CN102022983A - Method and device for measuring two-dimensional displacement by matching contrast serving as characteristic frame - Google Patents

Abstract

A method and device for measuring two-dimensional displacement with contrast as a characteristic frame matching, which is composed of a computer camera and an ordinary computer; firstly extract an axial side direction data in the reference frame as the contrast feature of the reflected image of the measured object , by calculating the self-correlation coefficient of the edge direction data, check the texture details of the reflective surface of the measured object, select the appropriate comparison window pixel array; As the best matcher, the two-dimensional displacement and velocity are obtained; according to the results, the position of the comparison window is adjusted or the reference frame is updated, and the scale of the cross-correlation operator array is adjusted to ensure the measurement accuracy and reduce the amount of calculation. To a certain extent It overcomes the influence of ambient light changes on the measurement.

Description

Method and device for measuring two-dimensional displacement using contrast as characteristic frame matching

technical field

The invention belongs to the technical field of digital image measurement, in particular to a method and a device for measuring two-dimensional micro-displacement of an object by using a computer camera.

Background technique

Computer cameras have become popular, and their core is an array of photoelectric sensing units, which lays the material foundation for measuring the motion of objects. "Method and device for measuring small two-dimensional displacement using computer camera" (invention patent application number: 2009101042778) analyzed several domestic patents related to "camera" and "measurement", and referred to some foreign patents on the detection of displacement by scanners and optical mice The patent proposes a method and device for measuring the tiny displacement of an object by using a computer camera and using correlation matching technology. However, this application is only aimed at the light intensity of the image frame, which is suitable for lighting conditions and the situation where the optical image of the reflective surface of the object is relatively stable.

Contents of the invention

In order to give full play to the function of the photoelectric sensor array of the computer camera, the present invention provides a method and device for measuring two-dimensional displacement with contrast as the characteristic frame matching. In the environment, measure the tiny two-dimensional displacement vector and velocity vector of the object on the plane perpendicular to the optical axis of the camera.

The technical solution adopted by the present invention to solve its technical problems is: the object to be measured is focused and imaged by a computer camera, and the camera is connected to an ordinary computer through a USB interface, and the computer is equipped with a USB interface, memory, CPU, hard disk, display Card and display, keyboard and mouse, operating system, camera driver, and camera shooting and edge direction data frame matching measurement displacement program; this program embodies the method of the present invention to implement frame matching to measure the tiny two-dimensional displacement of an object characterized by contrast, including :

Step 1. Taking a frame of an image of the object under test in the format of a bitmap (M×N, M, N∈ positive integer) as a reference frame;

Taking the position of the first pixel in the upper left corner of the pixel array as the origin, the horizontal direction to the right is the x-axis direction, and the vertically downward direction is the y-axis direction;

Select an area in the central area of the pixel array, the size of which is m ₀ ×n ₀ , m ₀ , n ₀ ∈ positive integer, called the comparison window, and the comparison window is respectively separated from the pixel array in the horizontal direction and vertical direction There are h and v pixels in each edge pixel in the direction, namely: m ₀ +2h=M, n ₀ +2v=N, h, v ∈ positive integer;

Step 2. For the pixel array of the above reference frame, export the edge direction data along the X-axis row by row:

If the light intensity value of a pixel is smaller than the light intensity value of the second pixel behind it by an error tolerance value error, that is, if I(X, Y)<I(X+2, Y)-error is defined There is a positive edge between these two pixels;

If the light intensity value of a pixel is greater than the light intensity value of the second pixel behind it by an error tolerance value error, that is, if I(X, Y)>I(X+2, Y)+error is defined There is a negative edge between these two pixels;

If the light intensity value of a pixel is close to the corresponding light intensity value of the second pixel behind it, the difference does not exceed an error tolerance value error, that is, if I(X+2, Y)-error<I(X, Y )<I(X+2, Y)+error, it is considered that there is no "edge" between these two pixels, or it is called the third type of edge;

