Metal melting mark detection equipment, method, device and equipment and storage medium

Abstract

The embodiment of the invention provides metal melting mark detection equipment, method and device, equipment and a storage medium. The method comprises the steps that a base is arranged; the detection mechanism comprises a first joint, a threaded rod, a movable block, a camera module, a lighting module and a first driver, one end of the threaded rod is arranged on the first joint, the movable block is provided with a threaded through hole, the threaded through hole is in threaded connection with the threaded rod, and the camera module and the lighting module are both arranged on the side face of the movable block; the camera modules and the lighting modules are distributed in the circumferential direction of the movable block at intervals, the first driver is fixed to the movable block, and the first driver is used for driving the movable block to move on the threaded rod; the angle adjusting mechanism is fixed to the base, the first joint is connected with the angle adjusting mechanism, and the angle adjusting mechanism is used for adjusting the angle of the first joint; the first driver, the camera module and the lighting module are electrically connected with the control module. According to the embodiment of the invention, the electric lead in the movement limited space can be effectively shot.

Description

Technical Field

[0001] The present invention relates to the field of metal melting mark detection, and in particular to a metal melting mark detection device, method, apparatus, equipment and storage medium. Background Art

[0002] With the increase of electrical equipment in my country, the incidence of electrical fires is getting higher and higher. After a fire occurs, investigators will investigate the post-disaster scene to find evidence to determine the cause of the fire. Among them, the remaining electrical wires at the scene are important evidence. For example, the metal melting marks on the surface of the remaining electrical wires may include primary short circuits and secondary short circuits. Primary short circuits are usually determined to be the cause of the fire, and secondary short circuits are usually determined to be short circuits generated after the fire occurs.

[0003] At present, at the fire scene, there are cases where some residual electrical wires are located in restricted movement spaces. For example, restricted movement spaces include hidden spaces, narrow spaces, and closed spaces, etc. Workers are often unable to enter the restricted movement spaces, and are unable to use traditional photographic equipment to directly photograph the electrical wires located in the restricted movement spaces, resulting in the inability to perform metal melting mark detection on all electrical wires at the fire scene, which is not conducive to investigating the cause of the fire. Summary of the invention

[0004] The following is a summary of the subject matter described in detail herein. This summary is not intended to limit the scope of the claims.

[0005] The embodiments of the present application provide a metal melt mark detection device, method, apparatus, equipment and storage medium, which can effectively photograph electrical conductors located in a space with restricted movement.

[0006] To achieve the above-mentioned purpose, the first aspect of an embodiment of the present application proposes a metal melt mark detection device, comprising: a base; a detection mechanism, comprising a first joint, a threaded rod, a movable block, a camera module, a lighting module and a first driver, one end of the threaded rod is arranged on the first joint, the movable block is provided with a threaded through hole, the threaded through hole is threadedly connected to the threaded rod, the camera module and the lighting module are both arranged on the side of the movable block, the camera module and the lighting module are distributed at intervals along the circumference of the movable block, the first driver is fixed on the movable block, the first driver is used to drive the movable block to move on the threaded rod so that the movable block is close to or away from the first joint; an angle adjustment mechanism, fixed on the base, the first joint is connected to the angle adjustment mechanism, and the angle adjustment mechanism is used to adjust the angle of the first joint; a control module, the first driver, the camera module and the lighting module are electrically connected to the control module respectively.

[0007] In some embodiments, a first gear is provided on the movable block, and a second gear meshing with the first gear is provided at the driving end of the first driver, the axis of the first gear coincides with the axis of the threaded rod, and the first driver is used to drive the second gear to rotate, so as to drive the first gear to rotate around the axis of the threaded rod.

[0008] In some embodiments, the detection mechanism includes a second driver, which is disposed on the first joint and is used to drive the threaded rod to rotate around the axis of the threaded rod.

[0009] In some embodiments, a third gear is provided on the threaded rod, and a fourth gear meshing with the third gear is provided at the driving end of the second driver. The second driver is used to drive the fourth gear to rotate, thereby driving the third gear to rotate around the axis of the threaded rod.

[0010] In some embodiments, the angle adjustment mechanism includes a second joint, a support arm, a rotational driver, a third driver and a fourth driver. The rotational driver is fixed on the base, the driving end of the rotational driver is connected to the second joint, one end of the support arm is rotatably connected to the second joint, and the other end of the support arm is rotatably connected to the first joint. The rotational driver is used to drive the second joint to rotate, the third driver is used to drive the support arm to rotate, and the fourth driver is used to drive the first joint to rotate.

[0011] To achieve the above-mentioned purpose, the second aspect of an embodiment of the present application proposes a metal melt mark detection method, comprising: obtaining an image of a wire to be tested taken by a camera module about a target metal wire, wherein the surface of the target metal wire contains a target melt mark; inputting the image of the wire to be tested into a melt mark detection model for target recognition, and obtaining a target recognition result, wherein the target recognition result is used to indicate the type of the target melt mark.

[0012] In some embodiments, before inputting the image of the conductor to be tested into the melt mark detection model for target recognition and obtaining the target recognition result, the metal melt mark detection method further includes: obtaining a sample image of a sample metal conductor and an identification label of the sample image, wherein the surface of the sample metal conductor includes a sample melt mark, and the identification label is used to indicate the position and type of the sample melt mark; calling the melt mark detection model to perform multi-level feature extraction on the sample image to obtain a first image feature and a second image feature with successively decreasing dimensions; performing channel compression on the second image feature to obtain a compressed image feature, performing content encoding on the compressed image feature to obtain a plurality of initial recombinant kernels, and performing normalization processing on each of the initial recombinant kernels to obtain a plurality of target recombinant kernels; performing upsampling processing on the second image feature based on each of the target recombinant kernels to obtain a first intermediate feature, wherein the first intermediate feature has the same dimension as the first image feature; performing bounding box prediction and category prediction based on the channel cascade result of the first intermediate feature and the first image feature to obtain a predicted recognition result; determining a model loss according to the predicted recognition result and the identification label, and training the melt mark detection model based on the model loss.

