Automatic ski maintenance control system based on computer vision

Abstract

The invention belongs to the technical field of snowboards, and particularly relates to a snowboard automatic maintenance control system based on computer vision, which comprises a hand-eye calibration module used for carrying out hand-eye calibration of a depth camera and a collaborative robot, including eye outside the hand and eye inside the hand, so as to obtain a space conversion relationship between the camera and a mechanical arm, therefore, point clouds collected by the depth camera are converted to a mechanical arm base coordinate system for unified representation, and subsequent mechanical arm path planning is facilitated. And the snowboard fixing and initial control module is used for firstly placing the snowboard and then fixing the snowboard. According to the invention, precise modeling and path planning are carried out through a computer vision technology, hardware cooperation such as a mechanical arm and a PLC is combined, and automation and standardization of ski maintenance can be realized. Compared with manual maintenance, the efficiency is greatly improved, and the maintenance time of a single snowboard is shortened.

Description

Technical Field

[0001] The present invention relates to the technical field of skis, and in particular to a ski automatic maintenance control system based on computer vision. Background Art

[0002] With the increasing popularity of skiing, the frequency of skiboard use continues to increase, and the need for maintenance is becoming increasingly prominent. Traditional skiboard maintenance relies on manual operations, which are subject to problems such as low efficiency, poor precision, and high operator experience requirements. For example, when manually trimming the edge, different operators rely on feel and experience to control the force and trajectory, making it difficult to ensure consistent trimming angle and smoothness for every ski. During the waxing process, the uniformity of the wax application is also affected by the operator's technique, resulting in unstable ski performance, affecting the skiing experience, and even shortening the ski's lifespan due to improper maintenance.

[0003] Based on the above, the current problems in ski maintenance are as follows:

[0004] 1. Low manual maintenance efficiency: Manual ski maintenance is a tedious process involving multiple steps, including fixing the skis, trimming the edges, dewaxing, waxing, ironing, and polishing. Maintaining a single ski often takes a long time. During peak hours at ski resorts, for example, a large number of skis await maintenance, making manual operations difficult to quickly process. This leads to delayed ski supply and impacts ski resort operational efficiency.

[0005] 2. Poor maintenance consistency: Manual maintenance relies on experience, and different personnel, and even the same person performing different operations, can have significant variations in their control of key parameters such as edge angle and wax thickness. For example, during edge trimming, it's difficult for humans to precisely control the edge angle within a very tight tolerance, resulting in varying skiing stability. Uneven waxing can also lead to inconsistent friction on the ski surface, affecting smooth gliding.

[0006] 3. Adaptability to complex working conditions: There are many different types of snowboards (single and double), each with its own distinct structure and maintenance requirements. Double skis are narrower, requiring the board surface to be adapted to the centerline path; single skis also have their own unique size and shape. Furthermore, after long-term use, snowboards develop complex surface wear and stain distribution. Severe scratches and stubborn stains, for example, are difficult to address with targeted manual maintenance.

[0007] In summary, a computer vision-based automatic ski maintenance control system is invented. Summary of the Invention

[0008] To solve the above technical problems, according to one aspect of the present invention, the present invention provides the following technical solutions:

[0009] A computer vision-based automatic ski maintenance control system, comprising:

[0010] The hand-eye calibration module is used to perform hand-eye calibration of the depth camera and the collaborative robot, including eye-outside-hand and eye-inside-hand calibration, to obtain the spatial transformation relationship between the camera and the robotic arm, thereby converting the point cloud collected by the depth camera into the robotic arm base coordinate system (world coordinate system) for unified representation, which facilitates subsequent robotic arm path planning;

[0011] The ski fixing and initial control module is used to first place the skis and then fix the skis;

[0012] Image acquisition and processing module, used to first control the depth camera to acquire images and then process the images;

[0013] The snowboard model building and recognition module is used to first generate the point cloud, then segment the point cloud, and then identify the type of snowboard;

[0014] Maintenance trajectory planning module, used to plan different maintenance trajectories based on ski type and actuator function;

[0015] Maintenance execution module, used to execute maintenance plans sequentially through preset processes;

[0016] Finishing module for removing the skis after maintenance.

