A bearing deformation measurement method and device based on image recognition

Abstract

This invention relates to a method and apparatus for measuring bearing deformation based on image recognition. The apparatus includes a base plate, a trapezoidal lead screw adjustment assembly, a measuring assembly, and a support assembly. The method includes: acquiring panoramic images of the inner and outer rings of the bearing under clearance fit and interference fit conditions; calculating the image ratio; performing image preprocessing, establishing a circular ROI, and extracting sub-pixel edge feature points; aligning feature points through polar coordinate transformation and isoangular interpolation; constructing a quality evaluation model including edge gradient intensity, geometric fitting residual, and local contrast; calibrating the confidence coefficient using a genetic algorithm; calculating the joint probability confidence of feature points; weighted calculation of radial deformation; and finally obtaining the raceway bottom deformation by combining elasticity formulas. This invention has a simple structure and is easy to operate. Through multi-dimensional quality evaluation and weighted fusion, it achieves high-precision and robust non-contact measurement of the raceway bottom deformation after interference fit of the bearing.

Description

Technical Field

[0001] This invention relates to the field of bearing deformation measurement technology, and specifically to a bearing deformation measurement method and apparatus based on image recognition. Background Technology

[0002] The primary function of space robots is to replace or assist astronauts in performing tasks in the extremely harsh space environment. These robots need to achieve high-precision positioning, high-torque output, and reliable operation over long periods under conditions of high vacuum, microgravity, extreme temperature differences, and strong radiation. The realization of these functions depends on the performance of the bearings in their drive system. In space robot applications, bearings are typically installed with an interference fit to improve system rigidity, suppress vibration, and ensure accurate motion transmission. However, while this fit eliminates clearance and improves support rigidity, it also introduces complex mechanical effects: the applied interference causes the inner ring of the bearing to expand and the outer ring to contract, leading to elastic deformation of the raceway. This deformation not only changes the absolute dimensions of the raceway but also alters the relative position of its center of curvature. Therefore, the amount of change at the bottom of the raceway directly determines the residual clearance or preload state after installation. For space robot joints, this means that during startup or load changes, the axis trajectory will drift unpredictably, directly weakening the system's dynamic response characteristics and positioning stability. Simultaneously, raceway deformation can also disrupt the uniformity of load distribution within the bearing, inducing abnormal cage vibration and noise. In a microgravity environment, such vibrations cannot be attenuated by gravity damping and will be transmitted and amplified along the robotic arm structure, causing flutter in the entire joint system.

[0003] Traditionally, the detection of raceway bottom changes in bearings after interference fit relies primarily on coordinate measuring machines (CMMs). However, this method suffers from problems such as complex operation procedures, high inspection costs, and stringent requirements for measurement environment and usage conditions, making it difficult to meet the current production practices' widespread demands for inspection efficiency, ease of operation, and cost control.

[0004] Existing technology, Chinese invention patent publication number CN117308784A, entitled "High-Precision Visual Online Measurement Method for Key Bearing Parameters," discloses a high-precision visual online measurement method for key bearing parameters. This method acquires partial images of the bearing edge using a side-view camera and employs 3D reconstruction and perspective projection transformation techniques to eliminate the impact of geometric distortion on measurement accuracy. However, this invention does not consider the impact of light and shadow reflections, surface oil contamination, or texture interference in the actual measurement environment on the reliability of edge feature point extraction. This results in limitations in robustness and noise immunity when processing data with non-ideal image quality. Furthermore, because only a partial area of ​​the bearing edge is measured, any minute drift in camera calibration parameters will be geometrically amplified during local fitting, leading to a decrease in measurement accuracy.

[0005] To address the above technical problems, this invention provides a measurement method and device for measuring the change in raceway bottom after interference fit of a bearing for a space robot, which is simple in structure, easy to operate, and low in cost. This enables rapid and accurate detection of the change in raceway bottom after interference fit of the bearing, thereby ensuring the geometric accuracy of the bearing after assembly. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an image recognition-based bearing deformation measurement method. This method acquires panoramic images of the inner and outer rings of the bearing under different fit conditions using an industrial camera; image processing algorithms are used to identify edge contours and lock the region of interest; and sub-pixel edge localization and polar coordinate interpolation are used to achieve precise alignment of feature positions in the initial and interference states. Subsequently, a quality evaluation model integrating edge gradient intensity, geometric fitting residuals, and local contrast is established, and weighted fusion of the full-circumference radial displacement is performed using parameter collaborative optimization. Finally, combining elasticity mechanics, the actual deformation of the bearing raceway bottom is calculated from the surface deformation.

[0007] A bearing deformation measurement device based on image recognition is also provided, which has a simple structure, is easy to operate and has low operating cost.

