CN105043259A - Numerical control machine tool rotating shaft error detection method based on binocular vision - Google Patents

Abstract

The invention relates to a binocular vision-based detection method for the rotation axis error of a CNC machine tool, which belongs to the technical field of machine tool precision detection, and relates to a method for detecting and identifying geometric errors of a dual-rotation axis of a five-axis CNC machine tool. The present invention adopts a high-resolution binocular vision system to collect the position information of the marking points pasted on the surface of the machine tool turntable; and then realizes the two position errors and The detection and collection of two angle errors complete the rapid measurement of geometric parameters. This method uses circular marking points, which not only has simple image processing procedures and high feature extraction accuracy, but also has good robustness, fast and convenient measurement. At the same time, this method solves the problem of difficult detection and identification of the installation error of the rotation axis of the CNC machine tool, and provides a new direction for the error detection and identification technology of the machine tool.

Description

Detection method of rotation axis error of CNC machine tool based on binocular vision

technical field

The invention belongs to the technical field of machine tool precision detection, and relates to a method for detecting and identifying geometric errors of two rotating shafts of a five-axis numerically controlled machine tool.

Background technique

In the fields of aviation, spaceflight and defense industry, the requirements for high-efficiency and high-precision manufacturing are getting higher and higher. Especially for engine impellers with complex structures, mold manufacturing and other parts, five-axis CNC machine tools can realize flexible control of position and direction, which is a widely used technology at present. However, compared with the three-axis CNC machine tool, the five-axis CNC machine tool not only has the error of three linear axes, but also increases the error of two rotary axes, which leads to an increase in the error term of the machine tool, which is inevitable. As an important component of five-axis CNC machine tools, the rotary axis is the main source of quasi-static errors and dynamic errors of machine tools because of the lack of methods for precision calibration and error compensation. Therefore, regular inspection and calibration of the error of the rotary axis can not only maintain the accuracy of the machine tool, but also lay the foundation for precision manufacturing.

At present, the error detection technology of CNC machine tools mainly includes: physical reference measurement method, laser interferometer, ballbar, laser tracker and so on. The patent number invented by Dong Hai of Dalian Chuangda Technology Co., Ltd. is CN102476323A "New CNC Machine Tool Error Detector" invented an error detection instrument based on circular trajectory, which evaluates the performance of machine tools by analyzing the error of circular trajectory interpolation. However, this method is not convenient to realize the identification of various errors, which leads to difficulties in error compensation. The patent No. CN103143984A "Dynamic Compensation Method for Machine Tool Error Based on Laser Tracker" invented by Tao Guibao of Chongqing University and others invented a real-time detection technology for machine tool error based on laser tracker. Although this method is simple and convenient, it is expensive. The laser interferometer has high measurement accuracy, but the operation is more complicated. To sum up, the current methods are mostly used for the error detection of linear axes, and the cost is high, and they are not suitable for the detection and identification of rotary axis errors. Therefore, it is necessary to study a convenient, fast and low-cost rotary axis error detection and identification technology.

Contents of the invention

The technical problem to be solved by the present invention is to overcome the problems of the prior art and to invent a method for measuring the installation error of the rotating shaft of a machine tool based on binocular vision. Affix multiple sets of reflective coding marks on the surface of the rotating shaft to be tested, the numerical control system controls the rotating shaft of the machine tool to rotate at a fixed angle, and uses binocular vision to collect reflective coding marking points at each angle, and obtain the three-dimensional position information of the coding marking points. Based on the position information of the above marker points, the least square method is used to fit the space circle, and the position coordinates of the space circle are obtained, and compared with the ideal axis center position, the linear error of the center of the rotation axis to be measured is obtained; at the same time, the space plane is fitted by the position of the mark points, The normal vector of this plane is compared with the ideal shaft axis vector, and the angular error of the shaft installation can be measured. This method uses circular marking points, which not only has simple image processing procedures and high feature extraction accuracy, but also has good robustness, fast and convenient measurement. At the same time, this method solves the problem of difficult detection and identification of the installation error of the rotation axis of the CNC machine tool, and provides a new direction for the error detection and identification technology of the machine tool.