[0013] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes a metal melt mark detection device, comprising: an acquisition module, used to acquire an image of a wire to be tested taken by a camera module about a target metal wire, wherein the surface of the target metal wire contains a target melt mark; a prediction module, used to input the image of the wire to be tested into a melt mark detection model for target recognition, and obtain a target recognition result, wherein the target recognition result is used to indicate the type of the target melt mark.

[0014] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the metal melt mark detection method described in the second aspect when executing the computer program.

[0015] To achieve the above-mentioned purpose, the fifth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the metal melt mark detection method described in the second aspect is implemented.

[0016] The embodiments of the present application include at least the following beneficial effects: by providing a threaded through hole on the movable block, and threading the threaded through hole with the threaded rod, it is equivalent to movably connecting the movable block with the threaded rod, and fixing a first driver on the movable block. Under the action of the controller, the movable block can be driven to move on the threaded rod by the first driver, so that the movable block is close to or away from the first joint. Since the camera module is provided on the side of the movable block, when the movable block moves on the threaded rod, the camera module can move synchronously with the movable block, so as to flexibly adjust the shooting position of the camera module, effectively improve the adaptability of the camera module to the space with restricted movement, and under the action of the controller, the camera module can effectively shoot the electrical wires located in the space with restricted movement; and a lighting module is also provided on the side of the movable block, so the lighting module will also move synchronously with the camera module. Since the camera module and the lighting module move along the movable block The first joint is arranged at intervals in the circumferential direction, so the light emission direction of the lighting module will not be directly facing the camera module, so that the lighting module can provide sufficient light to the space where the electric wire is located without affecting the shooting of the camera module, thereby improving the shooting quality of the camera module; in addition, the first joint is connected to the angle adjustment mechanism, and the angle adjustment mechanism is fixed on the base. Under the action of the angle adjustment mechanism, the angle of the first joint can be flexibly adjusted. When the angle of the first joint changes, the angle of the camera module can be driven to change, so the shooting angle of the camera module can be flexibly adjusted, further improving the adaptability of the camera module to the restricted movement space, so that the camera module can cope with more complex restricted movement space, ensuring that the camera module can effectively shoot the electric wires located in the restricted movement space, so that all the electric wires at the fire scene can be detected for metal melting marks, which is helpful for investigating the cause of the fire.

[0017] Other features and advantages of the present application will be described in the following description, and partly become apparent from the description, or understood by practicing the present application. The purpose and other advantages of the present application can be realized and obtained by the structures specifically pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide further understanding of the technical solution of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solution of the present application and do not constitute a limitation on the technical solution of the present application.

[0019] Figure 1 A schematic diagram of an optional structure of a metal melting mark detection device provided in an embodiment of the present application;

[0020] Figure 2 An optional explosion diagram of the detection mechanism provided in the embodiment of the present application;

[0021] Figure 3 An optional exploded schematic diagram of the angle adjustment mechanism provided in an embodiment of the present application;

[0022] Figure 4 An optional exploded schematic diagram of the angle adjustment mechanism provided in an embodiment of the present application;

[0023] Figure 5 An optional system block diagram of a control module provided in an embodiment of the present application;

[0024] Figure 6 A schematic diagram of an optional process of a metal melting mark detection method provided in an embodiment of the present application;

[0025] Figure 7 An optional flow chart of a method for training a melt mark detection model provided in an embodiment of the present application;

[0026] Figure 8 An optional flow chart of the training process of the melt mark detection model provided in the embodiment of the present application;

[0027] Fig. 9 An optional flowchart of the processing process of the convolution layer provided in the embodiment of the present application;

[0028] Fig.10 An optional flowchart of the processing process of the multi-scale feature extraction module provided in the embodiment of the present application;

[0029] Fig.11 An optional flowchart of the processing process of the reference convolution module provided in the embodiment of the present application;

[0030] Fig.12 An optional flowchart of the upsampling process provided in the embodiment of the present application;

[0031] Fig.13 An optional schematic diagram of a lightweight convolution process provided in an embodiment of the present application;

[0032] Fig.14 An optional structural schematic diagram of a metal melt mark detection device provided in an embodiment of the present application.

[0033] Fig.15 A schematic diagram of an optional hardware structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0035] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the characteristics of the target object such as target object attribute information or attribute information set, the permission or consent of the target object will be obtained first, and the collection, use and processing of these data will comply with relevant laws, regulations and standards. Among them, the target object can be a user. In addition, when the embodiment of the present application needs to obtain the target object attribute information, the target object's separate permission or separate consent will be obtained by means of a pop-up window or jumping to a confirmation page, etc., and after the separate permission or separate consent of the target object is clearly obtained, the necessary target object-related data for enabling the normal operation of the embodiment of the present application will be obtained.

[0036] In the description of this application, “several” means one or more, “more” means more than two, “greater than”, “less than”, “exceed”, etc. are understood to exclude the number, and “above”, “below”, “within”, etc. are understood to include the number.

