by the camera 11, and its video signal is converted into gray level image data by the A / D converter 12. The converted gray level image data is stored in the image memory 13. The gray level image data is converted into a video signal by the D / A converter 12, and the model image represented by the video signal is displayed on the video monitor 16. FIG. 5 shows an example of the model image displayed on the video monitor 16.

The operator looks at the model image (FIG. 5) displayed on the video monitor 16 and operates the cursor on the keyboard 25 to specify a window to be registered as model image data. The window has two diagonal points, P0 (x0,
y0) and P1 (x1, y1) are specified, and this is a rectangular area represented by the two diagonal points.

The CPU 21 registers the gray level image data in the window designated by the operator in the image memory 13 as model image data. The window reference position of the window including the registered model image data is the point P0 (x0, y0) designated by the operator. The model image data may be automatically registered by the device.

When the model image data is registered, the CPU 21 calculates the reference position of the model image for the registered model image in units lower than the pixels. The reference position determination method has the following five modes.

3.2 Method of Using Center of Gravity of Binary Image Data of Model Image as Reference Position

The CPU 21 converts the gray level image data of the registered model image into binary image data. Figure 6
Shows an example of a binary model image represented by the binary image data of the model image shown in FIG. The CPU 21 calculates the binary barycentric position of this binary image data. The binary center of gravity position MG (MGx, MGy) is calculated according to equations (9) and (10).

[0054]

[Equation 5]

The binary centroid position is calculated in units lower than the pixel. This binary centroid position is used as the reference position of the model image. In FIG. 6, the coordinate MG (MGx, MGy) is the reference position. The CPU 21 registers the calculated reference position of the model image.

With this method, an accurate reference position can be calculated when stable binary image data is obtained for the model image.

3.3 Method Using Center Position of Contour of Model Image as Reference Position

This method does not calculate the binary center of gravity position by using all the binary image data of the model image, but
The position of the center of gravity of the contour of the model image is calculated using the binary image data located on the contour of the model image.

The position of the center of gravity of the contour is M in equations (9) and (10).
It can be calculated by using (xi, yi) as binary image data of pixels located on the contour of the model image instead of the binary image data. The position of the center of gravity of the contour is calculated in units lower than the pixel.

The position of the center of gravity of the contour thus calculated is used as the reference position. This method can calculate an accurate reference position when stable binary image data cannot be obtained inside the model, for example, when the model is a character.

3.4 Method of Using Center of Gravity of Gray Level Image Data of Model Image as Reference Position

The CPU 21 calculates the barycentric position of the gray level image data of the registered model image. The position of the center of gravity can be calculated by using M (xi, yi) as gray level image data instead of binary image data in the equations (9) and (10). The binary centroid position is calculated in units lower than the pixel.

The center-of-gravity position thus calculated is used as the reference position of the model image. This method can calculate a more accurate reference position than the reference position based on the binary image data.

3.5 Method Using Intermediate Position of Contour Position of Specified Model Image as Reference Position

The CPU 21 D / A converts the registered model image data into a video signal and displays the model image represented by the model image data on the video monitor 16. FIG. 7A shows an example of the display screen of the video monitor 16.

The operator operates the cursor on the keyboard 25 while looking at the model image (FIG. 7 (A)) displayed on the video monitor 16 to specify a line (X line) parallel to the X axis of the model image. . The CPU 21 creates a gray level distribution of the model image data on the X line designated by the operator and displays it on the video monitor 16. FIG. 7B shows an example of the gray level distribution.
Displaying the gray level distribution of the image data on the specified X line on the video monitor 16 is called line bright display.

The operator looks at the gray level distribution displayed in line bright (FIG. 7 (B)) and operates the cursor on the keyboard 25 to set two edges (X) at both ends of the model.
0 and X1) Specify. The operator similarly designates two edges Y0 and Y1 in the Y direction.

In the CPU 21, the operator inputs X of the model image.
When two edges are designated for each of the direction and the Y direction, the intermediate position of each designated edge is calculated. The CPU 21 calculates the intermediate position according to the equations (11) and (12). The intermediate position is calculated in units lower than the pixel.

[0069]

[Equation 6]

The intermediate position thus calculated is used as the reference position of the model image. By displaying the line blind, the reference position based on the edge of the model image designated by the operator can be calculated.

3.6 Method of Using Reference Point at Intersection Point of Two Straight Lines Representing Contour Line of Model Image

This method can calculate an accurate reference position when the model image has a polygonal shape, for example, a triangular shape (see FIG. 8A).

The CPU 21 converts the registered model image data into binary image data. FIG. 8A shows an example of a model image represented by binary image data. FIG. 8B is an enlarged view of the contour line of the binary model image shown in FIG. In FIG. 8 (B), the pixels of the contour line of the model image are shown by double hatching.

