Power transmission line insulator front end intelligent identification and automatic detection method and system

Abstract

The invention belongs to the technical field of power transmission line inspection, and provides an intelligent identification and automatic detection method and system for the front end of a power transmission line insulator, and the method comprises the steps: analyzing a preset route in response to an inspection task, and flying to a target waypoint according to waypoint information in the preset route; a video stream of a target waypoint is obtained, image frames in the video stream are identified by using a pre-trained insulator string identification model, and an insulator string is detected; adjusting the posture to enable the target insulator string to be centered in front of the target insulator string; distance information of a front detected insulator steel cap is obtained, the three-dimensional coordinate position of a center point is calculated in combination with fuselage parameters and external parameter calibration data, and based on the three-dimensional coordinate position of the center point, the pixel-level offset of the center point from the center position of a video frame is calculated; and according to the relationship among the image coordinate system, the camera coordinate system and the world coordinate system, carrying out fine adjustment on the position of the machine body so as to carry out piece-by-piece contact type detection. According to the invention, the safety, precision and efficiency of power transmission equipment detection are significantly improved.

Description

Technical Field

[0001] This invention belongs to the field of power transmission line inspection technology, specifically relating to a method and system for intelligent identification and automatic detection of insulator front ends of transmission lines. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] As a critical component in transmission lines, the health of insulator strings directly affects the safe operation of the power system. Because insulators are exposed to high temperature, high humidity, and high voltage environments for extended periods, they are prone to faults such as flashover, cracks, and breakdown, resulting in insulators with zero health. Therefore, regular inspection and condition assessment of insulators are crucial aspects of power grid operation and maintenance.

[0004] Traditional insulator inspection methods still rely on manual tower climbing or remote observation from the ground using telescopes or telephoto cameras. These methods have the following significant drawbacks: low work efficiency, high difficulty, and the risk of falls and electric shock during manual tower climbing; dependence on manual target identification makes the results susceptible to subjective judgment and difficult to form quantitative indicators; and difficulty in achieving piece-by-piece inspection. To address these issues, drone-based auxiliary inspection methods have emerged in recent years.

[0005] Traditional drone inspection systems mostly remain at the stage of "image acquisition + offline recognition," which has many problems: images acquired by drones need to be transmitted back to the background for manual analysis, making it impossible to make timely judgments on the work site; insulator strings are identified as a whole target, making it impossible to accurately identify individual insulators; they cannot be linked with the flight control system to perform automatic alignment and precise positioning; and they cannot achieve contact-type electrical testing to accurately diagnose zero-value insulators, etc.

[0006] Some literature identifies insulator strings by constructing an insulator instance segmentation model, but it can only output the selected area of ​​the insulator string and cannot identify and number individual insulators. Furthermore, it does not involve spatial positioning, depth ranging, or three-dimensional geometric analysis of insulators.

[0007] Some literature describes methods that enhance an initial insulator image, extract feature maps of each surface defect layer, and then perform aggregation, decoding fusion, and feature map comparison analysis to obtain the detection results. However, this method is limited to image-level appearance recognition, lacks real-time capability, requires offline processing of defect recognition results without front-end feedback, and can only identify external defects, not perform contact-based detection.

[0008] In summary, none of the existing technical solutions have solved the following technical problems: High-precision identification and numbering of each insulator piece; closed-loop mapping of identification results to spatial coordinates and flight control actions; analysis of spatial positioning of high-altitude insulator pieces; spatial compensation for positional offset of the detection mechanism and high-precision contact detection; mission linkage capability of UAVs to perform automatic flight control in complex attitudes and environments; and a truly automatic detection and result closed-loop confirmation mechanism. Summary of the Invention

[0009] To address the aforementioned problems, this invention proposes an intelligent identification and automatic detection method and system for insulator front-ends in transmission lines. This invention is applicable to insulator condition detection scenarios in high-voltage transmission lines. By integrating an image recognition model, a lidar module, and a front-contact detection structure on a drone platform, it achieves insulator piece-by-piece identification, positioning, and automatic detection, significantly improving detection efficiency and accuracy while reducing the risks associated with manual operations.

