Rope groove abrasion line structured light detection system and quantitative evaluation method

Abstract

The invention discloses a rope groove abrasion line structured light detection system and a quantitative evaluation method, and belongs to the technical field of rope groove abrasion detection. The rope groove abrasion line structure light detection system comprises a power and transmission unit, a supporting unit and a data acquisition unit. The power and transmission unit is used for driving the traction sheave to simulate rotation under actual working conditions, the supporting unit is used for realizing height adjustment and stable supporting of the detection platform, and the data acquisition unit is used for acquiring a linear structure light image on the surface of the traction sheave. The method comprises the following steps: selecting a detected traction wheel to rotate at a constant speed, projecting a laser beam on the surface of the traction wheel through a linear structured light sensor, and performing linear structured light image acquisition on the surface of the traction wheel by using an industrial CCD (Charge Coupled Device) camera; a light strip center is obtained in a mode of combining multi-modal fusion with a texture suppression algorithm, and a complete three-dimensional model is constructed and the rope groove depth is calculated based on a light strip center line. The method can be better matched with the actual working condition of traction sheave rope groove abrasion, and detection precision and high robustness are both considered.

Description

Technical Field

[0001] This invention relates to the field of rope groove wear detection technology, and more specifically, to a structured light detection system and quantitative evaluation method for rope groove wear lines. Background Technology

[0002] Traction sheaves are important load-bearing and transmission components, mainly used in many fields including robots, vehicles and home appliances. Their performance directly determines the safety, stability and service life of the machine.

[0003] The mating parts between the traction sheave, bushing, and wire rope require continuous lubrication to reduce friction. Current technology often uses grease lubrication, but this is susceptible to clumping or loss due to external dust and humidity, leading to lubrication failure. After prolonged exposure to the rated load of machine components, grease can cause wear on the rope grooves due to friction. These problems not only accelerate rope groove wear but also cause wire rope slippage, increasing safety hazards during machine operation.

[0004] Due to the confined operating environment of traction sheaves, when studying the dynamic characteristics of traction sheaves, establishing a simulation-based detection system can reduce experimental costs when measuring traction sheave wear and better match actual working conditions, thus facilitating the study of the dynamic characteristics of traction sheave wear. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to provide a structured optical detection system for rope groove wear lines and a quantitative evaluation method, so as to at least facilitate the study of the dynamic characteristics of traction wheel wear and ensure the accuracy and robustness of the detection.

[0006] To solve the above technical problems, according to one aspect of the present invention, a structured optical detection system for rope groove wear lines is provided, comprising a power and transmission unit, a support unit, and a data acquisition unit; The power and transmission unit includes a servo motor, a drive shaft, a bearing housing, and at least two traction wheels; one end of the drive shaft is directly or indirectly connected to the output shaft of the servo motor, and the other end is installed in the bearing housing; the traction wheels are selectively installed on the drive shaft, wherein the inner hole of at least one traction wheel is connected to the eccentric bushing through an interference fit, and the drive shaft is connected to the inner hole of the eccentric bushing through a clearance fit. The support unit includes a bracket, guide columns, a lifting drive device, and a lifting platform; the guide columns are located at the four corners of the bracket, the lifting platform is connected to the output end of the lifting drive device, and the four corners of the lifting platform are machined with motion holes that are adapted to the guide columns; the servo motor and bearing housing in the power and transmission unit are mounted on the surface of the lifting platform. The data acquisition unit includes a controller and a line structured light sensor, an industrial CCD camera, and an incremental encoder connected thereto. The line structured light sensor projects a laser beam onto the surface of the rotating traction wheel. The industrial CCD camera acquires line structured light images of the traction wheel surface. The incremental encoder is coaxially connected to the drive shaft and outputs a trigger signal when the traction wheel rotates a predetermined angle. The controller receives the trigger signal and synchronously triggers the line structured light sensor and the industrial CCD camera to acquire images.

[0007] In a preferred embodiment, the lifting drive device includes a geared motor and a worm gear screw jack; two worm gear screw jacks are symmetrically arranged on both sides of the support, and the power input end of each worm gear screw jack is connected to a common transmission shaft driven by a geared motor through a coupling.

[0008] In a preferred embodiment, each guide column has a linear motion track machined on its inner and outer sides, and the sidewall of the motion hole of the lifting platform has a spherical ball groove that fits with the linear motion track.

[0009] In a preferred embodiment, the servo motor driver adjusts the servo motor speed through a built-in speed loop algorithm. The feedback signal of the speed loop comes from an incremental encoder and is used to simulate different speed conditions of the traction sheave.