The technical solution adopted in the present invention is a binocular vision-based method for detecting the rotation axis error of a CNC machine tool, which is characterized in that the present invention adopts a high-resolution binocular vision system to collect and affix the circular code mark points on the surface of the machine tool turntable. Position information, every time the turntable rotates through a certain angle, the vision system collects it once until it rotates once. Finally, through camera calibration, image segmentation, marker point extraction, and machine tool rotation axis error identification model, the detection and collection of two position errors and two angle errors of the machine tool rotation axis are realized, and four installation errors of the machine tool rotation axis are obtained, and the geometric parameters are completed. Rapid measurement; the specific steps of the detection method are as follows:

(1) Camera calibration

The present invention adopts the binocular vision calibration method based on high-precision checkerboard targets proposed by Zhang Zhengyou and others;

Firstly, use Zhang’s calibration method to determine the internal and external parameters of the two cameras, then perform three-dimensional reconstruction of the corner points of the checkerboard target, and establish a function f(x) according to the deviation between the reconstructed coordinates of the corner points and the actual coordinates, and optimize the internal and external parameters as a whole ;As follows:

f(x)＝(x _p -x _i ) ² +(y _p -y _i ) ² +(z _p -z _i ) ² (1)

Among them: x _p , y _p , z _p are the actual coordinates of each corner point, and x _i , y _i , _zi are the reconstructed coordinates of each corner point, then the objective function F(x) is established as follows:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; min</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$

in, is the sum of squares of all point deviation functions, and the LM method is used to optimize the objective function F(x) to obtain the global optimal solution of internal and external parameters;

(2) Image feature segmentation

First, noise reduction and filtering are performed on the image, and then all target features are preliminarily separated from the background by using the gray threshold method. The corresponding formula of the gray threshold method is:

$<span\; class="notranslate">\; \{</span>\begin{array}{cc}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

Among them, g(x, y) is the gray value corresponding to the pixel of the image (x, y), T represents the selected gray threshold value, G ₁ and G ₂ are the background set and feature mark set; then, the The feature mark set is used to mark connected regions, and use the region area as a threshold value to remove uninteresting connected regions in the image. The corresponding formula is as follows:

$\left\{\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; S</span>}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; 0</span>}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; S</span>}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; 1</span>}\end{array}\right.\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>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$

Among them, i=1,2....n are n connected regions, g _i (x,y) is the area of the i-th connected region, S is the threshold value of the connected region area; if the connected region area is smaller than S, then Set this connected region as the background;

(3) Extraction of feature marks

1) Code point center extraction:

First, 8 connected regions are used to mark the connected regions in the image, and then the curvature constraints are used to remove the uninteresting connected regions with larger and smaller curvatures. The corresponding formula is as follows:

$\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right.\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2....</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$

Among them, i=1,2....n are n connected regions, gt(i) is the eccentricity of the i-th connected region, e ₁ and e ₂ are the eccentricity thresholds, L(i)=0 Indicates that the i-th connected domain is set as the background; in this way, an accurate coded marker image can be obtained; then, using the centroid algorithm, the coordinates of the coded marker center are obtained;

2) Code point identification:

The present invention adopts the circular coding marking point, the circular coding center is the circular marking point 1, and the surrounding of the marking point is a segmented ring area concentric with it, which is used to represent the identity information of the circular coding, which is called the coding zone 2; The ring is divided into 15 parts on average according to the angle, and each part is 24 degrees, which is equivalent to a binary bit; each bit takes the foreground color as white and the back color as black, and the corresponding binary codes are "1" and "0"; Starting from the center of the circle, scan the solid and hollow coding tapes in a certain direction, white represents solid, black represents hollow, and the solid code tape is scanned as 1, and the hollow code tape is recorded as 0; if no code tape is scanned, start scanning again from the center ; After scanning for one week, the code value sequence of the entire code point is read out to form a binary sequence, and each binary sequence corresponds to a decimal integer, thereby obtaining the identity information of each code point;

After decoding, according to the identity information sent by different coded points, the pixel coordinates of the same coded point at each angle are stored in a file, and the pixel coordinates of the left and right images of the even marked points are sequentially obtained; and then obtained by Zhang’s calibration method The internal and external parameters of the camera are used to reconstruct the three-dimensional coordinates of each marked point;

