CN103337067B - The visible detection method of single needle scan-type screw measurement instrument probe X-axis rotating deviation - Google Patents

Abstract

The invention provides a visual detection method for the X-axis rotation deviation of a probe of a single-needle scanning thread measuring instrument, and relates to the technical field of image processing. The following process is included: Step 1, initialization calibration; Step 2, real-time detection of the sharp point of the probe; Step 3, calculation of the rotation deviation angle. The invention can measure and adjust the X-axis rotation positioning deviation of the probe through image processing means, realize automatic measurement, and improve efficiency under the premise of ensuring installation accuracy.

Description

Visual inspection method for X-axis rotation deviation of probe of single-needle scanning thread measuring instrument

technical field

The invention provides a visual detection method for the X-axis rotation deviation of a probe of a single-needle scanning thread measuring instrument, and relates to the technical field of image processing.

Background technique

Threaded connectors are mechanical parts widely used in the machinery industry. Its manufacturing accuracy directly affects the reliability, assembly accuracy and interchangeability of machine parts. Especially in the design and manufacture of aerospace vehicles, a large number of threaded connections are used. More than 50% of the connection between parts is realized by thread fit, and the quality of the connection determines the reliability and life of industrial products such as aircraft.

For ordinary threads, standard ring gauges (plug gauges) are often used to judge the screw-in. However, for threads with high precision requirements, such as the measurement of ring gauges (plug gauges), more accurate instruments are needed for detection. At present, there are not many mature products at home and abroad in terms of high-precision thread measuring instruments and equipment. Abroad, there is the MSXP thread measuring instrument produced by the Dutch IAC company, which adopts the stylus scanning contour method. There is no special manufacturer producing contact thread measuring instruments in China. Haliang Group has a stylus-type surface profile instrument, but it is not specifically for thread detection, and the calculation and error compensation of thread comprehensive parameters are not yet perfect. Harbin Institute of Technology, Changchun University of Science and Technology, Zhejiang University, etc. use laser scanning technology to obtain thread profile data for measurement, but from the currently published data, the measurement accuracy is not as good as contact scanning. Tianjin University, Nanjing University of Aeronautics and Astronautics, etc. have done research on thread parameter detection based on image vision. The method has greatly improved the detection efficiency, but the image vision measurement accuracy is not as good as stylus scanning, and it is difficult to solve the problem of accurate internal thread detection. Measurement.

In the stylus scanning thread measurement, in order to overcome the measurement errors caused by probe wear and other environmental changes, the relative measurement method is usually used. That is, the standard thread is used for calibration measurement first, and then the threaded workpiece to be measured is measured. And this requires the instrument to have extremely high positioning accuracy, otherwise, when the size difference between the standard thread and the thread to be measured is large, the error of the measurement result will be magnified.

Using the homogeneous transformation method of spatial coordinates, it can be judged that the positioning errors that have the greatest impact on the measurement results of pitch diameter parameters mainly include the deviation in the Y-axis direction and the rotation deviation between the upper and lower needle points of the probe and the X-axis (Fig. 1 in the coordinate system shown). However, the existence of actual processing errors and assembly errors makes it difficult to control these two deviations within the allowable range. The axial deviation of the Y-axis can be calculated through compensation, while the rotational deviation of the X-axis needs to be controlled through an effective detection method. At present, no relevant information has been found to show that the deviation can be automatically detected and controlled.

Contents of the invention

The object of the present invention is to provide a visual detection method for X-axis rotation deviation of a probe probe of a single-needle scanning thread measuring instrument capable of automatic detection and control.

