Method for measuring sliding-rolling ratio of cylindrical roller bearing based on image registration
By using image registration technology on cylindrical roller bearings, the problems of low accuracy and low efficiency of traditional measurement methods are solved, and the accurate measurement of slip-roll ratio is achieved, data accuracy and reliability are ensured, and the operation status of the bearing is supported. Detailed analysis.
Patent Information
- Application Number
- CN202411030026.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-30
- Publication Date
- 2024-11-15
AI Technical Summary
The traditional cylindrical roller bearing slip ratio measurement method has low measurement accuracy, low efficiency and is susceptible to human factors, making it difficult to accurately measure the slip ratio of the bearing under actual working conditions.
Using an image registration method, speckle patterns are made on the inner ring of the bearing and the end surface of the roller, the camera collects images for feature point matching and angle transformation, calculate the slip-roll ratio, reduce human error and improve measurement accuracy.
It realizes accurate measurement of the sliding and roll ratio of cylindrical roller bearings, ensures data accuracy, provides high-resolution image recording, supports detailed analysis of the bearing operation process, and improves the accuracy and reliability of the sliding and roll ratio measurement.
Smart Images

Figure CN118962174A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of cylindrical roller bearings, in particular to a method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration. Background Art
[0002] The slip-to-roll ratio of a cylindrical roller bearing refers to the ratio of the relative sliding speed of the inner and outer rings of the bearing to the rolling speed of the roller. It is an important parameter for measuring the friction performance of the bearing. A slip-to-roll ratio that is too large or too small will cause increased wear, noise and temperature rise in the bearing, affecting the life and reliability of the bearing. Therefore, accurate measurement of the slip-to-roll ratio of the bearing is one of the keys to ensuring the performance and life of the bearing. The traditional method of measuring the slip-to-roll ratio of a bearing mainly relies on a combination of manual measurement and instrument measurement. This measurement method has the disadvantages of low measurement accuracy, low efficiency, and susceptibility to human factors.
[0003] The invention patent with publication number CN110118656B discloses a device and method for measuring the friction force and slide-roll ratio of a bearing steel ball and inner and outer rings. The device and method utilize piezoelectric ceramics, multi-dimensional force sensors and laser sensors to measure the friction force and slide-roll ratio of a steel ball in an annular ring under different loads and different rotation speeds. However, the device and method require manual intervention, including manual setting and adjustment of parameters, which increases the subjectivity of the measurement and may introduce human errors.
[0004] With the development of computer vision and digital image processing technology, the measurement method of the sliding-rolling ratio of bearings based on computer vision has gradually become a research hotspot. However, in actual working conditions, since the rollers of the bearings rotate and revolve simultaneously, it poses a considerable challenge to the current image processing technology. Therefore, it is an important research direction to study the measurement method of cylindrical roller bearing rotating components based on image registration and improve the accuracy of the calculation of the sliding-rolling ratio of cylindrical roller bearings. Summary of the invention
[0005] In order to solve the deficiencies in the prior art, the present invention provides a method for measuring the slide-to-roll ratio of a cylindrical roller bearing based on image registration, which can effectively reduce the influence of the rotation angle of the inner ring and roller of the cylindrical roller bearing and improve the measurement accuracy and reliability of the slide-to-roll ratio.
