Distribution network tree distribution intelligent patrol method and system based on high-resolution remote sensing image
Through intelligent inspection methods based on high-score remote sensing images, the problems of inefficiency and high cost of traditional inspection methods are solved, efficient and accurate identification of the distribution of trees around distribution network lines are achieved, and the safety and reliability of the power grid are improved.
Patent Information
- Application Number
- CN202411992679.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-25
AI Technical Summary
The traditional distribution network barrier inspection method is inefficient and expensive, making it difficult to meet the efficient, accurate and comprehensive inspection requirements, especially under complex terrain and extreme weather conditions.
Using an intelligent patrol method based on high-score remote sensing images, the preprocessing of satellite remote sensing images, building a tree distribution sample library, and training an intelligent identification model is achieved to achieve efficient and accurate identification of tree distribution around distribution network lines.
It improves the accuracy and efficiency of tree distribution identification, can maintain stable identification performance under complex terrain and extreme weather conditions, reduces patrol costs, and improves the safety and reliability of the power grid.
Smart Images
Figure CN119888518A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of intelligent inspection of power grids, and in particular relates to an intelligent inspection method and system for tree distribution in distribution networks based on high-resolution remote sensing images. Background Art
[0002] The distribution network, as an indispensable key link in the power system, is directly related to all aspects of residents' daily life, including industrial production, commercial operations and household electricity consumption. Therefore, ensuring the continuous and stable operation of the distribution network is one of the top priorities of the power industry. However, in actual operation, the growth of trees around the distribution lines has become a safety hazard that cannot be ignored. As time goes by, the branches and leaves of the trees may gradually extend to the vicinity of the power lines, and even come into contact with the wires under the action of natural forces such as wind and rain, thereby causing power failures such as short circuits and tripping; more seriously, when encountering strong winds, heavy rains or geological disasters, the trees may fall and directly break the lines, causing large-scale power outages, seriously affecting the normal social order and people's lives. In view of this, regular, accurate and efficient inspections of the distribution of trees around the distribution lines, timely detection and proper handling of potential tree obstacles are of vital strategic significance for maintaining power grid security and improving power supply quality and reliability.
[0003] The traditional distribution network tree barrier inspection method mainly relies on manual on-site surveys combined with drone flight inspections. These two methods can indeed meet the daily inspection needs to a certain extent, but they also expose many limitations. Although manual inspections can intuitively judge the tree barrier situation, they are inefficient and difficult to cover all areas. There are also difficulties in inspecting high or inaccessible trees. At the same time, the labor cost is high, and long-term maintenance requires a lot of resources. Although drone inspections have improved the efficiency and scope of inspections and can overcome terrain restrictions to a certain extent, their operation technology requirements are high, and they are greatly affected by weather conditions (such as strong winds, rain and snow) and environmental factors (such as electromagnetic interference). In addition, the flight stability is poor in complex terrain or dense forest areas, and there may be blind spots, making it difficult to ensure the comprehensiveness and accuracy of the inspection. Especially in complex terrains such as remote areas, mountainous areas, and forest areas, or when encountering extreme weather conditions, the difficulty and cost of implementing traditional inspection methods have risen sharply, making it difficult to meet the urgent needs of modern power grid management for efficient, accurate, and all-weather inspections. Therefore, exploring and applying more intelligent and automated inspection technologies has become a key direction for improving the level of distribution network tree barrier management. Summary of the invention
[0004] The purpose of the present invention is to provide a distribution network tree distribution intelligent inspection method and system based on high-resolution remote sensing images to solve the problem that the difficulty and cost of traditional inspection methods are significantly increased and it is difficult to meet the requirements of efficient, accurate and comprehensive inspections.
[0005] To achieve the above object, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a distribution network tree distribution intelligent inspection method based on high-resolution remote sensing images, comprising: Collect satellite remote sensing images for patrolled distribution network lines, and pre-process the satellite remote sensing images to obtain fused images; Construct a distribution line tree distribution sample library based on fused images; Construct and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The image to be identified is input into the trained intelligent identification model to obtain the intelligent identification result of tree distribution.