[0017] As a preferred solution of the computer vision-based automatic ski maintenance control system described in the present invention, in the hand-eye calibration module, if the eye is outside the hand, the eye is directly calibrated. If the eye is in the hand, Then you will get the robot arm Afterwards, it will pass Calculate the spatial transformation relationship

[0018] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the steps of the ski fixing and initial control module are as follows:

[0019] S21: Place the skis upright on the ski bracket to provide a stable foundation for maintenance operations;

[0020] S22: The host computer controls the PLC through the TCP protocol to open the bracket suction cup solenoid valve, and use the suction cup to firmly fix the skis to prevent the skis from moving during maintenance and affecting the accuracy.

[0021] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the steps of the image acquisition and processing module are as follows:

[0022] S31, Image Acquisition: The host computer controls the depth camera to acquire RGB-D images, synchronously obtains the current RGB image and the corresponding depth image, and comprehensively records the appearance and spatial information of the skis;

[0023] S32, image processing: performing D2C alignment, hole filling filtering, truncation filtering, denoising and cropping on the image.

[0024] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the image processing steps are as follows:

[0025] S321, D2C alignment: aligns depth images and RGB images to ensure accurate correspondence between color and depth information, providing a consistent data foundation for subsequent processing;

[0026] S322, hole filling filter: fills the depth loss caused by reflection and occlusion in the depth image, repairs the image integrity, and in the case of reflective areas on the surface of the skis, supplements the missing depth value to ensure the accuracy of model construction;

[0027] S323, truncation filtering: based on the preset height from the camera to the ski plane, filter and retain the depth values ​​within the ski height range, remove background interference, and focus on the effective area of ​​the ski;

[0028] S324, Denoising: Removes noise points from the depth map, optimizes image quality, and prevents noise from affecting subsequent point cloud construction and analysis, such as eliminating false depth data caused by ambient light fluctuations;

[0029] S325, cropping: By presetting the working range mask, only the depth and image information of the working area are retained (without changing the resolution), narrowing the data processing range and improving computing efficiency, such as retaining only the data of the necessary operating area in the skis and surrounding areas.

[0030] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the steps of the ski model construction and recognition module are as follows:

[0031] S41, generate point cloud: using the camera internal parameters (factory parameters), convert the processed RGB-D image into point cloud data, presenting the snowboard shape in the form of three-dimensional space points, providing intuitive data for subsequent model analysis;

[0032] S42, point cloud segmentation: Use the DBSCAN clustering algorithm (search radius 0.02m, minimum number of cluster points 1000) to find the largest connected component and segment the snowboard point cloud. You can also combine deep learning methods to accurately extract the snowboard point cloud, eliminate interference from the bracket and background, and focus on the snowboard data itself.

[0033] S43, Identify ski type: Calculate the Y-axis range in the point cloud data. If it is greater than 0.15m, it is determined to be a single ski, and if it is less than 0.15m, it is determined to be a double ski. This provides a basis for subsequent differentiated maintenance path planning and adapts to the structural requirements of different skis.

[0034] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the maintenance trajectory planning module is specifically as follows:

[0035] S51, trimming path (edge ​​and bottom edge):

[0036] S511: Divide the X-axis into N segments according to a preset segment spacing of 0.05 m. Use the maximum and minimum Y coordinate points in each segment as the upper and lower edges, respectively, to obtain two sets of edge points [x, y, z] containing N points.

[0037] S512: Splice the preset actuator angles [rx, ry, rz] to form a total of 2N edge maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0038] S513: Using the spatial transformation relationship obtained by calibration Get the edge maintenance trajectory in the robotic arm coordinate system Guide the robotic arm to precisely trim the blade;

[0039] S52, waxing path (center line):

[0040] S521: Divide the Y-axis range Δ into two equal parts, recording it as Δ1 / 2;

[0041] S522: Calculate the bisector of Y, which is Y min +Δ 1 / 2 ;

[0042] S523: Sample at a preset interval of 0.1m to obtain N path points (if they do not exist, find the nearest real point nearby) [x, y, z]

