Cable surface detection device and method
By combining a cable surface inspection device with a convolutional neural network, accurate inspection of the cable surface is achieved, solving the problems of low efficiency and low accuracy of manual quality inspection and improving the level of automation and intelligence of inspection.
Patent Information
- Application Number
- CN202511286093.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-10
- Publication Date
- 2025-11-18
AI Technical Summary
In existing technologies, cable surface defect detection relies on manual quality inspection, which is inefficient, difficult to guarantee accuracy, costly, susceptible to environmental and personal factors, and inconvenient for data management. Furthermore, 2D optical inspection is insufficient in processing complex geometric shapes and depth information.
By employing a cable surface inspection device, combined with a six-laser profilometer and a convolutional neural network, feature extraction and fusion are performed on point cloud data and RGB images to achieve accurate detection of the cable surface.
It improves the accuracy and efficiency of cable surface defect detection, overcomes the shortcomings of manual quality inspection, enhances the automation and intelligence of the inspection, and ensures the high precision and reliability of the inspection results.
Smart Images

Figure CN120971453A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of cable surface defect detection technology, specifically to a cable surface detection device and method. Background Technology
[0002] Currently, in the industrial manufacturing sector, especially in the production of high-voltage cables and optical fibers, surface defect detection mainly relies on manual labor. However, this method has several limitations:
[0003] ① Traditional manual quality inspection is inefficient and difficult to guarantee accuracy: It requires dedicated personnel to monitor the product surface continuously 24 hours a day to detect defects in a timely manner. However, due to factors such as personnel fatigue and distraction, problems such as untimely warnings and inability to guarantee detection accuracy occur frequently.
[0004] ② The cost of manual quality inspection is high: In order to meet the needs of 24-hour continuous operation of the production line, companies usually need to configure at least three full-time quality inspectors to work in shifts, which significantly increases labor costs and imposes an economic burden on the company's operations.
[0005] ③ Influenced by external environmental and personal factors: For example, changes in temperature and lighting conditions in the work environment, as well as the emotional state and sense of responsibility of quality inspectors, can all affect the final detection rate. Especially in some extreme or dangerous work environments, manual inspection is neither safe nor feasible.
[0006] ④ The problems in data storage and management are also very prominent: the manual recording of defect information is prone to data loss, omission and incomplete storage, which greatly limits the possibility of subsequent data analysis and utilization, reflecting the low level of digitization in the whole process.
[0007] ⑤ Although traditional 2D optical inspection technology can improve inspection efficiency and accuracy to a certain extent, it has obvious shortcomings in handling complex geometries, surface textures and depth information, making it difficult to achieve a comprehensive assessment of product surface defects.
[0008] Therefore, how to capture subtle morphological changes on the surface of cables and achieve more accurate and comprehensive detection of surface defects is a technical problem that urgently needs to be solved. Summary of the Invention
[0009] The technical objective of this invention is to provide a cable surface inspection device and method to address the problem of how to capture subtle morphological changes on the cable surface and achieve more accurate and comprehensive detection of cable surface defects.
[0010] The technical objective of this invention is achieved as follows: 1. A cable surface inspection device, comprising a chassis, a display mounted on the top of the chassis, an industrial control computer mounted on the bottom of the chassis, cable inlets and cable outlets respectively mounted on the side walls of the chassis, the cable inlets and cable outlets being located in the lower middle part of the chassis, an image acquisition component being mounted at the cable outlet; the image acquisition component comprising an image acquisition bracket, one side of the image acquisition bracket being fixedly connected to the inner side wall of the chassis, and the other side of the image acquisition bracket being provided with a plurality of adjustable camera mounting mechanisms evenly distributed in a circular array, a laser profilometer being mounted on the adjustable camera mounting mechanisms; a plurality of cooling fans arranged in a circular array being provided at the cable inlet and cable outlet respectively, the cooling fans being correspondingly mounted to the laser profilometers and mounted on the side wall of the chassis;
[0011] An encoder wheel bracket is installed above the cable outlet. The encoder wheel bracket is installed in the upper middle part of the outer wall of the chassis and has an encoder on it. The encoder is connected to the laser profilometer via a camera array wiring method. The laser profilometer is connected to the industrial control computer via a network interface.