(4) Error identification of the rotary axis of the machine tool

There are two main error sources for the rotary axis error of CNC machine tools, namely connection error and volume error; the former has nothing to do with the command position of the machine tool, and is usually caused by the installation deviation of the rotary axis; the latter is related to the command position of the machine tool and is affected by the processing of machine tool parts. Influenced by precision; the present invention aims at the connection error of the machine tool's rotating shaft, and invents a method for detecting and identifying the machine's rotating shaft error based on binocular vision; there are 4 items of connection error, including 2 items of linear position error and 2 items of angle error;

According to the three-dimensional coordinates of the coding points in the visual coordinate system obtained at different angles, the least square method is used to fit the plane, and the plane equation is established:

Ax+By+Cz+D=0(6)

Among them, A, B, C, D are plane equation coefficients; after simplification, we can get:

$\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; B</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; D.</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

In order to achieve plane fitting, the objective function F(x) is established:

$\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$

in, $a_0=-\frac{A}{C},a_1=-\frac{B}{C},a_2=-\frac{\mathrm{D.}}{C},$ ( _xi , y _i , z _i )(i=1,2,3...n) are the three-dimensional coordinates of n coded marker points in the visual coordinate system; thus the fitted plane can be obtained, and the The normal vector of the plane. Comparing the normal vector of the fitting plane and the normal vector of the ideal plane, the two angle errors of the connection error of the rotation axis are solved;

In order to identify the linear position error of the connection error of the rotating shaft, according to the positional relationship of the coding points, every two points are connected to form a straight line L ₁ ; when the rotating shaft rotates at a certain angle, the straight line will follow the rotating shaft to rotate to form a straight line L ₂ , and The straight line L ₁ and the straight line L ₂ intersect at point P ₁ ; sequentially, the rotation axis rotates once to form n straight lines, and every two straight lines intersect at point P _i , i=1, 2,...n/2, Take the average value P of the coordinates of these points, and regard P as the center of the actual circle; comparing the coordinates of the actual center of the circle with the ideal center of the circle can obtain the linear position error of the connection error of the rotary axis of the machine tool:

er(x)=P(x)-P _ideal (x)(9)

er(y)=P(y) _-Pideal (y)(10)

Among them, er(x), er(y) are the linear position error of the rotation axis in the X and Y directions respectively, P(x), P(y) are the X and Y coordinates of the actual center of the rotation axis, P _ideal (x), P _ideal (y) is the X and Y coordinates of the ideal rotation axis circle center;

The beneficial effect of the present invention is that the method utilizes the reflective coding mark point to realize the detection and identification of the four connection errors of the rotary axis of the CNC machine tool, which is convenient, fast, robust, strong in noise resistance, and does not require laser alignment and other axes. Coordination and other advantages. This method effectively improves the efficiency of machine tool rotation axis error detection, avoids the cumbersome measurement process and complex identification model, and provides a fast and convenient method for error detection of CNC machine tools; at the same time, it provides a basis for other error detection of machine tools. direction.

Description of drawings

Figure 1 is a model diagram of the machine tool error detection device. Among them, 1‐left camera, 2‐right camera, 3‐reflective coding marking point, 4‐rotary platform of machine tool, 5‐NC machine tool.

Fig. 2 is a diagram of marking points of circular coding. Among them, 1‐circle marking point, 2‐coding band.

Figure 3 is a schematic diagram of the machine tool rotation angle axis error identification. 1‐ideal turntable plane, 2‐actual turntable plane, 3‐actual normal vector of turntable, 4‐coded mark point, ε ₁ ‐angular error between machine tool rotation axis and actual turntable normal vector 3, ε ₂ ‐machine tool rotation axis and The angular error of the Z axis.

Figure 4 is a schematic diagram of the position error identification of the rotary axis of the machine tool. 1‐ideal turntable, 2‐actual turntable, 3‐actual turntable center, 4‐coded mark point, 5‐ideal turntable center, δ ₁ ‐linear position error between machine tool rotation axis and X axis, δ ₂ ‐machine tool rotation axis and Y The linear position error of the axis.