[0006] Furthermore, collecting satellite remote sensing images of the distribution network lines to be inspected includes: For the distribution network lines to be inspected, determine the resolution, band, width, and cloud cover parameters of the satellite remote sensing image, and then collect the satellite remote sensing image at the specified time.
[0007] Furthermore, the preprocessing of the satellite remote sensing image to obtain the fused image includes: The received satellite remote sensing data is preprocessed by radiation correction, geometric correction, and orthorectification to obtain a satellite remote sensing preprocessed image with geographic coordinates, information enhancement, and clarity. The panchromatic and multispectral images are fused to obtain a fused image with both high resolution and multispectral features.
[0008] Furthermore, the construction of a distribution line tree distribution sample library based on fused images includes: The fused image is annotated with tree samples by semi-automatic annotation combined with manual verification. The samples are uniformly stored according to the determined sample library organization form to build a sample library for intelligent identification of distribution network tree distribution. Manually annotate seed samples, visually interpret trees based on their color spectrum, shape structure, texture, and shadow features, distinguish trees from other vegetation and background objects, and manually identify and annotate the distribution information of tree objects; use manually annotated seed samples as training data, use deep learning models to automatically predict the labels of the remaining samples, and finally manually check and correct the prediction results; A structured database is established to crop and encode samples and original images, and samples and images are cropped into image patches of fixed size. The sample patches and image patches are made to correspond one to one through unified and standardized naming.
[0009] Furthermore, the intelligent recognition model is constructed and trained based on the distribution line tree distribution sample library to obtain a trained intelligent recognition model, including: Semantic segmentation is used for intelligent identification of tree distribution on distribution lines, and an improved U-Net model based on high-resolution detail fusion is constructed and trained for intelligent identification of tree distribution on distribution networks.
[0010] Furthermore, based on the fully convolutional network U-Net, the skip connection structure of the U-Net network is improved according to the principle of maintaining high-resolution information throughout the full convolutional network. Different weights are assigned to feature maps at different levels. The weights are obtained through learning, so that the network can adaptively select the most important features. Deep supervision is adopted when setting the loss function. For the tree extraction task, a weighted sum of segmentation loss functions such as cross entropy loss, Dice loss, or IoU loss is used.
[0011] Furthermore, the image to be identified is input into a trained intelligent identification model to obtain the tree distribution intelligent identification result, including: The image to be identified is input into the trained model, and prediction is performed based on the dual receptive field seamless sliding window to obtain the intelligent recognition result of tree distribution.
[0012] In a second aspect, the present invention provides a distribution network tree distribution intelligent inspection system based on high-resolution remote sensing images, comprising: The data acquisition module is used to collect satellite remote sensing images of the distribution network lines to be inspected, and pre-process the satellite remote sensing images to obtain fused images; A sample library construction module is used to construct a distribution line tree distribution sample library based on fused images; A model training module is used to build and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The prediction output module is used to input the image to be identified into the trained intelligent recognition model to obtain the intelligent recognition result of tree distribution.
[0013] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method for intelligent inspection of tree distribution in distribution networks based on high-resolution remote sensing images when executing the computer program.
[0014] In a fourth aspect, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method for intelligent inspection of tree distribution in distribution networks based on high-resolution remote sensing images.
[0015] Compared with the prior art, the present invention has the following technical effects: The present invention proposes an intelligent inspection method for distribution network tree distribution based on high-resolution remote sensing images, aiming to achieve efficient and accurate identification of tree distribution around distribution network lines through the combination of high-resolution satellite remote sensing technology and deep learning algorithm.
[0016] The present invention ensures the pertinence and timeliness of data by accurately setting the resolution, band, width, cloud cover and other parameters of satellite remote sensing images and collecting images at a specified time. The preprocessing steps (including radiation correction, geometric correction, and orthorectification) effectively remove distortion and interference in the images, improving the quality and accuracy of the images. The fusion of panchromatic and multispectral images takes into account both high resolution and rich spectral information, providing high-quality input data for subsequent tree identification.