[0012] Preferably, the adjustable camera mounting mechanism includes a camera bracket mounted on an image acquisition bracket. The camera bracket is U-shaped, and two parallel Z-axis up-down adjustment screws are arranged between the two sides of the U-shaped camera bracket. A slider is provided on the Z-axis up-down adjustment screw, and the slider slides with the Z-axis up-down adjustment screw. A Z-axis angle adjustment plate is provided on the side of the slider away from the camera bracket. Two Z-axis angle adjustment arc holes are symmetrically arranged on the Z-axis angle adjustment plate, and Z-axis angle adjustment locking bolts are provided in the Z-axis angle adjustment arc holes.
[0013] More preferably, a connecting plate is provided on one end face of the slider, and two X-axis angle adjustment arc holes are symmetrically arranged on the connecting plate, and X-axis angle adjustment locking bolts are provided in the X-axis angle adjustment arc holes;
[0014] A Y-axis angle adjustment plate is provided on the upper end face of the connecting plate. The Y-axis angle adjustment plate is symmetrically provided with Y-axis angle adjustment arc holes, and Y-axis angle adjustment locking bolts are provided in the Y-axis angle adjustment arc holes.
[0015] More preferably, an X-axis front and rear adjustment plate is provided on the lower side of the Y-axis angle adjustment plate, and an X-axis front and rear adjustment elongated hole is provided at each of the four corners of the X-axis front and rear adjustment plate. An X-axis front and rear adjustment locking bolt is provided in the X-axis front and rear adjustment elongated hole. The base end of the laser profilometer is detachably connected to the X-axis front and rear adjustment plate through the X-axis front and rear adjustment locking bolt.
[0016] Preferably, the rear side of the chassis is provided with a sliding door, and the upper part of the center of the front side of the chassis is provided with a multi-functional integrated panel, which is provided with a lifting button, a network cable interface, a USB interface and an emergency stop button in sequence.
[0017] The top of the chassis has an opening pull-out door on each side, with one end of the opening pull-out door hinged to the top side wall of the chassis; an audible and visual alarm is installed on one edge of the top of the chassis.
[0018] Preferably, an uninterruptible power supply is also provided at the bottom of the chassis, which is electrically connected to the industrial control computer to supply power; at least two cooling fans are provided on the two side walls at the bottom of the chassis, and the cooling fans on both sides are located outside the industrial control computer and the uninterruptible power supply, respectively.
[0019] The cable inlet and cable outlet are C-shaped.
[0020] The chassis has feet at the bottom.
[0021] Based on the above-mentioned cable surface inspection device, the cable surface inspection method is as follows:
[0022] Acquire point cloud data and the corresponding RGB image around the target object;
[0023] The point cloud data belonging to the side of the cylinder is extracted and mapped to a two-dimensional plane to generate a rectangular depth image without stretching or compression, wherein the mapping maintains the local distance unchanged.
[0024] The RGB image and the depth image are registered and sized.
[0025] The three-channel color information of the registered RGB image is fused with the depth image to form four-channel input data containing R, G, B, and Depth.
[0026] The four-channel input data is fed into the trained convolutional neural network model for feature extraction and fusion, and outputs the category, location and size information of the target object.
[0027] As a preferred method, the point cloud extraction method for the side surface of the cylinder is as follows: geometric fitting or semantic segmentation is performed on the original point cloud to identify the point set that conforms to the cylindrical surface model, and points on the top and bottom surfaces and irrelevant areas are removed.
[0028] Ideally, the four-channel input data is fused through channel concatenation and then fed into a unified convolutional neural network for joint feature learning.
