Automatic defect grinding device and method based on machine vision

Abstract

The invention relates to the technical field of defect polishing, in particular to an automatic defect polishing device based on machine vision and a polishing method.The automatic defect polishing device comprises a working platform, a material conveying assembly is arranged on the working platform, a portal frame is movably arranged on the working platform, and a polishing system capable of moving transversely and vertically is arranged on the portal frame; and a high-speed camera is mounted on the grinding system. According to the method, accurate positioning of different defects of the metal plate and accurate calculation of the grinding size can be achieved, and the problems that after the defects are ground, the sizes and shapes are irregular, and later unified repairing treatment is inconvenient are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of design defect polishing, and specifically to an automatic defect polishing device and polishing method based on machine vision. Background Art

[0002] Common metal sheets often have defects such as cracks and bubbles during production, processing and service. The existence of these defects will seriously reduce the service life of the metal sheet, and may cause huge economic losses in serious cases. Repair is needed to reduce costs and increase utilization.

[0003] Laser cladding is also called laser cladding or laser melting. This technology mainly places various functional materials on the surface of the substrate of the remanufactured parts by pre-powder feeding and synchronous powder feeding. After irradiation with a high-energy-density laser beam, the alloy powder and a thin layer on the substrate surface are melted and solidified quickly, forming a laser cladding layer with a very low dilution rate and metallurgical bonding with the substrate surface, thereby achieving the purpose of repair. While meeting the specific performance requirements of the material surface, it can save a lot of material costs. It is widely used in the repair of worn parts and the preparation of anti-corrosion coatings in aerospace, marine and offshore industries. Before using this technology for repair, the defects need to be polished in a regular manner to facilitate later repair.

[0004] In the prior art, defect polishing is mostly done manually, where defects are manually found and polished using tools such as grinding wheels. The size and shape of the defects after polishing are irregular, which is not convenient for unified repair in the later stage. Summary of the invention

[0005] The purpose of the present invention is to provide an automatic defect grinding device and grinding method based on machine vision to solve the above-mentioned problems, so as to achieve accurate positioning of different defects of metal plates and accurate calculation of grinding dimensions, and overcome the problem that the defects are irregular in size and shape after grinding, which is not convenient for unified repair and treatment later.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A defect automatic grinding device based on machine vision comprises a working platform, a material conveying component is arranged on the working platform, a gantry is movably arranged on the working platform, a grinding system which can move laterally and vertically is arranged on the gantry, and a high-speed camera is installed on the grinding system.

[0008] Preferably, the material conveying assembly includes a plurality of conveying rollers rotatably connected to the working platform, one end of the conveying rollers is axially connected to the output shaft of a conveying motor, and the conveying motor is fixedly connected to the working platform.

[0009] Preferably, two side surfaces of the working platform are fixedly connected with ground rails, the gantry is slidably connected to the two ground rails, two sides of the working platform are rotatably connected with gantry screws, the gantry is threadedly connected to the gantry screws, one end of the gantry screw is axially connected to a gantry motor, and the gantry motor is installed on the working platform.

[0010] Preferably, a transverse lead screw is rotatably connected to one side of the gantry, a transverse slide rail is fixedly connected to the gantry, a transverse slider is slidably connected to the transverse slide rail, a fixed plate is fixedly connected to one side of the transverse slider, the fixed plate is threadedly connected to the transverse lead screw, one end of the transverse lead screw is axially connected to a transverse transmission motor, and the transverse transmission motor is fixedly connected to the gantry.

[0011] Preferably, a vertical slide rail is slidably connected to the side of the horizontal slide rail away from the fixed plate, and the vertical slide rail is fixedly connected to the grinding bracket. A vertical lead screw is rotatably connected to the grinding bracket, and the vertical lead screw is threadedly connected to the fixed plate. One end of the vertical lead screw is axially connected to the output shaft of a transmission motor, and the transmission motor is installed on the grinding bracket.

[0012] Preferably, the grinding system comprises a grinding motor, the grinding motor is fixedly connected to the grinding bracket, the grinding motor is located on a side away from the fixed plate, and the output shaft of the grinding motor is connected to a grinding wheel.