The error tolerance value in the above formula can be preset to a small value according to the specific lighting conditions, for example: error=10; the edge obtained in this way is located at the position of the first pixel after this pixel, that is, it is located at the position participating in the comparison On the pixel (x, y) in the middle of the two pixels; along the X-axis direction, all positive edges, negative edges, and third-type edges form the edge direction data of this direction, respectively in 3-bit binary value 001 , 010 and 100 represent, recorded as reference (x, y), save it;

Step 3, calculating the self-correlation matching coefficient of the pixel array of the comparison window in the reference frame:

$\mathrm{<span\; class="notranslate">\; auto</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; correlation</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$

In the formula, x and y are the coordinates of the pixels in the comparison window respectively, and the operation symbol · represents the binary logical AND operation, and the operation result is either logic 0 or logic 1, and the square brackets "[]" represent the value corresponding to the logic function in it. The value, or the value 0, or the value 1, is calculated pixel by pixel, and the accumulated result is used as this auto_correlation coefficient auto_correlation(a, b), and the 3×3 correlation matching operator is used: a=-1, 0, 1, b=-1, 0, 1, therefore, a total of 9 autocorrelation coefficients are generated;

Step 4. Analyze the fineness of the structural features on the surface of the observed object, that is, analyze the distinguishability of the light and dark contrast between the nearest adjacent pixels, and check:

auto_correlation(a,b)≥auto_correlation(0,0)×similarity

In the formula, similarity describes the similarity between the comparison window and its neighboring pixel arrays of the same scale, for example, similarity=60%, which can be set in advance, or can be debugged and selected according to the lighting conditions and the texture of the surface of the measured object;

If there are more than 3 self-association matching coefficients satisfying the above checking inequality, it is necessary to gradually expand the scope of the comparison window for each step row and step column: let m=m ₀ +step, n=n ₀ +step, and recalculate the new comparison window The self-correlation matching coefficient of the adjacent pixel is analyzed again until the self-correlation coefficient satisfying the above analysis inequality is not more than 3, and the result is 2h=Mm, 2v=Nn, where step is a stepping parameter, the initial value is 1, the scale of the comparison window needs to be expanded by 1 each time; if it exceeds a predetermined range in the frame and no suitable comparison window has been found, it is considered that this part of the reflective surface of the object is not suitable for the measurement work of the device, and given Prompt warning;

If there are no more than 3 self-correlation matching coefficients satisfying the above checking inequality, it means that the surface structure features of the object to be photographed are fine enough, and the values between the nearest neighbor pixels can be distinguished, and further attempts can be made to narrow the range of the comparison window by each step Rows and step columns to reduce the amount of calculation: let m=m ₀ -step, n=n ₀ -step, recalculate the self-correlation matching coefficient of the comparison window, and conduct the adjacent pixel discrimination analysis again until the selected comparison window area The number of self-correlation matching coefficients satisfying the above check inequality is equal to 3, and it is considered that the optimal comparison window pixel array that can be used for matching comparison under the current object surface condition and lighting condition is found: m=m ₀ -step , n=n ₀ -step, 2h=Mm, 2v=Nn;

Step 5. After the above shooting, after a period of time Δt, take a second frame of bitmap as a sampling frame;

Determine the side direction data of the pixels in the sampling frame along the X-axis direction line by line, and use 3-bit binary values 001, 010, and 100 to represent the positive side, negative side, and the third type of side, which are recorded as comparison (x, y), and saved Of;

Step 6. Perform cross-correlation matching with the pixel array {reference (x, y)} in the comparison window in the reference frame in the range of the sampling frame {comparison (x, y)} according to the 9×9 correlation matching array, The specific algorithm is:

$\mathrm{<span\; class="notranslate">\; the\; cross</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; correlation</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; comparison</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$

Corresponding to a=-4, -3, -2, -1, 0, +1, +2, +3, +4 and b=-4, -3, -2, -1, 0, +1, +2 , +3, +4 combination, a total of 81 cross-correlation matching coefficients cross_correlation (a, b) are generated, that is, there may be 81 kinds of movement situations in the comparison window of the reference frame;