[0006] In order to achieve the above object, the specific scheme adopted by the present invention is: A method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration mainly comprises the following steps: S1, respectively making speckle patterns on the inner ring end face and any roller end face of the same end of the bearing, using a first camera to collect the inner ring speckle image and the roller speckle image, and recording the shooting time; using a second camera to track the number of rotations of the inner ring and the number of rotations of the roller; S2, selecting two images from the collected inner ring speckle images as inner ring images to be processed, the inner ring images to be processed comprising a randomly selected inner ring reference image and an inner ring image to be registered with a shooting time interval Δt1 from the inner ring reference image; selecting two images from the roller speckle images as roller images to be processed, the roller images to be processed comprising a randomly selected roller reference image and an inner ring image to be registered with a shooting time interval Δt2 from the roller reference image; S3, performing image matching on the inner ring image to be processed and the roller image to be processed respectively, to obtain a matching relationship between feature points in the inner ring reference image and the inner ring image to be registered, and between feature points in the roller reference image and the roller image to be registered; S4, performing image registration on the inner ring reference image and the inner ring image to be registered, and on the roller reference image and the roller image to be registered, respectively, according to the affine transformation matrix including the angle transformation value θ, and summing the angle transformation values in the image registration process, wherein the sum of the angle transformation values of the inner ring is recorded as the inner ring speckle image rotation angle Δθ1, and the sum of the angle transformation values of the roller is recorded as the roller speckle image rotation angle Δθ2; S5, respectively taking the shooting time of the inner ring reference image and the inner ring image to be registered as the time start point and time end point, selecting images of the time period Δt1 from the images taken by the second camera, performing feature point tracking, and measuring the number of inner ring rotations N1; respectively taking the shooting time of the roller reference image and the roller image to be registered as the time start point and time end point, selecting images of the time period Δt2 from the images taken by the second camera, performing feature point tracking, and measuring the number of roller rotations N2; S6. Compensate the inner ring rotation number N1 measured in step S5 according to the inner ring speckle image rotation angle Δθ1 obtained in step S4, and calculate the inner ring linear velocity v1; compensate the roller self-rotation number N2 measured in step S5 according to the roller speckle image rotation angle Δθ2 obtained in step S4, and calculate the roller linear velocity v2; finally, calculate the slip-roll ratio according to the inner ring linear velocity v1 and the roller linear velocity v2.
[0007] Further, in step S2, the position difference between the roller speckle in the roller image to be registered and the roller reference image is less than one tenth of the image width, and the position difference between the inner ring speckle in the inner ring image to be registered and the inner ring reference image is less than one tenth of the image width.
[0008] Furthermore, the shooting time difference between the inner ring reference image and the roller reference image is less than 0.02 s, and the shooting time difference between the inner ring image to be registered and the roller image to be registered is less than 0.02 s.
[0009] Further, in step S3, the inner ring image to be processed and the roller image to be processed are processed in the same way, wherein the specific method for processing the inner ring image to be processed is: S31, graying and filtering the inner circle reference image and the inner circle image to be registered, and then normalizing them; S32, according to the position and size of the speckle in the inner circle reference image, select the area containing the speckle in the image as the region of interest, and only perform feature extraction and matching on the image located in the region of interest in the inner circle reference image and the inner circle image to be registered; S33, use the SURF algorithm to detect the feature points in the region of interest in the inner circle reference image and the inner circle image to be registered, and extract the feature points with scale invariance; S34, assigning directions to the extracted feature points by using the DAISY algorithm to obtain feature points with rotation invariance; S35, for each feature point, calculate its position coordinates and response value in the image, adopt non-maximum suppression, and remove those feature points with similar positions and low response values according to the set threshold, and only retain the feature points with dispersed positions and high response values; S36, using FLANN matching for rough matching; S37, using the PROSAC algorithm to eliminate mismatched points in the rough matching process, and selecting a number of matched feature point pairs to generate a matching relationship between the feature points in the inner circle reference image and the inner circle image to be registered.
[0010] In step S3, the main direction of the feature point is selected by the SURF algorithm, and on this basis, the DAISY operator is used to replace the original SURF descriptor to obtain a DAISY descriptor with rotation invariance. This not only retains the fast calculation speed of the original SURF algorithm, but can also quickly obtain feature points with scale invariance, rotation invariance and noise resistance. At the same time, it has better positioning performance and strong robustness to illumination and affine transformations, thereby improving its correct matching rate for rotated images.