[0037] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims or the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0038] To facilitate understanding of the technical solutions provided in the embodiments of the present application, some key terms used in the embodiments of the present application are explained here:

[0039] Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0040] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. Basic artificial intelligence technologies generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operating / interactive systems, mechatronics and other technologies. Artificial intelligence software technologies mainly include computer vision technology, speech processing technology, natural language processing technology, as well as machine learning / deep learning, autonomous driving, smart transportation and other major directions.

[0041] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.

[0042] At present, with the increase of electrical equipment in my country, the incidence of electrical fires is getting higher and higher. After a fire occurs, investigators will investigate the post-disaster scene to find evidence to determine the cause of the fire. Among them, the remaining electrical wires at the scene are important evidence. For example, the metal melting marks on the surface of the remaining electrical wires may include primary short circuits and secondary short circuits. Primary short circuits are usually determined to be the cause of the fire, and secondary short circuits are usually determined to be short circuits generated after the fire occurs. At the fire scene, there are cases where some of the remaining electrical wires are located in restricted movement spaces. For example, restricted movement spaces include hidden spaces, narrow spaces, and closed spaces, etc. Workers often cannot enter the restricted movement spaces, nor can they use traditional photographic equipment to directly shoot the electrical wires located in the restricted movement spaces, resulting in the inability to detect metal melting marks on all electrical wires at the fire scene, which is not conducive to investigating the cause of the fire.

[0043] Based on this, the embodiments of the present disclosure provide a metal melt mark detection device, method, apparatus, equipment and storage medium, which can effectively photograph electrical conductors located in a space with restricted movement.

[0044] The metal melt mark detection equipment, method, device, equipment and storage medium provided in the embodiments of the present application are specifically illustrated through the following embodiments. First, the metal melt mark detection equipment in the embodiments of the present application is described.

[0045] The embodiments of the present application are further described below in conjunction with the accompanying drawings.

[0046] Reference Figures 1 to 5, the embodiment of the present application provides a metal melting mark detection device, comprising:

[0047] Base 100;

[0048] The detection mechanism 200 includes a first joint 210, a threaded rod 220, a movable block 230, a camera module 240, a lighting module 280 and a first driver 250. One end of the threaded rod 220 is arranged on the first joint 210. The movable block 230 is provided with a threaded through hole 231. The threaded through hole 231 is threadedly connected to the threaded rod 220. The camera module 240 and the lighting module 280 are both arranged on the side of the movable block 230. The camera module 240 and the lighting module 280 are distributed at intervals along the circumference of the movable block 230. The first driver 250 is fixed on the movable block 230. The first driver 250 is used to drive the movable block 230 to move on the threaded rod 220 so that the movable block 230 is close to or away from the first joint 210.

[0049] An angle adjustment mechanism 300 is fixed on the base 100, the first joint 210 is connected to the angle adjustment mechanism 300, and the angle adjustment mechanism 300 is used to adjust the angle of the first joint 210;

[0050] The control module 400 , the first driver 250 , the camera module 240 and the lighting module 280 are electrically connected to the control module 400 , respectively.

[0051] The root of the first joint 210 is provided with a toothed through hole, through which the rotating shaft can pass. The rotating shaft can drive the first joint 210 to rotate through the toothed through hole. The rotation of the first joint 210 can drive the entire detection mechanism 200 to rotate.

[0052] The camera module 240 includes at least one camera, and the lighting module 280 includes at least one lighting lamp, and the number of lighting lamps may be in multiple relationship with the number of cameras. For example, assuming that the cross section of the movable block 230 is a hexagon, that is, the movable block 230 includes six sides, the camera module 240 includes two cameras, and the lighting module 280 includes four lighting lamps. The two cameras are respectively arranged on two opposite sides of the movable block 230, and the four lighting lamps are arranged on the remaining sides of the movable block 230, so that the two lighting lamps close to one of the cameras can provide sufficient light for the shooting environment of the camera, and the two lighting lamps close to the other camera can provide sufficient light for the shooting environment of the camera, thereby improving the overall shooting quality of the camera module 240.

[0053] Among them, the first driver 250 can be fixed on the movable block 230 in a variety of ways. For example, the first driver 250 is fixedly connected to the movable block 230 through mechanical fixing parts. Specifically, the first driver 250 and the mechanical fixing parts can be welded and fixed, and then the mechanical fixing parts are welded to the movable block 230. For another example, the first driver 250 is glued and fixed to the movable block 230 by chemical reagents.

[0054] Based on this, by setting a threaded through hole 231 on the movable block 230, and threading the threaded through hole 231 with the threaded rod 220, it is equivalent to movably connecting the movable block 230 with the threaded rod 220, fixing the first driver 250 on the movable block 230, and under the action of the controller, the movable block 230 can be driven by the first driver 250 to move on the threaded rod 220, so that the movable block 230 is close to or away from the first joint 210. Since the camera module 240 is set on the side of the movable block 230, when the movable block 230 is When the threaded rod 220 moves, the camera module 240 can move synchronously with the movable block 230, so as to flexibly adjust the shooting position of the camera module 240, effectively improving the adaptability of the camera module 240 to the limited movement space. Under the action of the controller, the camera module 240 can effectively shoot the electric wires located in the limited movement space; and a lighting module 280 is also provided on the side of the movable block 230, so the lighting module 280 will also move synchronously with the camera module 240. The modules 280 are distributed at intervals along the circumference of the movable block 230, so the light emission direction of the lighting module 280 will not face the camera module 240, so that the lighting module 280 can provide sufficient light to the space where the electric wire is located without affecting the shooting of the camera module 240, thereby improving the shooting quality of the camera module 240; in addition, the first joint 210 is connected to the angle adjustment mechanism 300, and the angle adjustment mechanism 300 is fixed on the base 100. Under the action of the angle adjustment mechanism 300, the angle of the first joint 210 can be flexibly adjusted. When the angle of the first joint 210 changes, the angle of the camera module 240 can be driven to change, so the shooting angle of the camera module 240 can be flexibly adjusted, further improving the adaptability of the camera module 240 to the restricted movement space, so that the camera module 240 can cope with more complex restricted movement spaces, ensuring that the camera module 240 can effectively shoot the electric wires located in the restricted movement space, so that all electric wires at the fire scene can be detected for metal melting marks, which is helpful for investigating the cause of the fire.