[0010] According to some embodiments, the present invention adopts the following technical solution: A method for intelligent identification and automatic detection of insulator front ends in transmission lines includes the following steps: In response to the inspection mission, the system analyzes the preset route and flies to the target waypoint based on the waypoint information within it. The video stream of the target waypoint is acquired, and the image frames are identified using a pre-trained insulator string recognition model to detect the insulator strings. Adjust the orientation so that the target insulator string is centered directly in front of it; Switch to the piece-by-piece recognition mode to identify the steel cap part of each insulator piece and obtain its center point position in the form of pixel coordinates; Obtain the distance information of the steel cap of the insulator being tested in front, and calculate the three-dimensional coordinate position of the center point by combining the machine body parameters and external parameter calibration data; Based on the three-dimensional coordinates of the center point, the pixel-level offset from the center of the video frame is calculated. The position of the camera body is finely adjusted according to the relationship between the image coordinate system, the camera coordinate system, and the world coordinate system to perform piece-by-piece contact detection.

[0011] As an alternative implementation, the insulator string identification model includes an input end, a backbone network, and an output end. The input end incorporates data augmentation and random affine transformation. The backbone network adopts the C2f structure in the YOLOv8-CSP module, and the Neck structure uses bidirectional feature fusion with FPN and PAN. The Dual Head multi-task head is divided into a rotating frame detection head and a segmentation detection head, which are used to identify insulator strings and insulator caps, respectively. The output end uses a joint loss function for end-to-end training.

[0012] As a further implementation method, the process of end-to-end training using a joint loss function at the output includes: the total loss is:

[0013] in, , , These are the classification loss, position regression loss, and angle loss, respectively, used for the detection branch loss of insulator string identification; , Separately segment DCE and Dice losses for semantic segmentation loss in insulator piece recognition to improve the ability to preserve small target boundaries; , These are the loss weights, used to adjust task balance during training.

[0014] As an alternative implementation, during the training process of the insulator string identification model, the error between the target true value and the network predicted value is calculated by forward propagation, and then the error is continuously minimized by backpropagation to update and optimize the network parameters.

[0015] As an alternative implementation, the process of obtaining the center point position in the form of pixel coordinates includes determining the arrangement direction of the target insulator string and numbering it sequentially, and calculating the center pixel coordinates of each insulator.

[0016] As an alternative implementation method, the process of obtaining the distance information of the steel cap of the insulator to be tested in front, and calculating the three-dimensional coordinate position of the center point by combining the machine body parameters and external parameter calibration data includes: obtaining the spatial depth of the center point, and using the pinhole camera model to convert the two-dimensional pixels in the image into three-dimensional coordinates based on the camera coordinate system by combining the machine body parameters and external parameter calibration data.

[0017] As an alternative implementation, the process of fine-tuning the position of the fuselage based on the relationship between the image coordinate system, the camera coordinate system and the world coordinate system includes obtaining the coordinates of the target point based on the camera coordinate system according to the three-dimensional coordinate position of the center point, and calculating the distance of the steel cap based on the coordinates of the target point. The camera attitude is obtained as the forward and downward unit vectors to obtain the flight target point in the camera coordinate system. The flight target point based on the camera coordinate system is converted to the fuselage coordinate system. Based on the UAV's positioning attitude, the target point coordinates are converted from the fuselage coordinate system to the world coordinate system. Based on the position of the detection device relative to the fuselage, the target point is calibrated to obtain the final target point.

[0018] As a further implementation, the coordinates of the detected target point are obtained based on the camera coordinate system. :

[0019] in, , These are the horizontal and vertical coordinates of the target point in the camera coordinate system, respectively. represents the depth coordinates of the target point in the camera coordinate system; x and y are the horizontal and vertical pixel coordinates of the target point in the image, respectively. , These are the horizontal and vertical pixel coordinates of the image center, respectively; , These are the camera's horizontal and vertical focal lengths, respectively. for To determine if correction is needed, if the ranging point is on the inter-disc spacing, no correction is needed; if the ranging point is on the front surface of the insulator disk, correction is needed. , where R is the radius of the insulator skirt and D is the spatial depth of the target point.