[0017] The present invention adopts a method of semi-automatic labeling combined with manual verification, which not only ensures the labeling efficiency but also ensures the accuracy of labeling. The deep learning model automatically predicts the labels of the remaining samples and manually verifies and corrects them, which greatly reduces the workload of manual labeling. The establishment of a structured database enables efficient storage and retrieval of samples and images, providing strong data support for model training.
[0018] The present invention selects semantic segmentation as the technical route for tree distribution recognition, and constructs an improved U-Net model based on high-resolution detail fusion, which fully considers the characteristics of the tree recognition task. By improving the skip connection structure of the U-Net network, maintaining high-resolution information, and adaptively selecting the most important features, the recognition accuracy of the model is improved. The loss function design of deep supervision is adopted to further optimize the training process of the model and ensure the accuracy of tree extraction.
[0019] The present invention inputs the image to be identified into the trained model and predicts it through a dual receptive field seamless sliding window, thus achieving comprehensive and accurate identification of tree distribution. This method not only improves the recognition efficiency, but also maintains stable recognition performance under complex terrain and severe weather conditions, breaking through the limitations of traditional inspection methods.
[0020] Compared with traditional manual on-site surveys and drone inspections, this method can quickly cover a large area and significantly improve inspection efficiency. It reduces the reliance on manpower and inspection equipment, and reduces inspection costs. Through the combination of deep learning algorithms and satellite remote sensing technology, the intelligent extraction of tree distribution information is realized, which improves the intelligence level of inspection. This method is not limited by terrain and weather conditions, and can maintain stable recognition performance in complex environments and extreme weather conditions. This method can be extended to other similar remote sensing monitoring tasks, such as vegetation cover monitoring, land use change detection, etc., and has broad application prospects.
[0021] In summary, this technical solution achieves efficient and accurate identification of tree distribution in distribution networks through the combination of high-resolution satellite remote sensing images and tree distribution semantic segmentation models, providing strong technical support for the safe and stable operation of power systems. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 It is a flow chart of the present invention.
[0023] Figure 2 This is a schematic diagram of the structure of the Chongqing distribution network tree distribution intelligent recognition model based on the improved U-Net with high-resolution detail fusion of the present invention.
[0024] Figure 3 This is an example of a tree sample.
[0025] Figure 4 Schematic diagram of the structure of the Chongqing distribution network tree distribution intelligent recognition model based on the improved U-Net with high-resolution detail fusion.
[0026] Figure 5 It is a dual receptive field seamless sliding window inference scheme.
[0027] Figure 6 This is an illustration of tree detection results in a standard false color image. DETAILED DESCRIPTION
[0028] The present invention is further described below in conjunction with the accompanying drawings: Example 1, please refer to Figure 1 The present invention provides a distribution network tree distribution intelligent inspection method based on high-resolution remote sensing images, including: Collect satellite remote sensing images for patrolled distribution network lines, and pre-process the satellite remote sensing images to obtain fused images; Construct a distribution line tree distribution sample library based on fused images; Construct and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The image to be identified is input into the trained intelligent identification model to obtain the intelligent identification result of tree distribution.
[0029] By collecting high-resolution satellite remote sensing images and combining them with preprocessing technology to obtain fused images, the present invention can capture more tree detail information, such as the outline, color, texture, etc. of the trees, thereby improving the accuracy of tree recognition.
[0030] The construction and training of intelligent recognition models have automated the tree distribution recognition process, greatly improving recognition efficiency and reducing manual intervention and recognition time.
[0031] The application of satellite remote sensing technology means that the inspection scope is no longer limited by geographical location and weather conditions, and can cover a wide range of distribution network line areas, including remote areas and complex terrain.