Detailed ways

The specific embodiments of the present invention will be described in detail below in conjunction with the technical solutions and accompanying drawings. Accompanying drawing 1 is a model diagram of a machine tool error detection device based on binocular vision. In the method, two left and right cameras 1 and 2 are used to collect coordinate information of coded marking points on the surface of the measured turntable, and after processing, the link error of the rotation axis is resolved and identified.

Install the measurement device first, install the left and right cameras 1 and 2 above the rotation axis, fix the left and right high-speed cameras 1 and 2, and adjust the position so that the measurement field of view is in the common field of view of the left and right high-speed cameras 1 and 2 Inside, adjust the brightness of the light source to increase the brightness of the measurement space; then, affix the reflective markers 3 on the surface of the turntable 4 arbitrarily, and control the machine tool to rotate at a certain angle. Every time it rotates, the left and right cameras 1 and 2 take pictures once until the turntable 4 rotates One week; at the end, the graphics workstation will perform tasks such as binocular camera calibration, image segmentation, feature extraction, and error identification.

The present invention adopts two high-resolution cameras 1 and 2 to shoot the motion of the object. The two cameras are VA-29M cameras with a resolution of 6576×4384, a target surface size of 35 mm, and a frame rate of 5 fps. The lens model is Canon EF24‐70mmf/2.8LIIUSM zoom lens, the parameters are as follows, lens focal length: f=24‐70, maximum aperture: F2.8, lens weight: 805g, lens size: 88.5mm×113mm. The shooting conditions are as follows: the pixel of the picture is 6576×4384, the focal length of the lens is 50mm, the object distance is 460mm, and the field of view is about 200mm×200mm.

(1) Carry out the calibration of the high-speed camera

The present invention adopts the camera calibration method based on the two-dimensional planar checkerboard target proposed by Zhang Zhengyou et al. to perform calibration to obtain the internal parameter K, external parameter [RT], and distortion coefficient δ of the two high-speed cameras, and then apply Levenberg-Marquardt( The LM) method optimizes formula (2), and the global optimal solution of the internal and external parameters of each camera in the binocular vision system can be obtained. The calibration results are shown in Table 1:

Table 1 Calibration results

(2) Image feature segmentation

The gray threshold method is used to preprocess the collected images. According to the formula (3), the coded marker points are initially separated from the background through the gray threshold. Then, mark the connected regions on the coded label set, and use the region area as the threshold value to remove the uninteresting connected regions in the image, and finally realize the image segmentation.

(3) Extraction of feature marks

8-connected regions are used to mark the connected regions in the image, and then the curvature constraints are used to remove the uninteresting connected regions with larger and smaller curvatures to obtain a clear image. At the same time, the centroid algorithm is used to obtain the central coordinates of the coded mark points; Figure 2 is a circular coded mark point diagram, which shows that the circle mark point 1 is white, and the surrounding ring area indicates that the identity information of the coded mark point is coded band 2. Starting from the center of the circle marking point, scan the solid and hollow coding tapes in a clockwise direction. Among them, white represents solid, black represents hollow, and the solid code tape is recorded as 1, and the hollow code tape is recorded as 0. After scanning for one week, read out the code value sequence of the entire code point to form a binary sequence 001100111011111, and convert it into a decimal integer 6607 to obtain the identity information of each code point. In addition, after binocular reconstruction, the 3D coordinates of the coded marker points can be obtained.

(4) Error identification of the rotary axis of the machine tool

In the present invention, the machine tool is controlled to rotate once every 5°, a total of 360°, and the left and right images are collected once at each angle, and the three-dimensional coordinates of the coding points are reconstructed. In order to realize the detection and identification of the connection error of the rotary axis of the machine tool, the plane fitting and the calculation of the center of the circle are carried out respectively. First, as shown in accompanying drawing 3, according to the three-dimensional coordinates of the coding point 4 in the visual coordinate system obtained at different angles, use the formulas (6), (7), and (8) to fit the plane of the marked point by the least square method 2, and obtain the normal vector 3 of the plane through calculation. Comparing the normal vector 3 of the fitting plane with the normal vector of the ideal plane, the Z-axis vector is regarded as the normal vector of the ideal plane, and the angle error ε ₁ between the machine tool rotation axis and the actual turntable normal vector 3 is obtained, and the machine tool rotation axis and The angular error of the Z axis ε ₂ two-term angular error.