[0055] Reference Figure 2An embodiment of the present application provides a metal melt mark detection device, wherein a first gear 232 is provided on a movable block 230, and a second gear 251 meshing with the first gear 232 is provided at a driving end of a first driver 250, and an axis of the first gear 232 coincides with an axis of a threaded rod 220, and the first driver 250 is used to drive the second gear 251 to rotate, so as to drive the first gear 232 to rotate around the axis of the threaded rod 220.

[0056] Based on this, the first driver 250 drives the second gear 251 to drive the first gear 232 to rotate, and the first gear 232 drives the movable block 230 to move on the threaded rod 220 or rotate around the axis of the threaded rod 220, so that the movable block 230 can flexibly adjust its position in a limited movement space, facilitate the adjustment of the shooting angle and the lighting angle, and obtain images with different effects.

[0057] Specifically, the movable block 230 may be driven in a variety of ways, which are not limited in the embodiment of the present application.

[0058] For example, in an optional driving mode, the first gear 232 rotates to drive the movable block 230 to rotate along the thread direction of the threaded rod 220, so that the threaded through hole 231 rotates and climbs or rotates and descends on the threaded rod 220. Since the movable block 230 is provided with a threaded through hole 231, the movable block 230 will move on the threaded rod 220 along with the threaded through hole 231, so that the movable block 230 is close to or away from the first joint 210.

[0059] For another example, in another optional driving mode, a rolling bearing or a sliding bearing is provided in the movable block 230, the inner ring of the rolling bearing or the sliding bearing is provided with a threaded through hole 231, and the sliding block is arranged on the outer ring of the rolling bearing or the sliding bearing. When the movable block 230 needs to move up and down on the threaded rod 220, the first gear 232 and the inner ring of the rolling bearing or the sliding bearing obtain a larger torque, so that the inner ring and the outer ring of the rolling bearing or the sliding bearing move relative to each other. At this time, the friction force obtained by the outer ring is not enough to drive the movable block 230 to rotate, so that the movable block 230 will not rotate during the movement. When the movable block 230 needs to rotate to adjust the shooting angle, the threaded rod 220 and the outer ring of the rolling bearing or the sliding bearing obtain a smaller torque, so that the inner ring and the outer ring of the rolling bearing or the sliding bearing are relatively still, so that the movable block 230 rotates around the axis of the threaded rod 220.

[0060] It can be seen that in the above two driving methods, since both use the gear drive method, the first gear 232 can obtain sufficient torque by setting the gear ratio of the first gear 232 and the second gear 251, so that the connecting part of the threaded through hole 231 can rotate faster, driving the movable block 230 to move on the threaded rod 220 more efficiently. This method can also be used in other situations where different torques are required. In addition, the torque of the first driver 250 can be adjusted by the control module 400 to make the first gear 232 and the threaded through hole 231 obtain different torques, which is not limited in the embodiments of the present application.

[0061] Reference Figure 3 An embodiment of the present application provides a metal melt mark detection device, wherein the detection mechanism 200 includes a second driver 260 , which is disposed on the first joint 210 , and the second driver 260 is used to drive the threaded rod 220 to rotate around the axis of the threaded rod 220 .

[0062] The second driver 260 does not directly drive the threaded rod 220 to rotate.

[0063] Based on this, by indirectly driving the threaded rod 220 through the second driver 260, different driving modes can be set according to different situations, which has greater flexibility and scene adaptability.

[0064] Reference Figure 3 An embodiment of the present application provides a metal melt mark detection device, a third gear 221 is provided on the threaded rod 220, a fourth gear 261 meshing with the third gear 221 is provided at the driving end of the second driver 260, and the second driver 260 is used to drive the fourth gear 261 to rotate, thereby driving the third gear 221 to rotate around the axis of the threaded rod 220.

[0065] It should be noted that when the threaded through hole 231 is fixedly connected to the movable block 230, since the movable block 230 will also rotate when moving up and down along the threaded rod 220, it is important for the movable block 230 to restore the shooting angle in time. At this time, the threaded rod 220 requires a larger torque, and the gear ratio of the third gear 221 and the fourth gear 261 can be adjusted so that the third gear 221 obtains a larger torque. This driving mode is characterized by a faster circumferential capture capability; when the threaded through hole 231 and the movable block 230 are connected by a rolling bearing or a sliding bearing, the gear ratio of the third gear 221 and the fourth gear 261 can be adjusted so that the third gear 221 obtains a smaller torque than the first gear 232, so that the inner ring and the outer ring of the rolling bearing or the sliding bearing are relatively stationary, driving the movable block 230 to rotate. This driving mode is characterized by a faster longitudinal capture capability.

[0066] Based on this, the second driver 260 drives the threaded rod 220 through gears, which improves the rotation accuracy of the threaded rod 220. At the same time, different gear combinations can be set to achieve different driving effects. For example, the first driving mode is more suitable for shooting a single metal wire with dense metal melt marks, and the second driving mode is more suitable for shooting multiple metal wires with sparse metal melt marks around. The staff can choose metal melt mark detection equipment with different driving modes according to the scene of the fire, which has good scene adaptability.

