Inventory method based on distribution network line equipment body identification model and storage medium

Abstract

The invention provides an inventory method based on a distribution network line equipment body identification model and a storage medium, an unmanned aerial vehicle technology, an intelligent image identification technology and an efficient inventory mode of an automatic processing flow are fused, whole-process online circulation of power line equipment inventory is realized, the inventory efficiency and accuracy are improved, and the manual operation difficulty and cost are reduced; the scheme comprises the following steps: step 1, formulating a route task and formulating an inventory task work order; step 2, performing on-site checking operation by the unmanned aerial vehicle; 3, when the intelligent model is used for identifying the equipment and returning the photo, automatically matching and associating the photo shot by the unmanned aerial vehicle and the equipment identified in the photo with the appraisal list; 4, returning an unmanned aerial vehicle checking result to the unmanned aerial vehicle inspection platform; step 5, manually confirming the inventory result, and auditing the inventory result; step 6, pushing checking information to the homologous maintenance tool, hooking the equipment photo by the homologous maintenance tool, closing the work order, and returning information; and 7, the unmanned aerial vehicle inspection platform updates the information state.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle inventory, and in particular to an inventory method based on a distribution network line equipment ontology recognition model. Background Art

[0002] At present, the inventory of power line equipment mainly relies on manually operated drones to collect data on site. The inventory personnel use drones to take photos of equipment on lines and towers, and copy the collected photos to the information intranet system offline. They manually identify the nameplates of the photos and the equipment data in the information intranet, modify the device names corresponding to the photos, and upload them. After the online review is correct, they are synchronized to the corresponding equipment of the same source maintenance tool. This method has exposed many shortcomings in actual application:

[0003] ·Manual inventory is difficult: It is difficult to operate drones for manual inventory of on-site equipment. This is especially true in complex distribution network equipment environments, with many obstructions, many crossing lines, densely populated houses, and a large number of devices that need to be counted at a single point. Manual inventory may result in the omission of some equipment, long operation time, and unclear photos.

[0004] Large workload of data verification: After collecting the inventory data, it is necessary to conduct a second data verification, which involves duplication of work and a large workload. It is necessary to manually recheck the correspondence between the equipment and photos in the intranet system. Since the appearance of the equipment is highly similar, there may be multiple types of equipment at the same location, which makes the verification task heavy.

[0005] System docking is cumbersome and time-consuming: It is necessary to synchronize the same source data and connect the two systems according to the standards corresponding to the same source data. It is necessary to establish corresponding process specifications, sort out and check whether there are any missing devices, process docking and other work that is cumbersome and time-consuming.

[0006] In view of the above-mentioned deficiencies in the prior art, it is of great significance to develop an online flow of the entire process of power line equipment inventory.

[0007] By introducing advanced image recognition technology and automated processing, we can achieve rapid and accurate identification and inventory of distribution network line equipment, improve inventory efficiency and accuracy, and reduce the difficulty and cost of manual operation. At the same time, by optimizing system docking and data synchronization processes, we can establish an efficient power line equipment inventory management model. Summary of the invention

[0008] The present invention aims to address the technical defects of the prior art and provide an inventory method and storage medium based on the distribution network line equipment body recognition model, which integrates drone technology, intelligent image recognition technology and automated processing flow to achieve an efficient inventory method, realize the online flow of the entire process of power line equipment inventory, improve inventory efficiency and accuracy, and reduce the difficulty and cost of manual operation.

[0009] The present invention provides an inventory method based on a distribution network line equipment body identification model, comprising the steps of:

[0010] Step 1: Prepare route tasks and inventory task work orders;

[0011] Step 2: On-site inventory operation by drone;

[0012] Step 3: When the intelligent model is used to identify the equipment and then transmits the photos back, the photos taken by the drone and the equipment identified in the photos are automatically matched and associated with the equipment list;

[0013] Step 4: The drone inventory results are transmitted back to the drone inspection platform;

[0014] Step 5: Manual confirmation and review of inventory results;

[0015] Step 6: Push the inventory information to the same-source maintenance tool, which mounts the equipment photo, closes the work order, and returns the information to the drone inspection platform;

[0016] Step 7: The drone inspection platform updates the information status.

