Shockproof hammer displacement identification method and device, computer equipment and storage medium

Abstract

The invention relates to a vibration damper displacement identification method and device, computer equipment and a storage medium. The method comprises the steps of performing instance segmentation on a to-be-processed image corresponding to a power transmission line based on a trained instance segmentation model to obtain a target segmentation map corresponding to the to-be-processed image; the object in the target segmentation image comprises a shockproof hammer, or the object in the target segmentation image comprises the shockproof hammer and a fitting; on the basis of the position information of the contour frames of the objects in the target segmentation map, determining the overlapping degree between the contour frames; determining a depth difference value of each object based on depth information of each object in the target segmentation map; the depth information is used for representing the distance of each pixel point in the target segmentation image; and if the overlapping degree between two objects in the objects is greater than a first threshold value and the depth difference value between the two objects is smaller than a second threshold value, determining that the displacement recognition result corresponding to the to-be-processed image is shifted. By adopting the method, the shift identification omission ratio can be reduced.

Description

Technical Field

[0001] The present application relates to the field of electric power safety technology, and in particular to a method, device, computer equipment, storage medium and computer program product for identifying the displacement of an anti-vibration hammer. Background Art

[0002] Transmission lines are an important part of the power transmission system. Ensuring the stable operation of transmission lines is crucial to the safety of the power grid. The anti-vibration hammer is a device installed on the transmission line to reduce the vibration of the conductor and prevent the conductor from fatigue fracture due to vibration. During long-term operation, the anti-vibration hammer may be displaced due to external forces or natural environmental influences, resulting in a weakened anti-vibration effect and even causing safety hazards.

[0003] The traditional method of monitoring the anti-vibration hammer relies on manual inspection, which is inefficient and has subjective judgment errors. With the development of image processing technology, computer vision technology can be used to automatically monitor the anti-vibration hammer to determine whether the anti-vibration hammer has shifted. However, in actual scenarios, there are various types of transmission line environments and displacement modes of the anti-vibration hammer, resulting in a high missed detection rate of the anti-vibration hammer displacement identification. Summary of the invention

[0004] Based on this, it is necessary to provide a method, device, computer equipment, computer-readable storage medium and computer program product for identifying the displacement of a shock-absorbing hammer, which can improve the displacement recognition rate, in order to address the above technical problems.

[0005] In a first aspect, the present application provides a method for identifying the displacement of a shock-absorbing hammer. The method comprises:

[0006] Based on the trained instance segmentation model, instance segmentation is performed on the image to be processed corresponding to the transmission line to obtain a target segmentation map corresponding to the image to be processed; the object in the target segmentation map includes the anti-vibration hammer, or the object in the target segmentation map includes the anti-vibration hammer and hardware;

[0007] Determine the overlap between the contour boxes based on the position information of the contour boxes of each object in the target segmentation map;

[0008] Determine the depth difference value of each object based on the depth information of each object in the target segmentation map; the depth information is used to represent the distance between each pixel point in the target segmentation map;

[0009] If the overlap degree between two objects among the objects is greater than a first threshold, and the depth difference between the two objects is less than a second threshold, then it is determined that the shift recognition result corresponding to the image to be processed is shifted.

[0010] In one embodiment, the instance segmentation model is a YOLOv8 model; before performing instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed, the method further includes:

[0011] Acquire a displacement data set corresponding to the shock-absorbing hammer; the displacement data set includes annotated images of various types of shock-absorbing hammers and various types of hardware in different environments and weather conditions in different contact states; the contact states include shock-absorbing hammers touching each other, shock-absorbing hammers touching hardware, and no contact;

[0012] The YOLOv8 model is trained based on the shifted data set and the preset hyperparameters until the evaluation result corresponding to the updated YOLOv8 model meets the training end condition, thereby obtaining a trained YOLOv8 model.

[0013] In one embodiment, obtaining a displacement data set corresponding to the shock-absorbing hammer includes:

[0014] Acquire sample images of each transmission line under different environments and weather conditions based on a photographing device; the sample images include various types of anti-vibration hammers and hardware, and anti-vibration hammers and hardware at various different angles;

[0015] Annotate the anti-vibration hammer and the hardware included in each of the sample images based on the annotation instruction to obtain an annotated sample image; the annotated sample image includes a contour frame, a category label and a touch status corresponding to the anti-vibration hammer and the hardware respectively;

[0016] The labeled sample images are subjected to data preprocessing to obtain a displacement data set corresponding to the shock-absorbing hammer; the data preprocessing includes at least one of uniform size, uniform format, data set division and quality inspection.

[0017] In one embodiment, the contour frame includes a rectangular frame; and determining the overlap between the contour frames based on the position information of the contour frames of each object in the target segmentation image includes:

[0018] For two random rectangular frames of the objects, based on the first coordinate values ​​of the diagonal points of the two rectangular frames, determine the second coordinate values ​​of the diagonal points of the overlapping portion between the two rectangular frames, and determine the union area of ​​the union area of ​​the two rectangular frames;

[0019] Based on the second coordinate value, determining the overlapping area of ​​the overlapping parts of the two rectangular frames;

[0020] The quotient of the overlapping area and the union area is determined as the degree of overlap between the rectangular frames of the two random objects.

[0021] In one of the embodiments, determining the depth difference value of each object based on the depth information of each object in the target segmentation map includes:

[0022] Using a monocular depth estimation strategy, depth estimation is performed on the image to be processed to obtain a depth map; the depth map contains depth information of each pixel in the image to be processed;

[0023] The position information of the outline frame of each object in the target segmentation map is aligned with the depth map to obtain the depth information of each object, and the difference between the depth information of each object is determined as the depth difference value.

[0024] In one embodiment, the method further comprises:

[0025] Based on the coordinates of the center points of the shock-absorbing hammers corresponding to the shock-absorbing hammers in the target segmentation diagram, a plurality of groups of adjacent shock-absorbing hammers are determined; the adjacent shock-absorbing hammers are two shock-absorbing hammers installed on the same conductor;

[0026] Determine a relative distance between the adjacent shock-absorbing hammers, and determine a third threshold corresponding to the relative distance based on a plurality of groups of relative distances of the adjacent shock-absorbing hammers;

[0027] If the relative distance between adjacent shock-absorbing hammers in the target segmentation map is greater than the third threshold, it is determined that the displacement recognition result corresponding to the image to be processed is displacement.