[0010] According to another aspect of the present invention, a method for quantitatively evaluating rope groove wear is provided, employing the above-described rope groove wear line structured light detection system, comprising: S1, System Calibration This includes coordinate transformation and optical plane calibration; S2, acquire line structured light image The selected traction sheave is rotated at a constant speed. A laser beam is projected onto the surface of the traction sheave using a line structured light sensor, and an industrial CCD camera is used to acquire line structured light images of the traction sheave surface. S3, Extract the center line of the light stripe S31: Extract features from the input line structured light image to obtain color feature map F1, depth feature map F2 and semantic feature map F3 respectively; S32: Weightedly fuse the color feature map F1, depth feature map F2, and semantic feature map F3 to obtain the fused feature map F. fused ; S33: For the fused feature map F fused After removing noise and suppressing texture interference, a redundant light stripe image I3 is obtained according to the majority voting rule; S34: Extract a single-pixel-wide laser stripe skeleton from the non-redundant light stripe image I3. S1 For laser stripe skeleton S 1 Weights are fused based on global and local attention weights, and a weight-optimized skeleton is output. S wegighted ; S35: Computational Optimization of the Skeleton S wegighted The second-order partial derivatives are used to construct... Hessian The matrix is ​​solved and the eigenvector corresponding to the smallest eigenvalue is obtained to determine the direction of the light stripe normal. A Gaussian function is fitted along the direction of the light stripe normal to obtain the sub-pixel center line of the single view. Then, the light stripe center line is output by fusing the multi-view majority voting rule. S4, Quantitative Evaluation Based on the centerline of the light stripe, a complete three-dimensional model is constructed and the depth of the rope groove is calculated.

[0011] In a preferred embodiment, in step S31, the input RGB image is converted to the HSI color space, and the HSI image is subjected to threshold screening pixel by pixel. Candidate pixels that meet the preset threshold conditions are assigned a value of 1, and the background is assigned a value of 0. After erosion and dilation optimization, a continuous and smooth color feature map is obtained. F 1 ; By combining the distance constraint between the laser emitter and the traction wheel, the range of light stripe depth values ​​is determined, and the world coordinates of the pixels are defined. X w , Y w , Z w Depth is determined, and pixels that meet the depth value range are marked as candidate pixels for light stripes and assigned a value of 1, while the rest are considered background and assigned a value of 0, thus generating a depth feature map. F 2 ; The RGB image is input into the trained U-Net model, which performs semantic segmentation of the light stripe based on a confidence threshold, generating a semantic feature map. F 3 .

[0012] In a preferred embodiment, the formula for weighted fusion in step S32 is: ; In the above formula, ( I, I () represents pixel coordinates; F fused ( I, I ) represents the value of the fused feature map at that pixel.

[0013] In a preferred embodiment, step S33 includes: fusing the feature mapsF fused A two-dimensional Fourier transform is performed, followed by noise removal using a Butterworth low-pass filter. Then, an improved Gabor filter is used to suppress texture interference. Finally, a redundant light stripe image is obtained based on the majority voting rule. I 3 , ;

[0014] In the above formula, K For the number of views, I 2 ( X,Y () is a depth feature map F 2 In pixel coordinates ( X,Y The value at ().

[0015] In a preferred embodiment, in step S34, the image without redundant light stripes is... I 3 Perform column and row projections, and determine the ROI region coordinates by taking the intersection of the projected ranges. X min ,X max ,Y min ,Y max ], for the coordinates within the ROI region I 3 Image pixels undergo 8-neighborhood analysis, based on the number of non-zero pixels. N ( P Excluding isolated points and internal pixels, the number of neighborhood jumps Z ( P To ensure pixels are located on continuous edges, two rounds of iterative selection are used to obtain a laser stripe skeleton with a single pixel width. S 1 .

[0016] In a preferred embodiment, in step S4, the sub-pixel coordinates of the light stripe centerline are extracted and the three-dimensional coordinates in the camera coordinate system are calculated; then, the three-dimensional coordinates are assembled into point cloud units according to the rotation angle interval of the traction wheel and translated; next, the particle swarm optimization algorithm is used to obtain the optimal rotation matrix through rotation, circle fitting and fitness evaluation; finally, based on the optimal rotation matrix, each point cloud unit is rotated and stacked to form a complete three-dimensional model.

[0017] The structured optical detection system for rope groove wear lines proposed in this invention can better match the actual working conditions of traction sheave rope groove wear, and can also better study the dynamic characteristics of traction sheave under different speed conditions. It reduces the restrictions on the position of measuring instruments and the measured object, and can effectively reduce the cost of the experimental stage for studying the dynamic characteristics of traction sheave wear.