[0011] Further, in step S4, the method for performing image registration of the inner ring to be processed image and the roller to be processed image is the same, wherein the specific method for performing image registration of the inner ring reference image and the inner ring to be registered image is: S41, calculating and obtaining an affine transformation matrix according to the coordinates of the matched feature point pairs selected in step S37, and obtaining an angle transformation value θ; S42, performing affine transformation on the inner circle image to be registered according to the affine transformation matrix, transforming the inner circle image to be registered into the coordinate system of the inner circle reference image to obtain a transformed registration image, superimposing the transformed registration image with the inner circle reference image, and if the coincidence degree between the two meets the coincidence accuracy requirement, the angle transformation value calculated in step S41 is the rotation angle Δθ1 of the inner circle speckle image; S43, if the overlap after superposition does not meet the overlap precision requirement, the transformed registration image replaces the original inner circle image to be registered and returns to step S3 for iteration until the overlap between the transformed registration image and the reference image meets the overlap precision requirement.
[0012] Further, in step S4, the affine transformation matrix H is composed of the rotation matrix R and the translation matrix t, and the affine transformation matrix H is expressed as: Among them, θ is the angle transformation value, Δx is the translation amount in the x direction, and Δy is the translation amount in the y direction.
[0013] Furthermore, in step S6, the sliding-rolling ratio ξ is calculated as follows: Among them, v1 is the inner ring linear velocity, and v2 is the roller linear velocity.
[0014] Furthermore, in step S6, the inner circle linear velocity v1 is calculated as follows: Among them, r1 is the inner ring raceway radius, N1 is the number of inner ring rotations, and Δθ1 is the rotation angle of the inner ring speckle image; The calculation method of roller linear velocity v2 is: Among them, r2 is the roller radius, N2 is the number of roller rotations, and Δθ2 is the rotation angle of the roller speckle image.
[0015] Beneficial effects:
[0016] (1) The present invention provides a method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration. The method based on image registration is used to achieve accurate measurement of the sliding-rolling ratio of a cylindrical roller bearing, ensuring synchronous measurement of the inner ring and roller of the cylindrical roller bearing, thereby ensuring the accuracy of the data in the subsequent calculation process;
[0017] (2) Through machine vision measurement, high-resolution images of each measurement are saved, and historical data can be traced back at any time after the measurement is completed. Image data of different time periods during the operation of the bearing can be analyzed in detail. By recording the motion trajectory of feature points, the complex rolling and sliding behavior of the roller under operating conditions can be comprehensively and intuitively analyzed.
[0018] (3) In order to improve the accuracy of the slip-to-roll ratio measurement, the present invention also introduces a compensation mechanism for the rotation angle of the inner ring and roller of the cylindrical roller bearing, and corrects the actual rotation angle of the inner ring and roller in the original measurement. The introduction of this compensation and correction step enables the accurate calculation of the slip-to-roll ratio when the cylindrical roller bearing is in operation; it provides a reliable technical means for the bearing performance evaluation under actual working conditions, and provides strong support for research and engineering applications in related fields. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] Figure 1 Flow chart of the measuring method of the present invention.
[0020] Figure 2 This is the schematic diagram of the image acquisition device.
[0021] Figure 3 Flowchart of image matching in the present invention.
[0022] Graphic markings: 1. Inner ring, 2. Roller, 3. First camera, 4. Computer, 5. Second camera.
[0023] Figure 2 , box A represents the field of view of the first camera, and box B represents the field of view of the second camera. DETAILED DESCRIPTION
[0024] The technical solution of the present invention will be clearly and completely described below in conjunction with specific embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0025] Reference Figures 1 to 3 The present invention provides a method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration, comprising the following steps:
[0026] S1. A speckle pattern is produced on the end face of the inner ring 1 and the end face of any roller 2 at the same end of the bearing, and a first camera 3 is used to collect the inner ring speckle image and the roller speckle image, and the shooting time is recorded; and a second camera 5 is used to track the number of rotations of the inner ring and the number of rotations of the roller.