Step 7. The cross-correlation coefficient with the highest degree of frame-frame correlation is the largest:

cross_correlation(a, b) → auto_correlation(0, 0)

Therefore, the direction in which the sampled frame moves relative to the reference frame and the magnitude of the movement are obtained:

Δx(i,j)=a, Δy(i,j)=b,

This is the relative displacement vector of the object in this sampling period, where i represents the sequence count of the measurement and shooting, and j represents the sequence count of the reference frame taken; during the measurement process, the total relative displacement vector of the object is:

Δx ₀ (i,j)=Δx ₀ (i-1,j)+Δx(i,j), Δy ₀ (i,j)=Δy ₀ (i-1,j)+Δy(i,j), The right side of the equation (Δx ₀ (i-1, j), Δy ₀ (i-1, j)) is the relative displacement vector accumulated by the object before;

Step 8. Velocity vector of object displacement:

Δv _x (i, j) = Δx (i, j)/Δt, Δv _y (i, j) = Δy (i, j)/Δt;

Step 9. If |Δx ₀ (i,j)-Δx ₀ (k,j-1)|≥2m/5, or |Δy ₀ (i,j)-Δy ₀ (k,j-1)|≥2n /5, wherein, k=max(i)|(j-1) represents the sequence count value of the last shot when the sequence count of the reference frame is j-1, that is, under the condition that the reference frame does not change , where the accumulation of the relative displacement of the comparison window has exceeded 2/5 of the magnitude of the comparison window, at this time, replace the reference frame with the latest sampling frame, and its comparison window is repositioned at the center of the new reference frame ;

If |Δx ₀ (i,j)-Δx ₀ (k,j-1)|<2m/5 and |Δy ₀ (i,j)-Δy ₀ (k,j-1)|<2n/5, do not Update the reference frame, but relatively shift the comparison window in the reference frame by Δx=a, Δy=b;

Step 10. If the reference frame is updated, calculate the self-correlation matching coefficient of the new reference frame as in step 3, check the surface structure features as in step 4, and re-determine the size of the optimal comparison window m×n;

If the reference frame is not updated, adjust the size of the cross-correlation matching operator array in step 6: if |a|<5 and |b|<5, change it to 7×7: a=-3, -2, - 1, 0, +1, +2, +3, b=-3, -2, -1, 0, +1, +2, +3, if |a|<3 and |b|<3, take Be 5*5: a=-2,-1,0,+1,+2, b=-2,-1,0,+1,+2, otherwise still take as 9*9 associative matching operator array;

Step 11. After the above-mentioned shooting, after a period of time Δt, take a third frame of bitmap as a new sampling frame;

Determine the side direction data of the pixels in the sampling frame along the X-axis direction line by line, where the positive side, negative side and the third type of side are respectively represented by 3-bit binary values 001, 010 and 100, which are recorded as comparison(x, y) , save it;

Step 12, according to the cross-correlation matching operator array determined in step 10, perform cross-correlation matching calculation on the pixel array in the comparison window in the reference frame in the range of the sampling frame, the specific algorithm is the same as step 6;

Step 13, skip to step 7 and continue measuring.

In the actual measurement process, measurement calibration can be further implemented to obtain direct measurement results.

The beneficial effect of the present invention is that it extracts the edge direction data as the contrast feature of the pixel array of the image frame, which effectively overcomes the influence of ambient light changes on the light intensity of image pixels; Measure the texture characteristics of the reflective surface of the object, and automatically adjust the size of the matching area to achieve the minimum calculation workload; it automatically adjusts the size of the cross-correlation operator according to the calculated displacement value to achieve the best search for the matching area; it automatically and timely The reference frame is updated accurately, and there is no preset threshold value that needs to intervene in the program, effectively ensuring the measurement accuracy.

Description of drawings

Fig. 1 is a block diagram of the measuring device of the present invention.

FIG. 2 is a schematic diagram of a pixel array generated after the photoelectric sensor chip undergoes photoelectric conversion.