[0017] Preferably, the route planning task in step 1 is based on point cloud UAV route planning, including the following steps:

[0018] Step 1-1, point cloud data acquisition: the drone collects high-precision point cloud data of the target area, and the point cloud data includes longitude, latitude, and altitude information of the target area;

[0019] Step 1-2, point cloud data calibration: manually mark the towers on the point cloud model according to the longitude, latitude and altitude information in the point cloud data, and mark the position of each tower;

[0020] Steps 1-3, route generation: Automatically generate the flight route of the drone based on the point cloud data and the manually calibrated tower positions. The route includes the aircraft's take-off and landing positions, hovering positions, camera shooting angles, and the number of photos taken.

[0021] Preferably, the intelligent model in step 3 adopts the YOLO11 target detection model.

[0022] Preferably, the training step of the intelligent model in step 3 includes:

[0023] Step 3-1, collect sample photos of materials related to on-site equipment;

[0024] Step 3-2, manually marking the device body;

[0025] Use LabelImg to manually label the collected photos and assign labels to the equipment bodies. The labels include tower head, cement pole, pole circuit breaker, pole transformer, cross arm, pole disconnector, steel pipe pole, and insulator.

[0026] Save the annotation results and generate an annotation file in XML format;

[0027] Step 3-3, export the sample data set;

[0028] Use Roboflow to convert the annotated files and photos into a YOLO format dataset, and divide the dataset into a training set and a test set;

[0029] Step 3-4, training and verification samples: train the training set and verify it on the verification set.

[0030] Preferably, the trained intelligent model is deployed on a flight controller terminal of the UAV.

[0031] Preferably, during the inventory operation, the recognition mode is turned on, and the intelligent model function will access the video stream in the flight controller for real-time recognition. When the corresponding device is recognized, a photo is taken and matched with the device body and the typical equipment list.

[0032] Preferably, in step 5, the photo quality and the completeness of the matching of the typical design list are manually reviewed, and if the review is passed, steps 6-7 are executed;

[0033] If the review fails, proceed to steps 2-4.

[0034] Preferably, the drone uses Beidou positioning and RTK modules to obtain the coordinates of the drone's photographing position.

[0035] Preferably, the drone is a DJI Mavic 3T / 3E equipped with a DJI Pilot 2 system.

[0036] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements an inventory method based on a distribution network line equipment ontology identification model.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] The present invention provides an inventory method based on the distribution network line equipment ontology recognition model, which is an efficient inventory method that integrates drone technology, intelligent image recognition technology and automated processing flow, and realizes rapid and accurate identification and inventory of power line equipment. It mainly includes the following steps:

[0039] Formulate route tasks and inventory task work orders: Before starting the inventory, it is necessary to formulate detailed route tasks and inventory work orders, which include determining the inventory area, inventory time, and the required drone equipment and personnel configuration. This step is the basis of the entire inventory process and ensures that the inventory work is carried out in an orderly manner;

[0040] On-site inventory operation by drone: Use drones to conduct on-site inventory operations on distribution network line equipment; drones can be equipped with high-definition cameras to take photos and videos of line equipment, providing raw data for subsequent intelligent identification;

[0041] When the intelligent model is used to identify the equipment and then transmits the photos back, the photos taken by the drone and the equipment identified in the photos are automatically matched and associated with the equipment list: through the intelligent model, the equipment in the photo is automatically identified by taking photos, and matched and associated with the equipment information in the equipment list, realizing automated equipment identification and inventory;

[0042] The drone inventory results are sent back to the drone inspection platform: After the inventory is completed, the drone will send the identification results and photo data back to the drone inspection platform, providing a basis for subsequent manual review and data processing;

[0043] Manual confirmation and review of inventory results: Although the intelligent model can automatically identify devices, manual confirmation and review of inventory results are still required to ensure the accuracy and reliability of the inventory;

[0044] Push inventory information to the same-source maintenance tool, which attaches equipment photos, closes the work order, and returns information to the drone inspection platform: After review, the inventory information will be pushed to the same-source maintenance tool, which will attach equipment photos, close the work order after processing, and feedback the processing results to the drone inspection platform;

[0045] ·UAV inspection platform updates information status: Finally, the UAV inspection platform will update the information status based on the feedback from the same-source maintenance tool, completing the online flow and data synchronization of the entire inventory process; through this process, the inventory work of power line equipment not only improves efficiency and accuracy, but also reduces the difficulty and cost of manual operation, and establishes an efficient power line equipment inventory management model;