[0013] An automatic grinding method The method is implemented based on the device of the above scheme, comprising the following steps:

[0014] The image of the defect on the plate is obtained by a high-speed camera, and the image is recognized based on machine vision to accurately locate the defect and obtain defect size information;

[0015] Obtaining defect position information of the to-be-polished portion according to the defect size information;

[0016] Based on the position information of the defect at the location to be polished, the polishing robot is moved to a suitable position, and based on the defect size information and the position information of the defect at the location to be polished, the defect is polished.

[0017] Preferably, obtaining an image of defects on the plate comprises the following steps:

[0018] Image acquisition, obtaining an image on the plate by means of the high-speed camera;

[0019] Image preprocessing, preprocessing the image obtained on the plate;

[0020] Defect location: determine the center coordinates and boundary coordinates of the defect based on the edge detection image processing algorithm;

[0021] The grinding system is controlled to grind the defects on the plate according to the pre-processed image and defect position.

[0022] Preferably, the images collected in the image collection step include: at least one of cracks, bubbles, slag inclusions, pits, undercuts, shrinkage cavities, and weld bumps;

[0023] The image preprocessing step includes:

[0024] Image preprocessing: reduce image noise through Gaussian filtering denoising algorithm; enhance noise interference contrast through single histogram equalization algorithm;

[0025] Feature extraction, extracting edge information in the image through the Canny edge detection algorithm, wherein the edge is the contour of the image collected in the image collection step;

[0026] Defect recognition uses a machine algorithm to identify the type of image captured in the image acquisition step.

[0027] Preferably, the defect identification step includes the following steps:

[0028] Using the YOLOv11 network structure, the YOLOv11 network is mainly composed of three parts: the backbone network (Backbone) is responsible for preliminary deep feature extraction; the neck (Neck) focuses on multi-scale feature fusion, and the detection head (Head) is used to output prediction results related to target detection; the backbone consists of an initial convolution layer of 7×7 and a stride of 2, 4 groups of 3 3×3 convolution layer modules and 3 2×2 maximum pooling layers, in which each convolution layer is equipped with a LeakyReLU activation function and a normalization layer. As the level deepens, the number of channels changes to 64, 128, 256, and 512; Neck uses an architecture similar to the feature pyramid network (FPN) to fuse features, splices high and low layer features through 2 bilinear interpolation upsampling, and integrates them through 3 1×1 convolution layers. At the same time, an improved module based on spatial pyramid pooling (SPP) was introduced, with three pooling kernels of 1×1, 5×5, and 9×9 processed in parallel, and the results were concatenated without changing the size of the feature map; the head uses three parallel prediction branches corresponding to tasks of different scales, starting with a 1×1 convolution layer, followed by two 3×3 convolution layers to refine the features. Finally, the branch outputs are integrated, regularized with a 1×1 convolution layer, and the confidence is limited to the 0-1 interval for screening through the Sigmoid function.

[0029] The present invention has the following technical effects:

[0030] 1. The present invention uses a visual device installed on the gantry bracket to identify the features of the position to be polished, and then uses an algorithm to identify and analyze the defect features to obtain the plane size information of the area to be polished. The system automatically adjusts the polishing size according to the defect size to perform accurate and targeted polishing of defects of different sizes. Finally, the robot uses the defect position identified by the visual device to drive the polishing device to complete the polishing work of several positions to be polished in sequence;

[0031] 2. The present invention connects the gantry and the grinding system by arranging four sliders on both sides of the fixed plate. The grinding system can be moved up and down and left and right through the motors arranged on the grinding bracket and the upper end of the gantry. The gantry can also move on the horizontal plane, so that the defects can be accurately positioned and ground. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0033] Figure 1 A schematic diagram of overall assembly distribution according to an embodiment of the present invention;

[0034] Figure 2 It is a schematic diagram of the assembly of a grinding system according to an embodiment of the present invention;

[0035] Figure 3 A schematic diagram of the connection between a grinding system and a gantry system according to an embodiment of the present invention;

[0036] Figure 4 It is a schematic diagram of the plane dimensions of the defective area to be polished according to an embodiment of the present invention.

