Anchor breaking equipment with variable chassis and visual auxiliary lock positioning function and method

Abstract

The invention relates to anchor breaking equipment with a variable chassis and a visual auxiliary lock positioning function and a method, and belongs to the technical field of anchor breaking devices. The anchor breaking equipment comprises a variable chassis, a hydraulic mechanism, an electric control mechanism, an anchor breaking mechanical arm, a visual auxiliary positioning mechanism and a case; the variable chassis is installed below the machine box, the hydraulic mechanism and the electric control mechanism are installed in the machine box, the anchor breaking mechanical arm is installed in front of the machine box, the visual auxiliary positioning mechanism is installed on the anchor breaking mechanical arm and the machine box, and the hydraulic mechanism is in hydraulic connection with the variable chassis and the anchor breaking mechanical arm and used for supplying liquid to the variable chassis and the anchor breaking mechanical arm. The electric control mechanism is electrically connected with the variable chassis, the hydraulic mechanism, the anchor breaking mechanical arm and the visual auxiliary positioning mechanism and used for controlling the variable chassis, the hydraulic mechanism, the anchor breaking mechanical arm and the visual auxiliary positioning mechanism to act. The track distance can be adjusted according to needs, operation of anchor breaking equipment is not affected by the track distance, use is flexible and convenient, and the application range is wide.

Description

Technical Field

[0001] The present invention relates to the technical field of anchor breaking devices, and in particular to an anchor breaking device and method with a variable chassis and a visually assisted lock positioning function. Background Art

[0002] Anchor breaking in coal mine roadways is an essential process for the normal operation of fully mechanized mining faces. In traditional technologies, anchor breaking in underground coal mine roofs is mostly done manually with handheld anchor breakers, which is labor-intensive, has a high safety risk factor, and has low anchor breaking efficiency. Existing anchor breaking vehicles have problems such as difficulty in docking the anchor cable lock with the anchor breaking mechanism and instability of the machine body when walking. The above situation is not conducive to the safe, efficient, and intelligent mining of fully mechanized coal mine faces.

[0003] Jiangyin Changli Technology Co., Ltd. has published a patent for "a mining anchor removal vehicle" (application number 202111231465.4). With this mining anchor removal vehicle, anchor removal is no longer manually controlled on-site at close range, but can also be remotely controlled at a longer distance to avoid threats to personal safety; the anchor removal vehicle carries various equipment and replaces manual travel in narrow mine tunnels, reducing labor intensity. However, the track distance between the two tracks of the mining anchor removal vehicle chassis is fixed, which limits its use scenarios due to the influence of the track distance. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides an anchor breaking device and method with a variable chassis and visual auxiliary lock positioning function. The technical solution of the present invention is as follows:

[0005] The present invention provides an anchor-breaking device with a variable chassis and a visually assisted lock positioning function, which includes a variable chassis, a hydraulic mechanism, an electric control mechanism, an anchor-breaking mechanical arm, a visually assisted positioning mechanism and a chassis;

[0006] The variable chassis is installed below the chassis, the hydraulic mechanism and the electric control mechanism are installed in the chassis, the anchor-breaking mechanical arm is installed in front of the chassis, the visual-assisted positioning mechanism is installed on the anchor-breaking mechanical arm and the chassis, the hydraulic mechanism is hydraulically connected to the variable chassis and the anchor-breaking mechanical arm and is used to supply fluid to the variable chassis and the anchor-breaking mechanical arm, and the electric control mechanism is electrically connected to the variable chassis, the hydraulic mechanism, the anchor-breaking mechanical arm and the visual-assisted positioning mechanism and is used to control the actions of the variable chassis, the hydraulic mechanism, the anchor-breaking mechanical arm and the visual-assisted positioning mechanism;

[0007] The variable chassis comprises a first crawler frame and a second crawler frame arranged in parallel and at intervals, a chassis frame is arranged between the first crawler frame and the second crawler frame, two first rectangular columns are fixedly connected to the front and rear sides of the first crawler frame, and the two first rectangular columns are slidably connected to one side of the chassis frame, two second rectangular columns are fixedly connected to the front and rear sides of the second crawler frame, and the two second rectangular columns are slidably connected to the other side of the chassis frame, and the cylinder bodies of the first push oil cylinder and the second push oil cylinder are fixedly connected to the middle of the first crawler frame and the second crawler frame respectively, and the first push oil cylinder The piston rods of the cylinder and the second push cylinder are respectively connected to the two sides of the chassis frame through pins, the first crawler frame and the second crawler frame are both installed with crawlers on the periphery, the chassis driving mechanism is installed at the rear ends of the first crawler frame and the second crawler frame, the chassis driving mechanism is electrically connected to the electronic control mechanism, the first push cylinder and the second push cylinder are hydraulically connected to the hydraulic mechanism, two parallel side plates are fixedly connected to the top surface of the chassis frame, the chassis bent plates are fixedly connected to the two side plates, the chassis is fixed on the chassis bent plates, and the anchor breaking mechanical arm is installed at the front of the chassis bent plates.

[0008] Optionally, the anchor-breaking mechanical arm includes a swivel assembly, a lifting arm and a telescopic arm. The swivel assembly is connected to the front of the chassis bent plate, a lifting arm base is installed on the top surface of the swivel assembly, the bottom end of the lifting arm is hinged to the bottom of the lifting arm base, a mounting seat is hinged to the top of the lifting arm, one end of the telescopic arm is hinged to the mounting seat, a mounting ear is fixedly connected to a position on one side of the telescopic arm near the mounting seat, a cylinder body of a lifting cylinder is fixedly connected to the top of the lifting wall base, the piston rod of the lifting cylinder is hinged to the mounting ear, and the cylinder bodies of the telescopic cylinder are fixedly connected on both sides of the telescopic arm A first sleeve is slidably connected in the telescopic arm, a second sleeve is slidably connected in the first sleeve, the piston rods of the two telescopic cylinders are respectively connected to the first sleeve and the second sleeve, an end of the second sleeve away from the telescopic arm is fixedly connected to an end flange, the end flange is fixedly connected to the shell of the first swing cylinder, the rotating shaft of the first swing cylinder is connected to a connecting flange, the connecting flange is fixedly connected to the shell of the second swing cylinder, the rotating shaft of the second swing cylinder is fixedly connected to a U-shaped connecting frame, and the unopened side of the U-shaped connecting frame is fixedly connected to an anchor breaker.

