Power grid hidden danger identification method and system

Abstract

The invention discloses a power grid hidden danger recognition method and system, and the method comprises the steps: detecting the geographic position range of a cloud layer shielding region through a multispectral satellite image of a target disaster region, screening out a plurality of schedulable unmanned aerial vehicle nests from a candidate unmanned aerial vehicle nest set according to the geographic position range of the cloud layer shielding region, and carrying out the recognition of the hidden danger of a power grid. And according to the geographic position range of the cloud layer shielding area, screening out an idle unmanned aerial vehicle closest to the cloud layer shielding area as an inspection unmanned aerial vehicle, generating a flight path of the inspection unmanned aerial vehicle covering the cloud layer shielding area, and collecting an image of the cloud layer shielding area along the flight path through the inspection unmanned aerial vehicle. Image splicing fusion is carried out on the image of the cloud layer shielding area and the image of the satellite non-shielding area, a suspected power grid hidden danger area is screened out from the spliced and fused images, and the position of the actual power grid hidden danger area is recognized by combining real-time operation data of the suspected power grid hidden danger area acquired by a ground end; therefore, the accuracy and efficiency of hidden danger identification are improved.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method and system for identifying potential power grid hazards. Background Technology

[0002] The power grid is a core infrastructure that ensures the lives of residents and industrial production. Its safe and stable operation is directly related to the normal order of society. Especially after sudden disasters such as typhoons, floods, and geological disasters, it is necessary to quickly ascertain the extent of the damage to the power grid in order to rapidly allocate repair personnel and materials and shorten the power outage time. In daily operation and maintenance, potential hazards caused by external forces around the power grid corridor, such as "tree obstacles" formed by trees too close to the lines, unauthorized illegal buildings, easily blown objects onto the lines, and geological disasters that may cause towers to tilt, also need to be identified and dealt with in a timely manner. Otherwise, it may easily cause line tripping and affect the safety of power supply.

[0003] Currently, existing methods for power grid inspection and emergency disaster relief consistently fall short of the requirements for wide-area coverage, rapid response, and accurate identification. On the one hand, disaster relief work faces difficulties during disasters, and routine monitoring also struggles to achieve comprehensive coverage. For example, after a typhoon makes landfall, large areas of the ground are flooded, roads are washed away, and manual inspections are simply impossible. While drone or helicopter inspections offer some flexibility, they cannot take off in strong winds and heavy rain. Even when the wind and rain subside to a safe operating period, drone inspections can only follow fixed routes, and quickly assessing the damage to all poles and substations in a given area still requires a significant amount of time. During routine monitoring, although cameras are installed in key areas, allowing drone inspections to check along the power lines, power grid lines typically span tens or even hundreds of kilometers, covering a wide and dispersed area. Existing methods are insufficient to meet monitoring needs, and many safety hazards such as tree obstructions and illegal structures are only discovered when they pose a serious threat to line safety.

[0004] On the other hand, methods with large-scale monitoring capabilities have slow response times. Satellite remote sensing is currently the only method capable of monitoring large areas simultaneously. Regardless of wind strength or rainfall, satellites can capture ground conditions, quickly identifying areas where tower collapses or water accumulation may occur. However, accessing satellite resources and acquiring satellite imagery data takes a considerable amount of time, and satellite images are easily obscured by clouds. Even when relevant data is obtained, it is difficult to quickly coordinate with aerial and manual inspections to form effective disaster assessment results. This results in an excessively long time interval between the availability of operational conditions and the acquisition of usable disaster assessment results after a disaster, hindering rapid decision-making for emergency power restoration. Manual inspections, machine inspections, and cameras can achieve rapid and accurate inspections under normal conditions or during the safety window after a disaster, but their coverage area is relatively small. Satellites can overcome the limitations of wind and rain to achieve wide coverage, but their response speed is slow and detailed information is blurred. As a result, during the safety window after a disaster, it is difficult to simultaneously meet the needs of wide coverage and rapid and accurate response, resulting in low disaster assessment efficiency. Ultimately, this leads to slow progress in power restoration work after disasters such as typhoons, which adversely affects the lives of residents. Summary of the Invention

[0005] In view of this, in order to solve the above-mentioned technical problems, the present invention provides a method and system for identifying potential power grid hazards.