[0043] S524: Splice the preset actuator angles [rx, ry, rz] to form a total of N maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0044] S525: Using the spatial transformation relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system Ensure waxing uniformity;

[0045] S53, wax removal, ironing, and polishing path (three equal lines):

[0046] S531: Divide the Y-axis range Δ into three equal parts, recording it as Δ1 / 3;

[0047] S532: Calculate the three equal lines of Y, which are Y min +Δ 1 / 3 and Y max -Δ 1 / 3 ;

[0048] S533: Each line is sampled at a preset interval of 0.1m, and each path point is obtained (if it does not exist, the nearest real point is found nearby) [x, y, z];

[0049] S534, stitching the preset actuator angles [rx, ry, rz] to form a total of 2N maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0050] S535, using the space conversion relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system To adapt to the corresponding maintenance procedures.

[0051] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the steps of the maintenance execution module are as follows (taking wax removal as an example):

[0052] S61: The host computer controls the robot arm to reach the preset position of the current actuator placement to prepare for actuator switching;

[0053] S62: Control the PLC to open the quick-change switch to release the current actuator to facilitate the replacement of the wax removal actuator;

[0054] S63: The robotic arm moves to the wax removal actuator switching point, and the host computer controls the PLC to close the quick-change switch, completing the actuator switching to ensure that the maintenance tool adapts to the process requirements;

[0055] S64: The robotic arm moves to the preset work starting point and prepares to perform the wax removal operation;

[0056] S65: According to the preset wax removal path and actuator angle, the robot arm moves and performs constant force control (generally maintaining a constant pressure of 10N) to accurately complete the wax removal operation. The trajectory and force are adjusted in real time through force sensor feedback to ensure uniform and thorough wax removal.

[0057] S66: After the wax removal task is completed, the robotic arm returns to the starting point and performs the bottom edge trimming, side edge trimming, waxing, ironing wax, and polishing processes in sequence until all maintenance processes are completed.

[0058] As a preferred solution of the computer vision-based automatic ski maintenance control system of the present invention, the steps of the finishing module are as follows:

[0059] S71: The host computer controls the PLC to close the ski bracket suction cup and release the ski.

[0060] S72: Remove the skis and complete the maintenance process.

[0061] Compared with existing technologies:

[0062] Through precise modeling and path planning using computer vision technology, combined with collaborative hardware such as robotic arms and PLCs, the system automates and standardizes ski maintenance. Compared to manual maintenance, this significantly improves efficiency and shortens the time required to maintain a single ski. Force feedback and precise trajectory control ensure consistent maintenance accuracy and enhance ski performance stability. Adapting to different ski types and complex surface conditions enhances system versatility and adaptability, the system provides efficient, high-quality ski maintenance solutions for ski resorts, ski repair shops, and other sectors, helping to upgrade equipment operations and maintenance in the ski industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the framework of the present invention. DETAILED DESCRIPTION

[0064] To make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0065] The present invention provides a computer vision-based automatic ski maintenance control system. Figure 1 ,include:

[0066] The hand-eye calibration module is used to calibrate the hand-eye of the depth camera and the collaborative robot, including the eye outside the hand and the eye inside the hand, to obtain the spatial transformation relationship between the camera and the robotic arm, so as to convert the point cloud collected by the depth camera into the robotic arm base coordinate system (world coordinate system) for unified representation, so as to facilitate the subsequent robotic arm path planning; among them, if the eye is outside the hand, it is directly calibrated. If the eye is in the hand, Then you will get the robot arm Afterwards, it will pass Calculate the spatial transformation relationship

[0067] The ski fixing and initial control module is used to first place the skis and then fix the skis;

[0068] Image acquisition and processing module, used to first control the depth camera to acquire images and then process the images;

[0069] The snowboard model building and recognition module is used to first generate the point cloud, then segment the point cloud, and then identify the type of snowboard;

[0070] Maintenance trajectory planning module, used to plan different maintenance trajectories based on ski type and actuator function;

[0071] Maintenance execution module, used to execute maintenance plans sequentially through preset processes;

[0072] Finishing module for removing the skis after maintenance.