Step 1. Initialize calibration:

Step 1-1, collect the RGB image of the first frame including the probe motion area, and convert the RGB image into a grayscale image according to the following formula;

Gray＝(R*19595+G*38469+B*7472)>>16

Among them, R, G, and B respectively represent the corresponding red, green, and blue color components in the RGB image, and Gray represents the gray value;

Step 1-2, using the SUSAN corner detection algorithm to detect the corners of the gray image to obtain all corner information including the sharp points of the upper and lower probes;

Step 1-3, manual operation, select the two corner points corresponding to the sharp points of the upper and lower probes;

Step 2. Real-time detection of the sharp point of the probe:

Step 2-1, collect the RGB image of the i+1th frame including the probe motion area, and convert it into a grayscale image, where i≥1;

Step 2-2. When i=1, the position of the sharp point of the probe in the previous frame image is the predicted sharp point, that is, P′ _i+1 =P _i ; when i≥2, according to the i-1th frame and The position of the sharp point of the probe in the grayscale image of the i frame, the position of the sharp point of the probe in the image of the i+1 frame is predicted by the following prediction formula;

P' _i+1 =P _i +(P _i -P _i-1 )=2P _i -P _i-1 ;

Wherein, P is the coordinate vector of the tip of the probe, P' is the predicted coordinate vector of the tip of the probe, and i is the frame number of the image;

Step 2-3, based on the predicted coordinate vector P' of the probe cusp, constructing a local region of interest on the grayscale image of the i+1th frame;

Step 2-4, by performing SUSAN corner point detection on the local region of interest, filter out the sharp point positions of the upper and lower probes;

Step 2-5, performing distance verification on the filtered upper and lower probe sharp points;

The distance check condition 1 is

|d(P _up ，P _down )-d ₀ (P _up ，P _down )|＜ε ₁

Among them, d(P _up , P _down ) represents the distance between the tip points of the upper and lower probes in k+1 frame images; d ₀ (P _up , P _down ) represents the upper and lower probe points in k frames of images The distance between sharp points; ε ₁ is the error threshold;

Step 2-6, judging the verification result of step 2-5, if condition 1 is met, then go to step 2-10, otherwise go to step 2-7;

Steps 2-7: Use the Canny operator to detect the edge information of the image in the local region of interest, use the Hough transform to extract the straight line contour, and further obtain the corresponding conic generatrices of the upper and lower cusps according to the following rules:

a) In the extracted straight line, the included angle between the two probe conical generatrices and the X-axis in the image is between 60° and 80°;

b) The angle between two conical generatrices satisfies θ ₁ , θ ₂ are two

The angle between the conical generatrix of the strip probe and the X-axis of the image;

c) The intersection point Q of the conical generatrices is the highest point of the intersection points of all successfully paired straight lines, and it is within the image;

Step 2-8, calculate the intersection point of the upper and lower contours according to the conical generatrices corresponding to the obtained upper and lower cusps, and perform distance verification;

The distance check condition 2 is

|d(Q _up ，Q _down )-d ₀ (Q _up ，Q _down )|＜ε ₂

Among them, d(Q _up , Q _down ) represents the distance between the upper and lower probe tip in the current frame image; d ₀ (Q _up , Q _down ) represents the upper and lower probe tip in the previous frame image The distance between points; ε ₂ is the error threshold;

Step 2-9, judging the verification result of step 2-8, if condition ₂ is satisfied, according to the distance between the ideal probe point _Q0 and P0 in the previous frame image and the ideal probe point Q in the current frame image, Calculate the actual tip point P of the probe in the current frame according to the formula.

P=Q+P ₀ -Q ₀ , turn to step 2-10;

If condition 2 is not satisfied, it is considered that the current frame image detection fails, and the measurement data points are not stored in the database, and go to step 2-1;

Step 2-10 records the positions of the sharp points of the upper and lower probes and the slope k of the connecting line;

Step 3. Calculate the rotation deviation angle

Step 3-1. Fit the positions of the upper and lower sharp points of the probe respectively, and obtain the motion straight line of the probe. The specific process is as follows;

In the video sequence captured by the probe’s up and down motion, if the number of effective probe tip images detected is N, the linear regression equation

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; N</span>}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; N</span>}{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

Find the straight line of the probe motion; where, k _M is the slope of the straight line of motion; x _i , y _i are the abscissa and ordinate of the tip point of the probe in the i-th frame image respectively;