[0037] Among them, 1. conveyor roller; 2. working platform; 3. steel plate; 4. high-speed camera; 5. grinding system; 6. conveyor roller; 7. conveyor motor; 8. ground rail; 9. gantry; 10. gantry screw; 11. grinding motor; 12. grinding wheel; 13. grinding bracket; 14. transmission motor; 15. vertical screw; 16. vertical slide rail; 17. horizontal slider; 18. slider; 19. horizontal slide rail; 20. horizontal screw; 21. horizontal transmission motor; 22. fixed plate. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] Reference Figures 1 to 4 As shown, this embodiment provides

[0041] A defect automatic grinding device based on machine vision includes a working platform 2, a material conveying component is provided on the working platform 2, a gantry 9 is movably provided on the working platform 2, a grinding system 5 which can move horizontally and vertically is provided on the gantry 9, and a high-speed camera 4 is installed on the grinding system 5.

[0042] According to a further optimization scheme, the material conveying assembly includes a plurality of conveying rollers 6 rotatably connected to the working platform 2 , one end of the conveying rollers 6 is axially connected to the output shaft of a conveying motor 7 , and the conveying motor 7 is fixedly connected to the working platform 2 .

[0043] A further optimized solution is that the two sides of the working platform 2 are fixedly connected with ground rails 8, the gantry 9 is slidably connected to the two ground rails 8, the two sides of the working platform 2 are rotatably connected with the gantry screws 10, the gantry 9 is threadedly connected to the gantry screws 10, one end of the gantry screw 10 is axially connected to the gantry motor, and the gantry motor is installed on the working platform 2.

[0044] A further optimized solution is that a transverse lead screw 20 is rotatably connected to one side of the gantry 9, a transverse slide rail 19 is fixedly connected to the gantry 9, a transverse slider 17 is slidably connected to the transverse slide rail 19, a fixed plate 22 is fixedly connected to one side of the transverse slider 17, the fixed plate 22 is threadedly connected to the transverse lead screw 20, a transverse transmission motor 21 is axially connected to one end of the transverse lead screw 20, and the transverse transmission motor 21 is fixedly connected to the gantry 9.

[0045] A further optimized solution is that a vertical slide rail 16 is slidably connected to one side of the horizontal slide rail 19 away from the fixed plate 22, and the vertical slide rail 16 is fixedly connected to the grinding bracket 13. A vertical lead screw 15 is rotatably connected to the grinding bracket 13, and the vertical lead screw 15 is threadedly connected to the fixed plate 17. One end of the vertical lead screw 15 is axially connected to the output shaft of the transmission motor 14, and the transmission motor 14 is installed on the grinding bracket 13.

[0046] According to a further optimization scheme, the grinding system 5 includes a grinding motor 11 , which is fixedly connected to a grinding bracket 13 . The grinding motor 11 is located on a side away from the fixed plate 22 , and an output shaft of the grinding motor 11 is connected to a grinding wheel 12 .

[0047] An automatic grinding method, which is implemented based on the above device, comprises the following steps:

[0048] The image of the defect on the plate is acquired by a high-speed camera, and the image is recognized based on machine vision to accurately locate the defect and obtain defect size information; the camera has a high frame rate and high resolution, and can clearly capture the image of the plate surface, especially on fast-moving or dynamic production lines.

[0049] Obtaining defect position information of the to-be-polished portion according to the defect size information;

[0050] Based on the position information of the defect at the location to be polished, the polishing robot is moved to a suitable position, and based on the defect size information and the position information of the defect at the location to be polished, the defect is polished.

[0051] Further optimizing the scheme, obtaining the image of the defect on the plate includes the following steps:

[0052] Image acquisition, obtaining an image on the plate through the high-speed camera 4;

[0053] Image preprocessing, preprocessing the image obtained on the plate;

[0054] Defect location: According to the edge detection image processing algorithm, the center coordinates and boundary coordinates of the defect are determined, providing accurate positioning basis for subsequent grinding operations.

[0055] The grinding system 5 is controlled to grind the defects on the plate according to the pre-processed image and defect position.