In order to identify the linear position error of the connection error of the rotary axis, according to the positional relationship of the code points, two code points are selected to form a straight line, which is denoted as the initial straight line L ₁ . The machine tool rotates every 5°, and a new straight line will be formed without turning once, and each initial straight line will form 71 new straight lines. The intersection of these straight lines is regarded as the center O, as shown in Figure 4, and finally passed Three sets of experiments were averaged to determine the exact center of the circle. Use formulas (9) and (10) to compare the coordinates of the actual center 3 and the ideal center 5, and finally obtain the linear position error δ ₁ between the machine tool rotation axis and the X axis, and the linear position error δ ₂ between the machine tool rotation axis and the Y axis.

The invention uses binocular vision to detect coded marking point information, and establishes a relatively simple error identification model to realize the error detection and identification of the rotation axis of the machine tool. This method has the advantages of convenience, quickness, good robustness, strong anti-noise ability, no need for laser alignment and cooperation with other axes, etc. It effectively improves the efficiency of machine tool rotation axis error detection, and at the same time provides the basis and direction for other error detection of machine tools .

Claims (1)

1. A method for detecting the error of the axis of rotation of a numerically controlled machine tool based on binocular vision, characterized in that the present invention adopts a high-resolution binocular vision system to collect the positional information attached to the circular code mark point on the surface of the turntable of the machine tool. After a certain angle, the vision system collects once until it rotates once. Finally, through camera calibration, image segmentation, marker point extraction, and machine tool rotation axis error identification model, the detection and collection of two position errors and two angle errors of the machine tool rotation axis are realized, and the rapid measurement of geometric parameters is completed. The specific steps of the detection method are as follows: (1) Camera calibration The present invention adopts the binocular vision calibration method based on high-precision checkerboard targets proposed by Zhang Zhengyou and others; Firstly, use Zhang’s calibration method to determine the internal and external parameters of the two cameras, then perform three-dimensional reconstruction of the corner points of the checkerboard target, and establish a function f(x) according to the deviation between the reconstructed coordinates of the corner points and the actual coordinates, and optimize the internal and external parameters as a whole ;As follows: f(x)＝(x _p -x _i ) ² +(y _p -y _i ) ² +(z _p -z _i ) ² (1) Among them: x _p , y _p , z _p are the actual coordinates of each corner point, and x _i , y _i , _zi are the reconstructed coordinates of each corner point, then the objective function F(x) can be established as follows: $\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; f</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>$ in, is the sum of squares of all point deviation functions, and the LM method is used to optimize the objective function F(x) to obtain the global optimal solution of internal and external parameters; (2) Image feature segmentation First, noise reduction and filtering are performed on the image, and all target features are initially separated from the background by using the gray threshold method. The corresponding formula of the gray threshold method is: $<span\; class="notranslate">\; \{</span>\begin{array}{cc}\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; T</span>}\\ \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; T</span>}\end{array}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$ Among them, g(x, y) is the gray value corresponding to the pixel of the image (x, y), T represents the selected gray threshold value, G ₁ and G ₂ are the background set and feature mark set; for feature mark The set is marked with connected regions, and the area of the region is used as the threshold value to remove uninteresting connected regions in the image. The corresponding formula is as follows: $<span\; class="notranslate">\; \{</span>\begin{array}{cc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; S</span>}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; 0</span>}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; \&Greater\; Equal;</span>\mathrm{<span\; class="notranslate">\; S</span>}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; i</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">\; 1</span>}\end{array}\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>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; )</span>$ Among them, i=1,2....n are n connected regions, g _i (x,y) is the area of the i-th connected region, S is the threshold value of the connected region area; if the connected region area is smaller than S, then Set this connected region as the background; (3) Extraction of feature marks 1) Code point center extraction: First, 8 connected regions are used to mark the connected regions in the image, and then the curvature constraints are used to remove the uninteresting connected regions with larger and smaller curvatures. The corresponding formula is as follows: $\left\{\begin{array}{cc}\mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; g</span>}\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ></span>{\mathrm{<span\; class="notranslate">\; e</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& \mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}\end{array}\right.