[0008] The technical problem solved by this invention is achieved through the following technical solution: A bearing deformation measurement method based on image recognition includes the following steps: Step 1: Obtain the initial panoramic image and the panoramic image after interference fit of the bearing inner ring. Using the initial panoramic image, calculate the image scale of the bearing inner ring. ; Step 2: Obtain the initial panoramic image and the panoramic image after interference fit of the bearing outer ring. Using the initial panoramic image, calculate the image scale of the bearing outer ring. ; Step 3: Perform image preprocessing on the initial panoramic image and the panoramic image after interference fit, establish a circular region of interest, extract the feature point set of the entire circumference of the bearing's inner / outer diameter through sub-pixel edge detection, and perform circle fitting; Step 4: Perform polar coordinate transformation on the feature point coordinates and align the initial image with the interference image using equal angle interpolation. Step 5: Construct edge gradient strength Geometric fitting residuals and local contrast Quality evaluation model; Step 6: Calculate the radial deformation of the bearing inner ring and bearing outer ring using weighted average. Steps three through six are used to process the measurement data of the bearing inner ring and the bearing outer ring, respectively; wherein, the object scale calculated in step one is used when calculating the bearing inner ring. The object scale calculated in step two is used when calculating the outer ring of the bearing. ; Step 7: Calculate the raceway bottom deformation of the inner and outer rings of the bearing based on the radial deformation.

[0009] Further, in step one, a simulated shaft with dimensions meeting the clearance fit requirements is selected, and the simulated shaft is installed on the bearing inner ring. The light source is turned on, and the handwheel is rotated to adjust the height of the industrial camera so that the entire bearing inner ring is imaged at the center of the field of view. An initial panoramic image containing the complete inner diameter outline of the bearing inner ring is captured using the industrial camera. The image ratio of the bearing inner ring is calculated using the nominal geometric dimensions of the bearing inner ring's inner diameter as a reference. ; With the industrial camera position unchanged, replace the simulation shaft with one whose size meets the set interference requirements, install the simulation shaft with the bearing inner ring, and take another panoramic image containing the complete inner diameter outline of the bearing inner ring at the same industrial camera station.

[0010] Further, in step two, a simulated shell with dimensions meeting the clearance fit requirements is selected, and the simulated shell is installed on the bearing outer ring. The light source is turned on, and the handwheel is rotated to adjust the height of the industrial camera so that the entire bearing outer ring is imaged at the center of the field of view. An initial panoramic image containing the complete outer diameter outline of the bearing outer ring is captured using the industrial camera. The image-to-object ratio of the outer ring is calculated using the nominal geometric dimensions of the bearing outer ring's outer diameter as a reference. ; With the industrial camera position unchanged, replace the simulation housing with one whose size meets the set interference fit requirements, install the simulation housing with the bearing outer ring, and take another panoramic image containing the complete outer diameter outline of the bearing outer ring at the same industrial camera station.

[0011] Furthermore, in step three, the initial panoramic image and the panoramic image after interference are denoised using an adaptive median filtering method, and the image is binarized using Otsu's adaptive threshold segmentation method to identify and extract the rough outline of the bearing edge. Based on this rough outline, an annular region of interest is automatically created around the edge; Within the annular region of interest, sub-pixel edge detection is performed to extract the sub-pixel feature point set of the entire circumference of the bearing's inner / outer diameter. The feature point set is then fitted to a circle using the least squares method to calculate the initial state of the fitted circle's center coordinates. Fitted circle center coordinates under interference condition .

[0012] Further, in step four, let i be the index of the extracted sub-pixel feature point (i=1, 2, 3...n), and use the center coordinates of the circles fitted by the initial image and the overlay image respectively. Using the mathematical origin, feature points in the Cartesian coordinate system in the initial image and the overlay image are respectively... Mapping to polar coordinates By aligning with equal-angle interpolation, the initial panoramic image and the overlay panoramic image are aligned at the same polar angle. Each has a corresponding polar diameter and ,in, and The calculation formulas are as follows: (1); (2); In the formula, Let x be the x-coordinate of the feature point in the rectangular coordinate system. The ordinate of the feature point in the rectangular coordinate system. To fit the x-coordinate of the circle's center in a rectangular coordinate system, To fit the ordinate of the circle's center in a rectangular coordinate system.