[0073] The steps of the ski fixing and initial control module are as follows:

[0074] S21: Place the skis upright on the ski bracket to provide a stable foundation for maintenance operations;

[0075] S22: The host computer controls the PLC through the TCP protocol to open the bracket suction cup solenoid valve, and use the suction cup to firmly fix the skis to prevent the skis from moving during maintenance and affecting the accuracy.

[0076] The steps of the image acquisition and processing module are as follows:

[0077] S31, Image Acquisition: The host computer controls the depth camera to acquire RGB-D images, synchronously obtains the current RGB image and the corresponding depth image, and comprehensively records the appearance and spatial information of the skis;

[0078] S32, image processing: performing D2C alignment, hole filling filtering, truncation filtering, denoising and cropping on the image;

[0079] The steps of image processing are as follows:

[0080] S321, D2C alignment: aligns depth images and RGB images to ensure accurate correspondence between color and depth information, providing a consistent data foundation for subsequent processing;

[0081] S322, hole filling filter: fills the depth loss caused by reflection and occlusion in the depth image, repairs the image integrity, and in the case of reflective areas on the surface of the skis, supplements the missing depth value to ensure the accuracy of model construction;

[0082] S323, truncation filtering: based on the preset height from the camera to the ski plane, filter and retain the depth values ​​within the ski height range, remove background interference, and focus on the effective area of ​​the ski;

[0083] S324, Denoising: Removes noise points from the depth map, optimizes image quality, and prevents noise from affecting subsequent point cloud construction and analysis, such as eliminating false depth data caused by ambient light fluctuations;

[0084] S325, cropping: By presetting the working range mask, only the depth and image information of the working area are retained (without changing the resolution), narrowing the data processing range and improving computing efficiency, such as retaining only the data of the necessary operating area in the skis and surrounding areas.

[0085] The steps of the ski model construction and recognition module are as follows:

[0086] S41, generate point cloud: using the camera internal parameters (factory parameters), convert the processed RGB-D image into point cloud data, presenting the snowboard shape in the form of three-dimensional space points, providing intuitive data for subsequent model analysis;

[0087] S42, point cloud segmentation: Use the DBSCAN clustering algorithm (search radius 0.02m, minimum number of cluster points 1000) to find the largest connected component and segment the snowboard point cloud. You can also combine deep learning methods to accurately extract the snowboard point cloud, eliminate interference from the bracket and background, and focus on the snowboard data itself.

[0088] S43, Identify ski type: Calculate the Y-axis range in the point cloud data. If it is greater than 0.15m, it is determined to be a single ski, and if it is less than 0.15m, it is determined to be a double ski. This provides a basis for subsequent differentiated maintenance path planning and adapts to the structural requirements of different skis.

[0089] The maintenance trajectory planning module is as follows:

[0090] S51, trimming path (edge ​​and bottom edge):

[0091] S511: Divide the X-axis into N segments according to a preset segment spacing of 0.05 m. Use the maximum and minimum Y coordinate points in each segment as the upper and lower edges, respectively, to obtain two sets of edge points [x, y, z] containing N points.

[0092] S512: Splice the preset actuator angles [rx, ry, rz] to form a total of 2N edge maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0093] S513: Using the spatial transformation relationship obtained by calibration Get the edge maintenance trajectory in the robotic arm coordinate system Guide the robotic arm to precisely trim the blade;

[0094] S52, waxing path (center line):

[0095] S521: Divide the Y-axis range Δ into two equal parts, recording it as Δ1 / 2;

[0096] S522: Calculate the bisector of Y, which is Y min +Δ 1 / 2 ;

[0097] S523: Sample at a preset interval of 0.1m to obtain N path points (if they do not exist, find the nearest real point nearby) [x, y, z]

[0098] S524: Splice the preset actuator angles [rx, ry, rz] to form a total of N maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0099] S525: Using the spatial transformation relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system Ensure waxing uniformity;

[0100] S53, wax removal, ironing, and polishing path (three equal lines):