[0028] In a second aspect, the present application also provides a device for identifying the displacement of a shock-absorbing hammer. The device comprises:

[0029] An instance segmentation module is used to perform instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed; the object in the target segmentation map includes the anti-vibration hammer, or the object in the target segmentation map includes the anti-vibration hammer and hardware;

[0030] An overlap determination module, configured to determine the overlap between the contour frames of the objects in the target segmentation map based on the position information of the contour frames of the objects;

[0031] A depth determination module, configured to determine a depth difference value of each object based on depth information of each object in the target segmentation map; the depth information is used to represent the distance between each pixel point in the target segmentation map;

[0032] The shift identification module is used to determine that the shift identification result corresponding to the image to be processed is shifted if the overlap between two objects in each of the objects is greater than a first threshold and the depth difference between the two objects is less than a second threshold.

[0033] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the method described in the first aspect are implemented.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0035] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the method described in the first aspect are implemented.

[0036] The above-mentioned anti-shock hammer displacement identification method, device, computer equipment, storage medium and computer program product perform instance segmentation on the image to be processed corresponding to the transmission line through a trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed. The trained instance segmentation model can identify the anti-shock hammer and the hardware, and add the contour boxes corresponding to each anti-shock hammer and the hardware on the basis of the image to be processed, so as to obtain a target segmentation map. Since there are many situations in which the anti-shock hammer is displaced, the displaced anti-shock hammer may come into contact with the hardware, or the displaced anti-shock hammer may come into contact with other anti-shock hammers. The server can obtain the position information of the contour boxes based on each object in the target segmentation map and determine the overlap between each contour box. And the server determines the depth difference value of each object based on the depth information of each object in the target segmentation map, and if there are two objects in each object whose overlap is greater than the first threshold and the depth difference value between the two objects is less than the second threshold, then the displacement identification result corresponding to the image to be processed is determined to be displaced. Since the overlap between the contour boxes can preliminarily indicate whether there is direct contact between the objects, and the depth information of each object can further determine the actual distance between the pixels corresponding to the objects, it is determined whether there is a depth error between the objects, and whether the objects in direct contact are real touches, and then it is determined whether the shock absorber has shifted, thereby improving the detection accuracy of identifying the displacement of the shock absorber. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the related technologies, the drawings required for use in the embodiments or the related technical descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1A diagram showing an application environment of a method for identifying the displacement of a shock-absorbing hammer in an embodiment;

[0039] Figure 2 A schematic diagram of a flow chart of a method for identifying the displacement of a shock-absorbing hammer in one embodiment;

[0040] Figure 3 A schematic diagram of a flow chart of a method for identifying the displacement of a shock-absorbing hammer in another embodiment;

[0041] Figure 4 A schematic diagram of the displacement of the anti-vibration hammer in one embodiment;

[0042] Figure 5 A schematic diagram of the displacement of the anti-vibration hammer in another embodiment;

[0043] Figure 6 is a structural block diagram of a shock-absorbing hammer displacement identification device in one embodiment;

[0044] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0045] 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.

[0046] The anti-vibration hammer displacement identification method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The server 104 can receive the image to be processed sent by the terminal 102. The image to be processed can be a wire image taken by a shooting device. Usually, the wire image can include a shockproof hammer and the hardware around the shockproof hammer. The server 104 performs instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model, identifies the shockproof hammer and hardware in the image to be processed, and marks them with contour boxes, so as to obtain the target segmentation map corresponding to the image to be processed. The server 104 determines the overlap between the two contour boxes through the position information of the contour boxes of each identified object in the image to be processed. The server 104 can determine the depth information of each object in the target segmentation map, and determine the depth difference value between the objects corresponding to the two contour boxes. Finally, according to the overlap between the two contour boxes and the depth difference value between the objects corresponding to the two contour boxes, it is judged whether each object in the image to be processed is displaced, and the displacement recognition result of the image to be processed is determined.

[0047] The terminal 102 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 104 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0048] In an exemplary embodiment, Figure 2 As shown, a method for identifying the displacement of a shock-absorbing hammer is provided. Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps S202 to S208. Among them:

[0049] Step S202: Based on the trained instance segmentation model, instance segmentation is performed on the image to be processed corresponding to the transmission line to obtain a target segmentation map corresponding to the image to be processed.

[0050] Specifically, the objects in the target segmentation map include anti-shock hammers, or the objects in the target segmentation map include anti-shock hammers and hardware. The instance segmentation model can identify various types of anti-shock hammers, as well as various types of hardware. The image to be processed may only include anti-shock hammers, or the image to be processed may include anti-shock hammers and hardware at the same time. The instance segmentation model can be trained based on real photos of anti-shock hammers and hardware on the wire taken at different angles, different lighting, backgrounds, and climatic conditions. The server can receive the image to be processed sent by the terminal, and perform instance segmentation on the image to be processed through the trained instance segmentation model to determine the contour box of each object in the image to be processed.

[0051] In one example, the image to be processed may contain two or more anti-vibration hammers. The server may identify the anti-vibration hammers and mark the contour boxes of the anti-vibration hammers in the image to be processed to obtain a target segmentation map. The image to be processed may contain at least one anti-vibration hammer and at least one hardware. The server may perform instance segmentation on the image to be processed, mark the contour boxes corresponding to the anti-vibration hammer and the hardware in the image to be processed to obtain a target segmentation map.

[0052] Step S204 : determining the overlap between the contour boxes based on the position information of the contour boxes of the objects in the target segmentation image.

[0053] Specifically, the location information may be the coordinates of the contour boxes. The server may determine the overlapping area and the union area between the contour boxes of each object according to the coordinates of the contour boxes of each object, and determine the degree of overlap between the two contour boxes according to the size of the overlapping area and the union area. Based on this, the server may obtain the degree of overlap between any two objects in the target segmentation map.

[0054] Step S206: determining a depth difference value of each object based on the depth information of each object in the target segmentation map.

[0055] The depth information is used to indicate the distance between each pixel in the target segmentation map. The depth difference value is used to indicate the distance between two objects. The larger the depth difference value, the greater the actual distance between the two objects, and the less likely they are to be in direct contact.

[0056] Specifically, the server can perform depth estimation on the pixel points of each object in the target segmentation map, determine the depth information of each pixel point, and merge the depth information of each pixel point to obtain the depth information of each object. Based on this, the server can determine the depth difference value between the two objects through the depth information corresponding to each of the two objects.

[0057] Step S208 : if the overlap between two objects among the objects is greater than the first threshold, and the depth difference between the two objects is less than the second threshold, then the shift recognition result corresponding to the image to be processed is determined to be shifted.