[0020] As a further implementation, the process of calculating the steel cap distance includes:

[0021] Correcting the offset of the front detection device, the distances of the center point of the detection device in front of and below the drone are respectively... x , h The target point is moved backward and upward by the corresponding distances to obtain the corrected distance.

[0022] As a further implementation method, the flight target point in the camera coordinate system is obtained. :

[0023] in, For the set target point, corresponding or The forward unit vector is The downward unit vector is .

[0024] As a further implementation, the flight target point based on the camera coordinate system is converted to the fuselage coordinate system:

[0025]

[0026] in, This is the external parameter relationship matrix between the gimbal attitude, camera, and drone; This is a rotation matrix used to describe the rotation of the camera coordinate system to align with the body coordinate system; It is a translation vector used to describe the position of the camera's optical center in the camera's coordinate system.

[0027] As a further implementation method, based on the UAV's positioning attitude, the target point coordinates are transformed from the fuselage coordinate system to the world coordinate system:

[0028]

[0029] in, The target point's world coordinates are the target input required by the flight control system. The transformation matrix for the fuselage to world coordinates; For the fuselage to face the rotation matrix, This is the drone's position matrix, representing the drone's position in the world coordinate system.

[0030] As a further implementation, the target point is calibrated based on the position of the detection device relative to the fuselage to obtain the final target point: .

[0031] in, The final target point in the world coordinate system; L This is a scalar measure of the distance between the detection device and the center of the machine body.

[0032] As an alternative implementation method, the process of fine-tuning the fuselage position includes: Calculate the error vector with the center of the steel cap as the target point:

[0033] in, This is the current location of the drone; Construct control vectors and set the maximum flight speed to... The flight control speed is:

[0034] in, The proportional gain controls the response speed; clip () is a function that limits the numerical range; This is the speed factor, used to control flight speed; This represents the distance between the current location and the target location. If both the position error and velocity meet the threshold constraints, the drone is deemed to have met the detection conditions, and detection is initiated.

[0035] in, and These are the position error and the velocity threshold, respectively.

[0036] A smart identification and automatic detection system for the front end of insulators of transmission lines includes: The drone is used to respond to inspection tasks, analyze the preset route, and fly to the target waypoint based on the waypoint information; adjust its attitude to center the target insulator string in front of it; The camera is used to acquire video streams of the target waypoint, and a pre-trained insulator string recognition model is used to identify the image frames and detect the insulator strings. The control module is used to switch to the piece-by-piece recognition mode, identify the steel cap part of each insulator piece, and obtain its center point position in the form of pixel coordinates; based on the three-dimensional coordinate position of the center point, calculate its pixel-level offset from the center position of the video frame, and make fine adjustments to the position of the camera body according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system. The laser module is used to acquire distance information of the steel cap of the insulator being tested in front; The calculation module is used to calculate the three-dimensional coordinate position of the center point based on distance information, combined with fuselage parameters and external parameter calibration data; The testing device is used for contact testing of each piece individually.

[0037] An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps in the method described above.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention innovatively provides a method for intelligent identification and automatic detection of insulators at the front end of transmission lines. It uses a lightweight YOLOv8 series model deployed at the front end to identify individual insulators, and outputs the center coordinates of each insulator through a spatial positioning algorithm, realizing individual insulator numbering and positioning, supporting subsequent automatic detection processes, and solving the problem that traditional solutions cannot achieve individual insulator identification and numbering.