[0029] More preferably, convolutional neural networks employ a dual-branch structure, with RGB images and depth images processed separately, and then feature fusion is performed through attention mechanisms or feature weighting strategies.
[0030] The cable surface inspection device and method of the present invention have the following advantages:
[0031] (I) This invention utilizes the geometric properties of a cylindrical surface as a developable surface (with zero Gaussian curvature). Through rigorous differential geometric analysis, an equidistant mapping is established from the parameter domain of the cylindrical surface to the rectangular plane. The equidistant mapping maintains the local distance of the point cloud data during the unfolding process, avoiding the stretching, compression, or deformation distortion commonly found in traditional projection methods. This fully preserves the spatial structure information of the original point cloud, improves the geometric fidelity of subsequent image processing, and thus achieves equidistant flattening of cylindrical point clouds while preserving geometric accuracy.
[0032] (ii) This invention can generate high-quality depth images and enhance spatial representation capabilities. By unfolding the three-dimensional point cloud along the side of the cylinder into a two-dimensional rectangular depth image, it realizes the normalization representation of non-planar structure point clouds. The depth image is compatible with standard image formats, which facilitates the subsequent use of mature two-dimensional convolutional neural networks for feature extraction, and significantly improves the utilization rate and processing efficiency of depth information.
[0033] (III) This invention constructs a four-channel fusion input to enrich the dimensions of perceptual information. It accurately registers and fuses the three-channel color information of the RGB image with the converted depth map to form four-channel input data containing R, G, B and Depth. The multimodal input method not only preserves the appearance texture features of the object, but also introduces accurate spatial depth information, enhancing the model's ability to understand the three-dimensional structure of the object.
[0034] (iv) This invention extracts features from RGB images and depth maps separately and then effectively fuses them, enabling the model to utilize both appearance features and geometric structure information simultaneously. This effectively addresses complex scenarios such as occlusion, lighting changes, and background interference, thereby improving the accuracy and robustness of 3D target detection. Experiments show that this invention can significantly improve the localization accuracy and category recognition accuracy of target detection in various environments. Furthermore, this invention is highly versatile and applicable to intelligent perception systems equipped with LiDAR cameras, demonstrating good scalability and promising engineering applications.
[0035] (V) This invention introduces six-laser profilometer 3D detection technology, combined with automated control and artificial intelligence algorithms, to achieve comprehensive and accurate detection of surface defects in cables, significantly improving detection efficiency and product quality, and overcoming the shortcomings of traditional manual quality inspection and 2D optical inspection methods;
[0036] (vi) Compared with 2D optical inspection, the 3D laser profilometer of the present invention can not only provide accurate dimensional measurement, but also capture subtle morphological changes on the surface of objects, thereby achieving more accurate and comprehensive detection of surface defects in cables. This effectively overcomes the problems of low efficiency, high cost, susceptibility to environmental and personal factors, and inconvenient data management in traditional methods, and greatly improves the quality and efficiency of cable production. At the same time, by integrating advanced artificial intelligence algorithms, it can automatically identify and record various types of surface defects, ensuring high accuracy and reliability of detection results, while improving the automation and intelligence level of the entire production process.
[0037] (vii) The present invention has six laser profilometers arranged around the cable to ensure that the cable can be accurately scanned and analyzed from all angles. At the same time, the data is processed and analyzed by the industrial control computer. This not only realizes 360-degree all-round detection of the cable, but also greatly improves the detection speed and accuracy, providing a reliable guarantee for the quality of cable production.
[0038] (viii) The top of the chassis of the present invention is designed with a convenient openable pull-out door, which provides a convenient maintenance path for the internal components, making maintenance and inspection easier;
[0039] (ix) An encoder wheel bracket and an encoder are provided on an outer side of the chassis of the present invention, which supports precise position or speed measurement and is suitable for application scenarios that require precise control.