[0058] Specifically, the server can traverse multiple objects in the target segmentation map, randomly determine two objects, and calculate the overlap between the two objects and the depth difference between the two objects. If the overlap between the two objects is greater than a preset first threshold, and at the same time the depth difference between the two objects is less than a second threshold, it is determined that the anti-shock hammer in the image to be processed is displaced. It should be understood that the objects in the target segmentation map can include at least one anti-shock hammer, and the other object can be an anti-shock hammer or a hardware. For example, the two objects can be a anti-shock hammer and another anti-shock hammer, or the two objects can be a anti-shock hammer and a hardware.

[0059] In addition, if the overlap between the two objects is less than a preset first threshold, it can be determined that the two objects are not in direct contact, and it can be determined that the two objects are not displaced. If the overlap between the two objects is greater than the preset first threshold, and at the same time the depth difference between the two objects is greater than the second threshold, it is determined that the two objects may be in direct contact in terms of overlap, but the two objects are not in direct contact in terms of depth, so it is determined that the two objects in the image to be processed are not displaced. It should be understood that, usually when the shock-absorbing hammer is not displaced, there will be no direct contact between the shock-absorbing hammer and the shock-absorbing hammer, or between the shock-absorbing hammer and the hardware. If the shock-absorbing hammer and the shock-absorbing hammer, or between the shock-absorbing hammer and the hardware are in direct contact, it can be determined that the shock-absorbing hammer is displaced.

[0060] In the above-mentioned anti-shock hammer displacement identification method, the image to be processed corresponding to the transmission line is instance segmented by a trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed. The trained instance segmentation model can identify the anti-shock hammer and the hardware, and add the contour boxes corresponding to each anti-shock hammer and the hardware on the basis of the image to be processed, so as to obtain the target segmentation map. Since there are many situations in which the anti-shock hammer is displaced, the displaced anti-shock hammer may come into contact with the hardware, or the displaced anti-shock hammer may come into contact with other anti-shock hammers. The image to be processed may only contain multiple anti-shock hammers, or the image to be processed may also contain at least one anti-shock hammer and at least one hardware, etc. The server can obtain the position information of the contour box based on each object in the target segmentation map, and determine the overlap between each contour box; and the server can determine the depth difference value of each object based on the depth information of each object in the target segmentation map, and if the overlap between two objects in each object is greater than the first threshold, and the depth difference value between the two objects is less than the second threshold, then the displacement identification result corresponding to the image to be processed is determined to be displaced. Since the overlap between the contour boxes can preliminarily indicate whether there is direct contact between the objects, and the depth information of each object can further determine the actual distance between the pixels corresponding to the objects, it is determined whether there is a depth error between the objects, and whether the objects in direct contact are real touches, and then it is determined whether the shock absorber has shifted, thereby improving the detection accuracy of identifying the displacement of the shock absorber.

[0061] In an exemplary embodiment, the instance segmentation model is a YOLOv8 model. Before the step of “performing instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed”, the method further includes:

[0062] Obtain a displacement dataset corresponding to the anti-vibration hammer. Train the YOLOv8 model based on the displacement dataset and preset hyperparameters until the evaluation result corresponding to the updated YOLOv8 model meets the training end condition, thereby obtaining a trained YOLOv8 model.

[0063] Among them, the displacement dataset includes labeled images of various types of shock-absorbing hammers and various types of hardware in different environments and weather conditions in different contact states; the contact states include the contact between shock-absorbing hammers, the contact between shock-absorbing hammers and hardware, and no contact.

[0064] Specifically, the server can collect data on power transmission lines under different environments and weather conditions through a drone equipped with a high-resolution camera, and determine that the collected power transmission lines contain various lighting, background and climate conditions. The transmission line may include a shockproof hammer and hardware, or the transmission line may only include a shockproof hammer. The shift data set contains multiple annotated images, each of which may include at least one shockproof hammer, that is, the image may include multiple different shockproof hammers; each annotated image may include at least one shockproof hammer, that is, the image may include multiple types of shockproof hammers. In addition, each annotated image may also include at least one hardware, or each annotated image may include at least one type of hardware. The shooting angle of the annotated image may be a preset angle, and the preset angle may include the front view and side view of each shockproof hammer and hardware.

[0065] In one example, a labeled image may include one anti-vibration hammer of type A, two anti-vibration hammers of type B, and the labeled image may include two hardware of type C, and three hardware of type D. A labeled image may include only one anti-vibration hammer of type A; a labeled image may include only one hardware of type C.

[0066] In addition, each labeled image in the shift data set can be labeled for four situations: the shock-absorbing hammer, the hardware, the shock-absorbing hammers touching each other, and the shock-absorbing hammer and the hardware touching each other. The labeled content can include the contour box of each object, the category label of each object, and the touch status. It can be seen that for the two objects of the shock-absorbing hammer and the hardware, the touch status can be the shock-absorbing hammer touching the shock-absorbing hammer, the shock-absorbing hammer touching the hardware, and no touching (no touching).

[0067] The server can use the YOLO (You Only Look Once) v8 model as an instance segmentation model and train the YOLOv8 model through the shifted data set. The server can create a YOLOv8 configuration file and set the hyperparameters of the YOLOv8 model according to the power transmission scenario to be identified. The hyperparameters include at least the learning rate, batch size, and number of training rounds. After the server loads the shifted data set, it trains the model according to the hyperparameters. In each round of training, the parameters of the YOLOv8 model can be adjusted according to the learning rate, thereby reducing the loss value of the loss function. After each round of training, an updated YOLOv8 model can be obtained. The server can determine the prediction accuracy index of the YOLOv8 model using the test set and validation set in the shifted data set for the updated YOLOv8 model. The server can train the YOLOv8 model multiple times according to the training end condition until the training rounds, prediction accuracy index, etc. of the YOLOv8 model meet the training end condition, thereby obtaining a trained YOLOv8 model.

[0068] In this embodiment, the YOLOv8 model is trained for multiple rounds using the displacement data set corresponding to the shock-absorbing hammer and preset hyperparameters until the YOLOv8 model meets the training end condition, thereby obtaining a trained YOLOv8 model, which can improve the training efficiency and prediction accuracy of the YOLOv8 model.