[0009] Optionally, the visually assisted positioning mechanism includes a first depth camera, a second depth camera and a video display. The first depth camera is installed below the anchor breaker, the second depth camera is installed on the top surface of one end of the telescopic arm, and the video display is installed on the rear side of the chassis. The first depth camera, the second depth camera and the video display are all electrically connected to the electronic control mechanism.

[0010] Optionally, the anchor breaking equipment with variable chassis and visually assisted lock positioning function also includes a lifting cylinder, the cylinder body of the lifting cylinder is connected to the middle of the inner bottom surface of the chassis, the piston rod of the lifting cylinder extends from the top surface of the chassis, and the lifting cylinder is hydraulically connected to the hydraulic mechanism.

[0011] Optionally, an electric bell and a gas sensor are installed side by side on the front side of the chassis, and a headlight is installed on the top surface of the chassis. The headlight, the electric bell and the gas sensor are all electrically connected to the electric control mechanism.

[0012] Optionally, an angle sensor is installed on the rotating assembly, a displacement sensor is installed on the lifting cylinder and the telescopic cylinder, and an angle sensor is installed on the first swing cylinder and the second swing cylinder, and each angle sensor and displacement sensor is electrically connected to the electronic control mechanism; proportional valves are installed on the oil circuits of the lifting cylinder, the telescopic cylinder, the first swing cylinder and the second swing cylinder, and each proportional valve is electrically connected to the electronic control mechanism.

[0013] The present invention also provides an anchor breaking method of an anchor breaking device with a variable chassis and a visual auxiliary lock positioning function, which comprises:

[0014] S1, the electric control mechanism obtains the broken anchor image collected by the visual auxiliary positioning mechanism;

[0015] S2, the electric control mechanism inputs the broken anchor image into a pre-trained anchor lock recognition network deployed on its edge computer;

[0016] S3, anchor lock identification network analysis and output of the anchor lock position;

[0017] S4, visual auxiliary positioning mechanism displays the position of the anchor cable lock;

[0018] S5, when the anchor breaking instruction of the anchor cable lock is obtained, the anchor cable lock is broken through the cooperation of the electric control mechanism, the hydraulic mechanism and the anchor breaking mechanical arm.

[0019] Optionally, the anchor lock identification network adopts an improved YOLOv8 model, and the improved YOLOv8 model includes a backbone network module, a neck network module and a head network module, the output end of the backbone network module is connected to the input end of the neck network module, and the output end of the neck network module is connected to the input end of the head network module;

[0020] The backbone network module of the improved YOLOv8 model replaces the C2f module in the backbone network module of the YOLOv8 model with a C2f-MLCA module;

[0021] The MLCA processes data including:

[0022] S3.1, locally pool the input feature map in the spatial dimension through LAP operation;

[0023] S3.2, the local pooled features are divided into two parallel branches, the first branch performs GAP operation and reshaping operation on the local pooled features in turn; the second branch performs reshaping operation on the local pooled features;

[0024] S3.3, perform one-dimensional convolution operation on the features after the reshaping operation of the two branches, perform UNAP operation on the features after the one-dimensional convolution operation of the first branch, perform another reshaping operation on the features after the one-dimensional convolution operation of the second branch, and then fuse them with the features after the UNAP operation of the first branch through addition;

[0025] S3.4, perform UNAP operation on the features after addition fusion, and multiply fuse the features after UNAP operation with the input feature map to obtain the output of MLCA.

[0026] Optionally, the neck network module of the improved YOLOv8 model replaces the PANet structure in the neck network module of the YOLOv8 model with a RepGFPN structure; the RepGFPN performs feature extraction and processing through CSPStage when processing data;

[0027] When performing feature extraction and processing, the CSPStage includes:

[0028] S3-1, perform a concatenation operation on the input features and divide the concatenated features into two branches;

[0029] S3-2, the features of the first branch are passed through a convolution layer with a convolution kernel size of 1; the features of the other branch are first passed through a convolution layer with a convolution kernel size of 1, and then processed multiple times by ELAM;

[0030] S3-3, after fusing the features of the two branches through the connection operation, a convolution layer with a convolution kernel size of 1 is used to merge the processed feature maps and output the final result.

[0031] Optionally, when S3 is implemented, it includes:

[0032] S31, the anchor lock recognition network divides the broken anchor image into S×S grids, and extracts the feature vector of each grid through its feature extraction network;

[0033] S32, predicting a bounding box and a confidence score for each grid using a fully connected layer of an anchor lock recognition network according to a feature vector of each grid;

[0034] S33, select the bounding box with the highest confidence, and remove the duplicate bounding boxes through the non-maximum suppression algorithm to obtain the target box where the anchor cable lock is located, and use the target box as the position of the anchor cable lock.

[0035] All the above optional technical solutions can be combined arbitrarily, and the present invention does not provide detailed descriptions of the structures after the combinations.

[0036] By means of the above scheme, the beneficial effects of the present invention are as follows:

[0037] By setting up a variable chassis including a first push cylinder and a second push cylinder, the first push cylinder and the second push cylinder can be extended and retracted under the control of the hydraulic mechanism, thereby realizing the change of the track distance between the first crawler frame and the second crawler frame. When the anchor-breaking equipment passes through a single pillar, in order to maintain good passability, the piston rods of the first push cylinder and the second push cylinder are retracted, and the track distance is kept at a minimum. When the anchor-breaking equipment is in an operating condition, the piston rods of the first push cylinder and the second push cylinder are extended, and the track distance is kept at a maximum distance to improve the stability of the whole machine. Therefore, the embodiment of the present invention can adjust the track distance as needed, so that the operation of the anchor-breaking equipment is not affected by the track distance, and it is flexible and convenient to use and has a wide range of applications.

[0038] By setting up a visual auxiliary positioning mechanism, the position of the anchor cable lock can be intelligently identified and displayed with the cooperation of the electronic control mechanism, reducing the difficulty of the anchor breaker to dock with the anchor cable lock, making it easier for the operator to locate the position of the anchor cable lock, thereby improving the efficiency of anchor breaking.

[0039] In summary, the anchor breaking equipment can reduce the labor intensity of workers, reduce the safety risk factor, and improve the efficiency of anchor breaking. The whole machine is stable and has a wide range of applications.

[0040] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention and implement it according to the contents of the specification, the following is a detailed description of the preferred embodiments of the present invention in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic structural diagram of the anchor breaking equipment provided by an embodiment of the present invention at one viewing angle.

[0042] Figure 2 It is a schematic structural diagram of the anchor breaking equipment provided by an embodiment of the present invention from another perspective.

[0043] Figure 3 It is a structural schematic diagram of the anchor breaking equipment provided by an embodiment of the present invention from another perspective.