[0018] The quantitative evaluation method proposed in this invention acquires line structured light images under different parameter conditions based on the designed rope groove wear line structured light detection system. The center of the light stripe is obtained by multimodal fusion combined with texture suppression algorithm, which takes into account both detection accuracy and strong robustness, and ensures the speed and integrity of 3D model reconstruction. Attached Figure Description

[0019] Figure 1 This is a schematic diagram of a structured optical detection system for rope groove wear lines; Figure 2 It is a structured light detection system for rope groove wear lines; Figure 3 These are line structured light images of the wear amount of the traction sheave rope groove collected at the experimental site. In the figure, 1 to 10 are 10 images of the wear amount of the traction sheave rope groove collected at the site. Figure 4 This is a comparison chart of the evaluation effect between the algorithm measurement values ​​and the actual values ​​obtained by the quantitative evaluation method of this invention. The horizontal axis represents 10 rope groove samples, and the vertical axis represents the rope groove wear amount. Figure 4 In the diagram, 'a' represents the global view, and 'b' represents a magnified view of point A within the global view.

[0020] In the diagram, 1-bracket, 2-gear motor, 3-worm gear screw jack, 4-lifting platform, 5-guide column, 6-servo motor, 7-shaft end retaining ring, 8-pair of wheels, 9-drive shaft, 10-traction sheave, 11-bearing housing, 12-incremental encoder, 13-L-shaped bracket, 14-linear motion track, 15-anti-rotation plate, 16-controller, 17-line structured light sensor, 18-industrial CCD camera. Detailed Implementation

[0021] The overall concept of this invention is to construct a structured light detection system for rope groove wear lines. Based on this system, structured light images are acquired, and after function optimization, pixel coordinates are obtained through coarse segmentation and sub-pixel center points are extracted. After coordinate transformation based on camera intrinsic parameters, a rotation matrix is ​​constructed using a particle swarm optimization (PSO) algorithm to achieve point cloud rotation stacking and high-precision 3D modeling.

[0022] Based on the above concept, a typical embodiment of the present invention provides a structured optical detection system for rope groove wear lines, including a power and transmission unit, a support unit, and a data acquisition unit.

[0023] 1. Power and transmission unit The power and transmission unit is used to drive the traction sheave to simulate the rotation under actual working conditions, such as... Figure 2As shown, the power and transmission unit includes a servo motor 6, a drive shaft 9, a bearing housing 11, and at least two traction wheels 10; one end of the drive shaft 9 is directly or indirectly connected to the output shaft of the servo motor 6, and the other end is installed in the bearing housing 11; the traction wheels 10 are selectively installed on the drive shaft 9, wherein the inner hole of at least one traction wheel 10 is connected to the eccentric bushing through an interference fit, and the drive shaft 9 is connected to the inner hole of the eccentric bushing through a clearance fit.

[0024] The output shaft of the servo motor 6 is machined with an A-type flat key and a threaded hole at its right end. The left half of the coupling wheel 8 is connected to the output shaft of the servo motor 6 via a flat keyway to achieve torque transmission. The right half of the coupling wheel 8 is machined with a concave keyway, and the left end of the drive shaft 9 is machined with a convex platform and a threaded hole. The coupling wheel 8 is slidably connected to the conventional shaft 9. The shaft end retaining rings 7 at both ends of the coupling wheel 8 are fixed to the threaded hole of the output shaft of the servo motor 6 at its right end and the threaded hole of the drive shaft 9 at its left end by fastening screws to prevent axial movement.

[0025] The eccentric bushing has a double-headed spiral oil groove on its outer diameter, and an annular oil storage groove is machined in the middle of its inner hole to store grease and reduce rotational friction torque. Both ends are chamfered to prevent dust from scratching the drive shaft. The inner hole of the traction sheave 10 is connected to the eccentric bushing via an interference fit to ensure connection stability. The drive shaft 9 is connected to the inner hole of the eccentric bushing via a clearance fit. Two D-shaped locking platforms are machined at the mating point between the drive shaft 9 and the eccentric bushing, and fixed with set screws to prevent axial movement.

[0026] During the test, a traction sheave 10 without an eccentric bushing was connected to the drive shaft 9 and the test was simulated. A second traction sheave 10 with an eccentric bushing was connected to the drive shaft 9 and the test was simulated again. The test simulated the wear of the traction sheave 10 under normal operation and the axial deflection of the traction sheave 10 due to wear.