[0069] In an exemplary embodiment, the specific implementation process of the step of "obtaining a displacement data set corresponding to the shock-absorbing hammer" includes:

[0070] Based on the shooting equipment, sample images of each transmission line under different environments and weather conditions are obtained; based on the labeling instructions, the anti-shock hammer and hardware contained in each sample image are labeled to obtain the labeled sample image; the labeled sample image is preprocessed to obtain the displacement data set corresponding to the anti-shock hammer.

[0071] The sample images include various types of shock-absorbing hammers and hardware, as well as shock-absorbing hammers and hardware at various different angles; the labeled sample images include the corresponding contour boxes, category labels and touch states of the shock-absorbing hammers and hardware; data preprocessing includes at least one of unified size, unified format, data set division and quality inspection.

[0072] Specifically, the server can obtain sample images of each transmission line under different environments and weather conditions through drones, surveillance cameras and other shooting equipment. The sample images may include multiple types of shockproof hammers and multiple types of hardware. For each sample image, at least one shockproof hammer may be included. The shockproof hammers may be shockproof hammers of the same type or shockproof hammers of different types. The number and type of shockproof hammers in each sample image are not specifically limited. Accordingly, for each sample image, at least one hardware fitting may be included. The hardware fittings may be hardware fittings of the same type or hardware fittings of different types. The number and type of hardware fittings in each sample image are not specifically limited.

[0073] In addition, the server can delete or enhance low-definition and low-resolution sample images to improve the usability of the sample images. The server can obtain the annotation instruction sent by the terminal, annotate the anti-vibration hammer and hardware included in the sample image, obtain the annotated sample image, and thus obtain the outline frame, category label and touch status of each anti-vibration hammer and hardware included in the sample image.

[0074] Based on this, the server can also rotate, crop, flip, scale, and other operations on the labeled sample images to ensure that the sample images have a uniform size. The server can also perform denoising on the sample images. In addition, the server can divide the sample images into training sets, validation sets, and test sets. In one example, 70% of the sample images are used for training, 15% of the sample images are used for validation, and 15% of the sample images are used for testing.

[0075] Optionally, the server can perform a quality check on the sample image to determine whether the format and clarity of the sample image meet the requirements and whether the annotation content of the sample image meets the actual situation. This improves the accuracy of annotation and the integrity of the data. The server can analyze the sample distribution and category distribution of the sample images in the shifted data set and return the analysis results to the terminal. The terminal can dynamically add appropriate sample images to the shifted data set based on the analysis results, thereby determining the balance of each sample image in the shifted data set.

[0076] In this embodiment, sample images of each transmission line under different environments and weather conditions are obtained by photographing equipment, and the shock-absorbing hammer and hardware contained in each sample image are labeled based on the labeling instructions to obtain the labeled sample images. The labeled sample images are subjected to data preprocessing to obtain the displacement data set corresponding to the shock-absorbing hammer, which can improve the data quality of the displacement data set, improve the training efficiency of the displacement data set, and improve the prediction accuracy of the YOLOv8 model.

[0077] In an exemplary embodiment, the contour box includes a rectangular box, and the specific implementation process of the step of "determining the overlap between the contour boxes based on the position information of the contour boxes of each object in the target segmentation image" includes:

[0078] For the rectangular frames of two random objects, based on the first coordinate values ​​of the diagonal points of the two rectangular frames, the second coordinate values ​​of the diagonal points of the overlapping part between the two rectangular frames are determined, as well as the area of ​​the union area of ​​the union area of ​​the two rectangular frames; based on the second coordinate values, the area of ​​the overlapping area of ​​the overlapping part of the two rectangular frames is determined; and the quotient of the overlapping area area and the union area area is determined as the degree of overlap between the rectangular frames of the two random objects.

[0079] The outline frame may be a rectangular frame. The area of ​​the union region is the area corresponding to the region contained in the union of two rectangular frames. The area of ​​the overlap region is the area corresponding to the region contained in the intersection of two rectangular frames.

[0080] Specifically, the server can obtain the rectangular boxes of two random objects and obtain the first coordinate values ​​of the diagonal points of the two rectangular boxes. For example, there are two rectangular boxes A and B, and the coordinates of the upper left corner and the lower right corner of rectangular box A can be , the coordinates of the upper left corner and lower right corner of the rectangular box B can be The server can determine the second coordinate values ​​corresponding to the diagonal points of the overlapping parts of the two rectangular frames according to the first coordinate values ​​of the rectangular frame A and the rectangular frame B. For example, the coordinates of the upper left corner and the lower right corner of the overlapping area of ​​the rectangular frame A and the rectangular frame B are determined as: , , , It should be understood that if or , it can be determined that the two rectangular boxes have no overlapping area.

[0081] The server can determine the area of ​​the overlapping region of the overlapping portion of the two rectangular frames based on the coordinates of the upper left corner and the lower right corner of the overlapping region. In an example, the area calculation formula of the overlapping region is:

[0082]

[0083] In addition, the server may also determine the area of ​​the union region of the two rectangular frames according to the first coordinate values ​​of the diagonal points of the two rectangular frames. The area calculation formula of the union region is:

[0084]

[0085]

[0086]

[0087] Based on this, the quotient of the overlapping area and the union area can be determined as the overlap between the rectangular boxes of two random objects. The formula is as follows:

[0088]

[0089] Optionally, the outline frame may also be a point frame of the object, and the point frame may be a point frame that marks the outline of the object through multiple points, and the outline frame is determined by connecting lines between the points. The server may determine the coordinate values ​​of the points of the overlapping part between the two point frames based on the coordinate values ​​of the points of the two point frames, and determine the area of ​​the union of the two point frames; determine the area of ​​the overlapping part of the two rectangular frames based on the coordinate values ​​of the points of the overlapping part between the two point frames; and finally, determine the quotient of the overlapping area area and the union area area as the degree of overlap between the point frames of the two random objects.

[0090] In this embodiment, the position information of the contour boxes of each object in the target segmentation image can be used to determine the overlapping area and the union area corresponding to the contour boxes, and the quotient of the overlapping area and the union area can be determined, which is the degree of overlap between the rectangular boxes of two random objects, thereby improving the accuracy of determining the degree of overlap.

[0091] In an exemplary embodiment, the specific implementation process of the step of “determining the depth difference value of each object based on the depth information of each object in the target segmentation map” includes:

[0092] Through the monocular depth estimation strategy, the depth of the image to be processed is estimated to obtain a depth map; the depth map contains the depth information of each pixel in the image to be processed; the position information of the contour box of each object in the target segmentation map is aligned with the depth map to obtain the depth information of each object, and the difference between the depth information of each object is determined as the depth difference value.