In Figure 1, 1. Computer camera, 2. Optical lens, 3. Photoelectric sensor chip, 4. USB interface, 5. Computer system, 6. USB interface, 7. CPU, 8. Memory, 9. Display card and monitor, 10. Hard disk, 11. Keyboard and mouse, 12. Operating system, 13. Camera driver, 14. Camera shooting and edge direction data frame matching measurement displacement program, 15. Lighting equipment.

In Fig. 2, 20. a frame of pixel array, 21. a comparison window, 22. a schematic example of an associated matching area in a sampling frame, and 23. a possible limit position of a comparison window in a reference frame.

Detailed ways

Fig. 1 is a block diagram of the measuring device of the present invention. First, the camera driver (13) distributed with the camera (1) is run on the computer system (5), and the camera (1) is connected to the computer (5) through the USB interfaces (4) and (6). Then, let the camera focus and image the object to be measured. The basic requirement for measuring ambient lighting conditions is that the light intensity imaged by the object to be measured can change to some extent, but the contrast between light and dark of the image is not allowed to change significantly. Based on this, the measurement environment is selected and the lighting equipment (15) is selected. Then, run the camera shooting and edge direction data frame matching measurement displacement program (14) to implement real-time measurement. This program (14) embodies the method for measuring two-dimensional displacement with contrast as the feature frame matching, specifically see the description in "Summary of the Invention", and the main points are explained as follows below:

In steps 1, 5 and 11, take a frame of the image of the measured object in bitmap format, the size of the bitmap frame is M×N, M, N∈ positive integer needs to be selected according to the specific model of the camera, and as far as possible It is possible to choose a frame rate with a faster shooting speed to ensure that the shooting speed is faster than the displacement speed of the measured object.

The common resolutions of digital cameras are QSIF(160*120), QCIF(176*144), SIF(320*240), CIF(352*288) and VGA(640*480), and the actual shooting speed corresponds to: 30fps (frames per second), 30fps, 20-26fps, 20-26fps and 10fps. Common resolutions are: QSIF(160*120), QCIF(176*144), SIF(320*240), CIF(352*288) and VGA(640*480), etc. The actual shooting speeds are : 30fps (frame per second), 30fps, 20-26fps, 20-26fps and 10fps, the video signal belongs to the digital signal mode and is output through the USB interface.

Fig. 2 shows the corresponding pixel array (20) generated after photoelectric conversion by the photosensor chip (3). The size of the pixel array (20) is M×N, M, N∈ integer, corresponding to the bitmap format, wherein each small square grid is a pixel, and its brightness value is 0-255. In the illustrated coordinate system, each small square grid or pixel is positioned as a pair of coordinate points (x, y), 0≤x≤M, 0≤y≤N, x, y∈integer. A comparison window (21) is opened in the central area in Fig. 2, and the size is taken as m×n, m, n∈ integer, and the edge pixels of the comparison window (21) and the horizontal direction and the vertical direction of the pixel matrix (20) are respectively The distance between h and v pixels is: m+2h=M, n+2v=N, h, v∈positive integer. Therefore, the coordinates of the first pixel square in the upper left corner of the comparison window are: x=h+1, y=v+1. The selection of m and n is related to the fineness of the structural features of the object's reflective surface, which is related to (matching) measurement accuracy, determines the calculation workload, and affects the response rate of the measurement device. For the above resolution and frame rate indicators, for example, initial values may be selected: m ₀ =80, n ₀ =80. After the object is displaced, the comparison window in the reference frame may move to a certain place in the sampling frame, such as the example 22 of the associated matching area in the sampling frame, but it is not allowed to move beyond the possible limit of the comparison window in the reference frame position 23.

The edge direction data described in steps 2, 5 and 11 reflect the light and dark contrast characteristics of the surface of the measured object. The recently submitted invention patent application "Peak and Valley Motion Detection Method and Device for Measuring Sub-pixel Displacement" provides detailed information on this. illustrate.