[0044] Figure 4 It is a structural schematic diagram of a variable chassis in an anchor breaking equipment provided by an embodiment of the present invention at a viewing angle.

[0045] Figure 5 It is a schematic structural diagram of a variable chassis in an anchor breaking equipment provided by an embodiment of the present invention from another perspective.

[0046] Figure 6 It is a schematic diagram of the structure of the anchor breaking mechanical arm in the anchor breaking equipment provided in an embodiment of the present invention.

[0047] Figure 7It is a schematic diagram of the partial structure of the anchor breaking mechanical arm in the anchor breaking equipment provided in an embodiment of the present invention.

[0048] Figure 8 It is a schematic diagram of the composition structure of the C2f-MLCA module in the anchor breaking method provided in an embodiment of the present invention.

[0049] Fig. 9 It is a schematic diagram of the process of Bottleneck and Bottleneck_MLCA ​​processing data in the anchor breaking method provided in an embodiment of the present invention.

[0050] Fig.10 It is a data processing flow chart of MLCA in the anchor breaking method provided in an embodiment of the present invention.

[0051] Fig.11 It is a schematic diagram of the network composition structure of RepGFPN in the anchor breaking method provided in an embodiment of the present invention.

[0052] Fig.12 It is a flowchart of feature extraction and processing performed by CSPStage in the anchor breaking method provided in an embodiment of the present invention.

[0053] The accompanying drawings are marked as follows: 1. variable chassis, 1-1. first crawler frame, 1-2. second crawler frame, 1-3. chassis frame, 1-4. first rectangular column, 1-5. second rectangular column, 1-6. first push cylinder, 1-7. second push cylinder, 1-8. crawler, 1-9. chassis driving mechanism, 1-10. side plate, 1-11. chassis bending plate, 2. hydraulic mechanism, 3. electric control mechanism, 4. anchor breaking mechanical arm, 4-1. slewing assembly, 4-2. lifting arm, 4-3. lifting cylinder, 4-4. telescopic arm , 4-5, telescopic cylinder, 4-6, first swing cylinder, 4-7, second swing cylinder, 4-8, anchor breaker, 4-9, U-shaped connecting frame, 4-10, connecting flange, 4-11, lifting wall base, 4-12, mounting seat, 4-13, mounting ear, 4-14, end flange, 4-15, first sleeve, 4-16, second sleeve, 5, lifting cylinder, 5-1, first depth camera, 5-2, second depth camera, 5-3, video display, 6, chassis, 7, car lights, 8, electric bell. DETAILED DESCRIPTION

[0054] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

[0055] like Figures 1 to 5As shown, the anchor-breaking equipment with a variable chassis and visually assisted lock positioning function provided by an embodiment of the present invention includes a variable chassis 1, a hydraulic mechanism 2, an electric control mechanism 3, an anchor-breaking mechanical arm 4, a visually assisted positioning mechanism and a chassis 6;

[0056] The variable chassis 1 is installed below the chassis 6, the hydraulic mechanism 2 and the electric control mechanism 3 are installed in the chassis 6, the anchor-breaking mechanical arm 4 is installed in front of the chassis 6, and the visual-assisted positioning mechanism is installed on the anchor-breaking mechanical arm 4 and the chassis 6. The hydraulic mechanism 2 is hydraulically connected to the variable chassis 1 and the anchor-breaking mechanical arm 4 and is used to supply fluid to the variable chassis 1 and the anchor-breaking mechanical arm 4. The electric control mechanism 3 is electrically connected to the variable chassis 1, the hydraulic mechanism 2, the anchor-breaking mechanical arm 4 and the visual-assisted positioning mechanism and is used to control the actions of the variable chassis 1, the hydraulic mechanism 2, the anchor-breaking mechanical arm 4 and the visual-assisted positioning mechanism.

[0057] The hydraulic mechanism 2 at least includes an oil tank, a hydraulic pump and a motor, which cooperate to enable the hydraulic mechanism 2 to supply fluid to the oil cylinders in the variable chassis 1 and the anchor-breaking mechanical arm 4. The electronic control mechanism 3 at least includes an edge computer and a human-computer interaction interface, which cooperate to control the actions of the variable chassis 1, the hydraulic mechanism 2, the anchor-breaking mechanical arm 4 and the visual auxiliary positioning mechanism.

[0058] When performing the anchor breaking operation, the hydraulic mechanism 2 and the electric control mechanism 3 are manually operated to make the variable chassis 1 move to a suitable position for breaking the anchor, and then the anchor breaking mechanical arm 4 is continuously operated to lift and rotate the anchor breaking mechanical arm 4 to a certain position to break the anchor.

[0059] In a specific embodiment, the variable chassis 1 includes a first crawler frame 1-1 and a second crawler frame 1-2 arranged in parallel and at intervals, a chassis frame 1-3 is arranged between the first crawler frame 1-1 and the second crawler frame 1-2, two first rectangular columns 1-4 are fixedly connected to the front and rear sides of the first crawler frame 1-1, and the two first rectangular columns 1-4 are slidably connected to one side of the chassis frame 1-3, two second rectangular columns 1-5 are fixedly connected to the front and rear sides of the second crawler frame 1-2, and the two second rectangular columns 1-5 are slidably connected to the other side of the chassis frame 1-3, and the cylinder bodies of the first push cylinder 1-6 and the second push cylinder 1-7 are fixedly connected to the middle of the first crawler frame 1-1 and the second crawler frame 1-2, respectively. The piston rods of 1-6 and the second push cylinder 1-7 are respectively connected to the two sides of the chassis frame 1-3 through pins, and the first crawler frame 1-1 and the second crawler frame 1-2 are both installed with crawlers 1-8 on the periphery, and the chassis driving mechanism 1-9 is installed at the rear ends of the first crawler frame 1-1 and the second crawler frame 1-2, and the chassis driving mechanism 1-9 is electrically connected to the electronic control mechanism 3. The first push cylinder 1-6 and the second push cylinder 1-7 are both hydraulically connected to the hydraulic mechanism 2, and two parallel side plates 1-10 are fixedly connected to the top surface of the chassis frame 1-3, and the chassis bent plates 1-11 are fixedly connected to the two side plates 1-10, and the chassis 6 is fixed on the chassis bent plates 1-11, and the anchor breaking mechanical arm 4 is installed in front of the chassis bent plates 1-11.