[0067] In a possible implementation, when the threaded through hole 231 is connected to the movable block 230 via a rolling bearing or a sliding bearing, the driving end of the second driver 260 is connected to one end of the transmission belt, and the other end of the transmission belt is connected to the threaded rod 220. Based on this, the transmission belt drive method with low transmission efficiency can effectively avoid giving too much instantaneous torque to the threaded rod 220, resulting in inconsistent rotation speeds of the movable block 230 and the threaded rod 220, which is conducive to relatively smooth rotation of the movable block 230, and when selecting the second driver 260, there can be more redundant space in the torque parameter, which facilitates the selection of the second driver 260, especially when it is easier to determine the replacement driver model during maintenance.

[0068] Reference Figure 3 The embodiment of the present application provides a metal melt mark detection device, the angle adjustment mechanism 300 includes a second joint 310, a support arm 320, a rotation driver 330, a third driver 340 and a fourth driver 350, the rotation driver 330 is fixed on the base 100, the driving end of the rotation driver 330 is connected to the second joint 310, one end of the support arm 320 is rotatably connected to the second joint 310, and the other end of the support arm 320 is rotatably connected to the first joint 210, the rotation driver 330 is used to drive the second joint 310 to rotate, the third driver 340 is used to drive the support arm 320 to rotate, and the fourth driver 350 is used to drive the first joint 210 to rotate.

[0069] Among them, the second joint 310 is provided with a through hole for allowing the rotating shaft to pass through, and the end of the support arm 320 close to the second joint 310 is provided with a toothed through hole, and the rotating shaft of the third driver 340 passes through the through holes of the second joint 310 and the support arm 320, so that the second joint 310 can drive the support arm 320 to rotate through the third driver 340; a through hole is also provided at the bottom center of the second joint 310 for rotationally connecting with the rotating driver 330.

[0070] Among them, the interior of the support arm 320 is hollow, and various electrical circuits can be placed in it, which can reduce the weight of the equipment. A through hole is set at the end of the support arm 320 connected to the first joint 210, and the fourth driver 350 connects the support arm 320 and the first joint 210 together through the through hole of the support arm 320 and the through hole of the first joint 210.

[0071] Based on this, the angle adjustment mechanism 300 can rotate 360° around the axis of the base 100, and the support arm 320 can rotate in the direction of the angle between the support arm 320 and the base 100, and the first joint 210 can rotate in the direction of the angle between the first joint 210 and the support arm 320, so that the entire metal melt mark detection equipment has multiple degrees of freedom, and can better adjust the position of the detection mechanism 200 in a restricted space, which is conducive to multi-angle shooting of metal wires in the ruins of a fire scene, and is more suitable for post-disaster detection scenarios of fires.

[0072] like Figure 6 As shown, Figure 6 An optional flow chart of a metal melt mark detection method provided in an embodiment of the present application, the metal melt mark detection method can be performed by a metal melt mark detection device, and the metal melt mark detection method includes but is not limited to the following steps S610 to S620:

[0073] Step S610, obtaining an image of the target metal wire to be tested taken by a camera module.

[0074] Step S620, inputting the image of the conductor to be tested into the melt mark detection model for target recognition to obtain a target recognition result.

[0075] The surface of the target metal wire includes a target melt mark, and the target recognition result is used to indicate the type of the target melt mark.

[0076] Based on this, by acquiring the image of the target metal wire to be tested taken by the camera module, the image of the target metal wire to be tested is input into the melt mark detection model for target recognition, and the target recognition result of the melt mark is obtained, thereby improving the objectivity, efficiency and accuracy of metal melt mark detection.

[0077] In addition, refer to Figure 7 In one embodiment, the image of the conductor to be tested is input into the melt mark detection model for target recognition. Before obtaining the target recognition result, the metal melt mark detection method further includes but is not limited to the following steps S710 to S760:

[0078] Step S710: acquiring a sample image of a sample metal wire and an identification label of the sample image.

[0079] Among them, the surface of the sample metal wire contains sample melt marks, and the identification label is used to indicate the location and type of the sample melt marks; the surface of the sample metal wire contains sample melt marks, and the identification label may specifically include a label boundary box and a label category. Since the sample melt marks in the sample image will be located within the label boundary box, the label boundary box can indicate the true location of the sample melt marks, and the label category can indicate the true type of the sample melt marks. The true type of the sample melt marks may be one of a variety of melt mark categories, and the melt mark categories may include primary short circuit and secondary short circuit. Generally, a sufficient number of sample images of sample metal wires need to be prepared in advance to achieve the target training effect. In addition, the original data set needs to be expanded by adjusting the brightness of each sample image, adding Gaussian noise, horizontally flipping, etc. to simulate the noise environment and complex application scenarios of different lighting conditions at the scene after the fire disaster. The number of images in the sample image set after data enhancement has doubled.

[0080] Based on this, the robustness of the subsequent training model was improved by collecting a sufficient number of sample images and performing special data enhancement on the sample images to simulate the post-fire scene.

[0081] Step S720, calling the melt mark detection model to perform multi-level feature extraction on the sample image, to obtain first image features and second image features with successively decreasing dimensions.