[0032] Once the intelligent recognition model is trained, it can identify tree distribution around the clock without time constraints and provide inspection results in a timely manner.
[0033] Traditional distribution network tree barrier inspections rely on manual on-site surveys and drone inspections, which are costly and inefficient. This invention replaces some manual work with intelligent technology, reducing the labor and time costs of inspections. Although the acquisition of satellite remote sensing images requires a certain cost, it is more cost-effective than long-term manual inspections and drone inspections, especially in large-scale inspection tasks.
[0034] Accurate tree distribution identification results help to promptly discover and deal with potential tree obstacles, avoid accidents such as short circuits caused by trees touching wires, and trees falling and breaking lines, thereby improving the safety and reliability of the power grid.
[0035] Through regular smart inspections, a tree growth profile around distribution lines can be established to provide data support for future grid planning and maintenance.
[0036] In summary, the intelligent inspection method for tree distribution in distribution networks based on high-resolution remote sensing images provided by the present invention has shown significant technical effects in improving recognition accuracy and efficiency, achieving large-scale all-weather inspections, reducing inspection costs, improving power grid security and reliability, and promoting intelligent and information development. This method provides new solutions and technical support for distribution network management in the power industry, and has broad application prospects and important practical value.
[0037] Embodiment 2, the present invention provides a distribution network tree distribution intelligent inspection method based on high-resolution remote sensing images, specifically comprising: Step 1: Acquire remote sensing images.
[0038] For the distribution network lines to be inspected, determine the appropriate key imaging parameters of satellite remote sensing images (resolution, band, width, cloud cover, etc.), then select the appropriate remote sensing satellite type from the existing remote sensing satellites, and collect satellite remote sensing images at the specified time.
[0039] Step 2: Satellite remote sensing image preprocessing.
[0040] The received satellite remote sensing data is preprocessed, including radiation correction, geometric correction, orthorectification and other steps, to obtain high-level satellite remote sensing preprocessing products with geographic coordinates, information enhancement, clarity and reliability. The panchromatic and multispectral images of the image are fused to obtain a fused image with both high resolution and multispectral features.
[0041] Step 3: Construction of distribution line tree distribution sample library. Use semi-automatic annotation combined with manual verification to annotate tree samples in the image, store them uniformly according to the determined sample library organization form, and build a sample library for intelligent identification of distribution network tree distribution.
[0042] Step 3.1: Manually label seed samples, select a small representative image area, visually interpret trees based on their color spectrum, shape structure, texture, and shadow features, distinguish trees, other vegetation, and background objects, and manually identify and label the distribution information of tree objects. Use a small number of manually labeled seed samples as training data, use a deep learning model to automatically predict the labels of the remaining samples, and finally manually check and manually correct the prediction results.
[0043] Step 3.2, establish a structured database, crop and encode the samples and original images. Crop the samples and images into fixed-size image patches, each of which is 256×256 pixels in size, and make the sample patches correspond to the image patches one by one through unified and standardized naming. The attribute fields in the sample database include serial number, category, text attribute, tree area coordinates and other information.
[0044] Step 4: Construction and training of intelligent identification model for tree distribution in distribution network Semantic segmentation technology is used to solve the problem of intelligent identification of tree distribution on distribution lines. An improved U-Net model based on high-resolution detail fusion is constructed and trained for intelligent identification of tree distribution on distribution networks.
[0045] In order to solve the problem of high-resolution information loss when using deep learning to extract trees, resulting in blurred edges and loss of details, and the inadequate use of spectrum, texture and other information of remote sensing images by deep learning networks, the advantages of multi-band and rich information of remote sensing images cannot be fully utilized. Based on the fully convolutional network U-Net, the symmetric encoding and decoding structure of the U-Net network is retained, such as Figure 2 As shown, the following two improvements have been made: (1) Based on the idea of maintaining high-resolution information throughout the full convolutional network, the skip connection structure of the U-Net network is improved. In order to better utilize the retained high-resolution information to train the network parameters (the copied or superimposed part in the figure), different weights are assigned to feature maps at different levels to more finely control the flow and fusion of information. The weights can be obtained through learning, allowing the network to adaptively select the most important features.