[0060] Among them, driven rollers and other components are installed at the front ends of the first crawler frame 1-1 and the second crawler frame 1-2. The chassis drive mechanism 1-9 includes at least a reducer, a gear and a hydraulic motor. It should be noted that the above content mainly introduces the relevant components of the variable chassis 1 for adjusting the track pitch. The specific components of the chassis drive mechanism 1-9 and other mechanisms for the variable chassis 1 to move and turn are not specifically limited in the embodiment of the present invention. For details, please refer to the relevant structures in the existing crawler mechanism.

[0061] In the embodiment of the present invention, the first push cylinder 1-6 and the second push cylinder 1-7 can be extended and retracted under the control of the hydraulic mechanism 2, thereby realizing the change of the track distance between the first track frame 1-1 and the second track frame 1-2. When the anchor-breaking equipment passes through the single pillar, in order to maintain good passability, the piston rods of the first push cylinder 1-6 and the second push cylinder 1-7 are retracted, and the track distance is kept at the minimum. When the anchor-breaking equipment is in an operating condition, the piston rods of the first push cylinder 1-6 and the second push cylinder 1-7 are extended, and the track distance is kept at the maximum distance to improve the stability of the whole machine. Therefore, the embodiment of the present invention can adjust the track distance as needed, so that the operation of the anchor-breaking equipment is not affected by the track distance, and it is flexible and convenient to use and has a wide range of applications.

[0062] In a specific embodiment, Figure 6 and Figure 7As shown, the anchor-breaking mechanical arm 4 includes a slewing assembly 4-1, a lifting arm 4-2 and a telescopic arm 4-4. The slewing assembly 4-1 is connected to the front part of the chassis bent plate 1-11, a lifting arm base 4-11 is installed on the top surface of the slewing assembly 4-1, the bottom end of the lifting arm 4-2 is hinged to the bottom of the lifting arm base 4-11, a mounting seat 4-12 is hinged to the top of the lifting arm 4-2, one end of the telescopic arm 4-4 is hinged to the mounting seat 4-12, a mounting ear 4-13 is fixedly connected to one side of the telescopic arm 4-4 near the mounting seat 4-12, a cylinder body of a lifting cylinder 4-3 is fixedly connected to the top of the lifting wall base 4-11, the piston rod of the lifting cylinder 4-3 is hinged to the mounting ear 4-13, and the two sides of the telescopic arm 4-4 are fixedly connected to the telescopic cylinder 4-5 Cylinder body, the first sleeve 4-15 is slidably connected in the telescopic arm 4-4, the second sleeve 4-16 is slidably connected in the first sleeve 4-15, the piston rods of the two telescopic cylinders 4-5 are respectively connected to the first sleeve 4-15 and the second sleeve 4-16, the second sleeve 4-16 is fixedly connected to the end away from the telescopic arm 4-4 with the end flange 4-14, the end flange 4-14 is fixedly connected to the shell of the first swing cylinder 4-6, the rotating shaft of the first swing cylinder 4-6 is connected with the connecting flange 4-10, the connecting flange 4-10 is fixedly connected to the shell of the second swing cylinder 4-7, the rotating shaft of the second swing cylinder 4-7 is fixedly connected with the U-shaped connecting frame 4-9, and the unopened side of the U-shaped connecting frame 4-9 is fixedly connected with the anchor breaker 4-8.

[0063] Among them, the rotating component 4-11 includes at least a rotating reducer and a hydraulic motor. Under the action of these components, the rotating component 4-11 can rotate in front of the chassis bent plate 1-11.

[0064] Specifically, when the anchor-breaking mechanical arm 4 is performing the anchor-breaking operation, the electric control mechanism 3 first controls the slewing assembly 4-11 to rotate to a certain position; then, the electric control mechanism 3 and the hydraulic mechanism 2 control the lifting cylinder 4-3 to lift to a suitable position, and further control the telescopic cylinder 4-5 to extend and retract a certain length, and control the first swing cylinder 4-6 and the second swing cylinder 4-7 to rotate a certain angle, so that the anchor breaker 4-8 is in a suitable anchor-breaking position, and the posture adjustment of the anchor breaker 4-8 when docking with the anchor cable lock is realized.

[0065] In a specific embodiment, the visual auxiliary positioning mechanism includes a first depth camera 5-1, a second depth camera 5-2 and a video display 5-3. The first depth camera 5-1 is installed below the anchor breaker 4-8, and can capture the position information of the anchor cable lock through the cutout of the anchor breaker 8, mainly solving the problem that it is difficult to manually observe the anchor cable lock into the anchor hole of the anchor breaker 8; the second depth camera 5-2 is installed on the top surface of one end of the telescopic arm 4-4, and is used to roughly capture the position information of the anchor cable lock on the top plate; the video display 5-3 is installed on the rear side of the chassis 6, and is used to display video data; the first depth camera 5-1, the second depth camera 5-2 and the video display 5-3 are all electrically connected to the electric control mechanism 3. The operator can observe the position information of the anchor cable lock on the video display 5-3. Using machine vision technology, after training, the first depth camera 5-1 and the second depth camera 5-2 can assist in the intelligent recognition of the anchor cable lock, and can display the position of the anchor cable lock on the video display 5-3.

[0066] In a specific embodiment, the anchor breaking equipment with a variable chassis and visually assisted lock positioning function further includes a lifting cylinder 5, the cylinder body of the lifting cylinder 5 is connected to the middle of the inner bottom surface of the chassis 6, the piston rod of the lifting cylinder 5 extends from the top surface of the chassis 6, and the lifting cylinder 5 is hydraulically connected to the hydraulic mechanism 2. Specifically, when breaking the anchor, the hydraulic mechanism 2 controls the piston rod of the lifting cylinder 5 to support the top plate of the roadway, thereby improving the stability of the whole machine during the anchor breaking operation.

[0067] In a specific embodiment, a bell 8 and a gas sensor are installed side by side on the front side of the chassis 6, and a headlight 7 is installed on the top surface of the chassis 6. The headlight 7, the bell 8 and the gas sensor are all electrically connected to the electric control mechanism 3. When the anchor breaking equipment is moving, the electric control mechanism 3 controls the bell 8 to ring to alert pedestrians. The gas sensor is used to measure gas data in real time and send it to the electric control mechanism 3. The headlight 7 is used to illuminate under the action of the electric control mechanism 3 when breaking the anchor.