[0006] The first aspect of this invention provides a method for identifying potential power grid hazards, comprising:

[0007] Acquire multispectral satellite imagery of the target disaster area, detect cloud-covered areas and boundary pixel coordinates of the cloud-covered areas based on the multispectral satellite imagery, and determine the geographical location range of the cloud-covered areas based on the boundary pixel coordinates of the cloud-covered areas.

[0008] Obtain the terrain and weather environment parameters of the cloud-covered area, construct environmental constraints, and based on the geographical location range of the cloud-covered area and the environmental constraints, select multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list.

[0009] Based on the geographical location of the cloud-covered area, the idle drones closest to the cloud-covered area are selected as inspection drones from the list of available drone nests. Combined with the preset compliant resolution parameters of the camera in the inspection drone, the flight trajectory of the inspection drone covering the cloud-covered area is generated.

[0010] The inspection drone collects images of the cloud-obscured area along the flight path and transmits the images of the cloud-obscured area back to the ground.

[0011] Acquire satellite images of the unobstructed area of ​​the target disaster area, and stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain a stitched and fused image;

[0012] By using a preset power grid hazard feature template, suspected power grid hazard areas are selected from the stitched and fused images. Combined with the real-time operation data of the suspected power grid hazard areas obtained from the ground terminal, the actual location of the power grid hazard areas is identified.

[0013] Preferably, the step of acquiring multispectral satellite imagery of the target disaster area, detecting cloud-covered areas and boundary pixel coordinates of the cloud-covered areas based on the multispectral satellite imagery, and determining the geographical location range of the cloud-covered areas based on the boundary pixel coordinates of the cloud-covered areas, includes:

[0014] Acquire multispectral satellite images of the target disaster area and preprocess the multispectral satellite images;

[0015] The band reflectance values ​​of each pixel in the preprocessed multispectral satellite image are obtained, and the pixels with band reflectance values ​​greater than a preset reflectance threshold are marked as potential cloud pixels. Multiple potential cloud pixels are then merged into a cloud pixel marker map.

[0016] Morphological dilation and erosion operations are performed on adjacent potential cloud pixels in the cloud pixel marker map to obtain the initial cloud occlusion region.

[0017] Extract the boundary pixel coordinates of each point on the boundary of the initial cloud-covered area, and perform coordinate transformation on the multiple boundary pixel coordinates to obtain the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area.

[0018] Based on the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area, the minimum bounding rectangle of the initial cloud-covered area is determined, and the geographical location range of the cloud-covered area is determined based on the minimum bounding rectangle.

[0019] Preferably, the step involves obtaining the terrain and weather environmental parameters of the cloud-covered area, constructing environmental constraints, and, based on the geographical location of the cloud-covered area and the environmental constraints, selecting multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list, including:

[0020] Based on the geographical location of the cloud-covered area, various terrain and weather environmental parameters within the cloud-covered area are determined, and environmental constraints are determined based on these various terrain and weather environmental parameters.

[0021] Obtain the position coordinates of each first candidate drone nest in the candidate drone nest set that satisfies the environmental constraints, and for each first candidate drone nest, determine the Euclidean distance from the position coordinates of the first candidate drone nest to the center of the cloud-covered area.

[0022] If the Euclidean distance is less than a preset distance threshold, the first candidate drone nest is determined to be the second candidate drone nest, and the second candidate drone nest is added to the initial available nest list.

[0023] Select second candidate drone nests from the initial list of available nests that have sufficient range for the round-trip flight distance relative to the cloud cover area; these will be the schedulable drone nests.

[0024] The multiple schedulable UAV nests are sorted according to their Euclidean distance to generate the list of available nests.

[0025] Preferably, the method further includes:

[0026] Obtain the current resolution parameters of the satellite's unobstructed area image, and determine whether the current resolution parameters meet the preset compliance resolution conditions;

[0027] If it is determined that the current resolution parameter meets the preset compliance resolution condition, then the current resolution parameter is used as the compliance resolution parameter of the drone's camera.

[0028] If it is determined that the current resolution parameter does not meet the preset compliance resolution condition, then the backup resolution parameter that meets the preset compliance resolution condition from the preset backup resolution parameter list is used as the compliance resolution parameter of the drone's camera.