Steps 3, 4 and 10 are to use the autocorrelation coefficient to analyze the fineness of the structural features on the surface of the measured object to see if the imaged frame under the current lighting environment can show enough details to be suitable for this measurement method. On the one hand, this is judged by analyzing whether each autocorrelation coefficient is close to the self-correlation coefficient of the comparison window itself; on the other hand, it is characterized by a parameter similarity that can be set in advance. This parameter describes the similarity between the comparison window and its adjacent pixel array of the same size, which can be adjusted and selected according to the lighting conditions and the texture of the surface of the measured object. The synthesis is the following analytical inequality:

auto_correlation(a,b)≥auto_correlation(0,0)×similarity

When "the number of autocorrelation coefficients satisfying the above analysis inequality is equal to 3", it is considered that the optimal comparison window pixel array for matching comparison has been found: m=m ₀ -step, n=n ₀ -step, 2h=Mm , 2v=Nn, otherwise, the scale of the comparison window will be enlarged or reduced automatically, which is beneficial to reduce the calculation workload. In extreme cases, the reflection characteristics of the surface of the measured object are not detailed enough, and the adjustment of the scale of the comparison window may exceed the range of the reference frame. Therefore, the adjustment range limit should be limited. Some reflective surfaces are not suitable for the measurement work of this method and device, and a warning will be given.

"The number of autocorrelation coefficients satisfying the above analysis inequality" does not necessarily have to be equal to 3, and can be adjusted according to the actual situation.

The basic idea in steps 6, 7 and 12 is to first use a large-scale cross-correlation matching operator matrix to search for the region that best matches the comparison frame in the reference frame, thereby determining the displacement of the pixel frame. Then, according to the displacement obtained, the size of the cross-correlation operator matrix is adjusted in order to reduce the amount of calculation. The size of the cross-correlation operator matrix is larger than the range of possible displacements.

Step 9 describes the method of moving the position of the built-in comparison window in the reference frame according to the obtained displacement measurement value, or updating the reference frame. There are many overlapping areas, which can reflect the measurement accuracy of less than one pixel unit, without accumulating measurement errors after multiple shooting measurements. "Method and device for measuring small two-dimensional displacement using computer camera" (invention patent application number: 2009101042778) made a specific analysis on this. As for "under the condition that the reference frame does not change, the accumulation of the relative displacement of the comparison window has exceeded 2/5 of the magnitude of the comparison window, at this time, replace the reference frame with the latest sampling frame, "The 2/5 overrun can be adjusted according to specific circumstances such as the size of the comparison window and the speed of the measured movement.

Claims (2)