[0027] Second, support unit The support unit is used to achieve height adjustment and stable support of the testing platform, such as... Figure 2 As shown, the support unit includes a bracket 1, a guide column 5, a lifting drive device, and a lifting platform 4; the guide column 5 is located at the four corners of the bracket 1, the lifting platform 4 is connected to the output end of the lifting drive device, and the four corners of the lifting platform 4 are machined with motion holes that are adapted to the guide column 5; the servo motor 6 and bearing seat 11 in the power and transmission unit are installed on the surface of the lifting platform 4.

[0028] like Figure 2 As shown, the bracket 1 is a convex cross-shaped rectangular bracket. Four vertical guide columns 5 with V-shaped cross sections are machined at the four corners of the bracket 1. A linear motion track 14 is machined on the inner and outer sides of each guide column 5. Nylon wear-resistant strips are set on the inner side of the linear motion track 14.

[0029] The lifting drive device includes a geared motor 2 and a worm gear screw jack 3; two worm gear screw jacks 3 are symmetrically arranged on both sides of the bracket 1, and the power input end of each worm gear screw jack 3 is connected to a common transmission shaft driven by the geared motor 2 through a coupling.

[0030] like Figure 2 As shown, a worm gear screw jack 3 is installed in the middle of the two guide columns 5 on one side. The worm gear screw jack 3 is vertically locked to the bracket 1 using cylindrical head screws to prevent axial displacement. The same device is also installed in the middle of the two guide columns 5 on the other side.

[0031] The power input shafts of the two worm gear screw jacks 3 are fixedly connected to one end of the corresponding couplings by interference fit, and the other end of the two couplings are connected by a common transmission shaft to ensure the coaxiality accuracy of the entire transmission shaft system.

[0032] The geared motor 2 is mounted on the bracket 1. The output shaft of the geared motor 2 and the power input shaft of the worm gear screw jack 3 are fixedly connected to the two ends of the corresponding coupling through interference fit. The coaxiality of the output shaft of the geared motor 2 and the power input shaft of the worm gear screw jack 3 is measured with a dial indicator.

[0033] The lifting platform 4 has a rectangular hollow structure. The four corners of the lifting platform 4 have V-shaped motion holes with spherical ball grooves, which are connected to the linear motion track 14 through clearance fit. The servo motor 6, bearing housing 11, and other devices are all locked in the corresponding positions of the lifting platform 4 with cylindrical head screws.

[0034] III. Data Acquisition Unit The data acquisition unit is used to acquire line structured light images of the surface of the traction wheel 10. The data acquisition unit includes a controller 16 and a line structured light sensor 17, an industrial CCD camera 18, and an incremental encoder 12 connected to it. The line structured light sensor 17 is used to project a laser beam onto the surface of the rotating traction wheel 10. The industrial CCD camera 18 is used to acquire line structured light images of the surface of the traction wheel 10. The incremental encoder 12 is coaxially connected to the drive shaft 9. The incremental encoder 12 is used to output a trigger signal when the traction wheel 10 rotates by a predetermined angle. The controller 16 receives the trigger signal and synchronously triggers the line structured light sensor 17 and the industrial CCD camera 18 to acquire images.

[0035] like Figure 2As shown, a threaded hole is machined at the lower end of the L-shaped bracket 13. Expansion bolts are used to securely fix the L-shaped bracket 13 to the lifting platform 4. A level is used to check the verticality and horizontality of the L-shaped bracket 13, and the shims are adjusted to ensure there is no tilt. One end of the anti-rotation plate 15 is inserted into the anti-rotation groove of the incremental encoder 12 housing, and the other end is bolted to the L-shaped bracket 13 to ensure that the incremental encoder 12 will not rotate with the shaft.

[0036] The anti-rotation plate 15 should be slightly pre-tightened on the encoder housing to avoid excessive tightness causing deformation of the incremental encoder 12 housing, or excessive looseness causing wobbling.

[0037] The end of the drive shaft 9 and the rotating shaft of the incremental encoder 12 are respectively machined with convex platforms. The two are connected by a coupling, and the coaxiality of the rotating shaft of the incremental encoder 12 and the drive shaft 9 is checked by a laser alignment instrument.

[0038] The line structured light sensor 17 and the industrial CCD camera 18 are respectively fixed on an independent adjustable bracket by precision fastening bolts. With the help of the micro-adjustment function of the independent adjustable bracket, the laser projection line can accurately cover the surface of the object being measured according to the detection requirements.

[0039] The controller 16 is installed in the electrical cabinet, and is connected to the line structured light sensor 17, the industrial CCD camera 18, and the incremental encoder 12 via cables. All cables are run through cable protection pipes and secured with cable ties to prevent pulling and bending.