[0046] Through the above steps, advanced image recognition technology and automated processing procedures are introduced to realize the online flow of the entire process of power line equipment inventory. This method not only improves the efficiency and accuracy of inventory, reduces the difficulty and cost of manual operation, but also establishes an efficient power line equipment inventory management mode by optimizing system docking and data synchronization processes;

[0047] In the future, with the continuous advancement of technology and in-depth application, this method will be further optimized and improved to provide more powerful support for the intelligent management of the power industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a flow chart of the present invention;

[0049] Figure 2 For UAV route planning based on point cloud;

[0050] Figure 3 Inventory photos of smart model recognition equipment for drones;

[0051] Figure 4 Set up a list interface for the dictionary;

[0052] Figure 5 Take screenshots for smart model training;

[0053] Figure 6 Complete the model file screenshot for intelligent training. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. All other embodiments obtained by ordinary technicians in the field without making creative work based on the embodiments of the present invention shall fall within the scope of protection of the present invention.

[0055] like Figures 1 to 6 As shown, it shows a specific embodiment of the present invention:

[0056] The inventory method based on the distribution network line equipment body identification model in the technical solution of the present invention comprises the steps of:

[0057] Step 1: Prepare route tasks and inventory task work orders;

[0058] Step 2: On-site inventory operation by drone;

[0059] Step 3: When the intelligent model is used to identify the equipment and then transmits the photos back, the photos taken by the drone and the equipment identified in the photos are automatically matched and associated with the equipment list;

[0060] Step 4: The drone inventory results are transmitted back to the drone inspection platform;

[0061] Step 5: Manual confirmation and review of inventory results;

[0062] Step 6: Push the inventory information to the same-source maintenance tool, which mounts the equipment photo, closes the work order, and returns the information to the drone inspection platform;

[0063] Step 7: The drone inspection platform updates the information status.

[0064] Preferably, the route planning task in step 1 is based on point cloud UAV route planning, including the following steps:

[0065] Step 1-1, point cloud data acquisition: the drone collects high-precision point cloud data of the target area, and the point cloud data includes longitude, latitude, and altitude information of the target area;

[0066] Step 1-2, point cloud data calibration: manually mark the towers on the point cloud model according to the longitude, latitude and altitude information in the point cloud data, and mark the position of each tower;

[0067] Steps 1-3, route generation: Automatically generate the flight route of the drone based on the point cloud data and the manually calibrated tower positions. The route includes the aircraft's take-off and landing positions, hovering positions, camera shooting angles, and the number of photos taken.

[0068] The high-precision point cloud data collected by drones in the target area is the basis for route planning. The point cloud data is obtained by scanning the target area with the laser radar system carried by the drone, such as Figure 2 As shown in Figure 1, it contains the longitude, latitude and altitude information of each point in the target area. These data usually exist in the form of three-dimensional coordinates and can accurately describe the terrain and landform characteristics of the target area.

[0069] When collecting point cloud data, the drone needs to fly according to the preset flight trajectory to ensure full coverage of the entire target area. The lidar system emits a laser beam and receives the reflected signal, and determines the three-dimensional coordinates of the target point by calculating the flight time and direction of the laser beam. After processing and analysis, these data can be obtained into high-precision and high-resolution point cloud data.

[0070] Point cloud data calibration is the process of marking key facilities such as poles and towers on the point cloud model. According to the longitude, latitude and altitude information in the point cloud data, combined with on-site surveys and field measurements, key facilities such as poles and towers are accurately marked on the point cloud model.

[0071] During the calibration process, special attention should be paid to the exact location and height of the tower. This information is crucial for subsequent route generation and drone inspections. If the calibration is inaccurate, the drone may deviate from the intended route or fail to accurately capture key facilities during the inspection.

[0072] Based on the point cloud data and the manually calibrated tower positions, the flight route of the drone can be automatically generated. Route generation is the core link of drone route planning, which requires comprehensive consideration of multiple factors such as the terrain and landform characteristics of the target area, the location and height information of the tower, and the performance parameters of the drone.

[0073] Route generation usually includes the following steps:

[0074] Determine the take-off and landing location: Determine the appropriate take-off and landing location based on the terrain and topographic features of the target area and the take-off and landing requirements of the drone. The take-off and landing location needs to be flat, open and safe to ensure that the drone can take off and land smoothly.