[0040] (x) The six laser profilometers of the present invention are arranged in a circular array on the adjustable camera mounting mechanism. The adjustable camera mounting mechanism can adjust the laser profilometers in front and behind, left and right, up and down and corresponding angles from the X-axis, Y-axis and Z-axis to ensure that the surface information of the target object is obtained from multiple angles and improve the comprehensiveness and accuracy of data acquisition.
[0041] (xi) A cooling fan 1 is provided on one side of the laser profilometer of the present invention, and a cooling fan 2 for cooling the industrial control computer and uninterruptible power supply is provided at the bottom of the chassis, so as to ensure effective heat dissipation management for the laser profilometer, industrial control computer and uninterruptible power supply.
[0042] (xii) The rear side of the chassis of the present invention is provided with a side-opening sliding door, which not only facilitates the replacement and debugging of the laser profilometer, but also provides a direct multi-functional integrated panel. The multi-functional integrated panel is equipped with lifting buttons, network cable interface, USB interface and emergency stop button, which simplifies the process of equipment start-up and daily inspection.
[0043] Therefore, this invention has the characteristics of reasonable design, simple structure, easy processing, small size, convenient use, and multiple uses, and thus has great value for promotion and use. Attached Figure Description
[0044] The invention will be further described below with reference to the accompanying drawings.
[0045] Appendix Figure 1 This is a schematic diagram of the cable surface inspection device.
[0046] Appendix Figure 2 For the appendix Figure 1 A schematic diagram of the structure of the multi-functional integrated panel;
[0047] Appendix Figure 3 This is a schematic diagram of the top of the chassis.
[0048] Appendix Figure 4 A schematic diagram of a chassis with a cable entry side;
[0049] Appendix Figure 5 A schematic diagram of a chassis with a cable outlet side;
[0050] Appendix Figure 6 A schematic diagram of the cable surface inspection device after the chassis casing has been removed;
[0051] Appendix Figure 7 A three-dimensional structural diagram of the adjustable camera mounting mechanism;
[0052] Appendix Figure 8 Rear view of the adjustable camera mounting mechanism;
[0053] Appendix Figure 9 Top view of the adjustable camera mounting mechanism;
[0054] Appendix Figure 10 This is a schematic diagram of the process of flattening a cylindrical surface.
[0055] In the diagram: 1. Chassis, 2. Monitor, 3. Industrial PC, 4. Cable inlet, 5. Cable outlet, 6. Image acquisition bracket, 7. Laser profilometer, 8. Cooling fan 1, 9. Encoder wheel bracket, 10. Encoder, 11. Camera bracket, 12. Z-axis up / down adjustment screw, 13. Slider, 14. Z-axis angle adjustment plate, 15. Z-axis angle adjustment arc hole, 16. Z-axis angle adjustment locking bolt, 17. Connecting plate, 18. X-axis angle adjustment arc hole, 19. X-axis angle adjustment locking bolt 20. Y-axis angle adjustment plate; 21. Y-axis angle adjustment arc hole; 22. Y-axis angle adjustment locking bolt; 23. X-axis front and rear adjustment plate; 24. X-axis front and rear adjustment elongated hole; 25. X-axis front and rear adjustment locking bolt; 26. Sliding door; 27. Multi-functional integrated panel; 28. Lifting button; 29. Network cable interface; 30. USB interface; 31. Emergency stop button; 32. Openable sliding door; 33. Uninterruptible power supply; 34. Two cooling fans; 35. Foot; 36. Audible and visual alarm. Detailed Implementation
[0056] The cable surface inspection device and method of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
[0057] In the description of this invention, it should be understood that the terms "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the invention and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.