[0068] In a specific embodiment, an angle sensor is installed on the rotating component 4-1, and the angle sensor is used to measure the rotation angle of the rotating component 4-1 in real time; displacement sensors are installed on the lifting cylinder 4-3 and the telescopic cylinder 4-5, and the displacement sensors are used to measure the telescopic displacement of the lifting cylinder 4-3 and the telescopic cylinder 4-5 in real time; angle sensors are installed on the first swing cylinder 4-6 and the second swing cylinder 4-7, and the angle sensors are used to measure the rotation angle of the first swing cylinder 4-6 and the second swing cylinder 4-7 in real time; each angle sensor and displacement sensor is electrically connected to the electric control mechanism 3; proportional valves are installed on the oil circuits of the lifting cylinder 4-3, the telescopic cylinder 4-5, the first swing cylinder 4-6 and the second swing cylinder 4-7, and each proportional valve is electrically connected to the electric control mechanism 3.

[0069] The lifting cylinder 4-3, telescopic cylinder 4-5, first swing cylinder 4-6 and second swing cylinder 4-7 are controlled by respective high-precision proportional valves, and each proportional valve and the above-mentioned actuators and sensors form a closed-loop feedback control circuit, thereby improving the control accuracy of the anchor-breaking mechanical arm 4.

[0070] The embodiment of the present invention further provides an anchor breaking method of an anchor breaking device with a variable chassis and a visually assisted lock positioning function, which comprises the following steps S1 to S5:

[0071] S1, the electric control mechanism 3 obtains the broken anchor image collected by the visual auxiliary positioning mechanism.

[0072] The broken anchor image is obtained by dividing the video data captured by the first depth camera 5-1 and / or the second depth camera 5-2 into frames.

[0073] S2, the electric control mechanism 3 inputs the broken anchor image into a pre-trained anchor lock recognition network deployed on its edge computer.

[0074] The specific composition structure of the anchor cable lock identification network is not specifically limited in the embodiment of the present invention. When it is implemented, a commonly used target recognition classification network can be used. In a specific embodiment, the anchor cable lock identification network uses an improved YOLOv8 model.

[0075] It should be noted that before inputting the broken anchor image into the pre-trained anchor cable lock recognition network, the anchor cable lock recognition network needs to be trained first. Specifically, when training the anchor cable lock recognition network, the following methods can be used:

[0076] S10. Collect data and create data sets.

[0077] Firstly, the first depth camera 5-1 and the second depth camera 5-2 installed underground are used to collect video data of the anchor cable lock, and the recorded video data is split into pictures frame by frame to form an original data set.

[0078] Due to the complex environment of the excavation working face, the real-time monitoring scene of the anchor breaking process is often under low illumination and dusty fog conditions, so the original data set needs to be expanded. However, it is difficult to obtain image data in the harsh environment of the mine, so the original data set is enhanced. First, the image synthesis fog enhancement algorithm is used to generate images in a dusty fog environment; then, the image brightness darkening processing algorithm is applied to generate images in a low illumination environment. Finally, these three data sets (original data set, dusty fog data set, low illumination data set) are integrated together to form a more diverse comprehensive data set.

[0079] Next, the comprehensive data set is annotated, and the anchor locks in the comprehensive data set images are annotated using Labelimg annotation software. After the annotation is completed, the system automatically generates an XML format file corresponding to each image. The XML format file content records the annotation category, image size, and the size and position of the annotation box in detail.

[0080] In order to facilitate subsequent model training, the XML files are uniformly converted to TXT format. The TXT file will be used as a label data set. The TXT file content contains several five-tuples (the number depends on the number of annotation boxes in the image). The first element of the five-tuple is the category information, and the next four elements are the coordinates of the two endpoints of the normalized annotation box. The specific position of the annotation box can be determined by these two endpoints.

[0081] Finally, the labeled dataset is divided into training set, validation set and test set in the ratio of 7:2:1.

[0082] S20. Build an improved YOLOv8 model as the anchor lock recognition network.

[0083] Specifically, the anchor lock identification network described in the embodiment of the present invention adopts an improved YOLOv8 model, which includes a backbone network module, a neck network module and a head network module. The output end of the backbone network module is connected to the input end of the neck network module, and the output end of the neck network module is connected to the input end of the head network module. The improvement scheme is: based on the YOLOv8 network model structure, the C2f module in the backbone network module is replaced with a C2f-MLCA module, and the PANet structure in the neck network module of the YOLOv8 model is replaced with a RepGFPN structure. The details are as follows:

[0084] S2.1. Improve C2f module.

[0085] The backbone network module of the YOLOv8 model is composed of a CBS module, a C2f module, and an SPPF module, wherein the C2f module contains many Bottleneck residual modules. These Bottleneck residual modules are connected in series to realize the extraction and fusion of features of different scales. However, the disadvantages of the Bottleneck residual module are also obvious. These Bottleneck residual modules cause the model to superimpose a large amount of information at the same position, and generate many redundant features in the process of obtaining key features, which may cause the loss of key information. Therefore, an embodiment of the present invention adds a lightweight attention mechanism module MLCA (mixed local channel attention) to the Bottleneck residual module, so that the key feature information extracted by it contains channel information, spatial information, local channel information and global channel information at the same time. The composition structure of the C2f-MLCA module is as follows: Figure 8 As shown. Fig. 9 As shown, (a) and (b) are schematic diagrams of the data processing process of Bottleneck and Bottleneck_MLCA ​​respectively.

[0086] MLCA aims to make up for the insufficiency of the traditional channel attention mechanism in capturing spatial feature information. Unlike the attention mechanism that only focuses on channel information, MLCA not only includes channel feature information, but also spatial feature information. The following is a detailed description of MLCA processing data.

[0087] In a specific embodiment, when the MLCA processes data, it includes:

[0088] S3.1, the input feature map is locally pooled in the spatial dimension through the LAP (local adaptive average pooling) operation.

[0089] Specifically, the input feature map of MLCA is a feature vector of shape C×W×H, where C is the number of channels, and W and H are the spatial dimensions (width and height). The input feature map is locally pooled in the spatial dimension through the LAP operation. After local pooling, the spatial dimension of the input feature map is compressed from W×H to ks×ks, and the shape of the output feature is C×ks×ks. This step retains the local spatial information and is used to capture fine-grained features.

[0090] S3.2, the local pooled features are divided into two parallel branches, the first branch performs GAP (global adaptive average pooling) operation and reshape operation on the local pooled features in turn; the second branch performs reshape operation on the local pooled features.

[0091] Specifically, the GAP operation converts the local pooled features into a vector of shape 1×C×ks×ks. The GAP operation retains the global feature information of the channel dimension while compressing the pooled spatial information into a more compact representation. The purpose of the reshaping operation is to adapt to the subsequent one-dimensional convolution operation. The second branch directly reshapes the local pooled features to ensure that the branch can retain the global features and prepare for the subsequent one-dimensional convolution operation.