[0075] Plan flight path: Plan the flight path of the drone based on point cloud data and tower location information. The flight path needs to avoid obstacles and dangerous areas as much as possible, while ensuring full coverage of the target area and key facilities.

[0076] Set the hovering position and shooting parameters: Set the appropriate hovering position and shooting parameters in the flight path. The hovering position needs to ensure that the drone can hover stably and take clear photos or videos; shooting parameters include camera shooting angle, exposure time, shutter speed, etc., which need to be adjusted according to shooting requirements and drone performance.

[0077] Determine the number of photos to be taken: Determine the number of photos to be taken based on the inspection requirements and the characteristics of the target area. The number of photos taken needs to be sufficient to ensure that the target area and key facilities are fully covered, while avoiding too many photos that will cause difficulties in data processing and analysis.

[0078] After the route is generated, strict testing and verification are required to ensure that the drone can fly and inspect according to the planned route. Testing and verification usually include multiple links such as simulated flight testing, field flight testing, and data comparison and analysis. Through these tests and verifications, problems and deficiencies in route planning can be discovered and corrected in a timely manner, ensuring that the drone can complete its mission safely and efficiently.

[0079] The 3D point cloud can provide the precise geographic location of the equipment, and the drone can take more accurate photos of the equipment based on this information. The visualization of the point cloud can optimize the various parts of the equipment and the shooting points more accurately, reducing the error caused by the jitter of the drone when the model is looking for the equipment.

[0080] In summary, point cloud-based UAV route planning is a complex and delicate task that requires comprehensive consideration of multiple factors and multiple steps of processing and analysis. Through scientific planning and management, it can ensure that the UAV can fly and inspect according to the predetermined route, improving the quality and efficiency of the inspection work.

[0081] Preferably, the intelligent model in step 3 adopts the YOLO11 target detection model.

[0082] YOLO11 is the latest real-time target detector launched by Ultralytics, which has significant improvements in accuracy, speed and efficiency.

[0083] Enhanced feature extraction: YOLO11 improves the accuracy of target detection by improving the Backbone and Neck architectures and adding components such as C3k2 and C2PSA.

[0084] Optimize efficiency and speed: YOLO11 redesigned the architecture, optimized the training process, and improved the processing speed.

[0085] Multi-environment adaptability: YOLO11 supports edge devices, cloud platforms, and NVIDIA GPUs, and has good environmental adaptability.

[0086] These properties of YOLO11 make it a versatile and efficient object detection model suitable for various computer vision tasks.

[0087] Preferably, the training step of the intelligent model in step 3 includes:

[0088] Step 3-1, collect sample photos of materials related to on-site equipment;

[0089] Step 3-2, manually marking the device body;

[0090] Use LabelImg to manually label the collected photos and assign labels to the equipment bodies. The labels include tower head, cement pole, pole circuit breaker, pole transformer, cross arm, pole disconnector, steel pipe pole, and insulator.

[0091] Save the annotation results and generate an annotation file in XML format;

[0092] Step 3-3, export the sample data set;

[0093] Use Roboflow to convert the annotated files and photos into a YOLO format dataset, and divide the dataset into a training set and a test set;

[0094] Step 3-4, training and verification samples: train the training set and verify it on the verification set.

[0095] like Figures 5-6 As shown, in step 3, the training of the intelligent model is the key to ensuring that the photos taken by the drone can be accurately identified.

[0096] Step 3-1: Collect sample photos of materials related to on-site equipment

[0097] Obtain a large number of photos containing distribution network line equipment for subsequent training and verification.

[0098] Collect photos covering various equipment types, angles, lighting and background conditions by taking photos with drones, taking photos on site or obtaining them from public datasets. A total of 7,931 photos have been collected from various sites in this plan.

[0099] Step 3-2: Manually mark the device body

[0100] Assign labels to device entities in photos so that the model can learn to recognize them.

[0101] Use labeling tools such as LabelImg for manual labeling. Open the LabelImg tool and import the photos that need to be labeled. Use the mouse to draw a rectangular frame on the photo to frame the equipment body. Assign labels to the rectangular frame, such as tower head, cement pole, pole-mounted circuit breaker, pole-mounted transformer, crossarm, pole-mounted disconnector, steel pipe pole, insulator, etc. Save the labeling results and generate XML formatted labeling files. These files contain the location and label information of the equipment in the photos. Ensure that the labeling is accurate to avoid missing or mislabeling. At the same time, in order to improve the generalization ability of the model, the number and types of labeled photos can be increased as much as possible.