[0058] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0059] Example 1:
[0060] As attached Figure 1 and 6 As shown, this embodiment provides a cable surface inspection device, the structure of which includes a chassis 1, a display 2 mounted on the top of the chassis 1, an industrial control computer 3 at the bottom of the chassis 1, cable inlets 4 and cable outlets 5 respectively opened on the two side walls of the chassis 1, the cable inlets 4 and cable outlets 5 are located in the lower middle part of the chassis 1, and an image acquisition component is installed at each of the cable inlets 4 and cable outlets 5; the image acquisition component includes an image acquisition bracket 6, one side of the image acquisition bracket 6 is fixedly connected to the inner side wall of the chassis 1, and several adjustable camera mounting mechanisms are mounted on the other side of the image acquisition bracket 6 in a circumferential array, and a laser profilometer 7 is mounted on the adjustable camera mounting mechanism; six cooling fans 8 in a circumferential array are installed at the cable inlets 4 and cable outlets 5 respectively, the cooling fans 8 are arranged one-to-one with the laser profilometers 7 and the cooling fans 8 are mounted on the side wall of the chassis 1, as shown in the attached figure. Figure 4 As shown. Among them, the cooling fan 8 is installed directly on the key heat-generating area of the laser profilometer 7, which can effectively remove heat from the area around the camera and keep its operating temperature within an ideal range. This is crucial for maintaining the accuracy and stability of the laser profilometer 7, especially in application scenarios where it operates continuously for a long time.
[0061] In this embodiment, a total of six laser profilometers 7 are mounted at different positions on the camera bracket 11, forming a surround scanning array. This layout ensures that surface information of the target object is acquired simultaneously from multiple angles, thereby improving the comprehensiveness and accuracy of data acquisition. Each laser profilometer 7 is calibrated to ensure consistent and high-precision output data. Grouping strategy: Each opposing laser profilometer group 7 is assigned a different trigger frequency. For example, taking six laser profilometers 7 as an example, these six laser profilometers 7 can be labeled A, B, C, D, E, and F, where A and D, B and E, and C and F are each a group. Each group of laser profilometers 7 is assigned a specific trigger frequency. For example, A and D use frequency f1, B and E use frequency f2, and C and F use frequency f3. The selection of frequency needs to consider factors such as ambient light conditions and camera sensor characteristics to ensure optimal performance; a central control system or synchronization signal generator is introduced to coordinate the triggering time of each camera, ensuring that cameras in different groups do not perform exposure operations simultaneously. The trigger frequency of each camera group is adjusted in real time according to the image quality and interference situation during actual operation. For example, if interference is found to still exist between a group of laser profilometers 7, the frequency difference between that group and other groups will be appropriately increased, the working status and output image quality of each camera will be continuously monitored, and the trigger frequency setting will be automatically optimized based on this information.
[0062] As attached Figure 3 As shown, an encoder wheel bracket 9 is installed above the cable outlet 5. The encoder wheel bracket 9 is located in the upper middle part of the outer wall of the chassis 1, and an encoder 10 is mounted on it. The encoder 10 is connected to the laser profilometer 7 via a camera array wiring method. The laser profilometer 7 is connected to the industrial control computer 3 via a network interface. The encoder wheel bracket 9 is preferably made of high-strength metal material to ensure sufficient mechanical strength and stability. The encoder wheel bracket 9 is designed with a modular structure, supporting quick disassembly, installation, and replacement for easy maintenance or upgrades. The encoder 10, mounted on the encoder wheel bracket 9, is a high-precision rotary sensor capable of accurately detecting and outputting position information or speed signals. The encoder 10 can be selected as an incremental or absolute encoder, depending on the requirements of the actual application scenario. For example, in applications requiring precise positioning, an absolute encoder with a resolution of thousands of pulses per revolution can be selected.