[0029] Preferably, the step of selecting the nearest idle drone to the cloud-covered area as the inspection drone based on the available drone nest list according to the geographical location range of the cloud-covered area, and generating the flight trajectory of the inspection drone covering the cloud-covered area by combining the preset compliance resolution parameters of the camera in the inspection drone, includes:

[0030] Based on the geographical location of the cloud-covered area, the available drones closest to the cloud-covered area are selected as inspection drones from the list of available drone nests.

[0031] Based on the preset compliant resolution parameters, focal length, and sensor size of the camera in the inspection drone, the image coverage area of ​​a single image captured by the inspection drone is determined.

[0032] Based on the image coverage area, the distribution of shooting points covering the cloud-covered area is determined by combining the preset shooting point intervals.

[0033] Based on the distribution of the shooting points, a gridded waypoint coordinate matrix covering the cloud-obscured area is determined;

[0034] Based on the gridded waypoint coordinate matrix, the positions of each shooting point are connected in a preset flight order to form a flight path sequence. The flight altitude of each shooting point is adjusted according to environmental constraints to generate a flight trajectory that includes the flight path sequence and flight altitude.

[0035] Preferably, the method further includes:

[0036] Based on the flight path sequence, feature points within the image coverage area of ​​each shooting point are extracted, and the number of feature points per unit area within the image coverage area is counted to obtain the feature point density value of the image coverage area.

[0037] For each image coverage area, the feature point density value is compared with a preset density threshold.

[0038] If the feature point density value of the image coverage area is lower than the preset density threshold, a new shooting point is added at the midpoint of the adjacent flight path in the image coverage area, and the flight height of the new shooting point is interpolated based on the flight height of the two adjacent shooting points to obtain the flight height of the new shooting point.

[0039] Based on the new shooting point and its flight altitude, the flight path sequence is updated to obtain a new flight trajectory.

[0040] Preferably, the step of collecting images of the cloud-obscured area along the flight path by the inspection drone and transmitting the images of the cloud-obscured area back to the ground includes:

[0041] The inspection drone collects images of the cloud-obscured area along the flight path and transmits the images of the cloud-obscured area back to the ground via a wireless transmission link.

[0042] The image of the cloud-obscured area received by the ground terminal is preprocessed, and the preprocessed image is sorted in time sequence according to the shooting timestamp to obtain the image time sequence.

[0043] The location information of each image in the image time sequence is obtained, and the projected area of ​​the region covered by all images in the image time sequence is determined based on the location information of each image.

[0044] The coverage rate is obtained by calculating the ratio of the projected area to the area of ​​the cloud-covered region.

[0045] If the coverage rate is greater than or equal to a preset coverage rate threshold, it is marked as complete coverage, and the image of the cloud-covered area is output as a complete image.

[0046] If the coverage rate is less than a preset coverage rate threshold, then the number of shooting points in the flight trajectory is increased, the flight trajectory is updated, and based on the updated flight trajectory, the process of collecting images of the cloud-obscured area by the inspection drone along the flight trajectory is repeated, and the images of the cloud-obscured area are transmitted back to the ground, until the coverage rate is greater than or equal to the preset coverage rate threshold, and the latest image of the cloud-obscured area is output as a complete image.

[0047] Preferably, the step of acquiring satellite images of the unobstructed area of ​​the target disaster area, and then stitching and fusing the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain a stitched and fused image includes:

[0048] Based on the multispectral satellite imagery and the cloud-covered areas, determine the unobstructed satellite imagery of the target disaster area;

[0049] Feature points are extracted from the satellite unobstructed area image and the cloud-obstructed area image, and matching feature point pairs between the satellite unobstructed area image and the cloud-obstructed area image are determined by a feature point matching algorithm;

[0050] Based on the matching feature point pairs, the transformation parameters between the satellite unobstructed area image and the cloud-obstructed area image are calculated. The transformation parameters are used to describe the geometric transformation relationship between the satellite unobstructed area image and the cloud-obstructed area image.

[0051] Geometric transformation is performed on the image of the cloud-covered area using transformation parameters to align the image of the cloud-covered area with the image of the satellite uncovered area in spatial position. The aligned image of the cloud-covered area and the image of the satellite uncovered area are then stitched together to obtain an initial stitched and fused image.