[0102] Step 3-3: Export sample dataset

[0103] Convert the annotated photos and annotation files into the format required for model training. Use dataset conversion tools such as Roboflow. Import the XML annotation files and corresponding photos into Roboflow.

[0104] Set the dataset format in Roboflow, such as YOLO format. YOLO is a popular target detection algorithm, and its dataset format includes photos, coordinates of annotation boxes, and label information.

[0105] Divide the dataset into a training set and a test set. In this solution, the samples are divided into a training set of 6717 and a test set of 1214. The training set is used to train the model, and the test set is used to verify the performance of the model. The division ratio can be adjusted according to actual needs, such as 80% for training and 20% for testing. Export the dataset to obtain the training set and test set in YOLO format.

[0106] Step 3-4: Training and validation samples

[0107] The model is trained using the training set, 200 batches of training samples are trained, and the performance of the model is verified on the validation set.

[0108] Choose appropriate deep learning frameworks and algorithms for training, such as TensorFlow, PyTorch, etc.

[0109] Configure the training environment, including installing the deep learning framework and configuring the GPU.

[0110] Load the training set and validation set, and select the appropriate model architecture and optimization algorithm. The model architecture can be selected according to the device type and characteristics, such as convolutional neural network (CNN). The optimization algorithm can be selected from stochastic gradient descent (SGD), Adam, etc.

[0111] Set hyperparameters such as learning rate, batch size, number of iterations, etc. These parameters have an important impact on the training effect of the model.

[0112] Start training the model. During the training process, the model will learn to recognize the device entity in the photo.

[0113] Verify the performance of the model on the validation set. Evaluate the performance of the model by calculating indicators such as accuracy and recall. If the performance is poor, you can adjust the hyperparameters or optimize the model architecture to improve it.

[0114] Through the above steps, an intelligent model that can accurately identify the distribution network line equipment can be trained. This model can be applied to drone inspections to achieve rapid and accurate identification and inventory of equipment.

[0115] Preferably, the trained intelligent model is deployed on a flight controller terminal of the UAV.

[0116] The trained intelligent model is deployed on the UAV flight controller terminal, and the process of directly identifying and taking pictures based on the captured video stream during the flight of the UAV, as well as associating the equipment and typical equipment lists is realized. In the training environment, the intelligent model is exported to a format suitable for embedded devices (such as UAV flight controllers).

[0117] The camera on the drone captures real-time video streams during flight. The flight controller needs to be able to obtain these video stream data and use them as raw data for model input.

[0118] like Figure 3 As shown in the figure, the flight controller uses an integrated intelligent model to perform real-time reasoning on the pre-processed video stream data. The model analyzes each frame of the image and detects and identifies the distribution network line equipment itself.

[0119] When the model recognizes the distribution network line equipment body, the flight controller will trigger the camera to take a photo. The timing and location of the photo can be determined based on the recognition results and preset rules to ensure that a clear and complete image of the equipment is captured.

[0120] The photographed photos are associated with the identified equipment type, and based on the equipment type and the identification result, the photographed photos are associated with a typical equipment list (ie, an equipment list or an inspection plan).

[0121] Through the above steps, the drone can perform real-time identification and take photos based on the captured video stream during flight, and associate the captured photos with the equipment and typical equipment list. This greatly improves the efficiency and accuracy of distribution network line inspection and provides strong data support for subsequent maintenance and management.

[0122] Preferably, during the inventory operation, the recognition mode is turned on, and the intelligent model function will access the video stream in the flight controller for real-time recognition. When the corresponding device is recognized, a photo is taken and matched with the device body and the typical equipment list.

[0123] The team pilots carry the drone to the vicinity of the target inventory route to complete the pre-takeoff preparations and choose to perform the inventory task. The drone automatically executes the route task preset for the current inventory task and autonomously completes actions such as flying and taking photos.

[0124] After taking the photo, the team members will manually check the equipment information or automatically perform an inventory and complete the information when the model is used to identify the equipment and photos are sent back.