[0027] Since the inner ring 1 and the roller 2 end faces lack obvious features, a speckle pattern is made on the inner ring 1 and one of the roller 2 end faces to improve the matching accuracy. Figure 2 As shown, a real-time image acquisition device is installed to acquire complete images of inner ring speckle and roller speckle when the bearing is running, and the shooting time of each frame of the image is recorded; Figure 2Schematic diagram of the image acquisition device, the first camera 3 is used to acquire the speckle reference image and the image to be registered under the stable running condition of the inner ring 1 and the measured roller 2, and transmit them to the computer 4 for subsequent image processing, and the second camera 5 is used to track and measure the number of rotations of the inner ring and the number of rotations of the roller;
[0028] S2. Select two images from the collected inner ring speckle images as inner ring images to be processed, wherein the inner ring images to be processed include a randomly selected inner ring reference image and an inner ring image to be registered whose shooting time is Δt1 with the shooting time interval Δt1 between the inner ring reference image and the inner ring; select two images from the roller speckle images as roller images to be processed, wherein the roller images to be processed include a randomly selected roller reference image and an inner ring image to be registered whose shooting time is Δt2 with the shooting time interval Δt2 between the roller reference image and the inner ring.
[0029] Since all images are acquired by the first camera 3, the length and width of all images are consistent. In step S2, in order to reduce the impact of the roller revolution on subsequent calculations, the position difference between the roller speckle in the roller image to be registered and the roller reference image is less than one tenth of the image width, and the position difference between the inner ring speckle in the inner ring reference image and the inner ring reference image is less than one tenth of the image width. In order to reduce the impact caused by the change of working conditions, the shooting time of the selected inner ring reference image and the roller reference image may be similar, and the time difference should be less than 0.02s. The shooting time of the selected inner ring image to be registered and the roller image to be registered should be as close as possible, and the time difference should be less than 0.02s. It is most preferred that both the reference image and the image to be registered contain both the inner ring speckle and the measured roller speckle.
[0030] S3. Process the inner ring image to be processed and the roller image to be processed respectively to obtain a matching relationship between feature points in the inner ring reference image and the inner ring image to be registered, and between feature points in the roller reference image and the roller image to be registered.
[0031] The following is a detailed description of the method for processing the inner circle reference image and the inner circle image to be registered: S31, preprocessing the input inner circle reference image and the inner circle image to be registered; S32. According to the position and size of the speckle in the reference image, determine a suitable circular region as a region of interest, extract and match the feature points in the region, and the region of interest should be included in the roller end face and cover enough speckle patterns. In the present invention, the circular region of interest (ROI) should include most of the speckle in the image, exclude irrelevant information outside the speckle region, avoid erroneous feature point matching, and keep the area and circumference unchanged without being affected by image rotation. S33, respectively detecting feature points in the reference image and the image to be registered by using the SURF algorithm, and extracting stable feature points therefrom; S34, using the DAISY operator to assign directions to the extracted feature points, and obtaining feature points with rotation invariance; S35, for each feature point, calculating its position coordinates and response value in the image, using non-maximum suppression to eliminate those feature points with similar positions and low response values, and only retaining feature points with dispersed positions and high response values; S36, using FLANN matching to perform fast approximate nearest neighbor search, using Euclidean distance to measure the similarity of feature vectors, matching the two sets of descriptors to obtain a rough matching pair; S37. Use the PROSAC algorithm to eliminate mismatched points in the rough matching process, and generate a matching relationship between feature points in the inner circle reference image and the inner circle image to be registered.
[0032] Among them, in step S31, image preprocessing includes graying and filtering the image, and then normalizing it. Specifically, graying is to convert a color image into a gray image, reduce the dimension of the image, and reduce the amount of calculation; filtering is to use methods such as Gaussian filters or median filters to remove noise in the image, smooth the image, and enhance the edge and texture of the image. Normalization is to map the gray value of the image to a fixed range, such as [0,255], eliminate the brightness difference of the image, and improve the contrast of the image.
[0033] The SURF algorithm in step S33 is a feature extraction and matching algorithm, which aims to find robust and efficient feature points in an image. Its main steps are: 1. Key point detection: By using the Hessian matrix of a box filter at different scales in the image to detect key points, the box filter is used instead of the Gaussian filter for fast convolution to accelerate the calculation of key point detection; 2. Direction assignment: The Haar wavelet response is calculated for the image area near each key point to find the main direction, thereby improving the rotation invariance of SURF to the image; 3. Descriptor calculation: A region is constructed around the key point, divided into small sub-regions, and the Haar wavelet response is calculated for each sub-region to form a vector, and all vectors are connected to form a SURF descriptor.