[0052] The stitching seams of the initially stitched and fused image are eliminated, and the pixel resolution of the initially stitched and fused image is unified to obtain the stitched and fused image.

[0053] Preferably, the step of filtering suspected power grid hazard areas from the stitched and fused image using a preset power grid hazard feature template, and identifying the actual location of the power grid hazard area by combining the real-time operational data of the suspected power grid hazard area obtained from the ground terminal, includes:

[0054] The stitched and fused image is divided into multiple sub-image regions under a sliding window.

[0055] For each of the sub-image regions, determine the similarity between the sub-image region and the preset power grid hidden danger feature template;

[0056] If the similarity exceeds a preset similarity threshold, the sub-image region is designated as a suspected power grid hazard region, and the location coordinates of the suspected power grid hazard region are extracted.

[0057] Based on the location coordinates of the suspected power grid hazard area, real-time operational data of the suspected power grid hazard area is obtained from the ground terminal;

[0058] The real-time operating data is compared with a preset normal operating data range. If the real-time operating data does not meet the preset normal operating data range, the location coordinates of the suspected power grid hazard area are determined to be the location of the actual power grid hazard area.

[0059] Secondly, the present invention also provides a power grid hazard identification system, comprising:

[0060] The obscured area determination module is used to acquire multispectral satellite images of the target disaster area, detect cloud-obscured areas and boundary pixel coordinates of the cloud-obscured areas based on the multispectral satellite images, and determine the geographical location range of the cloud-obscured areas based on the boundary pixel coordinates of the cloud-obscured areas.

[0061] The available drone nest determination module is used to obtain the terrain and weather environment parameters of the cloud-covered area, construct environmental constraints, and select multiple schedulable drone nests from the candidate drone nest set based on the geographical location range of the cloud-covered area and the environmental constraints, thereby obtaining an available drone nest list.

[0062] The flight trajectory generation module is used to select the idle drone closest to the cloud-covered area as the inspection drone based on the geographical location range of the cloud-covered area through the list of available drone nests, and generate the flight trajectory of the inspection drone covering the cloud-covered area by combining the preset compliance resolution parameters of the camera in the inspection drone.

[0063] The image acquisition and transmission module is used to acquire images of the cloud-obscured area along the flight path of the inspection drone, and transmit the images of the cloud-obscured area back to the ground.

[0064] The image stitching and fusion module is used to acquire satellite images of the unobstructed area of ​​the target disaster area, and to stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain the stitched and fused image.

[0065] The hidden danger area identification module is used to filter out suspected power grid hidden danger areas from the stitched and fused images using a preset power grid hidden danger feature template, and to identify the actual location of the power grid hidden danger area by combining the real-time operation data of the suspected power grid hidden danger area obtained by the ground terminal.

[0066] As can be seen from the above technical solution, this invention detects the geographical location of the cloud-covered area using multispectral satellite imagery of the target disaster area. Based on the geographical location of the cloud-covered area and environmental constraints, it selects multiple schedulable drone nests from a candidate drone nest set. Then, based on the geographical location of the cloud-covered area, it selects the idle drone closest to the cloud-covered area as the inspection drone, generates a flight path for the inspection drone covering the cloud-covered area, collects images of the cloud-covered area along the flight path, and transmits the images back to the ground. Finally, it analyzes the cloud-covered area... Imagery of the affected area and unobstructed satellite images are stitched together to ensure the integrity of the satellite imagery. Then, using a pre-defined power grid hazard feature template, suspected power grid hazard areas are selected from the stitched and fused imagery. Combined with real-time operational data from ground-based sources, the actual locations of these hazard areas are identified. This multi-source data fusion ensures complete imagery of the target disaster area and achieves efficient and accurate identification of power grid hazard areas. This not only overcomes the limitations of traditional satellite remote sensing technology under cloud cover but also leverages the flexibility and real-time nature of UAV inspections to acquire detailed imagery data of cloud-covered areas. Furthermore, the combination of real-time operational data from ground-based sources further verifies and confirms suspected hazard areas, significantly improving the accuracy and efficiency of hazard identification. Attached Figure Description