[0039] This invention innovatively provides a method for intelligent identification and automatic detection of insulator front ends in transmission lines. By combining a laser ranging module and an intrinsic parameter model, the identified image pixel coordinates and laser-measured depth information are converted into three-dimensional spatial coordinates. Combined with the spatial offset matrix of the UAV probe structure, the automatic generation of flight control navigation points and high-precision hovering control are achieved. This solves the problem of the disconnect between identification results and spatial positioning in the prior art.

[0040] This invention innovatively provides a method for intelligent identification and automatic detection of insulator front end of transmission lines. By connecting image recognition and laser ranging results to the flight control system, it realizes an integrated closed-loop control process from image → three-dimensional coordinates → flight target → hovering detection. It supports an automatic process control chain for automatic identification, approach, adjustment and detection, and solves the problem of the identification system and flight control system being disconnected.

[0041] This invention innovatively provides an intelligent identification and automatic detection system for the front end of transmission line insulators. It is equipped with a front-mounted contact detection device mounted under a drone. After precise hovering guided by flight control, the probes are lowered to both sides of the steel cap, realizing contact detection of each insulator and supplementing the ability to assess the functional status that cannot be evaluated by vision alone.

[0042] This invention innovatively provides an intelligent identification and automatic detection system for the front end of insulators in transmission lines. It supports automatic detection logic for the entire string, automatically identifies the position of the next insulator by number sequence, and completes position jump, attitude adjustment and detection device detection actions in conjunction with flight control commands. It achieves efficient continuous detection without human intervention and transmits data back to the remote controller in real time for confirmation, retesting or recording.

[0043] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0044] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0045] Figure 1 This is a schematic diagram of an embodiment of the intelligent identification and automatic detection process for the front end of a transmission line insulator. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0047] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0048] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0049] Where there is no conflict, the embodiments and features described in this application may be combined with each other.

[0050] Example 1 A method for intelligent identification and automatic detection of insulator front ends in transmission lines, such as Figure 1 As shown, it includes the following steps: In response to the inspection task, the system analyzes the preset route and flies to the target waypoint based on the waypoint information to initiate the visual recognition and detection process. This embodiment uses the front-facing camera of the drone to acquire the video stream of the target waypoint, and uses a pre-trained insulator string recognition model to identify the image frames and detect the insulator strings. Adjust the orientation so that the target insulator string is centered directly in front of it; Switch to the piece-by-piece recognition mode to identify the steel cap part of each insulator piece and obtain its center point position in the form of pixel coordinates; Obtain the distance information of the steel cap of the insulator being tested in front (in this embodiment, this can be achieved using the laser module of the machine body), and calculate the three-dimensional coordinate position of the center point by combining the machine body parameters and external parameter calibration data; Based on the three-dimensional coordinates of the center point, the pixel-level offset from the center of the video frame is calculated. The position of the fuselage is finely adjusted according to the relationship between the image coordinate system, the camera coordinate system, and the world coordinate system to perform piece-by-piece contact inspection. In this embodiment, by interfacing with the flight control system, the UAV is controlled to make minor position adjustments to achieve contact docking between the front contact inspection device and the target steel cap, thereby completing piece-by-piece contact inspection and transmitting inspection data and reports back in real time.

[0051] The insulator string identification model in this embodiment is a multi-task architecture based on the YOLOv8 backbone network. It has both rotating box target detection and semantic segmentation capabilities, and is suitable for the two-stage identification task in UAV insulator inspection, namely, long-distance detection of insulator strings and close-range identification of steel cap details.

[0052] Its training process includes: Sample library construction. A sample library was built based on drone inspection images of power transmission lines in a certain area. The open-source LabelMe software was used to directly convert the annotation information needed for target identification into JSON for storage. The sample library was divided into training, validation, and test sets in a 7:2:1 ratio. All JSON-formatted labels were converted to .txt format usable by the YOLO model and saved along with the corresponding images to the training, validation, and test sets.