[0034] The DAISY operator DAISY in step S34 is based on the gradient histogram. Each image area is described as the angle and amplitude of the gradient. This method allows dense sampling of features in the image and provides a rich description of the local structure of the image. The main steps are: 1. Construct a polar coordinate grid: Construct a polar coordinate grid divided into a series of sectors around the key points to capture local structural information; 2. Calculate the gradient histogram: Calculate the gradient histogram in each sector. The histogram contains the gradient amplitude in multiple directions, which is equivalent to describing the local texture direction; 3. Normalization: Perform L2 normalization on the histogram of each sector to improve the robustness to illumination changes, and connect all normalized histograms to form a DAISY descriptor.
[0035] In step S36, FLANN is a library for approximate nearest neighbor search, which is suitable for fast matching of large-scale data sets. In image processing, FLANN is often used to find similar feature points in feature space. Its main steps are: S361, indexing: constructing the feature vector into a FLANN index structure to accelerate the nearest neighbor search; S362, query nearest neighbor: for each feature vector to be matched, use Euclidean distance to measure the similarity of the feature vectors, and perform nearest neighbor search through FLANN index; S363, matching result: according to the result of the nearest neighbor search, a matching relationship between the feature vector in the image to be matched and the feature vector in the reference image is obtained to obtain a rough matching pair;
[0036] After the initial rough matching, there will still be a certain number of mismatched feature point pairs. In step S37, PROSAC is an improved version of the RANSAC algorithm. It gradually increases the number of samples and sorts them according to the contribution of each sample. It is used to fit the model in the presence of noise and outliers and estimate the transformation parameters in image registration. Its main steps are: S371, random sampling: randomly select 4 pairs of feature point pairs from the feature point set as samples, calculate the homography matrix and evaluate the quality of model fitting; S372, decision: calculate the projection errors between other matching pairs and the model, and decide whether to accept the current model parameters according to the quality of the model fitting; S373, update: if the error is less than the threshold, add the sample to the internal point set; S374, progressive sampling: gradually increase the number of samples k according to the quality and quantity of the internal points, and refit the model; S375, convergence judgment: when the number of internal points reaches the threshold or reaches the maximum number of iterations, the algorithm is stopped and the final model parameters are output.
[0037] S4, performing image registration on the inner ring reference image and the inner ring image to be registered, and on the roller reference image and the roller image to be registered, respectively, according to the affine transformation matrix including the angle transformation value θ, and summing the angle transformation values in the image registration process, wherein the sum of the angle transformation values of the inner ring is recorded as the inner ring speckle image rotation angle Δθ1, and the sum of the angle transformation values of the roller is recorded as the roller speckle image rotation angle Δθ2;
[0038] The affine transformation matrix H in step S4 is composed of the rotation matrix R and the translation matrix t. The affine transformation matrix H is expressed as:
[0039] Among them, θ is the angle transformation value, Δx is the translation in the x direction, and Δy is the translation in the y direction. Assuming A(x, y) and B(xˊ, yˊ) are a pair of feature points extracted from the match, the following transformation relationship exists:
[0040] The affine transformation has six degrees of freedom, and each pair of matching feature points provides two constraints. Therefore, at least three pairs of feature points are required to solve the parameters of the affine transformation. The selection of four pairs of feature points makes the algorithm more robust and over-determined while providing sufficient constraints.