[0013] Furthermore, in step five, the edge gradient intensity... The radial gradient magnitude is characterized by performing first-order difference calculations on the pixels in the neighborhood of the feature point along the polar radial direction in the polar coordinate system established in step four, and obtaining the radial gradient magnitude obtained by taking the maximum value of the slope of the gray-scale change curve. Geometric Fitting Residual The center and radius of the circle based on the ideal circle parameters fitted in step three Calculate the polar radius of each sub-pixel feature point i in step four. With fitted radius Substitute into the formula It can be concluded that; Local contrast Taking the sub-pixel coordinates of feature point i as the center, a fixed-size local window (7x7 pixel area) is selected, and the highest grayscale value of all pixels within this window is counted. With the lowest gray value Substitute into the formula It can be concluded that; Calculate the confidence score of the initial and over-exposed image quality for each feature point, and the confidence score of the i-th feature point. The formula is as follows: (3); In the formula, Let be the edge gradient intensity of the i-th feature point. Let be the geometric fitting residual of the i-th feature point. Let i be the local contrast of the i-th feature point. , , The coefficients are obtained through a genetic algorithm. First, multiple sets of bearing sample images with known actual deformation are acquired. The edge gradient intensity, fitting residual, and local contrast of each feature point are calculated, and the actual deformation is used as the true value. Then... The data is encoded as a real number vector and an initial population is randomly generated. The root mean square error between the estimated deformation and the true value is used as the fitness function. The data is iteratively evolved through selection, crossover, and mutation genetic operations until the fitness converges. Finally, the optimal combination of coefficients that minimizes the error is output.

[0014] The initial panoramic image and the overexposed panoramic image are at the same polar angle The paired feature points are associated, and the joint probability confidence of each feature point is calculated. Joint probability confidence The formula is as follows: (4); In the formula, The initial image feature point confidence score for this feature point. The confidence level of the feature point in the over-exposed image.

[0015] Further, in step six, the local radial displacement at the i-th polar angle point is calculated. The formula is: (5); In the formula, Let the polar radius be the interference pattern of the i-th feature point. The polar radius of the initial image of the i-th feature point; Using joint probability confidence The final radial deformation is obtained by weighted averaging of the deformation at all points around the circumference, and the deformation of the inner or outer diameter is then calculated. The formula is: (6); In the formula, k is the image scale, used when calculating the inner ring of the bearing. When calculating the outer ring of a bearing, use .

[0016] Furthermore, in step seven, based on the bearing inner diameter deformation... and outer diameter deformation Calculate the corresponding change in the bottom of the inner raceway. Changes in the bottom of the outer raceway The formula is: (7); (8); In the formula, d is the inner diameter of the bearing, and D is the outer diameter of the bearing. The diameter of the inner raceway bottom. The diameter of the outer raceway bottom. The Poisson's ratio of the inner and outer ring materials of the bearing. This represents the deformation of the bearing's inner diameter. This represents the deformation of the bearing's outer diameter.

[0017] A bearing deformation measuring device based on image recognition includes a base plate, a trapezoidal lead screw adjustment assembly, a measuring assembly, and a support assembly. The trapezoidal lead screw adjustment assembly includes a lead screw support plate fixed to the base plate. An EF support seat and an EK support seat are fixed to the lead screw support plate. A trapezoidal lead screw is rotatably connected between the EF support seat and the EK support seat. A handwheel is provided at the top of the trapezoidal lead screw. A trapezoidal lead screw nut is provided on the trapezoidal lead screw, and the nut is fixed in a nut seat. A protrusion on one side of the nut seat is disposed within a sliding hole in the lead screw support plate. The measuring assembly... The component includes an adapter plate fixed to the other side of the nut seat, with an industrial camera mounted on the lower end of the adapter plate; the support assembly includes an inner ring support fixed to the base plate and an outer ring support fixed to the inner ring support. The inner ring support is provided with a simulated shaft that engages with a conical surface, and the simulated shaft is provided with a bearing inner ring. The outer ring support is provided with a simulated shell that engages with a conical surface, and the simulated shell is provided with a bearing outer ring. The initial panoramic image and the panoramic image after interference fit of the bearing inner ring and bearing outer ring are acquired by the industrial camera to measure the bearing deformation.

[0018] Furthermore, the industrial camera is a CMOS camera, which is equipped with a telecentric lens and a parallel coaxial light source is fixed at the front end of the lens.

[0019] The advantages and positive effects of this invention are: 1. The bearing deformation measurement device based on image recognition of the present invention performs non-contact measurement of the deformation of the inner and outer rings of the bearing under interference fit, avoiding the risk of additional stress or scratches on the measured surface caused by traditional contact measurement.

[0020] 2. The bearing deformation measuring device based on image recognition of the present invention adopts a trapezoidal lead screw adjustment component and a conical surface fit design between the simulated shaft and the simulated shell. It can easily replace the fitting parts of different sizes to realize the comparative measurement of clearance fit and interference fit. The overall device has low cost and is easy to deploy on the production site.