[0082] Among them, reference Figure 8 , the sample image is input into the feature extraction network for multi-level feature extraction. The feature extraction network includes a plurality of sequentially connected convolutional layers. For example, the feature extraction network may include 10 sequentially connected convolutional layers. The sample image is input into the first convolutional layer in the feature extraction network, and then the output of the sixth convolutional layer in the feature extraction network is used as the first initial feature, the output of the eighth convolutional layer in the feature extraction network is used as the second initial feature, and the output of the tenth convolutional layer in the feature extraction network is used as the third initial feature. The second initial feature is subjected to feature extraction to obtain the first image feature, the third initial feature is subjected to feature extraction to obtain the front second image feature, and the front second image feature is input into the convolutional layer CBL14 for convolution to obtain the second image feature.

[0083] Among them, reference Fig. 9 ,The first convolutional layer is the convolutional layer CBL0, which is composed of a convolutional module, a batch normalization module, and an activation module connected in sequence, while the other nine convolutional layers are composed of depthwise convolution and point-by-point convolution.

[0084] Among them, reference Figure 8 , the 2nd to 10th convolutional layers are all depth-wise separable convolutions.

[0085] It should be noted that the feature extraction operation for the third initial feature is different from other feature extraction operations of the metal melt mark detection model. The multi-scale feature extraction module performs this process. Specifically, refer to Fig.10 , the third initial feature is input into the convolution layer CBL1 for convolution to obtain a first convolution result, and then the first convolution result is input into three maximum pooling layers for maximum pooling, and then the first convolution result and the three maximum pooling results are spliced ​​to obtain a first splicing result, the third initial feature is input into the convolution layer CBL2 for convolution to obtain a second convolution result, the first splicing result is input into the convolution layer CBL3 for convolution to obtain a third convolution result, the second convolution result and the third convolution result are spliced ​​and input into the convolution layer CBL4 for convolution to obtain a fourth convolution result, at this time, the fourth convolution result is determined as the pre-second image feature.

[0086] Based on this, multi-level feature extraction of sample images can promote model feature fusion, improve model performance, and enhance the generalization ability of the model. The specific design of 10 convolutional layers of the melt mark detection model can reduce the number of model parameters and computational complexity, and improve the running speed of the model.

[0087] Step S730, channel compression is performed on the second image feature to obtain a compressed image feature, content encoding is performed on the compressed image feature to obtain a plurality of initial recombinant cores, and normalization is performed on each of the initial recombinant cores to obtain a plurality of target recombinant cores.

[0088] Specifically, refer to Fig.12 , performing channel compression on the second image feature to obtain a compressed image feature, performing content encoding on the compressed image feature to obtain a plurality of initial recombinant kernels, and performing normalization processing on each initial recombinant kernel to obtain a plurality of target recombinant kernels.

[0089] It can be seen that the dimension of the second image feature is compressed by a 1×1 convolution to obtain the image compression feature, and the compressed image feature is divided into blocks. The compressed image feature is divided into twenty-five groups according to the width and height, and each group is divided into four blocks according to the depth. Every four blocks of features are seamlessly arranged and spliced ​​to form an initial recombinant kernel with a width and height twice as large. Each initial recombinant kernel is normalized to obtain a recombinant kernel.

[0090] Step S740: up-sample the second image features based on each target reorganization kernel to obtain a first intermediate feature.

[0091] The first intermediate feature has the same dimension as the first image feature.

[0092] Specifically, refer to Fig.12The up-sampling operation is specifically as follows: multiplying the recombinant kernel by the second image feature region at the corresponding position and the same size as the recombinant kernel, and combining the results to obtain the first intermediate feature.

[0093] Based on this, by processing the second image features to obtain the recombinant kernel corresponding to each position, the second image features can be subsequently upsampled with the help of each recombinant kernel, so that the model can aggregate contextual information in a larger receptive field, enhance the feature fusion capability of the network, and improve the detection capability of the model while increasing the number of parameters.

[0094] Step S750, performing bounding box prediction and category prediction based on the channel cascade result of the first intermediate feature and the first image feature to obtain a predicted recognition result.

[0095] Among them, the channel cascade result is obtained by channel cascading the first intermediate feature and the first image feature. Channel cascading refers to splicing the first intermediate feature and the first image feature in the channel dimension. The predicted recognition result includes a predicted bounding box and a predicted probability distribution corresponding to the predicted bounding box. The predicted bounding box is used to indicate the predicted position of the sample melt mark. The predicted probability distribution includes the predicted probabilities of multiple melt mark categories. Since the melt mark category with the highest prediction probability can be used as the prediction type, the predicted type of the sample melt mark can be determined through the predicted probability distribution.

[0096] Based on this, since the channel cascade result is obtained by channel cascading the first intermediate feature and the first image feature, the channel cascade result can simultaneously contain the information of the first intermediate feature and the first image feature, that is, the channel cascade result can contain more comprehensive image information of the sample image, and when bounding box prediction and category prediction are performed based on the channel cascade result, the recognition accuracy of the predicted recognition result can be effectively improved.

[0097] In one possible implementation, reference Figure 8 , performing bounding box prediction and category prediction based on the channel cascade result of the first intermediate feature and the first image feature to obtain a predicted recognition result, specifically, extracting the first initial feature to obtain a reference image feature, and the dimensions of the reference image feature, the first image feature, and the second image feature are reduced in sequence;

[0098] Then, the first intermediate feature and the first image feature are channel-concatenated to obtain a channel-concatenated result;

[0099] Then, the channel cascade result is input into the reference convolution module for convolution to obtain a second intermediate feature, and then the second intermediate feature is convolved and upsampled to obtain a third intermediate feature, wherein the upsampling process here can be consistent with the above-mentioned upsampling process;

[0100] Then, the third intermediate feature is concatenated with the reference image feature through channels and then convolved to obtain a fourth intermediate feature. For example, the convolution module having the same structure as the reference convolution module can be used for convolution.