[0093] Specifically, the server can estimate the depth corresponding to the image to be processed through a monocular depth estimation strategy. The server aligns the position information of the contour box of each object in the target segmentation map with the depth map, obtains the depth information of the corresponding position information in the depth map, and thus obtains the depth information corresponding to each object. In an example, the server can obtain the depth information of multiple pixels corresponding to the contour box of the object, and determine the mean or median of the depth information of each pixel in the contour box as the depth information of the object. The server can calculate the difference in depth information of two objects and determine it as a depth difference value.

[0094] In this embodiment, a monocular depth estimation strategy is used to determine the depth information of each pixel in the image to be processed, and the depth information of each object is determined in combination with the target segmentation map, thereby determining the depth difference value between the objects, which can improve the efficiency and accuracy of determining the depth difference value.

[0095] In an exemplary embodiment, the method further comprises:

[0096] Based on the coordinates of the center points of the shock-absorbing hammers corresponding to the shock-absorbing hammers in the target segmentation map, multiple groups of adjacent shock-absorbing hammers are determined; the adjacent shock-absorbing hammers are two shock-absorbing hammers installed on the same conductor; the relative distance between the adjacent shock-absorbing hammers is determined, and based on the relative distances of the multiple groups of adjacent shock-absorbing hammers, a third threshold value corresponding to the relative distance is determined; if there is a relative distance between the adjacent shock-absorbing hammers in the target segmentation map that is greater than the third threshold value, then the shift recognition result corresponding to the image to be processed is determined to be shifted.

[0097] Specifically, in addition to causing the shock-absorbing hammer and the shock-absorbing hammer, and the shock-absorbing hammer and the hardware to touch each other, the displacement of the shock-absorbing hammer can also be determined by the distance between multiple groups of adjacent shock-absorbing hammers. The server can obtain the coordinates of the center points of each shock-absorbing hammer in the target segmentation diagram, and determine multiple groups of adjacent shock-absorbing hammers based on the extension direction of the wire. The server can determine the relative distance between each group of adjacent shock-absorbing hammers. In one example, the server can determine the mode, mean, or median of the relative distance as the third threshold.

[0098] In addition, if the relative distances between each group of adjacent anti-shock hammers in the image to be processed are all less than the third threshold value, it is determined that the anti-shock hammers in the image to be processed have not shifted. If the relative distances between any group of adjacent anti-shock hammers in the image to be processed are greater than the third threshold value, it is determined that a certain anti-shock hammer in the group of adjacent anti-shock hammers has shifted, and the shift recognition result corresponding to the image to be processed is determined to be shifted.

[0099] In this embodiment, by determining multiple groups of adjacent shock-absorbing hammers and determining whether adjacent shock-absorbing hammers are displaced based on the relative distances between the adjacent shock-absorbing hammers, the judgment range of whether the shock-absorbing hammers are displaced can be further improved, thereby improving the comprehensiveness and accuracy of judging the displacement.

[0100] like Figure 3 As shown, the specific execution process of the above-mentioned anti-vibration hammer displacement identification method is described in detail below in combination with a specific embodiment.

[0101] 1. Regarding the construction of the transmission anti-vibration hammer displacement dataset: The image data of the anti-vibration hammer, the anti-vibration hammer touching, the anti-vibration hammer touching the hardware and the hardware are collected by drones in multiple scenes, and accurate annotation and multi-modal data fusion are performed to build a high-quality transmission anti-vibration hammer displacement dataset. This dataset can significantly improve the performance and robustness of the anti-vibration hammer displacement detection model and ensure the safe and stable operation of the transmission system.

[0102] 2. Regarding the identification of the displacement of the power transmission anti-vibration hammer: YOLOv8 is used to segment the anti-vibration hammer, hardware, anti-vibration hammer touching, and anti-vibration hammer touching hardware. The YOLOv8 instance segmentation algorithm is used to accurately identify the anti-vibration hammer, hardware, and the contact between the anti-vibration hammer and the hardware. YOLOv8 has high precision and real-time detection capabilities, and can accurately detect these targets in complex environments to ensure the stable operation of the power system.

[0103] 3. Regarding geometric distance judgment: The IoU of the detection frames between the anti-vibration hammer and the hardware, and between the anti-vibration hammer and the anti-vibration hammer is used to judge the touch. The detection frames of the anti-vibration hammer and the hardware obtained after YOLOv8 detection are used to calculate the intersection over union (IoU) between them. If the IoU value exceeds the set threshold, it is determined that the anti-vibration hammer and the hardware or the anti-vibration hammer are in touch. This method uses the high-precision detection results of YOLOv8 to achieve touch judgment through geometric analysis, effectively reducing misjudgments and missed detections.

[0104] 4. Regarding monocular depth estimation: Monocular depth estimation is used to eliminate misalignment. For the segmented touch targets, the monocular depth estimation algorithm is combined to estimate the depth information of the identified targets. If the depth values ​​of the two targets are similar, it is determined to be a real touch, otherwise it is a misalignment caused by the shooting angle. The combination of monocular depth estimation and YOLOv8 further improves the accuracy of detection and reduces misjudgment caused by angle problems.

[0105] 5. Regarding the relative distance between adjacent targets: The relative distance between the shock-absorbing hammers on adjacent conductors is used to determine whether they are displaced. YOLOv8 is used to detect the positions of the shock-absorbing hammers on adjacent conductors and calculate the relative distance between adjacent shock-absorbing hammers. By analyzing these distances, it is determined whether the shock-absorbing hammers are displaced. If the relative distance exceeds the normal range, it is determined that the shock-absorbing hammers are displaced. This method achieves efficient detection of shock-absorbing hammer displacement through geometric distance calculation, thereby improving the maintenance efficiency of the power system.

[0106] Combining the above five points, the problems of high cost, high false detection and missed detection existing in related technologies can be solved, and the reliability and stability of the system can be significantly improved.

[0107] Specifically, the anti-vibration hammer displacement identification method includes the following steps:

[0108] Step 1: Data collection and dataset construction. Function: Through multi-scenario data collection and precise annotation, a high-quality transmission anti-vibration hammer displacement dataset is constructed to provide accurate basic data for subsequent model training and evaluation. Principle: Use drones to collect image data containing anti-vibration hammers and hardware from different angles and environments to ensure that the data covers various actual situations, and use annotation tools to accurately annotate the targets in the image to construct a dataset containing rich scenarios and various conditions.