[0041] The method for image registration of the inner ring image to be processed and the roller image to be processed is the same, wherein the specific method for image registration of the inner ring reference image and the inner ring image to be registered is: S41, calculating and obtaining an affine transformation matrix according to the coordinates of the matched feature point pairs selected in step S371, and obtaining an angle transformation value θ; S42, performing affine transformation on the inner circle image to be registered according to the affine transformation matrix, transforming the inner circle image to be registered into the coordinate system of the inner circle reference image to obtain a transformed registration image, superimposing the transformed registration image with the inner circle reference image, and if the coincidence degree between the two meets the coincidence accuracy requirement, the angle transformation value calculated in step S41 is the rotation angle Δθ1 of the inner circle speckle image; S43, if the overlap degree after superposition does not meet the overlap precision requirement, the transformed registered image replaces the original image to be registered and returns to step S3 for iteration until the overlap degree between the transformed registered image and the reference image meets the overlap precision requirement.
[0042] S5. Use the shooting time of the inner ring reference image and the image to be registered as the starting time and the ending time respectively, select the images of the time period Δt1 from the images taken by the second camera, perform feature point tracking, and measure the number of inner ring rotations N1; use the shooting time of the roller reference image and the image to be registered as the starting time and the ending time respectively, select the images of the time period Δt2 from the images taken by the second camera, perform feature point tracking, and measure the number of roller rotations N2.
[0043] Step S5 performs feature point tracking, i.e., performs continuous feature point registration of adjacent frames of the image captured by the second camera 5. By recording the motion trajectory of the feature points changing with time, the complex rolling and sliding motion state of the roller 2 under the operating condition is intuitively judged and the number of rotations of the roller 2 is measured. At the same time, the rolling and sliding motion information of the roller is more comprehensively analyzed by calculating the motion transformation matrix between adjacent frames.
[0044] S6. Compensate the inner ring rotation number N1 measured in step S5 according to the inner ring speckle image rotation angle Δθ1 obtained in step S4, and calculate the inner ring linear velocity v1; compensate the roller self-rotation number N2 measured in step S5 according to the roller speckle image rotation angle Δθ2 obtained in step S4, and calculate the roller linear velocity v2; finally, calculate the slip-roll ratio according to the inner ring linear velocity v1 and the roller linear velocity v2.
[0045] Specifically, substitute Δt1, Δt2, N1, N2, Δθ1 and Δθ2 into the slide-roll ratio calculation formula: Among them, v1 is the inner ring linear velocity, v2 is the roller linear velocity, r1 is the inner ring raceway radius, r2 is the roller radius, and ξ is the sliding-rolling ratio.
[0046] When the sliding-rolling ratio ξ=0, the bearing is in a pure rolling state; when the sliding-rolling ratio ξ=1, the bearing is in a pure sliding state; when the rolling-slip ratio 0<ξ<1, the bearing is in a rolling-sliding motion state.
[0047] The technical solution of the present invention is described in detail below in conjunction with specific embodiments.
[0048] Example 1
[0049] Taking NTN NU406 cylindrical roller bearing (inner ring raceway radius r1 is 22.5mm, roller radius r2 is 15mm) as an example for testing. A method for measuring the sliding-rolling ratio of cylindrical roller bearings based on image registration mainly includes the following steps: S1, respectively making speckle patterns on the inner ring end face and any roller end face of the same end of the bearing, collecting the inner ring speckle images and roller speckle images during a part of the bearing operation by a high-speed first camera 3, and recording the shooting time; using a second camera 5 to track the number of inner ring rotations and roller rotations; S2. Select two images from the collected inner ring speckle images as inner ring images to be processed, the inner ring images to be processed include a randomly selected inner ring reference image and an inner ring image to be registered with a shooting time interval Δt1=4.873s from the shooting time of the inner ring reference image; select two images from the roller speckle images as roller images to be processed, the roller images to be processed include a roller reference image and a roller image to be registered with a shooting time interval Δt2=4.886s from the shooting time of the roller reference image; S3. Initialize the SURF feature detector using the cv2.SURF_create function in the computer vision processing library OPENCV, extract feature points from the input image by calling detectAndCompute through the SURF feature detector, re-describe the feature points using the feature.daisy function to obtain the direction of the feature points, perform FLANN feature point matching on the inner circle image to be processed and the roller image to be processed, select the two most similar feature points in the image to be registered and the reference image, determine whether the feature vector distance ratio is less than a threshold of 0.7, and use the two feature points less than the threshold of 0.7 as feature point pairs between the reference image and the image to be registered; randomly