[0053] Images are input into the network, and training begins with setting the number of label categories, names, and iteration counts. At the input end, Mosaic data augmentation and randomized affine transformations are introduced to enhance the model's robustness to multi-scale, small targets. The backbone network uses the C2f structure from the YOLOv8-CSP module to build a lightweight network with a good speed-accuracy balance. The Neck structure employs FPN+PAN bidirectional feature fusion to adapt to targets of varying sizes. The Dual Head multi-task head consists of a rotating bounding box detection head and a segmentation detection head, used to identify insulator strings and insulator caps, respectively. At the output end, a joint loss function is used for end-to-end training, with the total loss defined as follows:

[0054] in, , , These are the classification loss, position regression loss, and angle loss, respectively, used for the detection branch loss of insulator string identification; , Separating DCE and Dice losses separately for semantic segmentation loss in insulator piece recognition can improve the ability to preserve the boundaries of small targets. , To compensate for the loss weight coefficients, the task balance can be adjusted during training to achieve the best results for the multi-task model.

[0055] Model training calculates the error between the true target value and the network's predicted value through forward propagation, and then continuously minimizes the error through backpropagation to update and optimize the network parameters, enabling the model to have a good recognition rate for the target categories included in the training set. During training, the convergence and overfitting risk of the trained model are periodically verified on the validation test set. The loss curve, validation accuracy (mAP, mIoU), and early stopping and fine-tuning strategies are recorded, and the hyperparameters are updated accordingly.

[0056] Model Evaluation: For the insulator string identification module, the trained model is used to identify the test set, and the detection rate is calculated. As an evaluation indicator, its formula is as follows:

[0057] in, This represents the number of correctly identified samples, specifically the number of predicted bounding boxes that the model successfully detected (IoU ≥ 0.5) with an angle deviation ≤ 15°. This indicates the number of unidentified samples.

[0058] After successful identification, the insulator strings are segmented, and the Intersection over Union (IoU) index is used for evaluation.

[0059] in, P G represents the mask region predicted by the model; G represents the actual mask region.

[0060] To standardize the evaluation criteria for multi-task models, a comprehensive evaluation score is defined:

[0061] in, , These are weighting coefficients used to balance detection and segmentation tasks.

[0062] The model is integrated into the image correction and automatic exposure system. Once the model's recognition rate meets the requirements on the test set, it is integrated into the system for subsequent image correction.

[0063] The process of determining the spatial coordinates of the insulator disc in this embodiment includes: Target recognition. After the UAV begins autonomous inspection, it reaches the waypoint based on the route parsing information. The UAV sends video frames to the remote control via the image transmission module, obtains real-time image frame callbacks through the data interface, activates the insulator string recognition model, selects the insulator strings to be detected, determines the arrangement direction of the selected strings through an algorithm, numbers them sequentially, and calculates the center pixel coordinates (x, y) of each insulator. i ,y i ).

[0064] Spatial positioning: In this embodiment, a laser ranging module is used to obtain the spatial depth D (m) of the target point (corresponding to the center of the left steel cap of the insulator). A pinhole camera model is used to convert the two-dimensional pixels in the image into three-dimensional coordinates based on the camera coordinate system. That is, the three-dimensional coordinates corresponding to the center of the recognition box. As shown in the following formula:

[0065] in, , These are the horizontal and vertical coordinates of the target point in the camera coordinate system, respectively. x represents the forward (depth) coordinates of the target point in the camera coordinate system; x and y are the horizontal and vertical pixel coordinates of the target point in the image, respectively. , These are the horizontal and vertical pixel coordinates of the image center (camera principal point); , These are the horizontal and vertical focal lengths of the camera, respectively. They are also often used in the in-camera parameter matrix, with the unit being px (pixels).

[0066] Target point calculation: Calculate the pixel-level offset of the final target frame from the center of the video frame. Based on the relationship between the image coordinate system, camera coordinate system and world coordinate system, fine-tune the aircraft position to adjust the final target frame to the center of the image.

[0067] Output the coordinates of the detected target points based on the camera coordinate system. :

[0068] for It is necessary to determine whether correction is needed: if the ranging point is at the center of the assembly (on the inter-disc spacing), no correction is needed; if the ranging point is on the front surface of the insulator disk, correction is needed, i.e. , where R is the radius of the insulator skirt.