[0056] In a further optimization scheme, the image collected in the image collection step includes at least one of: cracks, bubbles, slag inclusions, pits, undercuts, shrinkage cavities, and weld nodules;

[0057] The image preprocessing step includes:

[0058] Image preprocessing, using Gaussian filtering denoising algorithm to reduce image noise; due to factors such as uneven lighting and camera noise in the actual production environment, there may be noise in the image. The Gaussian filtering denoising algorithm can reduce image noise, improve image quality, and provide a clear image basis for subsequent processing; the single histogram equalization algorithm can enhance the noise interference contrast and make the defect features more obvious, which is convenient for subsequent feature extraction and recognition.

[0059] Feature extraction, through the Canny edge detection algorithm, extract edge information in the image, the edge is the outline of the image collected in the image acquisition step; the edge is an important manifestation of defect features, such as the edge of a crack, the outline of a bubble, etc. Through edge detection, the approximate position and shape of the defect can be preliminarily located.

[0060] Defect recognition uses a machine algorithm to identify the type of image captured in the image acquisition step.

[0061] Further optimizing the solution, the defect identification step includes the following steps:

[0062] Using the YOLOv11 network structure, the YOLOv11 network is mainly composed of three parts: the backbone network (Backbone) is responsible for preliminary deep feature extraction; the neck (Neck) focuses on multi-scale feature fusion, and the detection head (Head) is used to output prediction results related to target detection; the backbone consists of an initial convolution layer of 7×7 and a stride of 2, 4 groups of 3 3×3 convolution layer modules and 3 2×2 maximum pooling layers, in which each convolution layer is equipped with a LeakyReLU activation function and a normalization layer. As the level deepens, the number of channels changes to 64, 128, 256, and 512; Neck uses an architecture similar to the feature pyramid network (FPN) to fuse features, splices high and low layer features through 2 bilinear interpolation upsampling, and integrates them through 3 1×1 convolution layers. At the same time, an improved module based on spatial pyramid pooling (SPP) is introduced, with three pooling kernels of 1×1, 5×5, and 9×9 processed in parallel, and the results are concatenated without changing the size of the feature map; the head uses three parallel prediction branches corresponding to tasks of different scales, starting with a 1×1 convolution layer, followed by two 3×3 convolution layers to refine the features. Finally, the branch outputs are integrated, regularized with a 1×1 convolution layer, and the confidence is limited to the 0-1 interval for screening through the Sigmoid function.

[0063] In the description of the present invention, it should be understood that the terms "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside" and "outside" etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0064] The embodiments described above are only descriptions of the preferred modes of the present invention, and are not intended to limit the scope of the present invention. Without departing from the design spirit of the present invention, various modifications and improvements made to the technical solutions of the present invention by ordinary technicians in this field should all fall within the protection scope determined by the claims of the present invention.

Claims

1. A defect automatic polishing device based on machine vision, characterized in that: The invention comprises a working platform (2), wherein a material conveying assembly is provided on the working platform (2), a gantry (9) is movably provided on the working platform (2), a grinding system (5) which can move horizontally and vertically is provided on the gantry (9), and a high-speed camera (4) is installed on the grinding system (5).

2. The automatic defect polishing device based on machine vision according to claim 1 is characterized in that: The material conveying assembly comprises a plurality of conveying rollers (6) rotatably connected to the working platform (2), one end of the conveying rollers (6) is axially connected to the output shaft of a conveying motor (7), and the conveying motor (7) is fixedly connected to the working platform (2).

3. The automatic defect polishing device based on machine vision according to claim 1 is characterized in that: The two side surfaces of the working platform (2) are fixedly connected to the ground rails (3), the gantry (9) is slidably connected to the two ground rails (8), the two sides of the working platform (2) are rotatably connected to the gantry screws (10), the gantry (9) is threadedly connected to the gantry screws (10), one end of the gantry screw (10) is axially connected to the gantry motor, and the gantry motor is installed on the working platform (2).