select four pairs of feature point pairs, and use the findHomograph function of OPENCV to calculate the homography matrix of the two images, where the threshold of the projection error is set to 5.0, the number of iterations is 1000, and the matching relationship between the feature points in the inner circle reference image and the inner circle image to be registered, and between the feature points in the roller reference image and the roller image to be registered is obtained by iteratively calculating the optimal homography matrix; S4. Calculate the angle transformation value θ of the affine transformation matrix through the function np.arctan2, and call the perspective transformation function cv2.warpPerspective of OPENCV at the same time. According to the affine transformation matrix, transform the image after inner circle registration to the coordinate system of the inner circle reference image, and transform the image after roller registration to the coordinate system of the roller reference image to obtain the registered image, superimpose the transformed registered image with the reference image, return to step S3 for 3 iterations, and the overlap between the transformed registered image and the reference image meets the overlap accuracy requirement. Sum the angle transformation values obtained by iteration to obtain the rotation angle Δθ1=114.7° of the inner circle speckle image and the rotation angle Δθ2=84.6° of the roller speckle image; S5, respectively taking the shooting time of the inner ring reference image and the inner ring image to be registered as the time start point and time end point, selecting the image of the time period Δt1 from the images taken by the second camera 5, performing feature point tracking, and measuring the number of inner ring rotations N1=478; respectively taking the shooting time of the roller reference image and the roller image to be registered as the time start point and time end point, performing feature point tracking on the image of the time period Δt2 from the images taken by the second camera 5, and measuring the number of roller rotations N2=362; S6, according to the rotation angle obtained in step S4, the number of rotations of the inner ring and the number of rotations of the roller are compensated respectively, and the linear velocities v1 and v2 of the inner ring and the roller on the contact surface are calculated by substituting them into formulas (4) and (5);
[0050] Substituting the above results into formula (3), the sliding-rolling ratio of the bearing is solved:
[0051] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Any equivalent changes or modifications made according to the essence of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration, characterized in that: The main steps are as follows: S1. Create speckle patterns on the inner ring end face and any roller end face of the same end of the bearing, use a first camera to capture the inner ring speckle image and the roller speckle image, and record the shooting time; use a second camera to track the number of rotations of the inner ring and the number of rotations of the roller; S2, selecting two images from the collected inner ring speckle images as inner ring images to be processed, the inner ring images to be processed comprising a randomly selected inner ring reference image and an inner ring image to be registered with a shooting time interval Δt1 from the inner ring reference image; selecting two images from the roller speckle images as roller images to be processed, the roller images to be processed comprising a randomly selected roller reference image and an inner ring image to be registered with a shooting time interval Δt2 from the roller reference image; S3, performing image matching on the inner ring image to be processed and the roller image to be processed respectively, to obtain a matching relationship between feature points in the inner ring reference image and the inner ring image to be registered, and between feature points in the roller reference image and the roller image to be registered; S4, performing image registration on the inner ring reference image and the inner ring image to be registered, and on the roller reference image and the roller image to be registered, respectively, according to the affine transformation matrix including the angle transformation value θ, and summing the angle transformation values in the image registration process, wherein the sum of the angle transformation values of the inner ring is recorded as the inner ring speckle image rotation angle Δθ1, and the sum of the angle transformation values of the roller is recorded as the roller speckle image rotation angle Δθ2; S5, respectively taking the shooting time of the inner ring reference image and the inner ring image to be registered as the time start point and time end point, selecting images of the time period Δt1 from the images taken by the second camera, performing feature point tracking, and measuring the number of inner ring rotations N1; respectively taking the shooting time of the roller reference image and the roller image to be registered as the time start point and time end point, selecting images of the time period Δt2 from the images taken by the second camera, performing feature point tracking, and measuring the number of roller rotations N2; S6. Compensate the inner ring rotation number N1 measured in step S5 according to the inner ring speckle image rotation angle Δθ1 obtained in step S4, and calculate the inner ring linear velocity v1; compensate the roller self-rotation number N2 measured in step S5 according to the roller speckle image rotation angle Δθ2 obtained in step S4, and calculate the roller linear velocity v2; finally, calculate the slip-roll ratio according to the inner ring linear velocity v1 and the roller linear velocity v2.
2. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 1, characterized in that: In step S2, the position difference between the roller speckle in the roller image to be registered and the roller reference image is less than one tenth of the image width, and the position difference between the inner ring speckle in the inner ring image to be registered and the inner ring reference image is less than one tenth of the image width.
3. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 1, characterized in that: The shooting time difference between the inner ring reference image and the roller reference image is less than 0.02s, and the shooting time difference between the inner ring image to be registered and the roller image to be registered is less than 0.02s.
4. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 1, characterized in that: In step S3, the inner ring image to be processed and the roller image to be processed are processed in the same way, wherein the specific method for processing the inner ring image to be processed is: S31, graying and filtering the inner circle reference image and the inner circle image to be registered, and then normalizing them; S32, according to the position and size of the speckle in the inner circle reference image, select the area containing the speckle in the image as the region of interest, and perform feature extraction and matching only on the image located in the region of interest in the inner circle reference image and the inner circle image to be registered; S33, using the SURF algorithm to respectively detect feature points in the region of interest in the inner circle reference image and the inner circle image to be registered, and extracting feature points with scale invariance; S34, assigning directions to the extracted feature points by using the DAISY algorithm to obtain feature points with rotation invariance; S35, for each feature point, calculate its position coordinates and response value in the image, adopt non-maximum suppression, and remove those feature points with similar positions and low response values according to the set threshold, and only retain the feature points with dispersed positions and high response values; S36, using FLANN matching for rough matching; S37, using the PROSAC algorithm to eliminate mismatched points in the rough matching process, and selecting a number of matched feature point pairs to generate a matching relationship between the feature points in the inner circle reference image and the inner circle image to be registered.
5. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 4, characterized in that: In step S4, the method for image registration of the inner ring to-be-processed image and the roller to-be-processed image is the same, wherein the specific method for image registration of the inner ring reference image and the inner ring to-be-registered image is: S41, calculating and obtaining an affine transformation matrix according to the coordinates of the matched feature point pairs selected in step S37, and obtaining an angle transformation value θ; S42, performing affine transformation on the inner circle image to be registered according to the affine transformation matrix, transforming the inner circle image to be registered into the coordinate system of the inner circle reference image to obtain a transformed registration image, superimposing the transformed registration image with the inner circle reference image, and if the coincidence degree between the two meets the coincidence accuracy requirement, the angle transformation value calculated in step S41 is the rotation angle Δθ1 of the inner circle speckle image; S43, if the overlap after superposition does not meet the overlap precision requirement, the transformed registration image replaces the original inner circle image to be registered and returns to step S3 for iteration until the overlap between the transformed registration image and the reference image meets the overlap precision requirement.
6. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 1, characterized in that: In step S4, the affine transformation matrix H is composed of the rotation matrix R and the translation matrix t. The affine transformation matrix H is expressed as: Among them, θ is the angle transformation value, Δx is the translation amount in the x direction, and Δy is the translation amount in the y direction.
7. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 1, characterized in that: In step S6, the sliding-rolling ratio ξ is calculated as follows: Among them, v1 is the inner ring linear velocity, and v2 is the roller linear velocity.
8. The method for measuring the sliding-rolling ratio of a cylindrical roller bearing based on image registration according to claim 7, characterized in that: In step S6, the inner circle linear velocity v1 is calculated as follows: Among them, r1 is the inner ring raceway radius, N1 is the number of inner ring rotations, Δθ1 is the rotation angle of the inner ring speckle image; the calculation method of the roller linear velocity v2 is: Among them, r2 is the roller radius, N2 is the number of roller rotations, and Δθ2 is the rotation angle of the roller speckle image.