[0109] In one example: 1. Use a drone equipped with a high-resolution camera to select transmission lines in different environments and weather conditions for data collection to ensure that the data set covers various lighting, background and climate conditions. For example, sunny days, cloudy days, rainy days, etc. 2. Ensure that the images are taken from different angles, especially the front and side views of the anti-vibration hammer and hardware to improve the comprehensiveness of the data set. 3. Ensure that the collected images are clear and the resolution is high enough to facilitate subsequent target detection and depth estimation. 4. Use professional data annotation tools for image annotation, here using Labelme. 5. Accurately annotate the anti-vibration hammer, hardware, anti-vibration hammer touching, and anti-vibration hammer touching hardware in the image. The annotation content includes the boundary contour, category label and touch status of the target. 6. Adjust the image to a uniform size and format for model training. Perform operations such as rotating, cropping, flipping, and scaling the image to increase the diversity of the data and improve the robustness of the model. Denoise the image to ensure the accuracy of the annotation and the clarity of the image. 7. Divide the data set into a training set, a validation set, and a test set. Typically, 70% is used for training, 15% for validation, and 15% for testing. 8. Perform a quality check on the dataset to ensure the accuracy of the annotations and the integrity of the data. Check whether the annotations are consistent with the actual situation and ensure that there are no omissions or errors. 9. Analyze the sample distribution and category ratio of the dataset to ensure the balance of the dataset in each category.

[0110] In this embodiment, through the above steps, a high-quality transmission shock-absorbing hammer displacement dataset is constructed to provide accurate and comprehensive basic data for subsequent instance segmentation and depth estimation, thereby ensuring the effectiveness of model training and evaluation.

[0111] Step 2: YOLOv8 target recognition. Function: Use the YOLOv8 model to perform instance segmentation on the anti-vibration hammer, hardware, anti-vibration hammer touching, and anti-vibration hammer touching hardware in the image, accurately identify and mark the outline of the target, and improve the accuracy of target detection. Principle: YOLOv8 is an advanced instance segmentation algorithm that can not only detect the location of the target, but also perform pixel-level segmentation and extract the precise outline of the target. YOLOv8 improves detection accuracy and processing speed by introducing improved network structure and algorithm.

[0112] In one example, 1. Select a YOLOv8 version suitable for instance segmentation tasks and use a pre-trained YOLOv8 model to accelerate the training process. 2. Convert the dataset constructed in step 1 to the format required by the YOLOv8 model. 3. Create a YOLOv8 configuration file and set the model's hyperparameters, including but not limited to learning rate, batch size, number of training rounds, etc., according to the power transmission scenario to be identified. 4. Load the training set and validation set data to ensure the correctness and integrity of the data. 5. Start the model training process. During the training process, gradually adjust the model parameters to minimize the detection error; monitor the training process in real time, check the loss function and accuracy indicators, and ensure that the training process is proceeding normally. 7. Use the test set to evaluate the trained YOLOv8 model. Calculate the model's accuracy indicators, analyze the model's false detection and missed detection on the test set, and identify the model's deficiencies. 9. According to the evaluation results, adjust the model's hyperparameters to optimize the model performance. Further enhance the data to improve the generalization ability and robustness of the model. 10. Use the new image for instance segmentation, use the trained YOLOv8 model to detect and segment the anti-vibration hammer, hardware and their contact conditions, and post-process the detection and segmentation results, including removing artifacts and merging overlapping areas to generate a clear target segmentation map. Visualize the segmentation results, check whether the segmentation contour is consistent with the actual target, and manually check the accuracy of the segmentation results to ensure the practicality and reliability of the model.

[0113] In this embodiment, through the above steps, the YOLOv8 model is used to perform instance segmentation of the shock-absorbing hammer and related targets, which can accurately identify the target and extract its contour information, providing detailed target information for subsequent touch judgment and displacement detection.

[0114] Step 3: Geometric analysis. Function: Through the detected rectangular frames, calculate the IoU between the shock-absorbing hammer and the shock-absorbing hammer, and between the shock-absorbing hammer and the hardware to determine whether a touch occurs. Principle: Determine whether a touch occurs by calculating the overlap of the detection frames between the shock-absorbing hammer and the hardware. Using the contour frames detected by the YOLOv8 model, analyze the intersection and union area of ​​these contour frames, and calculate the intersection over union (IoU) between the two targets. When the IoU value exceeds the set threshold, the system determines that a touch has occurred between the shock-absorbing hammer and the hardware or the shock-absorbing hammer. In this way, false detections are reduced and detection accuracy is improved.

[0115] In an example, 1. Extract the detection frame segmented by YOLOv8, and calculate the IoU values ​​between the anti-vibration hammer and the anti-vibration hammer, and between the anti-vibration hammer and the hardware:

[0116]

[0117] Given two rectangular boxes A and B, the coordinates of the upper left corner and the lower right corner are expressed as:

[0118]

[0119]

[0120] Calculate the coordinates of the upper left and lower right corners of the overlapping area:

[0121]

[0122]

[0123]

[0124]

[0125] It should be understood that if or , it can be determined that the two rectangular boxes have no overlapping area, and the overlapping area is 0. Otherwise, the area calculation formula of the overlapping area is:

[0126]

[0127] The area calculation formula of the union area of ​​two contour boxes is:

[0128]

[0129]

[0130] The area of ​​the union region is:

[0131]

[0132] Based on this, the quotient of the overlapping area and the union area can be determined as the overlap between the rectangular boxes of two random objects. The formula is as follows:

[0133]

[0134] 2. Set a reasonable IoU threshold. For example, it can be set to 0.5 here. If the IoU value between two detection boxes exceeds the threshold, it is determined that a touch occurs.

[0135] Step 4: Depth estimation. Function: Through the monocular depth estimation algorithm, the depth of the shock-absorbing hammer and hardware in the image is estimated to determine whether it is a real touch, thereby reducing misjudgment and missed detection caused by viewing angle problems and improving detection accuracy. Figure 4Figure 1 shows a schematic diagram of the displacement of the anti-vibration hammer. Principle: The depth information of the anti-vibration hammer and hardware in the image is estimated through a monocular depth estimation algorithm. Combined with the target position detected by YOLOv8, the depth data of each target area is extracted, and the depth difference between adjacent targets is calculated. If the depth difference is less than the set threshold, it is determined to be a real touch, otherwise it is considered to be a visual misalignment caused by the shooting angle. This method effectively improves the accuracy of touch determination and reduces misjudgments and missed detections.