[0063] As attached Figure 7As shown, the adjustable camera mounting mechanism includes a camera bracket 11 mounted on an image acquisition bracket 6. The camera bracket 11 is U-shaped, and two parallel Z-axis up-down adjustment screws 12 are installed between the two sides of the U-shaped camera bracket 11. A slider 13 is installed on the Z-axis up-down adjustment screw 12. The slider 13 slides with the Z-axis up-down adjustment screw 12, and a Z-axis angle adjustment plate 14 is installed on the side of the slider 13 away from the camera bracket 11. Two Z-axis angle adjustment arc holes 15 are symmetrically opened on the Z-axis angle adjustment arc holes 15, and Z-axis angle adjustment locking bolts 16 are installed in the Z-axis angle adjustment arc holes 15.
[0064] As attached Figure 8 As shown, in this embodiment, a connecting plate 17 is installed on one end face of the slider 13. Two X-axis angle adjustment arc holes 18 are symmetrically opened on the connecting plate 17, and X-axis angle adjustment locking bolts 19 are installed in the X-axis angle adjustment arc holes 18.
[0065] As attached Figure 9 As shown, in this embodiment, a Y-axis angle adjustment plate 20 is installed on the upper surface of the connecting plate 17. The Y-axis angle adjustment plate 20 has symmetrically opened Y-axis angle adjustment arc holes 21, and Y-axis angle adjustment locking bolts 22 are installed in the Y-axis angle adjustment arc holes 21. An X-axis front and rear adjustment plate 23 is installed on the lower side of the Y-axis angle adjustment plate 20. X-axis front and rear adjustment elongated holes 24 are opened at the four corners of the X-axis front and rear adjustment plate 23, and X-axis front and rear adjustment locking bolts 25 are installed in the X-axis front and rear adjustment elongated holes 24. The base end of the laser profilometer 7 is detachably connected to the X-axis front and rear adjustment plate 23 through the X-axis front and rear adjustment locking bolts 25.
[0066] As attached Figure 2 As shown, in this embodiment, a sliding door 26 is installed on the rear side of the chassis 1, and a multi-functional integrated panel 27 is installed on the upper part of the center of the front side of the chassis 1. The multi-functional integrated panel 27 is sequentially equipped with a lifting button 28, a network cable interface 29, a USB interface 30, and an emergency stop button 31, optimizing the user experience and improving operational efficiency. An audible and visual alarm 36 is installed on one edge of the top of the chassis 1. The audible and visual alarm 36 includes a high-brightness LED light and a buzzer, used to attract the operator's attention by emitting a strong light signal and sound alarm when abnormal equipment operation occurs, ensuring timely handling.
[0067] In this embodiment, an openable sliding door 32 is installed on each of the two sides of the top of the chassis 1. One end of the openable sliding door 32 is hinged to the top side wall of the chassis 1, which facilitates the user to quickly open it for maintenance, inspection or replacement of internal components. In addition, for easy operation, the openable sliding door 32 is also designed with an easy-to-grip handle.
[0068] In this embodiment, the cable inlet 4 and cable outlet 5 are C-shaped; an uninterruptible power supply 33 is also installed at the bottom of the chassis 1, which is electrically connected to the industrial control computer 3 to supply power; as shown in the attached... Figure 5 As shown, two cooling fans 34 are installed on the two side walls of the bottom of the chassis 1, respectively. The cooling fans 34 are located on the outside of the industrial computer 3 and the uninterruptible power supply (UPS) 33. The bottom of the chassis 1 is equipped with feet 35. The cooling fans 34 are responsible for providing heat dissipation for the industrial computer 3 and the uninterruptible power supply (UPS) 33. Considering that the industrial computer 3 and the uninterruptible power supply 33 are usually one of the main heat sources in the system, ensuring that cool air is effectively guided through these components and the generated heat is quickly dissipated not only helps to reduce the operating temperature of the industrial computer and the UPS, but also extends the service life of the industrial computer 3 and the uninterruptible power supply 33.
[0069] Example 2:
[0070] This embodiment provides a cable surface inspection method based on the cable surface inspection device in Embodiment 1. The method is as follows:
[0071] S1. Acquire point cloud data and the corresponding RGB image around the target object;
[0072] S2. Extract the point cloud data belonging to the side of the cylinder from the point cloud data and map it to a two-dimensional plane to generate a rectangular depth image without stretching or compression, wherein the mapping keeps the local distance unchanged.