[0021] 3. The bearing deformation measurement method based on image recognition of the present invention reflects the influence of the environment on the image by establishing a quality evaluation model. The model integrates the edge gradient intensity, geometric fitting residual and local contrast of the initial image and the image after interference fit. When calculating the deformation, it can automatically reduce the weight of unreliable measurement points caused by local reflection, poor focus or edge blur, etc., and improve the measurement accuracy. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the bearing deformation measurement device based on image recognition according to the present invention; Figure 2 This is a flowchart of the bearing deformation measurement method based on image recognition according to the present invention.

[0023] In the picture: 1-Base plate, 2-EF support seat, 3-Trapezoidal lead screw, 4-Lead screw support plate, 5-Nut seat, 6-Trapezoidal lead screw nut, 7-EK support seat, 8-Handwheel, 9-Adapter plate, 10-Industrial camera, 11-Bearing outer ring, 12-Simulation shell, 13-Outer ring support seat, 14-Simulation shaft, 15-Bearing inner ring, 16-Inner ring support seat. Detailed Implementation

[0024] The present invention will be further described in detail below through specific embodiments. The following embodiments are merely descriptive and not limiting, and should not be used to limit the scope of protection of the present invention.

[0025] like Figure 2 As shown, a bearing deformation measurement method based on image recognition includes the following steps: Step 1: Obtain the initial panoramic image and the panoramic image after interference fit of the bearing inner ring. Using the initial panoramic image, calculate the image scale of the bearing inner ring. : Select a simulation shaft whose dimensions meet the clearance fit requirements, install the simulation shaft on the bearing inner ring, turn on the light source, and rotate the handwheel to adjust the height of the industrial camera so that the entire bearing inner ring is imaged at the center of the field of view. Use the industrial camera to capture an initial panoramic image containing the complete outline of the bearing inner ring's inner diameter. Using the nominal geometric dimensions of the bearing inner ring's inner diameter as a reference, calculate the image scale of the bearing inner ring. ; Keeping the industrial camera position unchanged, replace the simulation shaft with one whose size meets the set interference requirements, install the simulation shaft with the bearing inner ring, and take another panoramic image containing the complete inner diameter outline of the bearing inner ring at the same industrial camera station.

[0026] Step 2: Obtain the initial panoramic image and the panoramic image after interference fit of the bearing outer ring. Using the initial panoramic image, calculate the image scale of the bearing outer ring. : Select a simulation shell whose dimensions meet the clearance fit requirements, install the simulation shell on the bearing outer ring, turn on the light source, and rotate the handwheel to adjust the height of the industrial camera so that the entire bearing outer ring is imaged at the center of the field of view. Use the industrial camera to capture an initial panoramic image containing the complete outline of the bearing outer ring's outer diameter. Using the nominal geometric dimensions of the bearing outer ring's outer diameter as a reference, calculate the object-image ratio of the outer ring. ; Keeping the industrial camera in the same position, replace the simulation housing with one whose size meets the set interference fit requirements, install the simulation housing with the bearing outer ring, and take another panoramic image containing the complete outer diameter outline of the bearing outer ring at the same industrial camera station.

[0027] Step 3: Perform image preprocessing on the initial panoramic image and the panoramic image after interference fit; establish a circular region of interest; extract the feature point set of the entire circumference of the bearing's inner / outer diameter through sub-pixel edge detection and perform circle fitting. An adaptive median filtering method was used to denoise the initial panoramic image and the panoramic image after interference, and Otsu's adaptive threshold segmentation was used to binarize the image, thereby identifying and extracting the rough outline of the bearing edge. Based on this, a ring-shaped region of interest is automatically established around the edge using the rough outline as a reference. This aims to shield background noise and stray information in non-measured areas and eliminate interference for subsequent high-precision detection. Then, within the annular region of interest, sub-pixel edge detection is performed to extract the sub-pixel feature point set of the entire circumference of the bearing's inner / outer diameter. The feature point set is then fitted to a circle using the least squares method to calculate the coordinates of the initial fitted circle center. Fitted circle center coordinates under interference condition ; Specifically, subpixel edge detection can employ a Zernike moment-based subpixel edge detection algorithm. Within a circular region of interest, a set of orthogonal Zernike moment templates are convolved with the image to calculate the low-order rotational invariant moments of the local image patch. Based on the step model of the edge points within the template, the subpixel-level normal deviation distance of the edge is analyzed, thus correcting the traditional integer-level pixel coordinates into a high-precision subpixel feature point set with an accuracy of less than 0.1 pixels.