[0101] Then, the fourth intermediate feature is convolved with the second intermediate feature and channel-concatenated to obtain the fifth intermediate feature;

[0102] Then, the fifth intermediate feature is convolved to obtain a sixth convolution feature, for example, the convolution may be performed using a convolution module having the same structure as the reference convolution module;

[0103] Then, after performing multiple feature extractions on the sixth convolution feature, a first detection feature set is obtained;

[0104] Then, the sixth convolution feature is extracted and channel-concatenated with the second image feature to obtain the sixth intermediate feature;

[0105] Then, the sixth intermediate feature is convolved to obtain a seventh convolution feature, for example, the convolution may be performed using a convolution module having the same structure as the reference convolution module;

[0106] Finally, the seventh convolution feature is subjected to multiple feature extractions to obtain a second detection feature set.

[0107] Among them, the first detection feature set and the second detection feature set are both used to describe the detection frames, predicted recognition results, and accuracy corresponding to the metal melt marks in the sample image. The difference is that the detection frames of the first detection feature set are more in number and smaller in specifications than those of the second detection feature set, and are respectively used to detect small metal melt marks and large metal melt marks in the picture.

[0108] Among them, it is also necessary to screen the first detection feature set and the second detection feature set. For the overlapping high accuracy results and low accuracy results, the low accuracy results should be deleted, and the first detection feature set and the second detection feature set after multiple rounds of screening should be used as the predicted recognition results.

[0109] Based on this, multiple feature extractions and channel cascades are performed on the channel cascade results of the first intermediate feature and the first image feature, and multi-level features can be fused to output a first detection feature set. Multiple channel cascades and feature extractions are then performed together with the second image feature, and multi-level features can be fused to output a second detection feature set. By screening the first detection feature set and the second detection feature set, more refined and accurate results can be obtained.

[0110] It should be noted that the reference Fig.11The reference convolution module includes convolution layer CBL6, convolution layer CBL7, convolution layer CBL8, convolution layer CBL9 and convolution layer CBL10; taking the determination process of the second intermediate feature as an example, the processing process of the reference convolution module is described in detail below.

[0111] First, the channel cascade result is input to the sequentially connected convolutional layers CBL6, CBL7, and CBL8 to obtain the convolution result A, and the channel cascade result is input to the convolutional layer CBL6 to obtain the result B, and then the channel cascade result is input to the sequentially connected convolutional layers CBL6 and CBL7 to obtain the convolution result C, and the channel cascade result is input to the convolutional layer CBL9 to obtain the convolution result D;

[0112] Then, the convolution result A, the convolution result B, the convolution result C and the convolution result D are channel-concatenated to obtain the convolution result E;

[0113] Finally, the cascade result E is input into the convolution layer CBL10 to obtain the second intermediate feature.

[0114] In one possible implementation, reference Fig.13 For the convolution layers CBL6, CBL7, CBL8, CBL9 and CBL10 of the reference convolution module, the convolution operations involved are all lightweight convolution operations. Taking the determination process of the second intermediate feature as an example, it is necessary to first perform conventional convolution on the channel cascade result to obtain conventional convolution features, and then perform linear transformation on the conventional convolution features and splice the two output results in the channel direction to obtain lightweight convolution results. Based on this, by adding linear transformation and channel splicing to the convolution operation, the redundant information extraction of the melt mark detection model in the stacked convolution structure can be reduced, the feature extraction capability can be enhanced, and the model parameters and calculation amount can be reduced.

[0115] Step S760, determining the model loss according to the predicted recognition result and the recognition label, and training the melt mark detection model based on the model loss.

[0116] Specifically, since the predicted recognition result includes the predicted bounding box and the predicted probability distribution corresponding to the predicted bounding box, and the recognition label includes the label bounding box and the label category, the true probability distribution of multiple melt mark categories can be determined by the label category. For example, assuming that the multiple melt mark categories include type A, type B and type C, and the label category is type A, the true probability distribution can be [1,0,0], which is equivalent to the true probability value of type A being 1, and the true probability values ​​of type B and type C being 0.

[0117] Therefore, determining the model loss based on the predicted recognition result and the recognition label is equivalent to determining the first difference between the predicted bounding box and the label bounding box, and determining the second difference between the predicted probability distribution and the true probability distribution, and then determining the model loss based on the first difference and the second difference, and then backpropagating the model loss and adjusting the model melt mark detection model parameters to minimize the model loss, and repeating the process from step S710 to step S760 until the predetermined number of training times is reached or the target training effect is achieved.

[0118] Based on this, model loss is used to characterize the accuracy of target recognition by the melt mark detection model. Generally speaking, the smaller the model loss value, the higher the accuracy of target recognition by the melt mark detection model. During the training process, the model loss is determined and the model parameters are adjusted by back-propagating the model loss to continuously optimize the model loss, so that the melt mark detection model reaches the optimal state, thereby improving the recognition accuracy of the model.

[0119] In addition, refer to Fig.14 The present application also provides a metal melting mark detection device 1400, comprising:

[0120] An acquisition module 1410 is used to acquire an image of a target metal wire to be tested taken by a camera module, wherein a surface of the target metal wire includes a target melting mark;

[0121] The prediction module 1420 is used to input the image of the conductor to be tested into the melt mark detection model for target recognition to obtain a target recognition result, wherein the target recognition result is used to indicate the type of the target melt mark.

[0122] It can be understood that the specific implementation of the metal melt mark detection device 1400 is basically the same as the specific embodiment of the above-mentioned metal melt mark detection method, and will not be repeated here.