Step 3-2, according to the positions of the sharp points of the upper and lower probes in each frame of image, find the straight line direction of the probe in each frame of image; that is

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; no</span>}}}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; no</span>}}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2...</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

Among them, k _i is the slope of the line connecting the upper and lower sharp points of the probe in the i-th frame image; (x _up , y _up ) is the coordinates of the upper probe sharp point, (x _down , y _down ) is the lower probe tip point coordinates;

The average slope k of the line connecting the upper and lower sharp points of the probe is

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\mathrm{<span\; class="notranslate">\; N</span>}}$

Utilize the average slope k _L of the line connecting the upper and lower sharp points of the probe to obtain a fitting straight line of the upper and lower sharp points;

Step 3-3. Calculate the angle θ between the straight line of the probe motion and the fitting line of the upper and lower sharp points

$\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; the\; tan</span>}<span\; class="notranslate">\; |</span>\frac{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}}<span\; class="notranslate">\; |</span>$

θ is the rotation deviation angle.

The invention can measure and adjust the X-axis rotation positioning deviation of the probe through image processing means, realize automatic measurement, and improve efficiency under the premise of ensuring installation accuracy.

Description of drawings

Figure 1 is the definition of the workpiece coordinate system;

Figure 2 is the relative positional relationship between the camera and the probe;

Figure 3 is the corner detection result of the entire image;

Figure 4 is the corner detection result of the local region of interest;

Figure 5 shows the corner detection results of the three corner detection algorithms for the region of interest near the tip of the probe when there is background shadow interference. Among them, (a) is the result of Harris corner detection algorithm; (b) is the result of SUSAN corner detection algorithm; (c) is the result of FAST corner detection algorithm;

FIG. 6 is a schematic diagram of the positional relationship between the ideal probe tip and the actual probe tip. Where Q represents the ideal probe tip, and P is the actual probe tip;

Fig. 7 is the edge detection results of the probe partial image by three edge detection algorithms. Among them, (a) is the detection result of Sobel operator, (b) is the detection result of Prewitt operator, (c) is the detection result of Canny operator;

Fig. 8 is a schematic diagram of a non-maximum suppression method;

Figure 9 is a Hough line detection effect diagram;

Figure 10 is the detection result of the probe conical busbar;

Fig. 11 is a schematic diagram when the tip of the probe overflows the region of interest;

Figure 12 is a video sequence screenshot of needle point tracking;

Label names in the figure: 1. Imaginary cylinder of the workpiece, 2. Probe, 3. Mounting panel, 4. Camera

detailed description

In order to ensure the accuracy of the sharp point recognition of the probe, the present invention adopts three aspects of corner detection, contour recognition and motion prediction for comprehensive recognition, which greatly improves the accuracy of the sharp point of the probe and effectively suppresses the background interference caused by false detection of the tip of the probe. Then, calculate the included angle between the motion line (Z-axis of the measuring instrument) and the line connecting the upper and lower sharp points by fitting the straight line to the position of the sharp point.

As shown in Figure 2, the camera is placed in the positive direction of the X-axis of the probe, parallel to the direction of the Z-axis of the probe. The up and down movement of the probe is collected by the camera, and the connection and movement of the upper and lower sharp points of the probe are automatically captured through image recognition, and the rotation deviation angle is calculated. Then adjust the probe rotation through the rear device to control the rotation deviation. Since the image is two-dimensional data, in the following formula derivation of image processing, the y-axis is used instead of the z-axis for expression. The software program steps are as follows:

Step 1. Initialize calibration:

Step 1-1, collect an RGB image including the probe motion area, and convert the RGB image into a grayscale image according to the following formula;

Gray＝(R*19595+G*38469+B*7472)>>16

Among them, R, G, and B respectively represent the corresponding red, green, and blue color components in the RGB image, and Gray represents the gray value;

Step 1-2, using the SUSAN corner detection algorithm to detect the corners of the gray image to obtain all corner information including the sharp points of the upper and lower probes;