[0073] S3. Register and align the RGB image and the depth image to their dimensions;
[0074] S4. The three-channel color information of the registered RGB image is fused with the depth image to form four-channel input data containing R, G, B, and Depth.
[0075] S5. Input the four-channel input data into the trained convolutional neural network model, perform feature extraction and fusion, and output the category, position and size information of the target object.
[0076] In this embodiment, the point cloud extraction method for the cylindrical side surface in step S2 is as follows: geometric fitting or semantic segmentation is performed on the original point cloud to identify the point set that conforms to the cylindrical surface model, and points on the top and bottom surfaces and irrelevant areas are removed.
[0077] In this embodiment, the four-channel input data in step S5 are fused through channel concatenation and then input into a unified convolutional neural network for joint feature learning.
[0078] In step S5 of this embodiment, the convolutional neural network adopts a dual-branch structure, which is used to process RGB images and depth images respectively, and then feature fusion is performed through attention mechanism or feature weighting strategy.
[0079] As attached Figure 10 As shown, based on the cylindrical point cloud imaging, the side surface of this cylinder (excluding the top and bottom surfaces) is unfolded into a rectangular plane without stretching or compression, thus achieving equidistant mapping.
[0080] Use parameters (θ, z) to represent any point on the cylindrical surface.
[0081]
[0082] Calculate the partial derivatives:
[0083]
[0084] Through flattening mapping:
[0085]
[0086] get:
[0087]
[0088] A cylinder is a developable surface with zero Gaussian curvature, allowing it to be flattened into a plane without stretching or compression. The theoretical basis for this flattening is that its first fundamental form can be transformed into a standard planar form through coordinate transformation. The final flattened result is a rectangular planar point cloud image. A depth image is generated by flattening the point cloud image. By integrating the color information of the RGB image and the depth information provided by the transformed depth map, a four-channel input containing rich spatial details is formed, improving the accuracy and robustness of object detection. Data from an RGB camera and a depth sensor are received, and the RGB image and corresponding depth map are preprocessed to ensure consistency in size and viewpoint. A convolutional neural network is used to extract feature representations from the RGB image and depth map respectively, and then a specific fusion strategy is used to merge the feature maps into a four-channel feature representation. A 3D object detection model capable of processing four-channel input is trained, which can predict the position, size, and category of objects.
[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.
Claims
1. A cable surface inspection apparatus characterized by, The cabinet is provided with a display on the top and an industrial computer on the bottom, cable inlets and outlets are arranged on the two side walls, the cable inlets and outlets are located at the lower part of the middle of the cabinet, and an image acquisition assembly is arranged at the cable outlet; The image acquisition assembly comprises an image acquisition support, one side of the image acquisition support is fixedly connected with the inner side wall of the cabinet, and a plurality of adjustable camera mounting mechanisms are arranged on the other side of the image acquisition support in a circumferential array, and a laser profiler is arranged on the adjustable camera mounting mechanism; a plurality of heat dissipation fans are arranged at the cable inlets and outlets in a circumferential array, and the heat dissipation fans are arranged in one-to-one correspondence with the laser profilers and are mounted on the side walls of the cabinet; An encoder is arranged on the code wheel support which is mounted on the upper part of the middle of the outer side wall of the cabinet, the encoder is connected with the laser profiler through a camera array wiring mode, and the laser profiler is connected with the industrial computer through a network interface.