[0092] S3.3, perform one-dimensional convolution (Conv1d) operation on the features after the reshaping operation of the two branches, perform UNAP operation on the features after the one-dimensional convolution operation of the first branch, and reshape the features after the one-dimensional convolution operation of the second branch again and fuse them with the features after the UNAP operation of the first branch through addition.

[0093] Specifically, both branches are operated through one-dimensional convolution (Conv1d), which can reduce computational complexity and further extract sequence or channel information. The one-dimensional convolution operation of the first branch is used to capture the complex relationship between local space and channel features. The second branch directly applies one-dimensional convolution operation to the global features after the reshaping operation to further extract global information.

[0094] The features after the one-dimensional convolution operation of the first branch are UNAPed (unpooled) to restore to the original spatial resolution C × W × H. The UNAP operation ensures that the feature map of this branch matches the input feature map in the spatial dimension.

[0095] The features after additive fusion not only contain the global context information (the second branch), but also combine the local spatial features (the first branch), making the feature expression richer and more comprehensive.

[0096] S3.4, perform UNAP operation on the features after addition fusion, and multiply fuse the features after UNAP operation with the input feature map to obtain the output of MLCA.

[0097] Specifically, the UNAP operation restores the fused features to a higher spatial resolution for more effective feature expression. The role of multiplication fusion is to highlight key features and suppress redundant information. The features finally output by MLCA contain both channel attention information and integrate global and local spatial features, making the model pay more attention to important information.

[0098] In summary, MLCA combines local and global spatial information with channel attention through a branching architecture, which not only retains channel information, but also enhances the performance of the attention mechanism by introducing spatial features. It uses local pooling and global pooling, convolution and unpooling operations to improve the model's attention to spatial and channel information without significantly increasing the amount of computation. The data processing flow of MLCA is as follows: Fig.10 shown.

[0099] S2.2. Improve feature fusion network.

[0100] YOLOv8 uses PANet to fuse multi-scale feature maps in the neck network module. By adding an additional bottom-up path, accurate positioning information in the lower layer is used to enhance the entire feature hierarchy, thereby shortening the information path between high-level and low-level features. However, this bottom-up path design lacks the interaction between high-level semantic information and low-level spatial information, resulting in poor multi-scale detection results. To address this problem, an embodiment of the present invention proposes an improved efficient hierarchical aggregation network, RepGFPN, as an alternative for the neck network module.

[0101] RepGFPN solves the above problems through skip connection structure and cross-scale connection, and effectively integrates multi-scale features. Its network structure is as follows Fig.11 As shown in the figure, P3-in, P4-in, and P5-in represent the feature maps from the three input layers respectively. The processed intermediate feature maps are P4-hid and P5-hid, which are extracted and processed by the multi-scale fusion module (CSPStage). The feature maps of the output layer are P3-out, P4-out, and P6-out, which output features of different scales respectively and are suitable for multi-scale detection tasks. Fig.11 The upward slanting arrow in the figure indicates a 2x downsampling operation, which is used to extract more abstract semantic information at a higher level. The downward slanting arrow indicates a 2x upsampling operation, which is used to refine the spatial resolution and supplement the position information of the lower layer. The upsampling operation uses the nearest neighbor interpolation method to restore spatial details, thereby improving detection accuracy while maintaining computational efficiency.

[0102] In a specific embodiment, the RepGFPN performs feature extraction and processing through CSPStage when processing data; the CSPStage structure is as follows: Fig.12 As shown, its input is 2 or 3 feature layers, which are divided into two branches after concat (connection), one branch passes through 1×1 convolution down channel, and the other branch passes through efficient layer aggregation network ELAN (efficient layer aggregation networks), and the two branches are concat again, and the output finally passes through 1×1 convolution down channel to obtain the final output. Specifically, when performing feature extraction and processing, the CSPStage includes:

[0103] S3-1, perform a connection operation on the input features and divide the features after the connection operation into two branches.

[0104] Specifically, CspStage receives the input feature map and splits the input into two branches through a Concat operation.

[0105] S3-2, the features of the first branch pass through a convolution layer with a convolution kernel size of 1; the features of the other branch first pass through a convolution layer with a convolution kernel size of 1, and then are processed multiple times by ELAM.

[0106] Specifically, the features of the first branch directly pass through a convolution layer with a kernel size of 1 (Conv K = 1), which is equivalent to adjusting the number of channels without changing the spatial dimension. The features of the other branch pass through a convolution with a kernel size of 1 (Conv K = 1), and then are processed multiple times by ELAM (repeated N times).

[0107] After receiving a feature map input, ELAM first passes through a RepConv. After RepConv, it passes through a convolution layer with a convolution kernel size of 3 (Conv K=3) to extract deeper spatial features. Then, an additive jump connection is used, which means that the input features and the results of Conv K=3 are additively fused.

[0108] After RepConv receives the input feature map, it divides the input feature map into two paths: Path 1, passes through a convolution layer with a kernel size of 3 and a stride of 2 (Conv2d K=3, s=2, p=1), followed by a BatchNorm2d layer and a ReLU activation function. Path 2, passes through a convolution layer with a kernel size of 1 and a stride of 2 (Conv2d K=1, s=2, p=0), followed by a BatchNorm2d layer. Then, the outputs of these two paths are fused together through an addition operation to form the final output.

[0109] S3-3, after fusing the features of the two branches through the connection operation, a convolution layer with a convolution kernel size of 1 is used to merge the processed feature maps and output the final result.

[0110] S30, model training and optimization.

[0111] First, the training parameters are set. According to the characteristics of the dataset and the computing power and video memory of the hardware device, the training parameters are set as follows: images of 640×640 pixels, 16 images per batch, 300 iterations, Adam optimizer, initial learning rate of 0.001, momentum value of 0.937, Moasic data enhancement method strength of 0.7, and Mixup data enhancement method strength of 0.5.

[0112] Secondly, the training set is input into the improved YOLOv8 model for iterative training, and verification is performed at the completion of each iteration. The weight file of the current model and the best model weight file in the previous training are saved. When the training reaches the set number of iterations, the training is ended and the best model weight file is saved.

[0113] Finally, after the training is completed, the final improved YOLOv8 model is tested using the validation set to calculate its mAP (average precision), accuracy, recall and other indicators in the target detection task. The model weight performance under different iterations during the training process is compared, and the optimal model is selected for deployment. The detection accuracy of anchor locks is verified, and the model parameters or data enhancement strategies are further adjusted as needed to improve the generalization ability of the model and the actual application effect.

[0114] S3, anchor lock identification network analysis and output of the anchor lock position.