[0123] In addition, refer to Fig.15 , Fig.15 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0124] The processor 1501 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0125] The memory 1502 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1502 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1502, and the processor 1501 calls and executes the metal melting mark detection method of the embodiment of the present application, for example, executing the above-described Figure 6 Steps S610 to S620 of the method, Figure 7 Steps S2710 to S760 of the method;

[0126] Input / output interface 1503, used to implement information input and output;

[0127] Communication interface 1504, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0128] A bus 1505 that transmits information between the various components of the device (e.g., the processor 1501, the memory 1502, the input / output interface 1503, and the communication interface 1504);

[0129] The processor 1501 , the memory 1502 , the input / output interface 1503 and the communication interface 1504 are connected to each other in communication within the device via the bus 1505 .

[0130] The present application also provides a storage medium, which is a computer-readable storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned metal melt mark detection method, for example, to execute the above-described Figure 6 Steps S610 to S620 of the method, Figure 7 Method steps S2710 to S760.

[0131] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0132] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0133] It can be understood by those skilled in the art that Figures 1 to 15 The technical solutions shown in the figure do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than those shown in the figure, or a combination of certain steps, or different steps.

[0134] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0135] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0136] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0137] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0138] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0139] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0140] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0141] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0142] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A metal melting mark detection device, characterized in that: include: Pedestal; A detection mechanism, comprising a first joint, a threaded rod, a movable block, a camera module, a lighting module and a first driver, wherein one end of the threaded rod is arranged on the first joint, the movable block is provided with a threaded through hole, the threaded through hole is threadedly connected to the threaded rod, the camera module and the lighting module are both arranged on the side of the movable block, the camera modules and the lighting modules are distributed at intervals along the circumference of the movable block, the first driver is fixed on the movable block, and the first driver is used to drive the movable block to move on the threaded rod so that the movable block is close to or away from the first joint; An angle adjustment mechanism is fixed on the base, the first joint is connected to the angle adjustment mechanism, and the angle adjustment mechanism is used to adjust the angle of the first joint; A control module, wherein the first driver, the camera module and the lighting module are electrically connected to the control module respectively.

2. The metal melting mark detection device according to claim 1, characterized in that: The movable block is provided with a first gear, and the driving end of the first driver is provided with a second gear meshing with the first gear. The axis of the first gear coincides with the axis of the threaded rod. The first driver is used to drive the second gear to rotate, so as to drive the first gear to rotate around the axis of the threaded rod.

3. The metal melting mark detection device according to claim 1, characterized in that: The detection mechanism comprises a second driver, which is arranged on the first joint and is used for driving the threaded rod to rotate around the axis of the threaded rod.

4. The metal melting mark detection device according to claim 3, characterized in that: The threaded rod is provided with a third gear, the driving end of the second driver is provided with a fourth gear meshing with the third gear, and the second driver is used to drive the fourth gear to rotate, thereby driving the third gear to rotate around the axis of the threaded rod.

5. The metal melting mark detection device according to claim 1, characterized in that: The angle adjustment mechanism includes a second joint, a support arm, a rotation driver, a third driver and a fourth driver. The rotation driver is fixed on the base, and the driving end of the rotation driver is connected to the second joint. One end of the support arm is rotationally connected to the second joint, and the other end of the support arm is rotationally connected to the first joint. The rotation driver is used to drive the second joint to rotate, the third driver is used to drive the support arm to rotate, and the fourth driver is used to drive the first joint to rotate.

6. A method for detecting metal melting marks, characterized in that: Applied to the metal melting mark detection device according to any one of claims 1 to 5, the metal melting mark detection method comprises: Acquire an image of a target metal wire to be tested taken by a camera module, wherein a surface of the target metal wire includes a target melting mark; The image of the conductor to be tested is input into a melt mark detection model for target recognition to obtain a target recognition result, wherein the target recognition result is used to indicate the position and type of the target melt mark.

7. The metal melting mark detection method according to claim 6, characterized in that: Before inputting the image of the conductor to be tested into the melt mark detection model for target recognition and obtaining the target recognition result, the metal melt mark detection method further includes: Acquire a sample image of a sample metal wire and an identification label of the sample image, wherein a surface of the sample metal wire includes a sample melt mark, and the identification label is used to indicate a position and type of the sample melt mark; Calling the melt mark detection model to perform multi-level feature extraction on the sample image to obtain first image features and second image features with successively decreasing dimensions; Performing channel compression on the second image feature to obtain a compressed image feature, performing content encoding on the compressed image feature to obtain a plurality of initial recombinant cores, and performing normalization processing on each of the initial recombinant cores to obtain a plurality of target recombinant cores; performing upsampling processing on the second image feature based on each of the target reorganization kernels to obtain a first intermediate feature, wherein the first intermediate feature has the same dimension as the first image feature; Perform bounding box prediction and category prediction based on the first intermediate feature and the channel concatenation result of the first image feature to obtain a predicted recognition result; A model loss is determined according to the predicted recognition result and the recognition label, and the melt mark detection model is trained based on the model loss.

8. A metal melting mark detection device, characterized in that: include: An acquisition module, used for acquiring an image of a target metal wire to be tested taken by a camera module, wherein a surface of the target metal wire includes a target melting mark; The prediction module is used to input the image of the conductor to be tested into a melt mark detection model for target recognition to obtain a target recognition result, wherein the target recognition result is used to indicate the type of the target melt mark.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the metal melt mark detection method according to any one of claims 6 to 7 when executing the computer program.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the metal melting mark detection method according to any one of claims 6 to 7 is implemented.

Images