[0028] Step 4: Perform polar coordinate transformation on the feature point coordinates and align the initial image with the overlay image using equal-angle interpolation: Let i be the index of the extracted sub-pixel feature point (i=1, 2, 3...n), and let the center coordinates of the circle fitted by the initial image and the overlay image be the coordinates of the circle. Using the mathematical origin, feature points in the Cartesian coordinate system in the initial image and the overlay image are respectively... Mapping to polar coordinates Then, by equiangular interpolation alignment, the initial panoramic image and the overlay panoramic image are aligned at the same polar angle. Each has a corresponding polar diameter and .in, and The calculation formulas are as follows: (1); (2); In the formula, Let x be the x-coordinate of the feature point in the rectangular coordinate system. The ordinate of the feature point in the rectangular coordinate system. To fit the x-coordinate of the circle's center in a rectangular coordinate system, To fit the ordinate of the circle's center in a rectangular coordinate system; Step 5: Construct edge gradient strength Geometric fitting residuals and local contrast Quality evaluation model: For each feature point after initialization and over-interference, calculate the edge gradient intensity of the initial image and the over-interference image respectively. Geometric fitting residuals and local contrast These three metrics quantify the reliability of feature points from different dimensions: edge gradient strength. This reflects the focusing quality and edge sharpness of the optical system. A larger gradient value indicates a steeper black-and-white boundary, less influence of optical diffraction or imaging blur on feature points, and higher positioning accuracy; geometric fitting residuals This reflects the degree of conformity between the feature point and the geometric features of the bearing's circle; the smaller the residual, the closer the point is to the true physical edge, effectively identifying and reducing the weight of spurious feature points caused by surface scratches, metal burrs, or oil contamination; local contrast. This reflects the stability of the local lighting environment; higher contrast means that the image has a better signal-to-noise ratio, which can provide more stable signal support for sub-pixel algorithms, thereby suppressing random positioning fluctuations caused by electronic noise. Among them, edge gradient intensity The radial gradient magnitude is characterized by performing first-order difference calculations on the pixels in the neighborhood of the feature point along the polar radial direction in the polar coordinate system established in step four, and obtaining the radial gradient magnitude obtained by taking the maximum value of the slope of the gray-scale change curve. Geometric Fitting Residual Based on the center and radius of the circle obtained from step three, the circle parameters are fitted. Calculate the polar radius of each sub-pixel feature point i in step four. With fitted radius Substitute into the formula It can be concluded that; Local contrast Taking the sub-pixel coordinates of feature point i as the center, a fixed-size local window (7x7 pixel area) is selected, and the highest grayscale value of all pixels within this window is counted. With the lowest gray value Substitute into the formula The conclusion is as follows.

[0029] Calculate the confidence score of the initial and over-exposed image quality for each feature point, and the confidence score of the i-th feature point. The formula is as follows: (3); In the formula, Let be the edge gradient intensity of the i-th feature point. Let be the geometric fitting residual of the i-th feature point. Let i be the local contrast of the i-th feature point. , , The coefficients are obtained through a genetic algorithm. First, multiple sets of bearing sample images with known actual deformation are acquired. The edge gradient intensity, fitting residual, and local contrast of each feature point are calculated, and the actual deformation is used as the true value. Then... The data is encoded as a real number vector and an initial population is randomly generated. The root mean square error between the estimated deformation and the true value is used as the fitness function. The data is iteratively evolved through selection, crossover, and mutation genetic operations until the fitness converges. Finally, the optimal combination of coefficients that minimizes the error is output.

[0030] The initial panoramic image and the overexposed panoramic image are at the same polar angle The paired feature points are associated, and the joint probability confidence of each feature point is calculated. Joint probability confidence The formula is as follows: (4); In the formula, The initial image feature point confidence score for this feature point. The confidence level of the feature point in the over-exposed image.

[0031] Step 6: Calculate the radial deformation of the bearing inner ring and outer ring using weighted average methods. Calculate the local radial displacement at the i-th polar angle point. The formula is: (5); In the formula, Let the polar radius be the interference pattern of the i-th feature point. The polar radius of the initial image of the i-th feature point; Using joint probability confidence The final radial deformation is obtained by weighted averaging of the deformation at all points around the circumference, and the deformation of the inner or outer diameter is then calculated. The formula is: (6); In the formula, k is the image scale, used when calculating the inner ring of the bearing. When calculating the outer ring of a bearing, use ; Steps three through six are used to process the measurement data of the bearing inner ring and the bearing outer ring, respectively; wherein, the object scale calculated in step one is used when calculating the bearing inner ring. The object scale calculated in step two is used when calculating the outer ring of the bearing. .

[0032] Step 7: Calculate the raceway bottom deformation of the inner and outer rings of the bearing based on the radial deformation: Based on the bearing inner diameter deformation and outer diameter deformation Calculate the corresponding change in the bottom of the inner raceway. Changes in the bottom of the outer raceway The formula is: (7); (8); In the formula, d is the inner diameter of the bearing, and D is the outer diameter of the bearing. The diameter of the inner raceway bottom. The diameter of the outer raceway bottom. The Poisson's ratio of the inner and outer ring materials of the bearing. This represents the deformation of the bearing's inner diameter. This represents the deformation of the bearing's outer diameter.