1. The method and device for measuring two-dimensional displacement with contrast as characteristic frame matching, are made up of a computer camera and a common computer, and the camera is connected to the computer through its USB interface, and the computer is equipped with a USB interface, memory , CPU, hard disk, display card and display, keyboard and mouse, operating system and camera driver, it is characterized in that, described computer system is also equipped with camera shooting and edge direction data frame matching measurement displacement program. 2. The method and device for measuring the two-dimensional displacement according to claim 1, wherein the contrast is characterized in that the camera shooting and edge direction data frame matching measurement displacement program provides a method for using a computer camera to implement Contrast measures two-dimensional displacement for feature frame matching, including: Step 1. Taking a frame of an image of the measured object in the format of a bitmap (M×N, M, N∈ positive integer) as a reference frame; Taking the position of the first pixel in the upper left corner of the pixel array as the origin, the horizontal direction to the right is the x-axis direction, and the vertically downward direction is the y-axis direction; Select an area in the central area of the pixel array, the size of which is m ₀ ×n ₀ , m ₀ , n ₀ ∈ positive integer, called the comparison window, and the comparison window is respectively separated from the pixel array in the horizontal direction and vertical direction Each h and v pixels of edge pixels in the direction have: m ₀ +2h=M, n ₀ +2v=N, h, v∈ positive integer; Step 2. For the pixel array of the above reference frame, export the edge direction data along the X-axis row by row: If the light intensity value of a pixel is smaller than the light intensity value of the second pixel behind it by an error tolerance value error, that is, if I(X, Y)<I(X+2, Y)-error Then it is defined that there is a positive edge between these two pixels; If the light intensity value of a pixel is greater than the light intensity value of the second pixel behind it by an error tolerance value error, that is, if I(X, Y) > I(X+2, Y)+error Then it is defined that there is a negative edge between these two pixels; If the light intensity value of a pixel is close to the corresponding light intensity value of the second pixel behind it, the difference does not exceed an error tolerance value error, that is, if I(X+2, Y)-error<I(X, Y )<I(X+2,Y)+error It is considered that there is no "edge" between these two pixels, or it is called the third type of edge; The error tolerance value in the above formula can be preset to a small value according to the specific lighting conditions, for example: error=10; the edge obtained in this way is located at the position of the first pixel after this pixel, that is, it is located at the position participating in the comparison On the pixel (x, y) in the middle of the two pixels; along the X-axis direction, all positive edges, negative edges, and third-type edges form the edge direction data of this direction, respectively in 3-bit binary value 001 , 010 and 100 represent, recorded as reference (x, y), save it; Step 3, calculating the self-correlation matching coefficient of the pixel array of the comparison window in the reference frame: $\mathrm{<span\; class="notranslate">\; auto</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; correlation</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$ In the formula, x and y are the coordinates of the pixels in the comparison window respectively, and the operation symbol · represents the binary logical AND operation, and the operation result is either logic 0 or logic 1, and the square brackets "[]" represent the value corresponding to the logic function among them. The value, either the value 0 or the value 1, is calculated pixel by pixel, and the accumulated result is used as the auto_correlation coefficient auto_correlation(a, b) for this time, using a 3×3 correlation matching operator: a=-1, 0, 1, b=-1, 0, 1, therefore, a total of 9 autocorrelation coefficients are generated; Step 4. Analyze the fineness of the structural features on the surface of the observed object, that is, analyze the distinguishability of the light and dark contrast between the nearest adjacent pixels, and check: auto_correlation(a,b)≥auto_correlation(0,0)×similarity In the formula, similarity describes the similarity between the comparison window and its neighboring pixel arrays of the same scale, for example, similarity=60%, which can be set in advance, or can be debugged and selected according to the lighting conditions and the texture of the surface of the measured object; If there are more than 3 self-association matching coefficients satisfying the above checking inequality, it is necessary to gradually expand the scope of the comparison window for each step row and step column: let m=m ₀ +step, n=n ₀ +step, and recalculate the new comparison window The self-correlation matching coefficient of the adjacent pixel is analyzed again until the self-correlation coefficient satisfying the above analysis inequality is not more than 3, and the result is 2h=Mm, 2v=Nn, where step is a stepping parameter, the initial value is 1, the scale of the comparison window needs to be expanded by 1 each time; if it exceeds a predetermined range in the frame and no suitable comparison window has been found, it is considered that this part of the reflective surface of the object is not