There is no clear mathematical definition of the corner point, but it is generally believed that the corner point is the point where the brightness of the two-dimensional image changes sharply or the point where the curvature is the maximum value on the edge curve of the image. The detection of the sharp point of the probe can be borrowed from the image corner detection method, and there are three main types of image corner detection methods: ①The corner detection method based on the boundary curve. This type of method is mainly to find out the local maximum point of curvature on the edge curve of the image. The performance of the edge detection algorithm will directly affect the quality of the detected corners, for example, when the edge line breaks in the edge detection, it will lead to the detection of false corners. ②Template-based corner detection method. Firstly, a series of corner templates with different angles are established, and then the similarity between the image to be tested and the standard template is compared within a certain window to detect the corners in the image. Due to the complexity of the corner structure, it is impossible to design a template covering all directions and corners. This kind of corner detection method is computationally intensive and complicated. ③Corner detection methods based on image grayscale mainly perform corner detection by calculating differential geometric features of pixels, such as Harris algorithm, SUSAN algorithm, FAST corner detection algorithm, etc.

Step 1-3, manual operation, select the two corner points corresponding to the sharp points of the upper and lower probes;

Step 2. Real-time detection of the sharp point of the probe:

Step 2-1, collect the RGB image of the i+1th frame including the probe motion area, and convert it into a grayscale image, where i≥1;

Step 2-2. When constructing a local region of interest, only the position of the tip of the probe in the previous image frame is taken as the center, and a rectangular region is constructed as the region of interest. In the constructed local image, due to the inconsistency of the time required for different image detection, the position of the probe will fluctuate. When the probe moves faster, the tip of the probe may overflow the area of interest. As shown in Figure 11.

However, if the setting of the region of interest is increased, the calculation amount will increase, which is not conducive to the real-time processing of the image. Therefore, it is necessary to make a simple prediction of the position of the tip of the probe in the next frame of image based on the linear motion of the probe and the information of the tip of the probe that has been detected, so as to basically ensure that the tip of the probe is always in the sense Center of the region of interest.

Although the moving speed of the probe is variable, it can be processed at a uniform speed in a very short continuous image time. Then the cusp position of the next frame can be calculated by the following formula

When i=1, the position of the sharp point of the probe in the previous frame image is the predicted sharp point, that is, P′ _i+1 =P _i ; when i≥2, according to the grayscale of the i-1th frame and the i frame The position of the tip of the probe in the image, the position of the tip of the probe in the i+1 frame image is predicted by the following prediction formula

P' _i+1 =P _i (P _i -P _i-1 =2P _i -P _i-1 ;

Wherein, P is the coordinate vector of the tip of the probe, P' is the predicted coordinate vector of the tip of the probe, and i is the frame number of the image;

Step 2-3, based on the predicted coordinate vector P' of the probe cusp, constructing a local region of interest on the grayscale image of the i+1th frame;

Step 2-4, by performing SUSAN corner point detection on the local region of interest, filter out the sharp point positions of the upper and lower probes;

Corner detection is performed on the entire image, as shown in Figure 3, because the background is not a single color, there will be many detection corners, which brings difficulties to the search and screening of the sharp point P of the probe. Considering that there are only two probe sharp points in the image, a local region of interest is constructed, as shown in Figure 4.

Figure 4 can effectively reduce the interference of the background color by artificially specifying the correct sharp point P of the probe and performing corner detection in its adjacent area. However, during the up-and-down movement of the probe, due to the significant changes in the reflective properties of the material and the background, the conventional corner detection algorithm often cannot find the sharp point P of the probe.

The three images in Figure 5 are the processing results of the probe image at a certain moment by three corner detection algorithms, Harris, SUSAN and FAST.

It can be seen that none of the three algorithms detects the real probe cusp P. Analysis of the image reveals that there is a sudden change in the color of the background object, which coincides with the sharp point P of the probe. Therefore, for the detection of the sharp point of the probe, the corner detection method based on the gray scale of the image cannot obtain correct results. For this, we consider contour information again.