[0040] The driver of servo motor 6 adjusts the speed of servo motor 6 through a built-in speed loop algorithm. The feedback signal of the speed loop of servo motor 6 comes from incremental encoder 12. It uses PID algorithm for control and updates the q-axis current command in each current loop cycle, so that the steady-state error and dynamic tracking error of the actual speed of servo motor 6 and the speed command sequence are controlled within the set range, thereby continuously and smoothly simulating different speed conditions of traction wheel 10.

[0041] Whenever the traction sheave 10 rotates 1°, it triggers the incremental encoder 12, and the controller 16 simultaneously triggers the line structured light sensor 17 and the industrial CCD camera 18 to ensure data synchronization.

[0042] According to this embodiment, during operation, the geared motor 2 is used as the power source to drive the worm gear screw jack 3 to move, so that the bracket 1 moves along the linear motion track 14 set on the guide column 5 to a suitable height.

[0043] The traction sheave 10 to be tested is mounted on the drive shaft 9. The servo motor 6 is used as the drive source. The speed is adjusted by the speed loop algorithm built into the servo motor 6 driver, so as to realize the working conditions of the traction sheave 10 at different speeds.

[0044] After the traction sheave 10 begins to rotate, the controller 16 synchronously triggers the line structured light sensor 17 and the industrial CCD camera 18. The laser beam emitted by the line structured light sensor 17 is precisely projected onto the surface of the rope groove of the traction sheave 10 to achieve full coverage. At the same time, the industrial CCD camera 18 synchronously starts the line structured light image acquisition operation.

[0045] Based on the structured optical detection system for rope groove wear lines described above, another typical embodiment of the present invention provides a quantitative detection method for rope groove wear.

[0046] This embodiment selects an elevator traction sheave from a building in Shaanxi Province as the object of on-site measurement. According to the general national standard and inspection rules for elevator traction machines, the rope groove depth of the traction sheave of the HSGL-160L permanent magnet synchronous gearless traction machine is 7mm. Among them, the U-shaped groove of the rope groove sample No. 9 was completely filled with oil. According to the general national standard and inspection rules for elevator traction machines, the standard depth of this type of U-shaped groove is 5mm.

[0047] The quantitative detection method for groove wear provided in this embodiment includes the following steps.

[0048] Step 1: Perform coordinate transformation. Transform the coordinates of a point on the traction sheave into the world coordinate system. X w , Y w , Z w Transform the coordinates to pixel coordinates using the transformation formula. you, you Its conversion formula is derived based on camera intrinsic parameters, extrinsic parameters and related physical parameters, thereby achieving a precise correspondence between spatial coordinates and pixel coordinates.

[0049] Step 2: Calibrate the optical plane. Turn on the laser emitter and illuminate the calibration plate. Acquire images using the industrial CCD camera 18. Then, change the orientation of the calibration plate and collect multiple sets of data. Fit the plane using the least squares method, ensuring the fit minimizes the sum of the squared distances from each point to the plane. Finally, derive the equation of the optical plane in the world coordinate system.

[0050] Step three: Based on the coordinate transformation method and optical plane calibration technology described in the above steps, and combined with the line structure optical detection rope groove wear device described in this invention, the traction wheel 10 to be detected is fixed on the transmission shaft 9 of the experimental platform at a constant speed. v The rotation begins, and the laser beam from the laser of the line structured light sensor 17 is projected onto the surface of the traction wheel 10. The industrial CCD camera 18 then captures a line structured light image of the traction wheel 10 surface. The acquisition result is as follows: Figure 3 As shown, this provides the raw image data for subsequent centerline extraction.

[0051] Step four: Convert the line structured light image obtained in step three from the RGB color image to the HSI color space. The conversion formula is as follows:

[0052] In the above formula, H, S, and I represent pixel data components in HSI space, respectively; R, G, and B represent pixel data components in RGB space, respectively.

[0053] The converted HSI image is thresholded pixel by pixel. Pixels that meet the preset threshold are marked as candidate pixels for light stripes and assigned a value of 1, while the remaining pixels are marked as background and assigned a value of 0. After optimization by erosion and dilation operations, a continuous and smooth color feature map is obtained. F 1 The light stripe and the background are initially separated from each other based on color.

[0054] Step 5: Combine the line structured light image obtained in Step 3 with the distance constraint between the laser emitter and the traction wheel 10 to determine the depth constraint corresponding to the light stripe, and then determine the world coordinates of the pixels. （X w , Y w , Z w ） Depth determination is performed, and pixels that meet the depth value constraints are marked as candidate pixels for light stripes and assigned a value of 1, while the remaining pixels are marked as background and assigned a value of 0, thus generating a depth feature map. F 2 Further filtering of light stripe pixels is performed from the spatial depth dimension.