[0067] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0068] Figure 1 This is an application environment diagram of a power grid hazard identification method provided in an embodiment of the present invention;

[0069] Figure 2A flowchart of a power grid hazard identification method provided in an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of a power grid hazard identification system provided in an embodiment of the present invention. Detailed Implementation

[0071] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0072] The power grid hazard identification method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 101 communicates with server 102 via a network. A data storage system can store the data that server 102 needs to process. The data storage system can be integrated onto server 102 or placed on the cloud or other network servers. Terminal 101 or server 102 acquires multispectral satellite imagery of the target disaster area. Based on the multispectral satellite imagery, it detects cloud-covered areas and their boundary pixel coordinates. Based on the boundary pixel coordinates of the cloud-covered areas, it determines the geographical location range of the cloud-covered areas. It acquires terrain and weather environmental parameters of the cloud-covered areas, constructs environmental constraints, and, based on the geographical location range of the cloud-covered areas and the environmental constraints, selects multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list. Based on the geographical location range of the cloud-covered areas, it selects the idle drone closest to the cloud-covered areas from the available nest list. To inspect the drone, and based on the preset compliant resolution parameters of the camera in the drone, a flight path is generated to cover the cloud-obscured area. The drone then collects images of the cloud-obscured area along the flight path and transmits these images back to the ground. Satellite images of the unobscured area of ​​the target disaster area are acquired, and the cloud-obscured and satellite images are stitched together to obtain a fused image. Using a preset power grid hazard feature template, suspected power grid hazard areas are selected from the fused image. Combined with real-time operational data of the suspected power grid hazard areas obtained from the ground, the actual location of the power grid hazard area is identified.

[0073] Terminal 101 can be, but is not limited to, various personal computers, laptops, smartphones, and tablets.

[0074] Server 102 can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides cloud computing services.

[0075] like Figure 2 As shown, this application provides a method for identifying potential power grid hazards, which is applied to... Figure 1 Taking terminal 101 or server 102 as an example, the explanation includes the following steps S1 to S6. Wherein:

[0076] Step S1: Obtain multispectral satellite imagery of the target disaster area. Detect cloud-covered areas and their boundary pixel coordinates based on the multispectral satellite imagery. Determine the geographical location range of the cloud-covered areas based on the boundary pixel coordinates of the cloud-covered areas.

[0077] Multispectral satellite imagery is acquired by capturing images of the target disaster area using multispectral sensors mounted on satellites. This type of imagery can capture light information in different wavelengths, thus reflecting various characteristics of ground features. By analyzing and processing multispectral satellite imagery, cloud-covered areas can be accurately detected, and their boundary pixel coordinates can be determined. Based on these boundary pixel coordinates, the geographical location range of the cloud-covered area can be further calculated, providing basic data for subsequent drone inspections and hazard identification.

[0078] Step S2: Obtain the terrain and weather environmental parameters of the cloud-covered area, construct environmental constraints, and based on the geographical location of the cloud-covered area and the environmental constraints, select multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list.

[0079] Among these, terrain and weather environmental parameters include, but are not limited to, altitude, terrain undulation, wind speed, wind direction, temperature, humidity, and visibility. These parameters have a significant impact on the flight safety and mission execution efficiency of UAVs. By collecting and analyzing these parameters, flight constraints for UAVs can be constructed to adapt to different environmental conditions. Combining the geographical range of cloud-covered areas, multiple schedulable UAV nests that meet the environmental constraints are selected from a pre-defined set of candidate UAV nests, forming a list of available nests. This ensures that UAVs can perform inspection tasks under safe and feasible conditions, improving mission success rate and efficiency.

[0080] Each drone nest is equipped with a drone capable of taking pictures.

[0081] Step S3: Based on the geographical range of the cloud-covered area, select the idle drone closest to the cloud-covered area from the list of available drone nests as the inspection drone, and generate the flight trajectory of the inspection drone covering the cloud-covered area by combining the preset compliant resolution parameters of the camera in the inspection drone.

[0082] After filtering out the list of available drone nests, the distance between each available nest and the cloud-covered area is calculated, and the closest idle drone is selected to ensure that the drone can reach the target area in the shortest time and with the least energy consumption, thereby improving inspection efficiency.