Step 2-5, performing distance verification on the filtered upper and lower probe sharp points;

The distance check condition 1 is

|d(P _up ，P _down )-d ₀ (P _up ，P _down )|＜ε ₁

Among them, d(P _up , P _down ) represents the distance between the tip points of the upper and lower probes in the i+1 frame image; d ₀ (P _up , P _down ) represents the upper and lower probe points in the i frame image The distance between sharp points; ε ₁ is the error threshold;

Step 2-6, judging the verification result of step 2-5, if condition 1 is met, then go to step 2-10, otherwise go to step 2-7;

Steps 2-7: Use the Canny operator to detect the edge information of the image in the local region of interest, use the Hough transform to extract the straight line contour, and further obtain the corresponding conic generatrices of the upper and lower cusps according to the following rules:

a) In the extracted straight line, the included angle between the two probe conical generatrices and the X-axis in the image is between 60° and 80°;

b) The angle between two conical generatrices satisfies θ ₁ , θ ₂ are two

The angle between the conical generatrix of the strip probe and the X-axis of the image;

c) The intersection point Q of the conical generatrices is the highest point of the intersection points of all successfully paired straight lines, and it is within the image;

The ideal model of the probe head is a conical structure (the actual tip of the head is a sphere with a radius of 50um), and the shooting direction of the visual image is parallel to its centerline. If two conical generatrixes can be detected and their intersection point can be calculated, it can be regarded as the ideal probe point Q. During the movement of the probe, the distance between the ideal sharp point Q and the actual sharp point P remains constant. Figure 6 shows the difference between P and Q points.

In order to extract the contour of the probe, it is first necessary to perform edge detection on the image. The methods of gray image edge detection are mainly divided into two categories: the first-order differential image edge detection operator and the second-order differential image edge detection operator. The first-order differential edge detection operators include: Roberts operator, Sobel operator, Krisch operator, Prewitt operator, etc. The second-order differential edge detection operators mainly include: Laplacian operator, LOG operator; There are detection methods such as Canny, SUSAN, and statistical discrimination.

Figure 7 shows the results of several methods for detecting the edge of the local image of the probe.

From the effect point of view, the effects of the three operators are all good, but the edges detected by the Sobel operator and the Prewitt operator will have a multi-pixel width, and there will be a certain error in finding the intersection point of the conic generatrix with the subsequent contour detection. The edge of the Canny operator is clearer, which can precisely locate the outline of the probe. The basic idea of Canny operator is: first select a certain Gauss filter to smooth the image, and then use non-extreme value suppression technology to process to obtain the final edge image. The steps are;

a) Smooth the image with a Gauss filter.

Here a Gaussian function H(x,y) with omitted coefficients is used:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; exp</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span>$

G(x,y)=f(x,y)*H(x,y)

where f(x,y) is the image data.

b) Calculate the magnitude and direction of the gradient using finite differences of the first order partial derivatives.

Use the first-order difference convolution template:

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; =</span>\left|\begin{array}{cc}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right|$

${\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; =</span>\left|\begin{array}{cc}\mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}& <span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right|$

can get:

Amplitude:

direction:

c) Perform non-maximum suppression on the gradient magnitude.

Just getting the global gradient is not enough to identify edges. In order to determine the edge, the point with the largest local gradient must be retained, and the non-maximum value must be suppressed, that is, the non-local maximum value point is set to zero to obtain a refined edge.

As shown in Fig. 8, the 4 sectors are numbered from 0 to 3, corresponding to 4 possible combinations of 3×3 neighborhoods. At each point, the central pixel M of the neighborhood is compared to the two pixels along the gradient line. If the gradient value of M is not greater than the gradient values of two adjacent pixels along the gradient line, let M=0.

d) Detect and connect edges with a dual-threshold algorithm.