[0055] Step six: Input the RGB image from the line structured light image obtained in step three into the trained U-Net model, and perform semantic segmentation of the light stripes based on the confidence threshold to generate a semantic feature map. F 3 .

[0056] Step 7: The color feature map obtained in the above steps... F 1 Depth feature map F 2 Semantic feature map F 3 Weighted summation is performed to fuse multimodal features and obtain a fused feature map. F fused .

[0057] ; In the above formula, ( I, I () represents pixel coordinates; F fused ( I, IThe values ​​α1, α2, and α3 of the fused feature map at this pixel represent the color feature map, respectively. F 1 Depth feature map F 2 Semantic feature map F 3 The weighting coefficients.

[0058] Step 8: Combine the fused feature maps obtained in the above steps. F fused A two-dimensional Fourier transform is performed, followed by noise removal using a Butterworth low-pass filter. Then, an improved Gabor filter is used to suppress texture interference. Finally, a redundant light stripe image is obtained based on the majority voting rule. I 3 The light stripe area is purified to eliminate interference for subsequent ROI extraction; ; It can also be expressed as:

[0059] In the above formula, K For the number of views, I 2 ( X,Y () is a depth feature map F 2 In pixel coordinates ( X,Y The value at ().

[0060] Step 9: Obtain the redundant light stripe image from the above steps. I 3 Perform column and row projections, and determine the coordinates of the ROI region by taking the intersection of the projected ranges. [X min ,X max ,Y min ,Y max ], for the coordinates within the ROI region I 3 Image pixels undergo 8-neighborhood analysis, based on the number of non-zero pixels. N ( P Excluding isolated points and internal pixels, the number of neighborhood jumps Z ( P To ensure pixels are located on continuous edges, two rounds of iterative selection are used to obtain a laser stripe skeleton with a single pixel width. S 1 .

[0061] Step 10: Develop the single-pixel-wide laser stripe skeleton obtained in the previous steps. S 1Perform global-local attention refinement, calculate the mean of global features and the mean of local neighborhood features, and obtain the global attention weights of the skeleton points. A global ( i, j ) and local attention weights A local ( I, I ), and perform weighted fusion. A final ( I, I Output weights to optimize the skeleton. S wegighted ; ; In the above formula, F att To fuse feature maps F fused Crop the feature map corresponding to its ROI region; F global for F att The feature mean of all candidate pixel values ​​for light stripes; K 1 、K 2 These are the coefficients representing the proportions of global attention and local attention, respectively. For feature dimensions.

[0062] Step 11: Optimize the skeleton using the weights obtained in the above steps. S wegighted Calculate the second-order partial derivatives of its image. I xx 、 I yy 、I xy Construct a Hessian matrix H, solve for the eigenvector corresponding to the smallest eigenvalue of the Hessian matrix H, determine the direction of the light stripe normal, and fit a Gaussian function along the direction of the light stripe normal. G(x) Output single-view subpixel center line The center line of the subpixel in the single view Multi-view centerline fusion is performed, and a majority voting rule is used to output the final light bar centerline. C final ( x,y ).

[0063]

[0064] Step 12, the center line of the light stripe obtained in the above steps C final ( x,y Extract the sub-pixel coordinates of each center line. you, you ), sub-pixel coordinates ( you, you Convert to normalized image plane coordinates ( X n ，Y n ， 1) Solve for the three-dimensional coordinates of the centerline point in the camera coordinate system using the light plane equation obtained in step two. X c ，Y c ，Z c Using the inverse derivation of the transformation formula derived from the camera intrinsic parameters, extrinsic parameters, and related physical parameters in step one, the three-dimensional coordinates in the camera coordinate system are transformed ( X c ，Y c ，Z c Transform to world coordinate system ( X w ，Y w ，Z w In this process, we ensure that each sub-pixel center point corresponds to a unique three-dimensional spatial location.

[0065] Step thirteen, convert the three-dimensional coordinates obtained in the above steps ( X c ，Y c ，Z c Based on the rotation angle interval of the traction sheave, the set of three-dimensional coordinates of the centerline at each angle is defined as a point cloud unit. PCL K ( K =1, 2...360, corresponding to 360 rotation angles respectively), select the first point cloud unit. PCL any point in 1 a n = ( X an ，Y an ，Z an ), calculate its translation vector T The 3D coordinates of all point cloud units are uniformly translated into the reconstructed coordinate system and defined as follows: ;

[0066] In the above formula, The nth after translation K Point cloud unit, T=-α n .