suitable for the measurement work of the device, and given Prompt warning; If there are no more than 3 self-correlation matching coefficients satisfying the above checking inequality, it means that the surface structure features of the object to be photographed are fine enough, and the values between the nearest neighbor pixels can be distinguished, and further attempts can be made to narrow the range of the comparison window by each step Rows and step columns to reduce the amount of calculation: let m=m ₀ -step, n=n ₀ -step, recalculate the self-correlation matching coefficient of the comparison window, and conduct the adjacent pixel discrimination analysis again until the selected comparison window area The number of self-correlation matching coefficients satisfying the above check inequality is equal to 3, and it is considered that the optimal comparison window pixel array that can be used for matching comparison under the current object surface condition and lighting condition is found: m=m ₀ -step , n=n ₀ -step, 2h=Mm, 2v=Nn; Step 5. After the above shooting, after a period of time Δt, take a second frame of bitmap as a sampling frame; Determine the side direction data of the pixels in the sampling frame along the X-axis direction line by line, and use 3-bit binary values 001, 010, and 100 to represent the positive side, negative side, and the third type of side, which are recorded as comparison (x, y), and saved Of; Step 6. Carry out cross-correlation matching with the pixel array {reference(x, y)} in the comparison window in the reference frame in the range of the sampling frame {comparison(x, y)} according to the 9×9 correlation matching array, The specific algorithm is: $\mathrm{<span\; class="notranslate">\; the\; cross</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; correlation</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; v</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; m</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; [</span>\mathrm{<span\; class="notranslate">\; reference</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; comparison</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; a</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$ Corresponding to a=-4, -3, -2, -1, 0, +1, +2, +3, +4 and b=-4, -3, -2, -1, 0, +1, +2 , +3, +4 combination, a total of 81 cross-correlation matching coefficients cross_correlation (a, b) are generated, that is, there may be 81 kinds of movement situations in the comparison window of the reference frame; Step 7. The cross-correlation coefficient with the highest degree of frame-frame correlation is the largest: cross_correlation(a, b) → auto_correlation(0, 0) Therefore, the direction in which the sampled frame moves relative to the reference frame and the magnitude of the movement are obtained: Δx(i,j)=a, Δy(i,j)=b, This is the relative displacement vector of the object in this sampling period, where i represents the sequence count of the measurement and shooting, and j represents the sequence count of the reference frame taken; during the measurement process, the total relative displacement vector of the object is: Δx ₀ (i,j)=Δx ₀ (i-1,j)+Δx(i,j), Δy ₀ (i,j)=Δy ₀ (i-1,j)+Δy(i,j), The right side of the equation (Δx ₀ (i-1, j), Δy ₀ (i-1, j)) is the relative displacement vector accumulated by the object before; Step 8. Velocity vector of object displacement: Δv _x (i, j) = Δx (i, j)/Δt, Δv _y (i, j) = Δy (i, j)/Δt; Step 9. If |Δx ₀ (i,j)-Δx ₀ (k,j-1)|≥2m/5, or |Δy ₀ (i,j)-Δy ₀ (k,j-1)|≥2n /5, wherein, k=max(i)|(j-1) represents the sequence count value of the last shot when the sequence count of the reference frame is j-1, that is, under the condition that the reference frame does not change , where the accumulation of the relative displacement of the comparison window has exceeded 2/5 of the magnitude of the comparison window, at this time, replace the reference frame with the latest sampling frame, and its comparison window is repositioned at the center of the new reference frame ; If |Δx ₀ (i,j)-Δx ₀ (k,j-1)|<2m/5 and |Δy ₀ (i,j)-Δy ₀ (k,j-1)|<2n/5, do not Update the reference frame, but relatively shift the comparison window in the reference frame by Δx=a, Δy=b; Step 10. If the reference frame is updated, calculate the self-correlation matching coefficient of the new reference frame as in step 3, check the surface structure features as in step 4, and re-determine the size of the optimal comparison window m×n; If the reference frame is not updated, adjust the size of the cross-correlation matching operator array in step 6: if |a|<5 and |b|<5, change it to 7×7: a=-3, -2, - 1, 0, +1, +2, +3, b=-3, -2, -1, 0, +1, +2, +3, if |a|<3 and |b|<3, take Be 5*5: a=-2,-1,0,+1,+2, b=-2,-1,0,+1,+2, otherwise still take as 9*9 associative matching operator array; Step 11. After the above-mentioned shooting, after a period of time Δt, take a third frame of bitmap as a new sampling frame; Determine the side direction data of the pixels in the sampling frame along the X-axis direction line by line, where the positive side, negative side and the third type of side are respectively represented by 3-bit binary values 001, 010 and 100, which are recorded as comparison(x, y) , save it; Step 12, according to the cross-correlation matching operator array determined in step 10, perform cross-correlation matching calculation on the pixel array in the comparison window in the reference frame in the range of the sampling frame, the specific algorithm is the same as step 6; Step 13, skip to step 7 and continue measuring.

Images