Using two thresholds T ₁ and T ₂ (T ₁ <T ₂ ), two threshold edge images N ₁ [i,j] and N ₂ [i,j] can be obtained. Since N ₂ [i,j] is obtained with a high threshold, it contains few false edges, but there are discontinuities (not closed). The double-threshold method needs to connect the edges into a contour in N ₂ [i, j]. When the end point of the contour is reached, the algorithm searches for edges that can be connected to the contour at the 8 adjacent points of N ₁ [i, j]. , so that the algorithm keeps collecting edges in N ₁ [i, j] until N ₂ [i, j] is connected. T ₂ is used to find each line segment, and T ₁ is used to extend in both directions of these line segments to find edge breaks and connect these edges.

After using the Canny operator to get the edge of the image, we need to extract the probe outline we care about, that is, the two conic generatrices. Generally, the method of linear contour extraction is mainly based on Hough transform. Straight line Hough transform uses the idea of "voting" to detect straight lines or line segments in digital images. It is a classic algorithm for image processing and line detection. Any straight line in the plane can be determined by two parameters ρ and θ. Among them, ρ determines the distance from the line to the origin, and θ determines the orientation of the line. Its functional relationship is

ρ=xcosθ+ysinθx∈[0,π]

Each point ( _xi , y ₎ in the image space is mapped to a set of accumulators C(ρ, θ) in the Hough space, which is the so-called voting process. ) of pixels. After the voting is over, each local maximum of C(ρ, θ) corresponds to a straight line segment, that is, the corresponding ρ and θ can uniquely determine this straight line. Figure 9 is the result after Hough transformation.

Analyze the characteristics of the local image: a) In the extracted straight line, the angle between the two probe conical generatrices and the X-axis in the image is basically the same, and the size is between 60° and 80°, which means that θ ₁ , θ ₂ ∈ [π/3,4π/9]; b) The angle between the two conical bus bars is about 45° according to the factory report, that is, the slope is approximately 1. Set the detection conditions according to the angle formula:

$\mathrm{<span\; class="notranslate">\; 0.9</span>}<span\; class="notranslate">\; <</span><span\; class="notranslate">\; |</span>\frac{{\mathrm{<span\; class="notranslate">\; tan\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; tan\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; tan\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; tan\&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 1.5</span>}$

It can be filtered to obtain the two conical generatrixes sought. Then extend the extracted conical generatrix to find the ideal probe point Q, as shown in Figure 10. Of course, a verification condition needs to be added c) The intersection point Q of the conical generatrices is the highest point of the intersection points of all paired successful straight lines, and it is within the image.

So far, the ideal probe tip Q has been detected. Ideally, the line connecting the upper and lower needle points of the probe can be represented by L(P _up , P _down ) or L(Q _up , Q _down ), and the two straight lines are coincident. However, due to actual processing errors, the two straight lines cannot be completely coincident. In the measurement process, the contact point is the actual sharp point, which makes the reliability of the P point higher than that of the Q point. Therefore, in the cusp detection, the strategy of L(P _up , P _down ) as the main and L(Q _up , Q _down ) as the auxiliary is also adopted.

Step 2-8, calculate the intersection point of the upper and lower contours according to the conical generatrices corresponding to the obtained upper and lower cusps, and perform distance verification;

The distance check condition 2 is

|d(Q _up ，Q _down )-d ₀ (Q _up ，Q _down )|＜ε ₂

Among them, d(Q _up , Q _down ) represents the distance between the upper and lower probe tip in the current frame image; d ₀ (Q _up , Q _down ) represents the upper and lower probe tip in the previous frame image The distance between points; ε ₂ is the error threshold;

Step 2-9, judging the verification result of step 2-8, if condition ₂ is satisfied, according to the distance between the ideal probe point _Q0 and P0 in the previous frame image and the ideal probe point Q in the current frame image, Calculate the actual tip point P of the probe in the current frame according to the formula.