[0067] Step fourteen, within the rotation angle range Internal random generation M There are particles, and the position vector of each particle is... z i =[ β i ,γ i The initial velocity is v i = 0 Then, for each particle i According to its position vector z i Construct rotation matrix R ( i ); R ( i )= R x ( β i )• R y ( γ i ); In the above formula, R x ( β i ) for around X The rotation matrix of the axis. R y ( γ i ) for around Y The rotation matrix of the axis.

[0068] Step 15, convert the point cloud units obtained in Step 13 into... By rotation matrix R ( i Rotate the point cloud to obtain the rotated point cloud. Then, in each rotated point cloud unit of N Least-square circle fitting is performed on each cross section to obtain the radius sequence. And calculate the fitness function. ;

[0069] In the above formula, The standard radius of the rope groove.

[0070] Step sixteen: Record the individual optimal position of each particle. and global best position Through the speed update formula v k+1 and position update formula z k+1 Update particle positions while simultaneously decaying inertia weights. When the global optimal fitness value is reached... The iteration stops when ε is reached, where ε is a preset fitness threshold, and the optimal rotation angle is output. β opt and γ opt Thus, the optimal rotation matrix is ​​obtained. R opt Then, the optimal rotation matrix is ​​applied to each point cloud unit. R opt They are then rotated and stacked to form a complete three-dimensional model.

[0071]

[0072] In the above formula, w For inertial weights, c 1 、c 2 As a learning factor, v k For the first K At the next iteration, the particle's current velocity vector. z k For the first K During the nth iteration, the 1st i The current position of each particle. rand These are uniformly random numbers.

[0073] Based on the final 3D model of the traction sheave, the rope groove depth of each sample was calculated by analyzing the data at the defect points of the 3D point cloud model. The calculation results are shown in Table 1. To compare the accuracy of the algorithm, each of the 10 rope groove samples was measured manually five times, and the data obtained are recorded in Table 2.

[0074]

[0075]

[0076] To evaluate the accuracy of the algorithm's measurements, this study further calculated the relative error between the true values ​​and the algorithm's measurements. Comparative analysis revealed that the maximum relative error did not exceed 4.512%, a result that fully validates the effectiveness and reliability of the proposed quantitative evaluation method.

Claims

1. A rope groove wear line structured light detection system, characterized by, The power and transmission unit comprises a servo motor (6), a transmission shaft (9), a bearing seat (11) and at least two traction wheels (10); one end of the transmission shaft (9) is directly or indirectly connected with the output shaft of the servo motor (6), and the other end is installed in the bearing seat (11); the traction wheels (10) are selectively installed on the transmission shaft (9), and the inner hole of at least one of the traction wheels (10) is connected with the eccentric shaft sleeve through interference fit, and the transmission shaft (9) is connected with the inner hole of the eccentric shaft sleeve through clearance fit. The support unit comprises a support (1), guide columns (5), a lifting driving device and a lifting platform (4); the guide columns (5) are arranged at the four corners of the support (1), the lifting platform (4) is connected with the output end of the lifting driving device, and the four corners of the lifting platform (4) are processed with movement holes matched with the guide columns (5); the servo motor (6) and the bearing seat (11) in the power and transmission unit are installed on the surface of the lifting platform (4). The data acquisition unit comprises a controller (16) and a line structured light sensor (17), an industrial CCD camera (18) and an incremental encoder (12) connected with the controller (16); the line structured light sensor (17) is used for projecting a laser beam on the surface of the rotating traction wheel (10); the industrial CCD camera (18) is used for collecting a line structured light image on the surface of the traction wheel (10); the incremental encoder (12) is coaxially connected with the transmission shaft (9), and the incremental encoder (12) is used for outputting a trigger signal when the traction wheel (10) rotates by a predetermined angle, the controller (16) receives the trigger signal, and the line structured light sensor (17) and the industrial CCD camera (18) are synchronously triggered to collect images. The lifting driving device comprises a speed reducer (2) and a worm and gear screw rod elevator (3); two worm and gear screw rod elevators (3) are symmetrically arranged on the two sides of the support (1), and the power input ends of the worm and gear screw rod elevators (3) are respectively connected with a common transmission shaft driven by the speed reducer (2) through a shaft coupling.

2. The rope groove wear line structured light detection system of claim 1, wherein: The inner and outer sides of each guide column (5) are respectively processed with a linear motion track (14), and the side wall of the movement hole of the lifting platform (4) is processed with a ball-type ball groove matched with the linear motion track (14).

3. The groove wear line structured light detection system of claim 1 or 2, wherein: The driver of the servo motor (6) adjusts the rotating speed of the servo motor (6) through a built-in speed loop algorithm, and the feedback signal of the speed loop comes from the incremental encoder (12) and is used for simulating different speed conditions of the traction wheel (10).