The CPU 21 calculates a linear equation representing two sides of the contour line of the model image for the binary image data. The linear equation is calculated by the least squares method using the coordinate data of the pixels of the contour line of the model image (pixels indicated by double hatching in FIG. 8B). If the equation of the straight line is Y = AX + B, the coordinates of the pixel of the contour line of the model image are (xi, yi)
Then, it can be calculated by equations (13) and (14).

[0075]

[Equation 7]

After calculating the two linear equations, the CPU 21 calculates the coordinates of the intersection MG of these two straight lines. The coordinates of the intersection are calculated in units lower than the pixel. The intersection is used as the reference position. The intersection of two straight lines is calculated in units lower than the pixel. The straight line equation may be calculated based on the coordinates of any two pixels of the contour line of the model image.

With this method, an accurate reference position can be calculated when the model has, for example, a polygonal shape, especially a triangular shape, as described above.

[Brief description of drawings]

FIG. 1 is a block diagram showing an electrical configuration of a position measuring device by image processing.

FIG. 2 shows an example of a model image represented by registered model image data.

FIG. 3 shows an example of a target image represented by target image data obtained by capturing an image of a measurement target.

FIG. 4 is a graph showing the vicinity of the maximum normalized correlation value.

FIG. 5 shows an example of a model image represented by model image data obtained by capturing an image of a model.

FIG. 6 shows an example of a model image represented by binary image data obtained by binarizing registered model image data.

FIG. 7 shows an example of a line bright display.

FIG. 8 shows an example of a binary image represented by binary image data obtained by binarizing a registered model image.
It is an enlarged view of the outline portion of the binary image shown in (A).

[Explanation of symbols]

 10 Timing control circuit 11 Camera 12 A / D converter 13 Image memory 14 D / A converter 15 Character memory 16 Video monitor 21 CPU 22 ROM 23 RAM 24 I / O interface 25 Keyboard

Claims (6)

[Claims] 1. Model image data and its reference position are registered in advance, and target image data obtained by imaging a measurement target is scanned in a window containing the model image data, and a model is created at each position of the window. The calculation of the matching degree between the image data and the target image data is performed, the window position on the target image data with the highest matching degree is found, and the reference position of the model image data,
An image of a model is captured by a position measuring device that calculates the measurement position of the measurement target corresponding to the reference position of the model image data based on the reference position of the window containing the model image data and the window position found on the target image data. The center of gravity position of the model image data obtained by performing the calculation is calculated in units lower than the pixel unit, and this center of gravity position is used as the reference position of the model image data.
A method for determining a reference position of a model image in a position measuring device by image processing. 2. Model image data obtained by capturing an image of a model is binarized, and a barycentric position of the binary model image data is calculated in a unit lower than a pixel unit, and the barycentric position is a reference position of the model image data. The method for determining a reference position of a model image in a position measuring device by image processing according to claim 1, wherein: 3. In the model image data obtained by capturing an image of a model, a contour line of an image representing the model is extracted, the center of gravity position of this contour line is calculated in units lower than the pixel unit, and the center of gravity position is calculated. The method for determining a reference position of a model image in a position measuring device by image processing according to claim 1, characterized in that the reference position of the model image data is used. 4. A center of gravity position of model image data represented by a gray level obtained by capturing an image of a model is calculated in a unit lower than a pixel unit, and this center of gravity position is used as a reference position of the model image data. A method for determining a reference position of a model image in a position measuring device by image processing according to claim 1, characterized in that. 5. Model image data and its reference position are registered in advance, and target image data obtained by imaging a measurement target is scanned in a window containing the model image data, and a model is created at each position of the window. The calculation of the matching degree between the image data and the target image data is performed, the window position on the target image data with the highest matching degree is found, and the reference position of the model image data,
An image of a model is captured by a position measuring device that calculates the measurement position of the measurement target corresponding to the reference position of the model image data based on the reference position of the window containing the model image data and the window position found on the target image data. The model image data obtained by doing so are displayed on the display device, and the gray level distributions of the model image data are respectively created on the two cutting lines that are orthogonal to each other and are specified on this display screen. Image processing characterized in that coordinates of an intermediate point between two edges at both ends specified on the distribution are respectively calculated in units lower than the pixel unit, and the calculated coordinates are used as reference positions of model image data. Method for determining a reference position of a model image in a position measuring device according to. 6. Model image data and its reference position are registered in advance, and target image data obtained by imaging a measurement target is scanned in a window containing the model image data, and a model is created at each position of the window. The calculation of the matching degree between the image data and the target image data is performed, the window position on the target image data with the highest matching degree is found, and the reference position of the model image data,
An image of a model is captured by a position measuring device that calculates the measurement position of the measurement target corresponding to the reference position of the model image data based on the reference position of the window containing the model image data and the window position found on the target image data. By binarizing the model image data obtained by doing the above, two straight lines representing the outline of the image representing the model are calculated, and the coordinates of the intersection of these two straight lines are calculated in units lower than the pixel unit, A method for determining a reference position of a model image in a position measuring device by image processing, characterized in that the calculated coordinates are used as a reference position of model image data.