P=Q+P ₀ -Q ₀ , turn to step 2-10;

Go to steps 2-10;

If condition 2 is not satisfied, it is considered that the current frame image detection fails, and the measurement data points are not stored in the database, and go to step 2-1;

Step 2-10 records the positions of the sharp points of the upper and lower probes and the slope k of the connecting line;

Step 3. Calculate the rotation deviation angle

Step 3-1. Fit the positions of the upper and lower sharp points of the probe respectively, and obtain the motion straight line of the probe. The specific process is as follows;

In the video sequence captured by the probe's up and down motion, if the number of effective probe points detected is N, use the linear regression equation

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; )</span>}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; N</span>}\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; N</span>}{\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

Find the straight line of the probe motion; where, k _M is the slope of the straight line of motion; x _i , y _i are the abscissa and ordinate of the tip point of the probe in the i-th frame image respectively;

Step 3-2, according to the positions of the sharp points of the upper and lower probes in each frame of image, the direction of the straight line where the probe is located in each frame of image is obtained; that is

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; no</span>}}}{{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; u</span>}\mathrm{<span\; class="notranslate">\; p</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}\mathrm{<span\; class="notranslate">\; o</span>}\mathrm{<span\; class="notranslate">\; w</span>}\mathrm{<span\; class="notranslate">\; no</span>}}}<span\; class="notranslate">\; ,</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2...</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>$

Among them, k _i is the slope of the line connecting the upper and lower sharp points of the probe in the i-th frame image; (x _up , y _up ) is the coordinates of the upper probe sharp point, (x _down , y _down ) is the lower probe tip point coordinates;

The average slope k of the line connecting the upper and lower sharp points of the probe is

${\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}<span\; class="notranslate">\; =</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}}{\mathrm{<span\; class="notranslate">\; N</span>}}$

Utilize the average slope k _L of the line connecting the upper and lower sharp points of the probe to obtain a fitting straight line of the upper and lower sharp points;

Step 3-3. Calculate the angle θ between the straight line of the probe motion and the fitting line of the upper and lower sharp points

$\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; t</span>}\mathrm{<span\; class="notranslate">\; a</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; |</span>\frac{{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}}{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; L</span>}}{\mathrm{<span\; class="notranslate">\; k</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}}<span\; class="notranslate">\; |</span>$

θ is the rotation deviation angle.

Claims (1)

1., to a visible detection method for the X-axis rotating deviation of single needle scan-type screw measurement instrument measuring probe, it is characterized in that, comprise following process:

Step 1, initialization are demarcated:

Step 1-1, collection the 1st frame comprise the RGB image in probe motion region, according to following formula, RGB image are converted into gray level image;

Gray＝(R*19595+G*38469+B*7472)>>16

Wherein, R, G, B represent red, green, blue three color components of the correspondence in RGB image respectively, and Gray represents gray-scale value;

Step 1-2, utilize SUSAN Corner Detection Algorithm, Corner Detection is carried out to gray level image, obtains all angle point informations comprising upper and lower probe cusp;

Step 1-3, manual operation, choose two angle points that upper and lower probe cusp is corresponding;

Step 2, in real time detector probe cusp:

Step 2-1, collection the i-th+1 frame comprise the RGB image in probe motion region, are translated into gray level image, wherein i >=1;

Step 2-2, as i=1, with the position of previous frame image middle probe cusp for prediction cusp, i.e. P ' _i+1=P _i; When i>=2, according to the position of the gray level image middle probe cusp of the i-th-1 frame and i frame, by the probe position of cusp in following predictor formula prediction i+1 two field picture;

P′ _i+1＝P _i+(P _i-P _i-1)＝2P _i-P _i-1；

Wherein, P is the coordinate vector of probe cusp, the prediction coordinate vector that P ' is probe cusp, and i is the frame number of image;

Step 2-3, based on probe pinpoint point prediction coordinate vector P ', the i-th+1 frame gray level image builds local region of interest;

Step 2-4, by carrying out SUSAN Corner Detection to local region of interest, filter out upper and lower probe position of cusp;

Step 2-5, the upper and lower probe cusp filtered out carried out to distance verification;

Distance verification condition 1 is

|d(P _up，P _down)-d ₀(P _up，P _down)|＜ε ₁

Wherein, d (P _up, P _down) represent in i+1 two field picture between upper and lower probe cusp distance; d ₀(P _up, P _down) represent in i two field picture between upper and lower probe cusp distance; ε ₁for error threshold;