[0069] Secondly, the distance to the steel cap is calculated:

[0070] in, and These represent the distances between the steel caps on the left and right sides of the corresponding insulator disc.

[0071] Secondly, correct the offset of the front detection device: the distances of the center point of the detection device in front of and below the drone are respectively x , h By moving the target point backward (relative to the drone's direction of travel) and upward by the corresponding distances, the corrected distance can be obtained. The camera attitude is then used as the forward unit vector. With downward unit vector The flight target point in the camera coordinate system is obtained. :

[0072] in, For the set target point, corresponding or It is calibrated by the operator.

[0073] Subsequently, the flight target point based on the camera coordinate system is converted to the fuselage coordinate system:

[0074]

[0075] in: This is the external parameter relationship matrix between the gimbal attitude, camera, and drone; The rotation matrix, derived from the calibration of the gimbal lens-body, is used to describe the rotation of the camera coordinate system to align with the body coordinate system. It is a translation vector, derived from the installation calibration, used to describe the position of the camera's optical center in the camera body coordinate system.

[0076] Finally, based on the drone's positioning attitude, the target point coordinates are transformed from the drone's coordinate system to the world coordinate system:

[0077]

[0078] in: The target point's world coordinates are the target input required by the flight control system. The transformation matrix for the fuselage to world coordinates is provided in real time by the flight control system. The fuselage orientation rotation matrix is ​​derived from the flight control IMU attitude and is constructed using roll / pitch / yaw. This is the drone's position matrix, derived from RTK / GPS positioning, representing the drone's position in the world coordinate system.

[0079] Finally, based on the position of the detection device relative to the fuselage, the target point is calibrated to obtain the final target point:

[0080] in, The final target point in the world coordinate system; L This is a scalar measure of the distance between the detection device and the center of the machine body.

[0081] Next, we will perform the detection: Construct an error vector, taking the center of the steel cap as the target point, and calculate the error vector:

[0082] in, This shows the current location of the drone.

[0083] Error control and joystick mapping are used to construct a control vector, and the maximum flight speed is set to... Then the flight control speed is:

[0084] in, The proportional gain controls the response speed; clip () is a function that limits the numerical range; This is the speed factor, used to control flight speed; This represents the distance between the current location and the target location.

[0085] Iterative adjustment and termination judgment. When both position error and velocity simultaneously meet the threshold constraints, the UAV can be determined to have met the detection conditions, and detection can be performed.

[0086] in, and These are the position error and the velocity threshold, respectively.

[0087] This embodiment has a dual-task recognition function. It uses a multi-task model built on Yolov8 to perform dual-task recognition, calling different detection heads to identify insulator strings and single insulators respectively, with a recognition accuracy of over 96%. It has the function of accurately locating the target point. By docking the identification module and the laser ranging unit, it can accurately locate and perform detection.

[0088] It has an automatic contact electrical detection function, and through automatic flight control docking, it automatically flies to the monitoring point to perform contact detection.

[0089] The method provided in this embodiment is used for insulator testing of transmission lines, which realizes insulator identification, positioning and automatic testing piece by piece, significantly improving testing efficiency and accuracy and reducing the risk of manual operation.

[0090] Example 2 A smart identification and automatic detection system for the front end of insulators of transmission lines includes: The drone is used to respond to inspection tasks, analyze the preset route, and fly to the target waypoint based on the waypoint information; adjust its attitude to center the target insulator string in front of it; The camera is used to acquire video streams of the target waypoint, and a pre-trained insulator string recognition model is used to identify the image frames and detect the insulator strings. The control module is used to switch to the piece-by-piece recognition mode, identify the steel cap part of each insulator piece, and obtain its center point position in the form of pixel coordinates; based on the three-dimensional coordinate position of the center point, calculate its pixel-level offset from the center position of the video frame, and make fine adjustments to the position of the camera body according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system. The laser module is used to acquire distance information of the steel cap of the insulator being tested in front; The calculation module is used to calculate the three-dimensional coordinate position of the center point based on distance information, combined with fuselage parameters and external parameter calibration data; The testing device is used for contact testing of each piece individually.