[0125] The whole process of power line equipment inventory is realized online, and a list of typical line equipment (i.e., typical equipment list) is formulated. The typical equipment list is compiled by the State Grid according to the specifications, and the standardized typical equipment list is entered into the flight controller terminal to realize autonomous identification of drones and match and associate with the typical equipment list, thereby improving the efficiency of line equipment inventory. In terms of drone operation, point cloud-based drone route planning is realized to automate drone operation; in terms of equipment identification, a 3D point cloud-based equipment identification model is used to achieve more accurate identification and shooting; in terms of equipment maintenance, equipment typical equipment is provided, and the complete input information required for matching based on the identified equipment is provided; in terms of process control, online integrated operation is provided, and the required equipment maintenance system and homologous tools are connected to achieve integrated operation.

[0126] Preferably, in step 5, the photo quality and the completeness of the matching of the typical design list are manually reviewed, and if the review is passed, steps 6-7 are executed;

[0127] If the review fails, proceed to steps 2-4.

[0128] like Figure 4 As shown, you can review the equipment list, which mainly involves reviewing the quality of the photos taken by the drone and the completeness of the equipment list. Ensure that the photo quality meets the standards and that the photos are correctly matched with the equipment in the equipment list.

[0129] Review passed: If the photo quality and the matching of the design list meet the standards, the review is passed and proceed to the next steps (steps 6-7).

[0130] Failed the review: If the photo quality is poor or the property list is incomplete, the review will fail and you will need to redo the previous steps.

[0131] Through the manual review process, any potential problems can be discovered and corrected in a timely manner, ensuring the reliability and effectiveness of the entire inspection process.

[0132] Preferably, the drone uses Beidou positioning and RTK modules to obtain the coordinates of the drone's photographing position.

[0133] Relying on the Beidou positioning + RTK module of the drone to obtain high-precision coordinates of the drone's photo-taking location, and using coordinate comparison and model recognition equipment, the photos taken by the drone and the equipment identified in the photos are automatically matched and associated with the typical equipment list, reducing manual operation links.

[0134] Preferably, the drone is a DJI Mavic 3T / 3E equipped with a DJI Pilot 2 system.

[0135] In this solution, the DJI Mavic 3T / 3E drone is selected and equipped with the DJI Pilot 2 system, mainly considering the following aspects:

[0136] Portability and durability

[0137] Mavic 3T / 3E drones are known for their lightweight and compact design, which can be held in one hand, easy to carry and quickly deployed by one person. This feature makes them very suitable for front-line teams to carry out flexible operations in complex environments.

[0138] The drone's body is made of durable materials, which can withstand the challenges of long-term operations and ensure stable operation even in harsh environments.

[0139] HD shooting and zoom capabilities

[0140] Mavic 3T / 3E are equipped with high-resolution camera systems, including wide-angle camera and telephoto camera, supporting up to 56x hybrid zoom. This enables the drone to capture clear and detailed images at long distances.

[0141] The wide-angle camera of the Mavic 3E uses a 4 / 3 CMOS sensor with 20 million pixels, which can significantly improve the accuracy of mapping. The Mavic 3T is additionally equipped with a thermal imaging camera, which is suitable for firefighting, search and rescue, inspection, night operations and other scenes.

[0142] Battery life and fast charging technology

[0143] The drone has a flight time of up to 45 minutes and a larger single-flight operating radius, effectively improving operational efficiency and coverage.

[0144] It supports 100W high-energy fast charging and can charge a group of batteries at the same time. Charging for 50 minutes can meet the needs of the intelligent flight battery to continue working. In addition, 88W direct fast charging can be achieved directly through the drone body, ensuring flexible and continuous operation outdoors.

[0145] Intelligent flight and obstacle avoidance

[0146] The drone is equipped with an advanced flight control system that supports a variety of route types, such as "waypoint flight", "mapping aerial photography", "oblique photography", "strip flight", etc., to meet diverse operational needs.

[0147] Equipped with the APAS 5.0 obstacle avoidance system, it can achieve safe flight in multi-directional obstacle operation scenarios. At the same time, the fisheye lens and real-time 3D terrain reconstruction function further improve the safety and stability of flight.

[0148] Integration and Compatibility

[0149] Mavic 3T / 3E supports the PSDK unified interface, which facilitates the integration of third-party hardware devices.

[0150] The DJI Pilot 2 system provides an intuitive and easy-to-use user interface. UAV control and camera operation can be achieved with one click. Mission information and flight status are clearly presented, making operations more efficient.