[0136] In an example, 1. Select a suitable monocular depth estimation algorithm that can infer depth information from a single image. 2. Preprocess the input image, such as scaling and normalization, to ensure that it is consistent with the input requirements of the depth estimation model. 3. Process the image through the depth estimation model to generate a depth map for each pixel. The depth map contains the depth information of each point in the image. 4. Use the YOLOv8 segmentation result to extract the target area of ​​the shock absorber and hardware. Align the segmented target area with the depth map. 5. Extract the depth value of the shock absorber and hardware from the depth map. Calculate the mean or median depth in each target area as the depth information of the target. 6. Compare the depth information of the segmented shock absorber and hardware. Calculate the depth difference between the targets:

[0137]

[0138] 7. Set a threshold for depth difference to determine whether the target is actually touched. If the depth difference is less than the set threshold, it is considered a real touch, otherwise it is considered a false detection caused by perspective misalignment. 8. Combine the depth estimation with the instance segmentation result of YOLOv8 to generate the final touch determination result. Analyze the errors that may be introduced by the depth estimation, and adjust the threshold and model parameters to improve the determination accuracy.

[0139] In this embodiment, through the above steps, the depth estimation technology can be used to effectively eliminate misjudgments caused by viewing angle problems and accurately identify real touch situations, thereby reducing false detections and missed detections and improving the reliability of transmission shock-absorbing hammer displacement detection.

[0140] Step 5: Calculate the relative distance. Function: Use the relative distance between the shock-absorbing hammers on adjacent wires to determine whether they are displaced and improve the detection rate. Principle: Figure 5 As shown in the figure, for the displacement captured from this perspective, there is no reference object such as a shock absorber or touch, so it is easy to miss the detection. The YOLOv8 model is used to detect the position of the shock absorbers on the adjacent wires, record the coordinates of the center point of the detection frame, calculate the distance between the adjacent shock absorbers on the two adjacent wires, compare their distances, and compare them with the set normal distance threshold to determine whether the shock absorbers are displaced.

[0141] In an example, 1. After the shock absorbers are detected by instance segmentation of the image using the YOLOv8 model, the coordinates of the center point of each detection frame are recorded; 2. The shock absorbers with similar horizontal coordinates of the center points are selected as adjacent shock absorbers; 3. The relative distance between adjacent shock absorbers is calculated using the geometric distance formula:

[0142]

[0143] in, and They are the center point coordinates of two adjacent shock-absorbing hammers.

[0144] 4. Analyze the calculated relative distance and set a threshold value of a normal range. If the relative distance between adjacent shock-absorbing hammers exceeds the normal range, it is determined that the shock-absorbing hammers are displaced.

[0145] In this embodiment, through the above steps, the displacement of the shock-absorbing hammer can be effectively detected, potential safety hazards of the power transmission line can be discovered and dealt with in a timely manner, and the maintenance efficiency and stability of the power system can be improved.

[0146] In summary, the embodiments of the present application use computer vision-based recognition technology and use drones and fixed cameras to collect image data, which greatly reduces costs and improves economic benefits. In addition, by using geometric analysis (IoU calculation) and monocular depth estimation, it is possible to effectively distinguish between real touches and false touches caused by shooting angles, reduce false detection rates, and reduce the burden of manual inspections; relative distance calculation can significantly reduce missed detections. The comprehensive application of the above-mentioned technical means can achieve accurate identification and real-time alarm of the displacement of the transmission shock-absorbing hammer, effectively improving the maintenance efficiency and safety of the transmission line.

[0147] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0148] Based on the same inventive concept, the embodiment of the present application also provides a shock-absorbing hammer displacement identification device for implementing the shock-absorbing hammer displacement identification method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more shock-absorbing hammer displacement identification device embodiments provided below can refer to the limitations of the shock-absorbing hammer displacement identification method above, and will not be repeated here.

[0149] In an exemplary embodiment, Figure 6 As shown, a shock-absorbing hammer displacement identification device 600 is provided, comprising: an instance segmentation module 601, an overlap determination module 602, a depth determination module 603 and a displacement identification module 604, wherein:

[0150] The instance segmentation module 601 is used to perform instance segmentation on the to-be-processed image corresponding to the transmission line based on the trained instance segmentation model, and obtain a target segmentation map corresponding to the to-be-processed image; the object in the target segmentation map includes the anti-vibration hammer, or the object in the target segmentation map includes the anti-vibration hammer and the hardware;

[0151] An overlap determination module 602 is used to determine the overlap between the contour boxes based on the position information of the contour boxes of the objects in the target segmentation map;

[0152] A depth determination module 603 is used to determine the depth difference value of each object based on the depth information of each object in the target segmentation map; the depth information is used to represent the distance between each pixel in the target segmentation map;

[0153] The shift identification module 604 is configured to determine that the shift identification result corresponding to the image to be processed is shifted if the overlap between two objects among the objects is greater than a first threshold and the depth difference between the two objects is less than a second threshold.

[0154] Furthermore, the instance segmentation model is a YOLOv8 model, and the device also includes a model training module, which is specifically used to: obtain a displacement data set corresponding to the shock-absorbing hammer; the displacement data set includes labeled images of various types of shock-absorbing hammers and various types of hardware in different touch states under different environments and weather conditions; the touch state includes shock-absorbing hammers touching each other, shock-absorbing hammers touching hardware and no touching; the YOLOv8 model is trained based on the displacement data set and preset hyperparameters until the evaluation result corresponding to the updated YOLOv8 model meets the training end condition, thereby obtaining a trained YOLOv8 model.

[0155] Furthermore, the model training module is specifically used to: obtain sample images of each transmission line under different environments and weather conditions based on a shooting device; the sample images include multiple types of shockproof hammers and hardware, as well as shockproof hammers and hardware at multiple different angles; annotate the shockproof hammers and hardware contained in each sample image based on the annotation instructions to obtain annotated sample image; the annotated sample image includes the contour box, category label and touch status corresponding to the shockproof hammer and the hardware; perform data preprocessing on the annotated sample images to obtain a displacement data set corresponding to the shockproof hammer; the data preprocessing includes at least one of unified size, unified format, data set division and quality inspection.

[0156] Furthermore, the contour frame includes a rectangular frame, and the overlap determination module 602 is specifically used to: for the rectangular frames of two random objects, based on the first coordinate values ​​of the diagonal points of the two rectangular frames, determine the second coordinate values ​​of the diagonal points of the overlapping part between the two rectangular frames, and determine the union area of ​​the union area of ​​the two rectangular frames; based on the second coordinate value, determine the overlapping area of ​​the overlapping part of the two rectangular frames; determine the quotient of the overlapping area area and the union area area, which is the overlap between the rectangular frames of the two random objects.