[0046] (2) The idea of deep supervision is adopted when setting the loss function (the red layers in the figure are all used for loss function output calculation), which encourages the network to produce meaningful outputs in the intermediate layers, thereby accelerating the training process and improving the generalization ability of the model. For the tree extraction task, the weighted sum of commonly used segmentation loss functions such as cross entropy loss, Dice loss or IoU loss can be considered.
[0047] Step 5: Input the image to be identified into the trained model, perform prediction based on the dual receptive field seamless sliding window, and obtain the intelligent identification result of tree distribution.
[0048] This invention uses high-resolution satellite remote sensing images and tree distribution semantic segmentation models to achieve intelligent extraction of tree distribution information, breaking through the limitations of traditional inspection methods, greatly improving inspection efficiency, reducing manpower input and inspection costs, and enhancing the level of intelligence.
[0049] Example 3: The technical solution of the present invention has been implemented in 50 lines including 10kV five-pair branch lines in a certain area, and is introduced by taking the inspection in a certain area as an example.
[0050] Step 1: Acquire remote sensing images.
[0051] Based on the 50 designated high-fault lines, determine the appropriate key imaging parameters of satellite remote sensing images (including resolution better than 0.5 meters, width, imaging angle, etc.), then select the appropriate remote sensing satellite type from the existing remote sensing satellites, and collect satellite remote sensing images at the specified time.
[0052] Step 2: Satellite remote sensing image preprocessing.
[0053] The received satellite remote sensing data is preprocessed, including radiation correction, geometric correction, orthorectification and other steps, to obtain high-level satellite remote sensing preprocessing products with geographic coordinates, information enhancement, clarity and reliability. The panchromatic and multispectral images of the image are fused to obtain a fused image with both high resolution and multispectral features.
[0054] Step 3: Construct a sample library of tree distribution on a distribution line. Focusing on the tree distribution inspection of a designated distribution line in a city, the images are labeled with tree samples by using semi-automatic labeling combined with manual verification. The samples are uniformly stored according to the determined sample library organization form, and a sample library suitable for intelligent identification of tree distribution in a distribution network in a certain area is constructed.
[0055] Step 3.1: Manually label seed samples, select a small representative image area, visually interpret trees based on their color spectrum, shape structure, texture, and shadow features, distinguish trees, other vegetation, and background objects, and manually identify and label the distribution information of tree objects. Use a small number of manually labeled seed samples as training data, use a deep learning model to automatically predict the labels of the remaining samples, and finally manually check and manually correct the prediction results.
[0056] Step 3.2, in the sample storage stage, the samples and original images need to be cropped and encoded first. The samples and images are cropped into fixed-size image blocks (patches), each patch is 256×256 pixels in size, and the sample patches are in one-to-one correspondence with the image patches through unified and standardized naming. The organization of sample storage is shown in the following table: Table 1 Description of sample library attribute fields
[0057] Table 2 Organization of sample library
[0058] Step 4: Construction and training of a tree distribution intelligent recognition model for a distribution network Semantic segmentation technology is used to solve the problem of intelligent identification of tree distribution on distribution lines. An improved U-Net model based on high-resolution detail fusion is constructed and trained for intelligent identification of tree distribution in a distribution network.
[0059] Among them, in order to solve the problem of high-resolution information loss when using deep learning to extract trees, resulting in blurred edges and loss of details, and the inadequate use of spectrum, texture and other information of remote sensing images by deep learning networks, and the inability to give full play to the advantages of multi-band and rich information of remote sensing images, on the basis of the fully convolutional network U-Net, the symmetric encoding and decoding structure of the U-Net network is retained, and the following two improvements are made (such as Figure 4 shown): (1) Based on the idea of maintaining high-resolution information throughout the full convolutional network, the skip connection structure of the U-Net network is improved. In order to better utilize the retained high-resolution information to train the network parameters (the copied or superimposed part in the figure), different weights are assigned to feature maps at different levels to more finely control the flow and fusion of information. The weights can be obtained through learning, allowing the network to adaptively select the most important features.