[0115] In a specific embodiment, the S3, when implemented, includes:

[0116] S31, the anchor lock recognition network divides the broken anchor image into S×S grids, and extracts the feature vector of each grid through its feature extraction network.

[0117] Specifically, the broken anchor image is divided into S×S grids to decompose the broken anchor image into multiple small areas, and each grid is responsible for detecting one or more anchor locks that may exist in the broken anchor image. If the center point of the anchor lock falls into a certain grid, the grid will be responsible for predicting the location of the anchor lock. Through this division method, the anchor lock recognition network can detect multiple anchor locks in parallel and ensure good detection results even in complex scenes.

[0118] The feature extraction network is a convolutional neural network, which is responsible for extracting different levels of features from the broken anchor image. These feature vectors not only contain the boundary information of the anchor lock, but also capture the detailed features of the anchor lock, such as color, texture, and shape. The continuous application of convolutional layers enables the model to recognize objects in the image from local to global while maintaining computational efficiency.

[0119] S32, predicting a bounding box and a confidence level for each grid using a fully connected layer of an anchor lock recognition network according to the feature vector of each grid.

[0120] For each grid, the anchor lock recognition network predicts one or more bounding boxes and outputs the following information: (1) Bounding box location: The model predicts the center position, width, and height of the anchor lock. This information is used to determine the exact location of the anchor lock in the image. (2) Category: Whether it is an anchor lock. (3) Confidence score: The confidence score indicates the possibility that the bounding box contains the anchor lock. The higher the score, the more confident the model is in the prediction result. This score consists of two parts: classification confidence and anchor lock existence confidence.

[0121] S33, select the bounding box with the highest confidence, and remove the duplicate bounding boxes through the non-maximum suppression algorithm to obtain the target box where the anchor cable lock is located, and use the target box as the position of the anchor cable lock.

[0122] This step can avoid repeated detection of the same anchor fixture, thereby ensuring that each anchor fixture is marked by only a unique bounding box.

[0123] S4, the visually assisted positioning mechanism displays the position of the anchor cable lock.

[0124] Specifically, the video display 5-3 in the visual auxiliary positioning mechanism displays the position of the anchor cable lock, so that the operator can observe the position of the anchor cable lock from the video display 5-3.

[0125] S5, when the anchor breaking instruction of the anchor cable lock is obtained, the anchor cable lock is broken through the cooperation of the electric control mechanism 3, the hydraulic mechanism 2 and the anchor breaking mechanical arm 4.

[0126] When the operator determines the location of the anchor cable lock, he starts to operate the anchor breaking equipment to break the anchor. Specifically, the electric control mechanism 3 controls the hydraulic mechanism 2 to move the variable chassis 1 to a suitable position, and then continues to control the anchor breaking mechanical arm 4 to move the anchor breaker 4-8 to break the anchor.

[0127] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the technical principles of the present invention, and these improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An anchor breaking device with a variable chassis and visually assisted lock positioning function, characterized in that: It comprises a variable chassis (1), a hydraulic mechanism (2), an electric control mechanism (3), an anchor-breaking mechanical arm (4), a visual auxiliary positioning mechanism and a chassis (6); The variable chassis (1) is installed below the chassis (6), the hydraulic mechanism (2) and the electric control mechanism (3) are installed in the chassis (6), the anchor-breaking mechanical arm (4) is installed in front of the chassis (6), and the visual auxiliary positioning mechanism is installed on the anchor-breaking mechanical arm (4) and the chassis (6); the hydraulic mechanism (2) is hydraulically connected to the variable chassis (1) and the anchor-breaking mechanical arm (4) and is used to supply fluid to the variable chassis (1) and the anchor-breaking mechanical arm (4); the electric control mechanism (3) is electrically connected to the variable chassis (1), the hydraulic mechanism (2), the anchor-breaking mechanical arm (4) and the visual auxiliary positioning mechanism and is used to control the movement of the variable chassis (1), the hydraulic mechanism (2), the anchor-breaking mechanical arm (4) and the visual auxiliary positioning mechanism; The variable chassis (1) comprises a first crawler frame (1-1) and a second crawler frame (1-2) which are arranged in parallel and at intervals, a chassis frame (1-3) being arranged between the first crawler frame (1-1) and the second crawler frame (1-2), two first rectangular columns (1-4) being fixedly connected to the front and rear sides of the first crawler frame (1-1), and the two first rectangular columns (1-4) are slidably connected to one side of the chassis frame (1-3), two second rectangular columns (1-5) being fixedly connected to the front and rear sides of the second crawler frame (1-2), and the two second rectangular columns (1-5) are slidably connected to the other side of the chassis frame (1-3), the cylinder bodies of the first push oil cylinder (1-6) and the second push oil cylinder (1-7) are fixedly connected to the middle of the first crawler frame (1-1) and the second crawler frame (1-2), and the first push oil cylinder (1-6) and the second push oil cylinder (1-7) are respectively connected to the cylinder bodies of the first push oil cylinder (1-6) and the second push oil cylinder (1-7). The piston rods of the oil shifting cylinders (1-7) are respectively connected to the two sides of the chassis frame (1-3) through pins; crawlers (1-8) are installed on the peripheries of the first crawler frame (1-1) and the second crawler frame (1-2); chassis drive mechanisms (1-9) are installed at the rear ends of the first crawler frame (1-1) and the second crawler frame (1-2); the chassis drive mechanisms (1-9) are electrically connected to the electric control mechanism (3); the first push oil cylinder (1-6) and the second push oil cylinder (1-7) are hydraulically connected to the hydraulic mechanism (2); two mutually parallel side plates (1-10) are fixedly connected to the top surface of the chassis frame (1-3); chassis bent plates (1-11) are fixedly connected to the two side plates (1-10); the chassis (6) is fixed on the chassis bent plate (1-11); and the anchor breaking mechanical arm (4) is installed at the front of the chassis bent plate (1-11).