Step 2-6, determining step 2-5 check results, as satisfied condition 1, then proceed to step 2-10, otherwise go to step 2-7;

Step 2-7, detect image edge information in local region of interest with Canny operator, utilize Hough transform to extract outline of straight line, obtain upper cusp and element of cone corresponding to lower cusp respectively according to following rule further:

A) in extracted straight line, in two probe element of cones and image, X-axis corner dimension is between 60 ° ~ 80 °;

B) angle between two element of cones meets θ ₁, θ ₂two probe element of cones and image X-axis angle respectively;

C) element of cone intersection point Q is the peak of all successful matching straight-line intersections, and in image;

Step 2-8, according to cusp on gained and element of cone corresponding to lower cusp, calculate upper and lower profile intersection point, row distance of going forward side by side verifies;

Distance verification condition 2 is

|d(Q _up，Q _down)-d ₀(Q _up，Q _down)|＜ε ₂

Wherein, d (Q _up, Q _down) represent in current frame image between upper and lower probe cusp distance; d ₀(Q _up, Q _down) represent in previous frame image between upper and lower probe cusp distance; ε ₂for error threshold;

Step 2-9, determining step 2-8 check results, as satisfied condition 2, then according to previous frame image middle ideal probe cusp Q ₀with P ₀distance and current frame image middle ideal probe cusp Q, be calculated as follows actual probes cusp P in present frame;

P=Q+P ₀-Q ₀, proceed to step 2-10;

If condition 2 does not also meet, then think that current frame image detects unsuccessfully, measurement data points stored in database, does not go to step 2-1;

Step 2-10 records position and the line slope k thereof of upper and lower probe cusp;

Step 3, calculating deviation angle

Step 3-1, the respectively upper and lower position of cusp of sniffing probe, obtain probe motion straight line, detailed process is as follows:

In the video sequence of probe once upper and lower capturing movement, if the effective probe pinpoint dot image number detected is N, utilize equation of linear regression

$k_M=\frac{\underset{i=1}{\overset{N}{\&Sigma;}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{N}{\&Sigma;}}{(x_i-\stackrel{\&OverBar;}{x})}^2}=\frac{\underset{i=1}{\overset{N}{\&Sigma;}}{x}_i{y}_i-N\stackrel{\&OverBar;}{xy}}{\underset{i=1}{\overset{N}{\&Sigma;}}{x}_i^2-N{\stackrel{\&OverBar;}{x}}^2}$

Obtain probe motion straight line; Wherein, k _mfor the slope of line of motion; x _i, y _ibe respectively horizontal stroke, the ordinate of the i-th two field picture middle probe cusp; $\stackrel{\&OverBar;}{x}=\frac{\underset{i=1}{\overset{N}{\&Sigma;}}{x}_i}{N},$ $\stackrel{\&OverBar;}{y}=\frac{\underset{i=1}{\overset{N}{\&Sigma;}}{y}_i}{N},$

Step 3-2, according to the upper and lower probe position of cusp in every two field picture, obtain each two field picture middle probe place rectilinear direction; Namely

$k_i=\frac{y_{up}-y_{down}}{x_{up}-x_{down}},i=1,2...,N$

Wherein, k _ibe in the i-th two field picture, the line slope of the upper and lower cusp of probe; (x _up, y _up) be upper probe pinpoint point coordinate, (x _down, y _down) be lower probe pinpoint point coordinate;

The average of probe upper and lower cusp line slope k is

$k_L=\frac{\underset{i=1}{\overset{N}{\&Sigma;}}{k}_i}{N}$

Utilize probe upper and lower cusp line average gradient k _lobtain the fitting a straight line of a upper and lower cusp;

The angle theta of step 3-3, calculating probe motion straight line and upper and lower cusp fitting a straight line

$\mathrm{\&theta;}=atan\left|\frac{k_L-k_M}{1+k_L{k}_M}\right|$

θ is deviation angle.