[0033] To further verify the effectiveness of the present invention and demonstrate the specific calculation process, the following calculation example is provided, using an angular contact ball bearing of model 71910AC as an example. The basic parameters of the bearing are shown in Table 1: Table 1. Basic parameters of 71910AC angular contact ball bearings

[0034] A bearing deformation measurement method based on image recognition was used to acquire panoramic images of the inner and outer rings of the bearing under initial clearance and interference fit conditions with an interference fit of 0.01 mm. After image preprocessing, a ring-shaped region of interest was established. Feature points were aligned using sub-pixel edge localization and polar coordinate interpolation, and the joint probability confidence was calculated using edge gradient intensity, geometric fitting residual, and local contrast. Based on a weighted average, the system calculated that under the 0.01 mm interference fit condition, the actual deformation of the inner diameter was 0.0089 mm, and the actual deformation of the outer diameter was 0.0069 mm.

[0035] Based on the obtained deformation of the bearing's inner and outer diameters, and by substituting them into formulas (7) and (8) respectively, we can obtain that the deformation of the inner raceway is 0.0086 mm and the deformation of the outer raceway is 0.0070 mm.

[0036] The measurement results of this invention are compared with the theoretical calculation results, and the comparison results are shown in Table 2.

[0037] Table 2 Comparison of Results

[0038] As shown in Table 2, for the angular contact ball bearing model 71910AC, under experimental conditions with an interference fit of 0.01 mm, the inner raceway deformation measured by this invention is 0.0086 mm, and the outer raceway deformation is 0.0070 mm. These measurement data are in high agreement with the theoretical calculation results, fully verifying the accuracy and effectiveness of the measurement method of this invention.

[0039] like Figure 1As shown, a bearing deformation measuring device based on image recognition includes a base plate 1, a trapezoidal lead screw adjustment assembly, a measuring assembly, and a support assembly. The base plate 1 has several threaded holes. The trapezoidal lead screw adjustment assembly includes a lead screw support plate 4 bolted to the base plate 1. An EF support seat 2 and an EK support seat 7 are bolted to the lead screw support plate 4. A trapezoidal lead screw 3 is rotatably connected between the EF support seat 2 and the EK support seat 7. A handwheel 8 is provided at the top of the trapezoidal lead screw 3. A trapezoidal lead screw nut 6 is provided on the trapezoidal lead screw 3 and is bolted into a nut seat 5. A protrusion is provided on the side of the sliding hole in the lead screw support plate 4. The measuring component includes an adapter plate 9, which is fixed to the other side of the nut seat 5 by bolts. An industrial camera 10 is fixed to the lower end of the adapter plate 9 by bolts. The support component includes an inner ring support seat 16 fixed to the base plate 1 by bolts and an outer ring support seat 13 fixed to the inner ring support seat 16 by bolts. The inner ring support seat 16 is provided with a simulated shaft 14 that fits with a conical surface. The simulated shaft 14 is provided with a bearing inner ring 15. The outer ring support seat 13 is provided with a simulated shell 12 that fits with a conical surface. The simulated shell 12 is provided with a bearing outer ring 11. There are multiple sets of simulated shells 12 and simulated shafts 14 of different sizes available. The industrial camera 10 is a CMOS camera with a telecentric lens mounted on it. A parallel coaxial light source is fixed at the front end of the lens.

[0040] Working principle of the invention: Install the simulated shaft to the bearing inner ring, turn on the light source, and rotate the handwheel to adjust the height of the industrial camera so that the entire bearing inner ring is imaged in the center of the field of view. Use the industrial camera to take an initial panoramic image containing the complete inner diameter outline of the bearing inner ring. Keeping the position of the industrial camera unchanged, replace the simulated shaft with one whose size meets the set interference fit requirements, install the simulated shaft to the bearing inner ring, and take another panoramic image containing the complete inner diameter outline of the bearing inner ring at the same industrial camera station.

[0041] The simulated housing is installed on the bearing outer ring. The light source is turned on, and the handwheel is rotated to adjust the height of the industrial camera so that the entire bearing outer ring is imaged in the center of the field of view. An initial panoramic image containing the complete outer diameter contour of the bearing outer ring is captured using the industrial camera. Keeping the position of the industrial camera unchanged, a simulated housing with dimensions meeting the set interference fit requirements is replaced. The simulated housing is then installed on the bearing outer ring, and another panoramic image containing the complete outer diameter contour of the bearing outer ring is captured at the same industrial camera position. The bearing deformation measurement device based on image recognition is simple in structure, easy to operate, and has low operating costs.