[0151] To sum up, the DJI Mavic 3T / 3E drone is equipped with the DJI Pilot 2 system. This choice will significantly improve the efficiency and accuracy of inspection, surveying and mapping operations.

[0152] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements an inventory method based on a distribution network line equipment ontology identification model.

[0153] The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that includes one or more available media. The available medium may be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a digital video disc (DVD)), or a semiconductor medium (e.g., a solid state disk (SSD)).

[0154] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge scope of ordinary technicians in the field without departing from the purpose of the present invention. These changes involve related technologies well known to those skilled in the art, which all fall within the scope of protection of the patent of the present invention.

[0155] Many other changes and modifications may be made without departing from the concept and scope of the present invention.It should be understood that the present invention is not limited to the specific embodiments, and the scope of the present invention is defined by the appended claims.

Claims

1. An inventory method based on the distribution network line equipment ontology identification model, characterized in that: Includes steps: Step 1: Prepare route tasks and inventory task work orders; Step 2: On-site inventory operation by drone; Step 3: When the intelligent model is used to identify the equipment and then transmits the photos back, the photos taken by the drone and the equipment identified in the photos are automatically matched and associated with the equipment list; Step 4: The drone inventory results are transmitted back to the drone inspection platform; Step 5: Manual confirmation and review of inventory results; Step 6: Push the inventory information to the same-source maintenance tool, which mounts the equipment photo, closes the work order, and returns the information to the drone inspection platform; Step 7: The drone inspection platform updates the information status.

2. The inventory method based on the distribution network line equipment ontology identification model according to claim 1 is characterized in that: The route planning task in step 1 is based on the point cloud UAV route planning, including the following steps: Step 1-1, point cloud data acquisition: the drone collects high-precision point cloud data of the target area, and the point cloud data includes longitude, latitude, and altitude information of the target area; Step 1-2, point cloud data calibration: manually mark the towers on the point cloud model according to the longitude, latitude and altitude information in the point cloud data, and mark the position of each tower; Step 1-3, route generation: Automatically generate the flight route of the drone based on the point cloud data and the manually calibrated tower position. The route includes the aircraft's take-off and landing position, hovering position, camera shooting angle, and the number of photos taken.

3. The inventory method based on the distribution network line equipment ontology identification model according to claim 2 is characterized in that: The intelligent model in step 3 adopts the YOLO11 target detection model.

4. The inventory method based on the distribution network line equipment ontology identification model according to any one of claims 1 to 3 is characterized in that: The training steps of the intelligent model in step 3 include: Step 3-1, collect sample photos of materials related to on-site equipment; Step 3-2, manually marking the device body; Use LabelImg to manually label the collected photos and assign labels to the equipment bodies. The labels include tower head, cement pole, pole circuit breaker, pole transformer, cross arm, pole disconnector, steel pipe pole, and insulator. Save the annotation results and generate an annotation file in XML format; Step 3-3, export the sample data set; Use Roboflow to convert the annotated files and photos into a YOLO format dataset, and divide the dataset into a training set and a test set; Step 3-4, training and verification samples: train the training set and verify it on the verification set.

5. The inventory method based on the distribution network line equipment ontology identification model according to claim 4 is characterized in that: The trained intelligent model is deployed on the flight controller terminal of the UAV.

6. The inventory method based on the distribution network line equipment ontology identification model according to claim 5 is characterized in that: When taking inventory, turn on the recognition mode and the intelligent model function will access the video stream in the flight controller for real-time recognition. When the corresponding device is recognized, it will take a photo and match it with the device body and the equipment list.

7. The inventory method based on the distribution network line equipment ontology identification model according to claim 6 is characterized in that: In step 5, the quality of the photos and the completeness of the matching of the reference list are manually reviewed. If the review is passed, steps 6-7 are executed; If the review fails, proceed to steps 2-4.

8. The inventory method based on the distribution network line equipment ontology identification model according to claim 7 is characterized in that: The drone uses Beidou positioning and RTK modules to obtain the coordinates of the drone's photo-taking location.

9. The inventory method based on the distribution network line equipment ontology identification model according to claim 1 is characterized in that: The drone is a DJI Mavic 3T / 3E equipped with a DJI Pilot 2 system.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, an inventory method based on a distribution network line equipment ontology identification model as described in any one of claims 1 to 9 is implemented.

Images