2. The cable surface inspection apparatus of claim 1, wherein The adjustable camera mounting mechanism comprises a camera support mounted on the image acquisition support, the camera support is in a U shape, two parallel Z-axis up-down adjustment screws are arranged between the two side walls of the U-shaped camera support, a sliding block is arranged on the Z-axis up-down adjustment screw, the sliding block is in sliding fit with the Z-axis up-down adjustment screw, a Z-axis angle adjustment plate is arranged on the side of the sliding block away from the camera support, two Z-axis angle adjustment arc-shaped holes are symmetrically arranged on the Z-axis angle adjustment plate, and a Z-axis angle adjustment locking bolt is arranged in the Z-axis angle adjustment arc-shaped hole.
3. The cable surface inspection apparatus of claim 2, wherein An end surface of the sliding block is provided with a connecting plate, two X-axis angle adjustment arc-shaped holes are symmetrically arranged on the connecting plate, and an X-axis angle adjustment locking bolt is arranged in the X-axis angle adjustment arc-shaped hole. A Y-axis angle adjustment plate is arranged at the upper end surface of the connecting plate, Y-axis angle adjustment arc-shaped holes are symmetrically arranged on the Y-axis angle adjustment plate, and Y-axis angle adjustment locking bolts are arranged in the Y-axis angle adjustment arc-shaped holes.
4. The cable surface inspection apparatus of claim 3, wherein A X-axis front-rear adjustment plate is arranged on the lower side of the Y-axis angle adjustment plate, X-axis front-rear adjustment long holes are arranged at the four corners of the X-axis front-rear adjustment plate respectively, X-axis front-rear adjustment locking bolts are arranged in the X-axis front-rear adjustment long holes, and the base end of the laser profiler is detachably connected with the X-axis front-rear adjustment plate through the X-axis front-rear adjustment locking bolts.
5. The cable surface inspection apparatus of claim 1, wherein, A pull door is arranged on the rear side of the cabinet, a multifunctional integrated panel is arranged at the upper part of the middle of the front side of the cabinet, and a lifting button, a network interface, a USB interface and an emergency stop button are sequentially arranged on the multifunctional integrated panel; An openable stretch door is arranged at each side of the top of the cabinet, and one end of the openable stretch door is hingedly connected with the top side wall of the cabinet. A sound and light alarm is arranged at the side edge of the top of the cabinet.
6. The cable surface inspection apparatus of claim 1, wherein An uninterruptible power supply is further arranged on the bottom of the cabinet, and the uninterruptible power supply is electrically connected with the industrial computer to supply power for the industrial computer; At least two heat dissipation fans are arranged at the two side walls of the bottom of the cabinet, and the heat dissipation fans are located at the outer sides of the industrial computer and the uninterruptible power supply respectively. The cable inlets and outlets are in a C shape. A footing is arranged on the bottom of the cabinet.
7. The cable surface inspection method based on the cable surface inspection apparatus according to any one of claims 1 to 6, characterized by, The method specifically comprises the following steps: Obtaining point cloud data and corresponding RGB images around a target object; The point cloud data belonging to the cylindrical side surface is extracted and mapped to a two-dimensional plane to generate a non-stretching and non-compressed rectangular depth image, wherein the mapping keeps the local distance unchanged; The RGB image and the depth image are registered and size-aligned; The three-channel color information of the registered RGB image is fused with the depth image to form four-channel input data containing R, G, B and Depth; The four-channel input data is input into the trained convolutional neural network model for feature extraction and fusion, and the class, position and size information of the target object are output.
8. The cable surface inspection method according to claim 7, wherein, The point cloud extraction method of the cylindrical side surface is to perform geometric fitting or semantic segmentation on the original point cloud, identify the point set conforming to the cylindrical surface model, and remove the points on the upper and lower bottom surfaces and the points in the non-related area.
9. The cable surface inspection method according to claim 7 or 8, characterized in that, The four-channel input data is fused by channel splicing and input into a unified convolutional neural network for joint feature learning.
10. The cable surface inspection method of claim 9, wherein, The convolutional neural network adopts a double-branch structure, which is used to process the RGB image and the depth image respectively, and then the features are fused through attention mechanism or feature weighting strategy.