[0101] S531: Divide the Y-axis range Δ into three equal parts, recording it as Δ1 / 3;

[0102] S532: Calculate the three equal lines of Y, which are Y min +Δ 1 / 3 and Y max -Δ 1 / 3 ;

[0103] S533: Each line is sampled at a preset interval of 0.1m, and each path point is obtained (if it does not exist, the nearest real point is found nearby) [x, y, z];

[0104] S534, stitching the preset actuator angles [rx, ry, rz] to form a total of 2N maintenance trajectory points (with posture) P in the camera coordinate system cam =[x,y,z,rx,ry,rz];

[0105] S535, using the space conversion relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system To adapt to the corresponding maintenance procedures.

[0106] The steps of the maintenance execution module are as follows (taking wax removal as an example):

[0107] S61: The host computer controls the robot arm to reach the preset position of the current actuator placement to prepare for actuator switching;

[0108] S62: Control the PLC to open the quick-change switch to release the current actuator to facilitate the replacement of the wax removal actuator;

[0109] S63: The robotic arm moves to the wax removal actuator switching point, and the host computer controls the PLC to close the quick-change switch, completing the actuator switching to ensure that the maintenance tool adapts to the process requirements;

[0110] S64: The robotic arm moves to the preset work starting point and prepares to perform the wax removal operation;

[0111] S65: According to the preset wax removal path and actuator angle, the robot arm moves and performs constant force control (generally maintaining a constant pressure of 10N) to accurately complete the wax removal operation. The trajectory and force are adjusted in real time through force sensor feedback to ensure uniform and thorough wax removal.

[0112] S66: After the wax removal task is completed, the robotic arm returns to the starting point and performs the bottom edge trimming, side edge trimming, waxing, ironing wax, and polishing processes in sequence until all maintenance processes are completed.

[0113] The steps of the closing module are as follows:

[0114] S71: The host computer controls the PLC to close the ski bracket suction cup and release the ski.

[0115] S72: Remove the skis and complete the maintenance process.

[0116] When used, the specific steps are as follows:

[0117] S1: The hand-eye calibration module is used to calibrate the depth camera and the collaborative robot's hand-eye, including the eye outside the hand and the eye inside the hand, to obtain the spatial transformation relationship between the camera and the robotic arm, so as to convert the point cloud collected by the depth camera into the robotic arm base coordinate system (world coordinate system) for unified representation, so as to facilitate the subsequent robotic arm path planning; among them, if the eye is inside the hand, it is Then you will get the robot arm Afterwards, it will pass Calculate the spatial transformation relationship

[0118] S2: The ski fixing and initial control module first places the ski upright on the ski fixing bracket to provide a stable foundation for maintenance operations. Then, the host computer controls the PLC through the TCP protocol to open the bracket suction cup solenoid valve, using the suction cup to firmly fix the ski to prevent the ski from moving during maintenance and affecting accuracy.

[0119] S3: The image acquisition and processing module first controls the depth camera to capture RGB-D images, synchronously obtaining the current RGB image and the corresponding depth image, and comprehensively recording the appearance and spatial information of the skis. The image is then subjected to D2C alignment, hole filling filtering, truncation filtering, denoising, and cropping.

[0120] S4: The ski model construction and recognition module first uses the camera's internal parameters (factory parameters) to convert the processed RGB-D image into point cloud data, presenting the ski shape in the form of three-dimensional spatial points, providing intuitive data for subsequent model analysis; then, the DBSCAN clustering algorithm (search radius 0.02m, minimum number of cluster points 1000) is used to find the largest connected component and segment the ski point cloud; deep learning methods can also be combined to accurately extract the ski point cloud, eliminating interference from brackets, background, etc., and focusing on the ski body data; then, the Y-axis range of the point cloud data is calculated, and a range greater than 0.15m is determined to be a single ski, and a range less than 0.15m is determined to be a double ski, providing a basis for subsequent differentiated maintenance path planning to adapt to the structural requirements of different skis;

[0121] S5: The maintenance trajectory planning module plans different maintenance trajectories based on the type of ski and actuator function, including the edge trimming path (edge ​​and bottom edge), waxing path (centerline), and wax removal, ironing, and polishing paths;