\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 2....</span>}\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 5</span>}<span\; class="notranslate">\; )</span>$ Among them, i=1,2....n are n connected regions, gt(i) is the eccentricity of the i-th connected region, e ₁ and e ₂ are the eccentricity thresholds, L(i)=0 Indicates that the i-th connected domain is set as the background; in this way, an accurate coded marker image is obtained; using the centroid algorithm, the coordinates of the coded marker center are obtained; 2) Code point identification: The present invention adopts the ring coding marking point, the ring coding center is the circular marking point (1), and the surrounding of the marking point is a segmented ring area concentric with it, which is used to represent the identity information of the ring coding, called the coding band (2 ); the ring is divided into 15 parts on average according to the angle, and each part is 24 degrees, which is equivalent to a binary bit; each bit takes the foreground color as white and the back color as black, and the corresponding binary codes are "1" and "0". ;Starting from the center of the marked point, scan the solid and hollow coding tapes according to a certain direction. When the solid code tape is scanned, it is recorded as 1, and the hollow code tape is recorded as 0. If the code tape is not scanned, start scanning again from the center; scan for one week After that, the code value sequence of the entire code point is read out completely to form a binary sequence, and each binary sequence corresponds to a decimal integer, thereby obtaining the identity information of each code point; After decoding, according to the identity information sent by different coded points, the pixel coordinates of the same coded point at each angle are stored in a file, and the pixel coordinates of the left and right images of the even marked points are sequentially obtained; and then obtained by Zhang’s calibration method The internal and external parameters of the camera are used to reconstruct the three-dimensional coordinates of each marked point; (4) Error identification of the rotary axis of the machine tool The invention detects and identifies the connection error of the rotary axis of the machine tool, including two linear position errors and two angular errors; According to the three-dimensional coordinates of the coding points in the visual coordinate system obtained at different angles, the least square method is used to fit the plane, and the plane equation is established: Ax+By+Cz+D=0(6) Among them, A, B, C, D are plane equation coefficients; after simplification, we can get: $\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; A</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; B</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; D.</span>}}{\mathrm{<span\; class="notranslate">\; C</span>}}<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$ In order to achieve plane fitting, the objective function F(x) is established: $\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; m</span>}\mathrm{<span\; class="notranslate">\; i</span>}\mathrm{<span\; class="notranslate">\; no</span>}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; no</span>}}{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; a</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; z</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 8</span>}<span\; class="notranslate">\; )</span>$ in, $a_0=-\frac{A}{C},a_1=-\frac{B}{C},a_2=-\frac{\mathrm{D.}}{C},$ ( _xi , y _i , z _i )(i=1,2,3...n) are the three-dimensional coordinates of n coded marker points in the visual coordinate system; thus the fitted plane can be obtained, and the The normal vector of the plane. Comparing the normal vector of the fitting plane and the normal vector of the ideal plane, the two angle errors of the connection error of the rotation axis are solved; In order to identify the linear position error of the connection error of the rotating shaft, according to the positional relationship of the coding points, every two points are connected to form a straight line L ₁ ; when the rotating shaft rotates at a certain angle, the straight line will follow the rotating shaft to rotate to form a straight line L ₂ , and The straight line L ₁ and the straight line L ₂ intersect at point P ₁ ; sequentially, the rotation axis rotates once to form n straight lines, and every two straight lines intersect at point P _i , i=1, 2, 3...n/2, for these The coordinates of the points take the average value P, and P is regarded as the center of the actual circle; comparing the coordinates of the actual center of the circle with the ideal center of the circle can obtain the linear position error of the connection error of the rotary axis of the machine tool: er(x)=P(x)-P _ideal (x)(9) er(y)=P(y) _-Pideal (y)(10) Among them, er(x), er(y) are the linear position error of the rotation axis in the X and Y directions respectively, P(x), P(y) are the X and Y coordinates of the actual center of the rotation axis, P _ideal (x), P _ideal (y) is the X and Y coordinates of the ideal rotation axis circle center.