2. The anchor breaking equipment with a variable chassis and visually assisted lock positioning function according to claim 1 is characterized in that: The anchor-breaking mechanical arm (4) comprises a slewing assembly (4-1), a lifting arm (4-2) and a telescopic arm (4-4); the slewing assembly (4-1) is connected to the front of the chassis bent plate (1-11); a lifting arm base (4-11) is installed on the top surface of the slewing assembly (4-1); the bottom end of the lifting arm (4-2) is hinged to the bottom of the lifting arm base (4-11); a mounting seat (4-12) is hinged to the top of the lifting arm (4-2); one end of the telescopic arm (4-4) is hinged to the mounting seat (4-12); a mounting ear (4-13) is fixedly connected to one side of the telescopic arm (4-4) near the mounting seat (4-12); a cylinder body of a lifting cylinder (4-3) is fixedly connected to the top of the lifting wall base (4-11); a piston rod of the lifting cylinder (4-3) is hinged to the mounting ear (4-13); both sides of the telescopic arm (4-4) are fixedly connected to the telescopic cylinder (4-5 ), a first sleeve (4-15) is slidably connected in the telescopic arm (4-4), a second sleeve (4-16) is slidably connected in the first sleeve (4-15), piston rods of the two telescopic oil cylinders (4-5) are respectively connected to the first sleeve (4-15) and the second sleeve (4-16), one end of the second sleeve (4-16) away from the telescopic arm (4-4) is fixedly connected to an end flange (4-14), the end flange (4-14) is fixedly connected to a shell of a first swing oil cylinder (4-6), a rotating shaft of the first swing oil cylinder (4-6) is connected to a connecting flange (4-10), the connecting flange (4-10) is fixedly connected to a shell of a second swing oil cylinder (4-7), a rotating shaft of the second swing oil cylinder (4-7) is fixedly connected to a U-shaped connecting frame (4-9), and an anchor breaker (4-8) is fixedly connected to the unopened side of the U-shaped connecting frame (4-9).

3. The anchor breaking equipment with a variable chassis and visually assisted lock positioning function according to claim 2 is characterized in that: The visual auxiliary positioning mechanism comprises a first depth camera (5-1), a second depth camera (5-2) and a video display (5-3); the first depth camera (5-1) is installed below the anchor breaker (4-8); the second depth camera (5-2) is installed on the top surface of one end of the telescopic arm (4-4); the video display (5-3) is installed on the rear side of the chassis (6); and the first depth camera (5-1), the second depth camera (5-2) and the video display (5-3) are all electrically connected to the electric control mechanism (3).

4. The anchor breaking equipment with a variable chassis and visually assisted lock positioning function according to claim 1 is characterized in that: It also includes a lifting cylinder (5), the cylinder body of the lifting cylinder (5) is connected to the middle of the inner bottom surface of the chassis (6), the piston rod of the lifting cylinder (5) extends from the top surface of the chassis (6), and the lifting cylinder (5) is hydraulically connected to the hydraulic mechanism (2).

5. The anchor breaking equipment with a variable chassis and visually assisted lock positioning function according to claim 1 is characterized in that: An electric bell (8) and a gas sensor are installed side by side on the front side of the chassis (6), and a headlight (7) is installed on the top surface of the chassis (6). The headlight (7), the electric bell (8) and the gas sensor are all electrically connected to the electric control mechanism (3).

6. The anchor breaking equipment with a variable chassis and visually assisted lock positioning function according to claim 3 is characterized in that: An angle sensor is installed on the rotary assembly (4-1), a displacement sensor is installed on the lifting cylinder (4-3) and the telescopic cylinder (4-5), and an angle sensor is installed on the first swing cylinder (4-6) and the second swing cylinder (4-7), and each angle sensor and displacement sensor is electrically connected to the electric control mechanism (3); proportional valves are installed on the oil circuits of the lifting cylinder (4-3), the telescopic cylinder (4-5), the first swing cylinder (4-6) and the second swing cylinder (4-7), and each proportional valve is electrically connected to the electric control mechanism (3).

7. An anchor breaking method using the anchor breaking equipment with a variable chassis and visually assisted lock positioning function as claimed in any one of claims 1 to 6, characterized in that: include: S1, the electric control mechanism (3) obtains the broken anchor image collected by the visual auxiliary positioning mechanism; S2, the electric control mechanism (3) inputs the broken anchor image into a pre-trained anchor lock recognition network deployed on its edge computer; S3, anchor lock identification network analysis and output of the anchor lock position; S4, visual auxiliary positioning mechanism displays the position of the anchor cable lock; S5, when the anchor breaking instruction of the anchor cable lock is obtained, the anchor cable lock is broken through the cooperation of the electric control mechanism (3), the hydraulic mechanism (2) and the anchor breaking mechanical arm (4).

8. The anchor breaking method according to claim 7, characterized in that: The anchor lock identification network adopts an improved YOLOv8 model, which includes a backbone network module, a neck network module and a head network module, wherein the output end of the backbone network module is connected to the input end of the neck network module, and the output end of the neck network module is connected to the input end of the head network module; The backbone network module of the improved YOLOv8 model replaces the C2f module in the backbone network module of the YOLOv8 model with a C2f-MLCA module; The MLCA processes data including: S3.1, locally pool the input feature map in the spatial dimension through LAP operation; S3.2, the local pooled features are divided into two parallel branches, the first branch performs GAP operation and reshaping operation on the local pooled features in turn; the second branch performs reshaping operation on the local pooled features; S3.3, perform one-dimensional convolution operation on the features after the reshaping operation of the two branches, perform UNAP operation on the features after the one-dimensional convolution operation of the first branch, perform another reshaping operation on the features after the one-dimensional convolution operation of the second branch, and then fuse them with the features after the UNAP operation of the first branch through addition; S3.4, perform UNAP operation on the features after addition fusion, and multiply fuse the features after UNAP operation with the input feature map to obtain the output of MLCA.

9. The anchor breaking method according to claim 7 or 8, characterized in that: The neck network module of the improved YOLOv8 model replaces the PANet structure in the neck network module of the YOLOv8 model with the RepGFPN structure; the RepGFPN performs feature extraction and processing through CSPStage when processing data; When performing feature extraction and processing, the CSPStage includes: S3-1, perform a concatenation operation on the input features and divide the concatenated features into two branches; S3-2, the features of the first branch are passed through a convolution layer with a convolution kernel size of 1; the features of the other branch are first passed through a convolution layer with a convolution kernel size of 1, and then processed multiple times by ELAM; S3-3, after fusing the features of the two branches through the connection operation, a convolution layer with a convolution kernel size of 1 is used to merge the processed feature maps and output the final result.

10. The anchor breaking method according to claim 7, characterized in that: When S3 is specifically implemented, it includes: S31, the anchor lock recognition network divides the broken anchor image into S×S grids, and extracts the feature vector of each grid through its feature extraction network; S32, predicting a bounding box and a confidence score for each grid using a fully connected layer of an anchor lock recognition network according to a feature vector of each grid; S33, select the bounding box with the highest confidence, and remove the duplicate bounding boxes through the non-maximum suppression algorithm to obtain the target box where the anchor cable lock is located, and use the target box as the position of the anchor cable lock.

Images