[0122] S6: Execute the maintenance plan in sequence using the preset process through the maintenance execution module;

[0123] Take wax removal as an example:

[0124] S61: The host computer controls the robot arm to reach the preset position of the current actuator placement to prepare for actuator switching;

[0125] S62: Control the PLC to open the quick-change switch to release the current actuator to facilitate the replacement of the wax removal actuator;

[0126] S63: The robotic arm moves to the wax removal actuator switching point, and the host computer controls the PLC to close the quick-change switch, completing the actuator switching to ensure that the maintenance tool adapts to the process requirements;

[0127] S64: The robotic arm moves to the preset work starting point and prepares to perform the wax removal operation;

[0128] S65: According to the preset wax removal path and actuator angle, the robot arm moves and performs constant force control (generally maintaining a constant pressure of 10N) to accurately complete the wax removal operation. The trajectory and force are adjusted in real time through force sensor feedback to ensure uniform and thorough wax removal.

[0129] S66: After the wax removal task is completed, the robotic arm returns to the starting point and performs the bottom edge trimming, side edge trimming, waxing, ironing, and polishing processes in sequence until all maintenance processes are completed;

[0130] S7: Through the finishing module, the host computer controls the PLC to close the ski bracket suction cup to release the ski fixation; then remove the ski, completing the entire maintenance process.

[0131] Although the present invention has been described above with reference to embodiments, various modifications may be made thereto and equivalent components may be substituted without departing from the scope of the present invention. In particular, as long as there are no structural conflicts, the various features of the embodiments disclosed herein may be combined with each other in any manner, and the omission of an exhaustive description of such combinations in this specification is solely for the sake of space and resource conservation. Therefore, the present invention is not limited to the specific embodiments disclosed herein, but includes all technical solutions falling within the scope of the claims.

Claims

1. A computer vision-based automatic ski maintenance control system, characterized in that: include: The hand-eye calibration module is used to perform hand-eye calibration of the depth camera and the collaborative robot, including eye-outside-hand and eye-inside-hand calibration, to obtain the spatial transformation relationship between the camera and the robotic arm. This allows the point cloud collected by the depth camera to be converted to the robotic arm base coordinate system for unified representation, facilitating subsequent robotic arm path planning. The ski fixing and initial control module is used to first place the skis and then fix the skis; Image acquisition and processing module, used to first control the depth camera to acquire images and then process the images; The snowboard model building and recognition module is used to first generate the point cloud, then segment the point cloud, and then identify the type of snowboard; Maintenance trajectory planning module, used to plan different maintenance trajectories based on ski type and actuator function; Maintenance execution module, used to execute maintenance plans sequentially through preset processes; Finishing module for removing the skis after maintenance.

2. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: In the hand-eye calibration module, if the eye is outside the hand, it is directly calibrated. If the eye is in the hand, Then you will get the robot arm Afterwards it will pass Calculate the spatial transformation relationship 3. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of the ski fixing and initial control module are as follows: S21: Place the skis upright on the ski bracket to provide a stable foundation for maintenance operations; S22: The host computer controls the PLC through the TCP protocol to open the bracket suction cup solenoid valve, and use the suction cup to firmly fix the skis to prevent the skis from moving during maintenance and affecting the accuracy.

4. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of the image acquisition and processing module are as follows: S31, Image Acquisition: The host computer controls the depth camera to acquire RGB-D images, synchronously obtains the current RGB image and the corresponding depth image, and comprehensively records the appearance and spatial information of the skis; S32, image processing: performing D2C alignment, hole filling filtering, truncation filtering, denoising and cropping on the image.

5. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of image processing are as follows: S321, D2C alignment: aligns depth images and RGB images to ensure accurate correspondence between color and depth information, providing a consistent data foundation for subsequent processing; S322, hole filling filter: fills the depth loss caused by reflection and occlusion in the depth image, repairs the image integrity, supplements the missing depth value, and ensures the accuracy of model construction; S323, truncation filtering: based on the preset height from the camera to the ski plane, filter and retain the depth values ​​within the ski height range, remove background interference, and focus on the effective area of ​​the ski; S324, Denoising: Remove noise points in the depth map to optimize image quality and prevent noise from affecting subsequent point cloud construction and analysis; S325, cropping: By presetting the working range mask, only the depth and image information of the working area are retained, which reduces the data processing range and improves the computing efficiency.

6. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of the ski model construction and recognition module are as follows: S41, generate point cloud: using camera intrinsic parameters, convert the processed RGB-D image into point cloud data, presenting the snowboard shape in the form of three-dimensional space points, providing intuitive data for subsequent model analysis; S42, segmentation point cloud: using the DBSCAN clustering algorithm to find the maximum connected component and segment the ski point cloud; S43, Identify ski type: Calculate the Y-axis range in the point cloud data. If it is greater than 0.15m, it is determined to be a single ski, and if it is less than 0.15m, it is determined to be a double ski. This provides a basis for subsequent differentiated maintenance path planning and adapts to the structural requirements of different skis.

7. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The maintenance trajectory planning module is as follows: S51, edge trimming path: S511: Divide the X-axis into N segments according to a preset segment spacing of 0.05 m. Use the maximum and minimum Y coordinate points in each segment as the upper and lower edges, respectively, to obtain two sets of edge points [x, y, z] containing N points. S512: Splice the preset actuator angles [rx, ry, rz] to form a total of 2N edge maintenance trajectory points P in the camera coordinate system cam =[x,y,z,rx,ry,rz]; S513: Using the spatial transformation relationship obtained by calibration Get the edge maintenance trajectory in the robotic arm coordinate system Guide the robotic arm to precisely trim the blade; S52, waxing path: S521: Divide the Y-axis range Δ into two equal parts, recording it as Δ1 / 2; S522: Calculate the bisector of Y, which is Y min +Δ 1 / 2 ; S523: Sample at a preset interval of 0.1m to obtain N path points [x, y, z] S524: Splice the preset actuator angles [rx, ry, rz] to form a total of N maintenance trajectory points P in the camera coordinate system cam =[x,y,z,rx,ry,rz]; S525: Using the spatial transformation relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system Ensure waxing uniformity; S53, wax removal, ironing, polishing path: S531: Divide the Y-axis range Δ into three equal parts, recording it as Δ1 / 3; S532: Calculate the three equal lines of Y, which are Y min +Δ 1 / 3 and Y max -Δ 1 / 3 ; S533: Each line is sampled at a preset interval of 0.1m to obtain N path points [x, y, z]; S534, stitching the preset actuator angles [rx, ry, rz] to form a total of 2N maintenance trajectory points = [x, y, z, rx, ry, rz] in the camera coordinate system; S535, using the space conversion relationship obtained by calibration Get the maintenance trajectory in the robotic arm coordinate system To adapt to the corresponding maintenance procedures.

8. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of the maintenance execution module are as follows: S61: The host computer controls the robot arm to reach the preset position of the current actuator placement to prepare for actuator switching; S62: Control the PLC to open the quick-change switch to release the current actuator to facilitate the replacement of the wax removal actuator; S63: The robotic arm moves to the wax removal actuator switching point, and the host computer controls the PLC to close the quick-change switch, completing the actuator switching to ensure that the maintenance tool adapts to the process requirements; S64: The robotic arm moves to the preset work starting point and prepares to perform the wax removal operation; S65: The robot arm moves and performs constant force control according to the preset wax removal path and actuator angle to accurately complete the wax removal operation. The trajectory and force are adjusted in real time through force sensor feedback to ensure uniform and thorough wax removal. S66: After the wax removal task is completed, the robotic arm returns to the starting point and performs the bottom edge trimming, side edge trimming, waxing, ironing wax, and polishing processes in sequence until all maintenance processes are completed.

9. The computer vision-based automatic ski maintenance control system according to claim 1, characterized in that: The steps of the closing module are as follows: S71: The host computer controls the PLC to close the ski bracket suction cup and release the ski. S72: Remove the skis and complete the maintenance process.

Images