[0157] Furthermore, the depth determination module 603 is specifically used to: perform depth estimation on the image to be processed through a monocular depth estimation strategy to obtain a depth map; the depth map contains the depth information of each pixel in the image to be processed; align the position information of the contour box of each object in the target segmentation map with the depth map to obtain the depth information of each object, and determine the difference between the depth information of each object, which is the depth difference value.

[0158] Furthermore, the device also includes an adjacent shift identification module, which is specifically used to: determine multiple groups of adjacent shock-absorbing hammers based on the center point coordinates corresponding to each shock-absorbing hammer in the target segmentation map; the adjacent shock-absorbing hammers are two shock-absorbing hammers installed on the same wire; determine the relative distance between adjacent shock-absorbing hammers, and based on the relative distances of multiple groups of adjacent shock-absorbing hammers, determine a third threshold corresponding to the relative distance; if there is a relative distance between adjacent shock-absorbing hammers in the target segmentation map that is greater than the third threshold, then determine that the shift identification result corresponding to the image to be processed is shifted.

[0159] Each module in the above-mentioned anti-vibration hammer displacement identification device can be implemented in whole or in part by software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.

[0160] In an exemplary embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as shown in FIG. Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store training image data for training an instance segmentation model. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for identifying the displacement of an anti-vibration hammer is implemented.

[0161] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0162] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0163] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0164] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0166] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0167] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0168] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for identifying the displacement of a shock-absorbing hammer, characterized in that: The method comprises: Based on the trained instance segmentation model, instance segmentation is performed on the image to be processed corresponding to the transmission line to obtain a target segmentation map corresponding to the image to be processed; the object in the target segmentation map includes the anti-vibration hammer, or the object in the target segmentation map includes the anti-vibration hammer and hardware; Determine the overlap between the contour boxes based on the position information of the contour boxes of each object in the target segmentation map; Determine the depth difference value of each object based on the depth information of each object in the target segmentation map; the depth information is used to represent the distance between each pixel point in the target segmentation map; If the overlap degree between two objects among the objects is greater than a first threshold, and the depth difference between the two objects is less than a second threshold, then it is determined that the shift recognition result corresponding to the image to be processed is shifted.

2. The method according to claim 1, characterized in that The instance segmentation model is a YOLOv8 model; before performing instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model and obtaining a target segmentation map corresponding to the image to be processed, the method further includes: Acquire a displacement data set corresponding to the shock-absorbing hammer; the displacement data set includes annotated images of various types of shock-absorbing hammers and various types of hardware in different environments and weather conditions in different contact states; the contact states include shock-absorbing hammers touching each other, shock-absorbing hammers touching hardware, and no contact; The YOLOv8 model is trained based on the shifted data set and the preset hyperparameters until the evaluation result corresponding to the updated YOLOv8 model meets the training end condition, thereby obtaining a trained YOLOv8 model.

3. The method according to claim 2, characterized in that The step of obtaining a displacement data set corresponding to the anti-vibration hammer includes: Acquire sample images of each transmission line under different environments and weather conditions based on a photographing device; the sample images include various types of anti-vibration hammers and hardware, and anti-vibration hammers and hardware at various different angles; Annotate the anti-vibration hammer and the hardware included in each of the sample images based on the annotation instruction to obtain an annotated sample image; the annotated sample image includes a contour frame, a category label and a touch status corresponding to the anti-vibration hammer and the hardware respectively; The labeled sample images are subjected to data preprocessing to obtain a displacement data set corresponding to the shock-absorbing hammer; the data preprocessing includes at least one of uniform size, uniform format, data set division and quality inspection.

4. The method according to claim 1, characterized in that: The contour frame includes a rectangular frame; and determining the overlap between the contour frames based on the position information of the contour frames of each object in the target segmentation map includes: For two random rectangular frames of the objects, based on the first coordinate values ​​of the diagonal points of the two rectangular frames, determine the second coordinate values ​​of the diagonal points of the overlapping portion between the two rectangular frames, and determine the union area of ​​the union area of ​​the two rectangular frames; Based on the second coordinate value, determining the overlapping area of ​​the overlapping parts of the two rectangular frames; The quotient of the overlapping area and the union area is determined as the degree of overlap between the rectangular frames of the two random objects.

5. The method according to claim 1, characterized in that The determining of the depth difference value of each object based on the depth information of each object in the target segmentation map includes: Using a monocular depth estimation strategy, depth estimation is performed on the image to be processed to obtain a depth map; the depth map contains depth information of each pixel in the image to be processed; The position information of the outline frame of each object in the target segmentation map is aligned with the depth map to obtain the depth information of each object, and the difference between the depth information of each object is determined as the depth difference value.

6. The method according to claim 1, characterized in that The method further comprises: Based on the coordinates of the center points of the shock-absorbing hammers corresponding to the shock-absorbing hammers in the target segmentation diagram, a plurality of groups of adjacent shock-absorbing hammers are determined; the adjacent shock-absorbing hammers are two shock-absorbing hammers installed on the same conductor; Determine a relative distance between the adjacent shock-absorbing hammers, and determine a third threshold corresponding to the relative distance based on a plurality of groups of relative distances of the adjacent shock-absorbing hammers; If the relative distance between adjacent shock-absorbing hammers in the target segmentation map is greater than the third threshold, it is determined that the displacement recognition result corresponding to the image to be processed is displacement.

7. A device for identifying the displacement of a shock-absorbing hammer, characterized in that: The device comprises: An instance segmentation module is used to perform instance segmentation on the image to be processed corresponding to the transmission line based on the trained instance segmentation model to obtain a target segmentation map corresponding to the image to be processed; the object in the target segmentation map includes the anti-vibration hammer, or the object in the target segmentation map includes the anti-vibration hammer and hardware; An overlap determination module, configured to determine the overlap between the contour frames of the objects in the target segmentation map based on the position information of the contour frames of the objects; A depth determination module, configured to determine a depth difference value of each object based on depth information of each object in the target segmentation map; the depth information is used to represent the distance between each pixel point in the target segmentation map; The shift identification module is used to determine that the shift identification result corresponding to the image to be processed is shifted if the overlap between two objects in each of the objects is greater than a first threshold and the depth difference between the two objects is less than a second threshold.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images