[0042] This invention relates to an image recognition-based bearing deformation measurement device. It performs non-contact measurement of the deformation of the inner and outer rings of a bearing under interference fit conditions, avoiding the risks of additional stress or scratches on the measured surface caused by traditional contact measurements. Employing a trapezoidal lead screw adjustment assembly and a conical fit design between the simulated shaft and the simulated housing, it allows for easy replacement of mating parts of different sizes, enabling comparative measurement of clearance and interference fits. The overall device is low-cost and easy to deploy on the production site.

[0043] Although embodiments and drawings of the present invention have been disclosed for illustrative purposes, those skilled in the art will understand that various substitutions, variations and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A method of bearing deformation measurement based on image recognition, characterized in that: Includes the following steps: Step one, obtaining the initial panoramic image and the interference panoramic image of the bearing inner ring, and calculating the object-image ratio of the bearing inner ring based on the initial panoramic image ; Step two, obtaining the initial panoramic image and the interference panoramic image of the bearing outer ring, and calculating the object-image ratio of the bearing outer ring based on the initial panoramic image ; Step 3: Perform image preprocessing on the initial panoramic image and the panoramic image after interference fit, establish a circular region of interest, extract the feature point set of the entire circumference of the bearing's inner / outer diameter through sub-pixel edge detection, and perform circle fitting; Step 4: Perform polar coordinate transformation on the feature point coordinates and align the initial image with the interference image using equal angle interpolation. Step five, constructing edge gradient strength , geometric fitting residual and local contrast quality evaluation model; Step 6: Calculate the radial deformation of the bearing inner ring and bearing outer ring using weighted average. The measurement data of the bearing inner ring and the bearing outer ring are processed respectively through the steps three to six; wherein the object-image ratio solved in the step one is adopted in the bearing inner ring calculation , and the object-image ratio solved in the step two is adopted in the bearing outer ring calculation . Step 7: Calculate the raceway bottom deformation of the inner and outer rings of the bearing based on the radial deformation.

2. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: The step one, selecting the simulation shaft which size meets the gap fit requirement, installing the simulation shaft and the bearing inner ring, opening the light source, rotating the hand wheel to adjust the height of the industrial camera, making the whole bearing inner ring image in the center of the field of view, using the industrial camera to shoot the initial panoramic image containing the inner diameter profile of the complete bearing inner ring, taking the nominal geometric size of the bearing inner ring as the reference, calculating the object-image ratio of the bearing inner ring ; With the industrial camera position unchanged, replace the simulation shaft with one whose size meets the set interference requirements, install the simulation shaft with the bearing inner ring, and take another panoramic image containing the complete inner diameter outline of the bearing inner ring at the same industrial camera station.

3. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: In step two, a simulated shell with dimensions meeting the clearance fit requirements is selected. The simulated shell is then installed on the bearing outer ring. The light source is turned on, and the handwheel is rotated to adjust the height of the industrial camera, so that the entire bearing outer ring is imaged at the center of the field of view. An initial panoramic image containing the complete outer diameter outline of the bearing outer ring is captured using the industrial camera. The image-to-object ratio of the outer ring is calculated using the nominal geometric dimensions of the bearing outer ring's outer diameter as a reference. ; With the industrial camera position unchanged, replace the simulation housing with one whose size meets the set interference fit requirements, install the simulation housing with the bearing outer ring, and take another panoramic image containing the complete outer diameter outline of the bearing outer ring at the same industrial camera station.

4. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: In step three, the initial panoramic image and the panoramic image after interference are denoised using an adaptive median filtering method, and the image is binarized using the Otsu method of adaptive threshold segmentation to identify and extract the rough outline of the bearing edge. Based on this rough outline, an annular region of interest is automatically created around the edge; Within the annular region of interest, sub-pixel edge detection is performed to extract the sub-pixel feature point set of the entire circumference of the bearing's inner / outer diameter. The feature point set is then fitted to a circle using the least squares method to calculate the initial state of the fitted circle's center coordinates. Fitted circle center coordinates under interference condition .

5. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: In step four, let i be the index of the extracted sub-pixel feature points (i=1, 2, 3...n), and use the center coordinates of the circles fitted by the initial image and the overlay image respectively. Using the mathematical origin, feature points in the Cartesian coordinate system in the initial image and the overlay image are respectively... Mapping to polar coordinates By aligning with equal-angle interpolation, the initial panoramic image and the overlay panoramic image are aligned at the same polar angle. Each has a corresponding polar diameter and ,in, and The calculation formulas are as follows: （1）； （2）； In the formula, Let x be the x-coordinate of the feature point in the rectangular coordinate system. The ordinate of the feature point in the rectangular coordinate system. To fit the x-coordinate of the circle's center in a rectangular coordinate system, To fit the ordinate of the circle's center in a rectangular coordinate system.

6. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: Step five, edge gradient intensity The radial gradient magnitude is characterized by performing first-order difference calculations on the pixels in the neighborhood of the feature point along the polar radial direction in the polar coordinate system established in step four, and obtaining the radial gradient magnitude obtained by taking the maximum value of the slope of the gray-scale change curve. Geometric Fitting Residual The center and radius of the circle based on the ideal circle parameters fitted in step three Calculate the polar radius of each sub-pixel feature point i in step four. With fitted radius Substitute into the formula It can be concluded that; Local contrast Taking the sub-pixel coordinates of feature point i as the center, a fixed-size local window (7x7 pixel area) is selected, and the highest grayscale value of all pixels within this window is counted. With the lowest gray value Substitute into the formula It can be concluded that; Calculate the confidence score of the initial and over-exposed image quality for each feature point, and the confidence score of the i-th feature point. The formula is as follows: （3）； In the formula, Let be the edge gradient intensity of the i-th feature point. Let be the geometric fitting residual of the i-th feature point. Let i be the local contrast of the i-th feature point. , , The coefficients are obtained through a genetic algorithm. First, multiple sets of bearing sample images with known actual deformation are acquired. The edge gradient intensity, fitting residual, and local contrast of each feature point are calculated, and the actual deformation is used as the true value. Then... The algorithm is encoded as a real number vector and an initial population is randomly generated. The root mean square error between the estimated deformation and the true value is used as the fitness function. The algorithm iterates through selection, crossover, and mutation genetic operations until the fitness converges. Finally, the optimal combination of coefficients that minimizes the error is output. The initial panoramic image and the overexposed panoramic image are at the same polar angle The paired feature points are associated, and the joint probability confidence of each feature point is calculated. Joint probability confidence The formula is as follows: （4）； In the formula, The initial image feature point confidence score for this feature point. The confidence level of the feature point in the over-extension image.

7. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: In step six, the local radial displacement at the i-th polar angle point is calculated. The formula is: （5）； In the formula, Let the polar radius be the interference pattern of the i-th feature point. The polar radius of the initial image of the i-th feature point; Using joint probability confidence The final radial deformation is obtained by weighted averaging of the deformation at all points around the circumference, and the deformation of the inner or outer diameter is then calculated. The formula is: （6）； In the formula, k is the image scale, used when calculating the inner ring of the bearing. When calculating the outer ring of a bearing, use .

8. The bearing deformation measurement method based on image recognition according to claim 1, characterized in that: Step seven, based on the bearing inner diameter deformation... and outer diameter deformation Calculate the corresponding change in the bottom of the inner raceway. Changes in the bottom of the outer raceway The formula is: （7）； （8）； In the formula, d is the inner diameter of the bearing, and D is the outer diameter of the bearing. The diameter of the inner raceway bottom. The diameter of the outer raceway bottom. The Poisson's ratio of the inner and outer ring materials of the bearing. This represents the deformation of the bearing's inner diameter. This represents the deformation of the bearing's outer diameter.

9. A bearing deformation measurement device based on image recognition, characterized in that: The system includes a base plate, a trapezoidal lead screw adjustment assembly, a measuring assembly, and a support assembly. The trapezoidal lead screw adjustment assembly includes a lead screw support plate fixed to the base plate. An EF support seat and an EK support seat are fixed to the lead screw support plate. A trapezoidal lead screw is rotatably connected between the EF support seat and the EK support seat. A handwheel is located at the top of the trapezoidal lead screw. The trapezoidal lead screw has a trapezoidal lead screw nut, which is fixed within a nut seat. A protrusion on one side of the nut seat is located within a sliding hole in the lead screw support plate. The measuring assembly includes an adapter plate. The adapter plate is fixed to the other side of the nut seat, and an industrial camera is installed at the lower end of the adapter plate; the support assembly includes an inner ring support fixed to the base plate and an outer ring support fixed to the inner ring support. The inner ring support is provided with a simulated shaft that is fitted with a conical surface, and the simulated shaft is provided with a bearing inner ring. The outer ring support is provided with a simulated shell that is fitted with a conical surface, and the simulated shell is provided with a bearing outer ring. The initial panoramic image and the panoramic image after interference fit of the bearing inner ring and bearing outer ring are acquired by the industrial camera to measure the bearing deformation.

10. The bearing deformation measuring device based on image recognition according to claim 9, characterized in that: The industrial camera is a CMOS camera, which is equipped with a telecentric lens and a parallel coaxial light source is fixed at the front end of the lens.

Images