[0091] Example 3 An electronic device includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps in the method described above.

[0092] For example, electronic devices are flight control devices.

[0093] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0094] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0095] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0096] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0097] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made by those skilled in the art without creative effort within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent identification and automatic detection of insulator front ends in transmission lines, characterized in that, Includes the following steps: In response to the inspection mission, the system analyzes the preset route and flies to the target waypoint based on the waypoint information within it. The video stream of the target waypoint is acquired, and the image frames are identified using a pre-trained insulator string recognition model to detect the insulator strings. Adjust the orientation so that the target insulator string is centered directly in front of it; Switch to the piece-by-piece recognition mode to identify the steel cap part of each insulator piece and obtain its center point position in the form of pixel coordinates; Obtain the distance information of the steel cap of the insulator being tested in front, and calculate the three-dimensional coordinate position of the center point by combining the machine body parameters and external parameter calibration data; Based on the three-dimensional coordinates of the center point, the pixel-level offset from the center of the video frame is calculated. The position of the camera body is finely adjusted according to the relationship between the image coordinate system, the camera coordinate system, and the world coordinate system to perform piece-by-piece contact detection.

2. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 1, characterized in that, The insulator string identification model includes an input end, a backbone network, and an output end. The input end incorporates data augmentation and random affine transformation. The backbone network adopts the C2f structure from the YOLOv8-CSP module, and the Neck structure uses bidirectional feature fusion with FPN and PAN. The Dual Head multi-task head is divided into a rotating frame detection head and a segmentation detection head, which are used to identify insulator strings and insulator caps, respectively. The output end uses a joint loss function for end-to-end training.

3. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 2, characterized in that, The end-to-end training process using a joint loss function at the output end includes: The total loss is: in, , , These are the classification loss, position regression loss, and angle loss, respectively, used for the detection branch loss of insulator string identification; , Separately segment DCE and Dice losses for semantic segmentation loss in insulator piece recognition to improve the ability to preserve small target boundaries; , These are the loss weights, used to adjust task balance during training.

4. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 1, characterized in that, During the training process of the insulator string identification model, the error between the target's true value and the network's predicted value is calculated through forward propagation, and then the network parameters are updated and optimized by continuously minimizing the error through backpropagation.

5. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 1, characterized in that, The process of obtaining the center point position in the form of pixel coordinates includes determining the arrangement direction of the target insulator string and numbering it in sequence, and calculating the center pixel coordinates of each insulator.

6. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 1, characterized in that, The process of obtaining the distance information of the steel cap of the insulator being tested in front, and calculating the three-dimensional coordinate position of the center point by combining the machine parameters and external parameter calibration data includes: obtaining the spatial depth of the center point, combining the machine parameters and external parameter calibration data, and using the pinhole camera model to convert the two-dimensional pixels in the image into three-dimensional coordinates based on the camera coordinate system.

7. The intelligent identification and automatic detection method for the front end of transmission line insulators as described in claim 1, characterized in that, The process of fine-tuning the position of the camera body based on the relationship between the image coordinate system, the camera coordinate system, and the world coordinate system includes obtaining the coordinates of the target point based on the three-dimensional coordinates of the center point and the camera coordinate system, and calculating the distance to the steel cap based on the coordinates of the target point. The camera attitude is obtained as the forward and downward unit vectors to obtain the flight target point in the camera coordinate system. The flight target point based on the camera coordinate system is converted to the fuselage coordinate system. Based on the UAV's positioning attitude, the target point coordinates are converted from the fuselage coordinate system to the world coordinate system. Based on the position of the detection device relative to the fuselage, the target point is calibrated to obtain the final target point.

8. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 7, characterized in that, The coordinates of the detected target point are obtained based on the camera coordinate system. : in, , These are the horizontal and vertical coordinates of the target point in the camera coordinate system, respectively. represents the depth coordinates of the target point in the camera coordinate system; x and y are the horizontal and vertical pixel coordinates of the target point in the image, respectively. , These are the horizontal and vertical pixel coordinates of the image center, respectively; , These are the camera's horizontal and vertical focal lengths, respectively. for To determine if correction is needed, if the ranging point is on the inter-disc spacing, no correction is needed; if the ranging point is on the front surface of the insulator disk, correction is needed. , where R is the radius of the insulator skirt and D is the spatial depth of the target point.

9. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 7, characterized in that, The process of calculating the distance between the steel caps includes: Correcting the offset of the front detection device, the distances of the center point of the detection device in front of and below the drone are respectively... x , h The target point is moved backward and upward by the corresponding distances to obtain the corrected distance.

10. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 7, characterized in that, Obtain the flight target point in the camera coordinate system : in, For the set target point, corresponding or The forward unit vector is The downward unit vector is .

11. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 7, characterized in that, Convert the flight target point based on the camera coordinate system to the fuselage coordinate system: in, This is the external parameter relationship matrix between the gimbal attitude, camera, and drone; This is a rotation matrix used to describe the rotation of the camera coordinate system to align with the body coordinate system; It is a translation vector used to describe the position of the camera's optical center in the camera's coordinate system.

12. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 7, characterized in that, Based on the drone's attitude positioning, the target point coordinates are transformed from the drone's coordinate system to the world coordinate system: in, The target point's world coordinates are the target input required by the flight control system. The transformation matrix for the fuselage to world coordinates; For the fuselage to face the rotation matrix, This is the drone's position matrix, representing the drone's position in the world coordinate system.

13. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 10, characterized in that, Based on the position of the detection device relative to the fuselage, the target point is calibrated to obtain the final target point: in, The final target point in the world coordinate system; L This is a scalar measure of the distance between the detection device and the center of the machine body.

14. The intelligent identification and automatic detection method for the front end of a transmission line insulator as described in claim 1, characterized in that, The process of fine-tuning the fuselage position includes: Calculate the error vector with the center of the steel cap as the target point: in, This is the current location of the drone; Construct control vectors and set the maximum flight speed to... The flight control speed is: in, The proportional gain controls the response speed; clip () is a function that limits the numerical range; This is the speed factor, used to control flight speed; This represents the distance between the current location and the target location. If both the position error and velocity meet the threshold constraints, the drone is deemed to have met the detection conditions, and detection is initiated. in, and These are the position error and the velocity threshold, respectively.

15. A smart identification and automatic detection system for the front end of insulators in transmission lines, characterized in that, include: Unmanned aerial vehicles (UAVs) are used to respond to inspection tasks, analyze preset routes, and fly to the target waypoint based on the waypoint information within them. Adjust the orientation so that the target insulator string is centered directly in front of it; The camera is used to acquire video streams of the target waypoint, and a pre-trained insulator string recognition model is used to identify the image frames and detect the insulator strings. The control module is used to switch to the piece-by-piece recognition mode, identify the steel cap part of each insulator piece one by one, and obtain the center point position in the form of pixel coordinates; Based on the three-dimensional coordinate position of the center point, the pixel-level offset from the center position of the video frame is calculated, and the position of the camera body is finely adjusted according to the relationship between the image coordinate system, the camera coordinate system and the world coordinate system. The laser module is used to acquire distance information of the steel cap of the insulator being tested in front; The calculation module is used to calculate the three-dimensional coordinate position of the center point based on distance information, combined with fuselage parameters and external parameter calibration data; The testing device is used for contact testing of each piece individually.

16. An electronic device, characterized in that, It includes a memory and a processor, as well as computer instructions stored in the memory and running on the processor, which, when executed by the processor, perform the steps of the method according to any one of claims 1-14.

Images