4. The automatic defect polishing device based on machine vision according to claim 1 is characterized in that: A transverse lead screw (20) is rotatably connected to one side of the gantry (9), a transverse slide rail (19) is fixedly connected to the gantry (9), a transverse slider (17) is slidably connected to the transverse slide rail (19), a fixed plate (22) is fixedly connected to one side of the transverse slider (17), the fixed plate (22) is threadedly connected to the transverse lead screw (20), one end of the transverse lead screw (20) is axially connected to a transverse transmission motor (21), and the transverse transmission motor (21) is fixedly connected to the gantry (9).

5. The automatic defect polishing device based on machine vision according to claim 4 is characterized in that: A vertical slide rail (16) is slidably connected to one side of the transverse slide rail (19) away from the fixed plate (22); the vertical slide rail (16) is fixedly connected to the grinding bracket (13); a vertical lead screw (15) is rotatably connected to the grinding bracket (13); the vertical lead screw (15) is threadedly connected to the fixed plate (22); one end of the vertical lead screw (15) is axially connected to the output shaft of a transmission motor (14); and the transmission motor (14) is installed on the grinding bracket (13).

6. The automatic defect polishing device based on machine vision according to claim 5 is characterized in that: The grinding system (5) comprises a grinding motor (11), wherein the grinding motor (11) is fixedly connected to the grinding bracket (13), the grinding motor (11) is located on a side away from the fixed plate (17), and the output shaft of the grinding motor (11) is connected to a grinding wheel (12).

7. An automatic polishing method, characterized in that: The method is implemented based on the device according to any one of claims 1 to 6, and comprises the following steps: The image of the defect on the plate is obtained by a high-speed camera, and the image is recognized based on machine vision to accurately locate the defect and obtain defect size information; Obtaining defect position information of the to-be-polished portion according to the defect size information; Based on the position information of the defect at the location to be polished, the polishing robot is moved to a suitable position, and based on the defect size information and the position information of the defect at the location to be polished, the defect is polished.

8. The automatic grinding method according to claim 7, characterized in that: Acquiring an image of a defect on the plate comprises the following steps: Image acquisition, obtaining an image on the plate through the high-speed camera (4); Image preprocessing, preprocessing the image obtained on the plate; Defect location: determine the center coordinates and boundary coordinates of the defect based on the edge detection image processing algorithm; The grinding system (5) is controlled according to the pre-processed image and defect position to grind the defects on the plate.

9. The automatic grinding method according to claim 8, characterized in that: The images collected in the image collection step include: at least one of cracks, bubbles, slag inclusions, pits, undercuts, shrinkage cavities, and weld bumps; The image preprocessing step includes: Image preprocessing: reduce image noise through Gaussian filtering denoising algorithm; enhance noise interference contrast through single histogram equalization algorithm; Feature extraction, extracting edge information in the image through the Canny edge detection algorithm, wherein the edge is the contour of the image collected in the image collection step; Defect recognition uses a machine algorithm to identify the type of image captured in the image acquisition step.

10. The automatic grinding method according to claim 9, characterized in that: The defect identification step includes the following steps: Using the YOLOv11 network structure, the YOLOv11 network is mainly composed of three parts: the backbone network is responsible for preliminary deep feature extraction; the neck focuses on multi-scale feature fusion, and the detection head is used to output prediction results related to target detection; Backbone consists of an initial 7×7 convolution layer with a stride of 2, 4 groups of 3 3×3 convolution layer modules and 3 2×2 maximum pooling layers, where each convolution layer is equipped with a LeakyReLU activation function and a normalization layer. As the level deepens, the number of channels changes to 64, 128, 256, and 512; Neck uses an architectural idea similar to a feature pyramid network to fuse features, and splices high and low layer features through 2 bilinear interpolation upsampling, and integrates them through 3 1×1 convolution layers. At the same time, a module based on spatial pyramid pooling improvement was introduced, with three pooling kernels of 1×1, 5×5, and 9×9 processed in parallel, and the results were concatenated without changing the size of the feature map; the head uses three parallel prediction branches corresponding to tasks of different scales, starting with a 1×1 convolution layer, followed by two 3×3 convolution layers to refine the features. Finally, the branch outputs are integrated, regularized with a 1×1 convolution layer, and the confidence is limited to the 0-1 interval for screening through the Sigmoid function.

Images