4. The rope groove wear line structured light detection system of claim 3, wherein: The rope groove wear line structured light detection system comprises:

5. A method for quantitatively evaluating a rope groove wear, characterized by, S1, system calibration including coordinate conversion and light plane calibration; S2, acquiring a line structured light image selecting a detected traction wheel to start rotating at a constant speed, projecting a laser beam on the surface of the traction wheel (10) through the line structured light sensor (17), and collecting a line structured light image on the surface of the traction wheel (10) by using the industrial CCD camera (18); S3, extracting a light strip center line ​ S31: feature extraction is performed on the input line structured light image to obtain color feature map F1, depth feature map F2 and semantic feature map F3 respectively; S32: the color feature map F1, the depth feature map F2 and the semantic feature map F3 are weighted and fused to obtain a fused feature map F fused ; S33: fusing the feature maps F fused Remove noise and suppress texture interference, get the non-redundant light strip image I3 according to the majority voting rule; S34: Extracting single-pixel wide laser fringe skeleton from the non-redundant light strip image I3 S 1 , the laser fringe skeleton S 1 Weight fusion according to global and local attention weights, output weight optimized skeleton S wegighted ; S35: calculating the optimized skeleton S wegighted the second-order partial derivative of the function, constructing Hessian a matrix and solving the eigenvector corresponding to the minimum eigenvalue to determine the light bar normal direction; fitting a Gaussian function along the light bar normal direction to obtain a single-view sub-pixel center line, and then fusing through a multi-view majority voting rule to output the final light bar center line; S4, quantitative evaluation Based on the light bar center line, a complete three-dimensional model is constructed and the depth of the rope groove is calculated.

6. The method of quantitatively evaluating the wear of a rope groove according to claim 5, characterized by: In step S31, the input RGB image is converted to HSI color space, and threshold screening is performed on the HSI image pixel by pixel. The light bar candidate pixels satisfying the preset threshold condition are assigned a value of 1, and the background is assigned a value of 0. After corrosion and expansion optimization, a continuous and smooth color feature map is obtained F 1 ; Combining the distance constraint between the laser emitter and the traction wheel (10), the range of light stripe depth values ​​is determined, and the world coordinates of the pixels are determined. X w , Y w , Z w Depth is determined, and pixels that meet the depth value range are marked as candidate pixels for light stripes and assigned a value of 1, while the rest are considered background and assigned a value of 0, thus generating a depth feature map. F 2 ; The RGB image is input into the trained U-Net model, and light bar semantic segmentation is completed based on a confidence threshold to generate a semantic feature map F 3 .

7. The rope groove wear quantitative evaluation method according to claim 5 or 6, characterized by: In step S32, the formula of weighted fusion is: ； In the above formula, (x, y) is the pixel coordinate; i,j F fused i,j is the value of the fused feature map at the pixel.​​ 8. The method of quantitatively evaluating the wear of a rope groove according to claim 7, characterized by: Step S33 includes: The fusion feature map is obtained F fused A two-dimensional Fourier transform is performed, noise is removed in combination with a Butterworth low-pass filter, texture interference is suppressed through an improved Gabor filter, and finally a redundant light bar image is obtained based on a majority voting rule I 3 , ； In the above formula, K For the number of views, I 2 ( X,Y () is a depth feature map F 2 In pixel coordinates ( X,Y The value at ().

9. The method of quantitatively evaluating the wear of a rope groove according to claim 8, characterized by: In step S34, Image without redundant light stripes I 3 Perform column and row projections, and determine the ROI region coordinates by taking the intersection of the projected ranges. X min ,X max ,Y min ,Y max ], for the coordinates within the ROI region I 3 Image pixels undergo 8-neighborhood analysis, based on the number of non-zero pixels. N ( P Excluding isolated points and internal pixels, the number of neighborhood jumps Z ( P To ensure pixels are located on continuous edges, two rounds of iterative selection are used to obtain a laser stripe skeleton with a single pixel width. S 1 .

10. The rope groove wear quantitative evaluation method according to claim 5 or 9, characterized by: In step S4, the sub-pixel coordinates of the light bar center line are extracted and the three-dimensional coordinates in the camera coordinate system are calculated; then the three-dimensional coordinates are collected into point cloud units and translated according to the rotation angle interval of the traction wheel; then the particle swarm optimization algorithm is used to obtain the optimal rotation matrix through rotation, circle fitting and fitness evaluation; finally, based on the optimal rotation matrix, the point cloud units are rotated and stacked to form a complete three-dimensional model.

Images