[0060] (2) The idea of deep supervision is adopted when setting the loss function (the red layers in the figure are all used for loss function output calculation), which encourages the network to produce meaningful outputs in the intermediate layers, thereby accelerating the training process and improving the generalization ability of the model. For the tree extraction task, the weighted sum of commonly used segmentation loss functions such as cross entropy loss, Dice loss or IoU loss can be considered.
[0061] Step 5: Input the image to be identified into the trained model, perform prediction based on the dual receptive field seamless sliding window, and obtain the intelligent identification result of tree distribution.
[0062] For each model prediction, due to the limitation of sliding window size, the effective receptive field that the model can actually perceive cannot extend beyond the sliding window, which leads to incomplete contextual information in the edge area of the sliding window, and then makes the prediction results unstable, such as errors, omissions, and truncation. Especially when multiple sliding windows are spliced together, the instability of the prediction results near the sliding window splicing line will be more serious due to the error accumulation effect, which has a great impact on the global prediction accuracy. In this regard, the prediction scheme of this technical solution adopts a seamless sliding window prediction scheme with dual receptive fields. The schematic diagram of the scheme is shown in the figure. Figure 5 As shown in the figure. The receptive field is divided into a basic receptive field and an extended receptive field. The basic receptive field is the inner area of the prediction window with complete context information. The extended receptive field is the outer area of the prediction window with incomplete context information. During the reasoning process, the window slides with a step size of (L-2αL), where L represents the sliding window size and α represents the overlap coefficient. Only the prediction results within the basic receptive field are retained and finally spliced into the final prediction results. The reasoning results within the basic receptive field have an extended perception of the background context within the extended receptive field, which can greatly improve the stability of the spliced prediction results and effectively avoid the instability near the window splicing line.
[0063] In yet another embodiment of the present invention, a distribution network tree distribution intelligent inspection system based on high-resolution remote sensing images is provided, which can be used to implement the above-mentioned distribution network tree distribution intelligent inspection method based on high-resolution remote sensing images. Specifically, the system includes: The data acquisition module is used to collect satellite remote sensing images of the distribution network lines to be inspected, and pre-process the satellite remote sensing images to obtain fused images; A sample library construction module is used to construct a distribution line tree distribution sample library based on fused images; A model training module is used to build and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The prediction output module is used to input the image to be identified into the trained intelligent recognition model to obtain the intelligent recognition result of tree distribution.
[0064] The division of modules in the embodiments of the present invention is schematic and is only a logical function division. There may be other division methods in actual implementation. In addition, each functional module in each embodiment of the present invention may be integrated into one processor, or may exist physically separately, or two or more modules may be integrated into one module. The above-mentioned integrated modules may be implemented in the form of hardware or in the form of software functional modules.
[0065] In another embodiment of the present invention, a computer device is provided, the computer device includes a processor and a memory, the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding function; the processor described in the embodiment of the present invention can be used for the operation of the distribution network tree distribution intelligent patrol method based on high-resolution remote sensing images.
[0066] In another embodiment of the present invention, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in the computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the intelligent patrol method for distribution network tree distribution based on high-resolution remote sensing images in the above embodiment.
[0067] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0068] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0069] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0070] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.
Claims
1. A distribution network tree distribution intelligent inspection method based on high-resolution remote sensing images, characterized in that: include: Collect satellite remote sensing images for patrolled distribution network lines, and pre-process the satellite remote sensing images to obtain fused images; Construct a distribution line tree distribution sample library based on fused images; Construct and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The image to be identified is input into the trained intelligent identification model to obtain the intelligent identification result of tree distribution.
2. The method for intelligent inspection of tree distribution in distribution network based on high-resolution remote sensing images according to claim 1 is characterized in that: The collection of satellite remote sensing images for the distribution network lines to be inspected includes: For the distribution network lines to be inspected, determine the resolution, band, width, and cloud cover parameters of the satellite remote sensing image, and then collect the satellite remote sensing image at the specified time.
3. The intelligent inspection method for tree distribution in distribution network based on high-resolution remote sensing images according to claim 1 is characterized in that: The preprocessing of the satellite remote sensing image to obtain the fused image includes: The received satellite remote sensing data is preprocessed by radiation correction, geometric correction, and orthorectification to obtain a satellite remote sensing preprocessed image with geographic coordinates, information enhancement, and clarity. The panchromatic and multispectral images are fused to obtain a fused image with both high resolution and multispectral features.
4. The method for intelligent inspection of tree distribution in distribution network based on high-resolution remote sensing images according to claim 1 is characterized in that: The method of constructing a distribution line tree distribution sample library based on fused images includes: The fused image is annotated with tree samples by semi-automatic annotation combined with manual verification. The samples are uniformly stored according to the determined sample library organization form to build a sample library for intelligent identification of distribution network tree distribution. Manually annotate seed samples, visually interpret trees based on their color spectrum, shape structure, texture, and shadow features, distinguish trees from other vegetation and background objects, and manually identify and annotate the distribution information of tree objects; use manually annotated seed samples as training data, use deep learning models to automatically predict the labels of the remaining samples, and finally manually check and correct the prediction results; A structured database is established to crop and encode samples and original images, and samples and images are cropped into image patches of fixed size. The sample patches and image patches are made to correspond one to one through unified and standardized naming.
5. The intelligent inspection method for tree distribution in distribution network based on high-resolution remote sensing images according to claim 1 is characterized in that: The intelligent recognition model is constructed and trained based on the distribution line tree distribution sample library to obtain a trained intelligent recognition model, including: Semantic segmentation is used for intelligent identification of tree distribution on distribution lines, and an improved U-Net model based on high-resolution detail fusion is constructed and trained for intelligent identification of tree distribution on distribution networks.
6. The intelligent inspection method for tree distribution in distribution network based on high-resolution remote sensing images according to claim 5 is characterized in that: Based on the fully convolutional network U-Net, the skip connection structure of the U-Net network is improved according to the principle of maintaining high-resolution information throughout the full convolutional network. Different weights are assigned to feature maps at different levels. The weights are obtained through learning, so that the network can adaptively select the most important features. Deep supervision is adopted when setting the loss function. For the tree extraction task, a weighted sum of segmentation loss functions such as cross entropy loss, Dice loss, or IoU loss is used.
7. The method for intelligent inspection of tree distribution in distribution network based on high-resolution remote sensing images according to claim 1 is characterized in that: The image to be identified is input into the trained intelligent identification model to obtain the tree distribution intelligent identification result, including: The image to be identified is input into the trained model, and prediction is performed based on the dual receptive field seamless sliding window to obtain the intelligent recognition result of tree distribution.
8. The intelligent inspection system for tree distribution in distribution network based on high-resolution remote sensing images is characterized by: include: The data acquisition module is used to collect satellite remote sensing images of the distribution network lines to be inspected, and pre-process the satellite remote sensing images to obtain fused images; A sample library construction module is used to construct a distribution line tree distribution sample library based on fused images; A model training module is used to build and train an intelligent recognition model based on a distribution line tree distribution sample library to obtain a trained intelligent recognition model; The prediction output module is used to input the image to be identified into the trained intelligent recognition model to obtain the intelligent recognition result of tree distribution.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method for intelligent inspection of tree distribution in distribution networks based on high-resolution remote sensing images as described in any one of claims 1 to 7 are implemented.
10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, the steps of the method for intelligent inspection of tree distribution in distribution networks based on high-resolution remote sensing images as described in any one of claims 1 to 7 are implemented.