[0083] Subsequently, by combining the preset compliant resolution parameters of the camera mounted on the inspection drone, a flight trajectory covering the entire cloud-obscured area is generated. The compliant resolution parameters ensure that the acquired image data meets the accuracy requirements of subsequent image processing and analysis, thereby guaranteeing the accuracy of hazard identification. The generation of the flight trajectory comprehensively considers the shape and size of the cloud-obscured area, as well as the drone's flight performance, ensuring that the drone can complete the image acquisition task comprehensively and efficiently.

[0084] Step S4: Collect images of the cloud-obscured area along the flight path using the inspection drone, and transmit the images of the cloud-obscured area back to the ground.

[0085] The inspection drone flies along a pre-generated flight path over areas obscured by clouds, using its onboard camera to capture images of the target area. During the acquisition process, the drone transmits the captured image data back to the ground in real time for subsequent image processing and analysis.

[0086] Step S5: Obtain satellite images of the unobstructed area of ​​the target disaster area, and stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain the stitched and fused image.

[0087] Among them, the satellite unobstructed area imagery is obtained by taking pictures of the unobstructed parts of the disaster-stricken area by satellite. After acquiring the satellite unobstructed area imagery, advanced image processing algorithms are used to precisely stitch and fuse the cloud-obstructed area imagery collected by the inspection drone with the satellite unobstructed area imagery, making the stitched imagery more visually continuous and consistent.

[0088] The stitched and fused image not only contains comprehensive information on unobstructed satellite areas but also supplements detailed data on cloud-obstructed areas, providing a more complete and accurate image foundation for subsequent hazard identification. This image stitching and fusion technology effectively overcomes the limitations of cloud obstruction on satellite remote sensing imagery, improving image utilization and identification accuracy, thereby more accurately identifying potential power grid hazards.

[0089] Step S6: Using the preset power grid hazard feature template, select suspected power grid hazard areas from the stitched and fused images, and combine the real-time operation data of the suspected power grid hazard areas obtained from the ground to identify the actual location of the power grid hazard areas.

[0090] The preset power grid hazard feature template is built based on the learning and analysis of a large amount of power grid hazard image data. The template covers various features that power grid equipment may exhibit when hazards occur, such as abnormal equipment shape, color change, and texture change.

[0091] By comparing the stitched and merged image with a preset template of potential power grid hazards, areas with high similarity to the template features can be quickly and accurately identified; these areas are then designated as suspected power grid hazard areas. After identifying these areas, real-time operational data is acquired from the ground station. This data is then used for further analysis and verification. Real-time operational data includes, but is not limited to, key parameters of power grid equipment such as current, voltage, and temperature, which directly reflect the operating status of the equipment. By comparing the real-time operational data with a preset range of normal operating data, it can be determined whether an anomaly actually exists in the suspected power grid hazard area. If the real-time operational data exceeds the range of normal operating data, the suspected power grid hazard area is determined to be an actual power grid hazard area.

[0092] It should be noted that, in this embodiment, multispectral satellite imagery of the target disaster area is used to detect the geographical location of the cloud-covered area. Based on the geographical location of the cloud-covered area and environmental constraints, multiple schedulable drone nests are selected from a set of candidate drone nests. Furthermore, based on the geographical location of the cloud-covered area, the idle drone closest to the cloud-covered area is selected as the inspection drone. A flight path is generated for the inspection drone covering the cloud-covered area. The inspection drone collects images of the cloud-covered area along the flight path and transmits these images back to the ground. Finally, the cloud-covered area... Image stitching and fusion are performed on satellite images and unobstructed satellite images to ensure the integrity of the satellite imagery. Then, using a pre-set power grid hazard feature template, suspected power grid hazard areas are selected from the stitched and fused images. Combined with real-time operational data of these suspected hazard areas acquired from the ground, the actual locations of the hazard areas are identified. This multi-source data fusion ensures complete imagery of the target disaster area and achieves efficient and accurate identification of power grid hazard areas. This not only overcomes the limitations of traditional satellite remote sensing technology under cloud cover but also leverages the flexibility and real-time nature of UAV inspections to acquire detailed image data of cloud-covered areas. Furthermore, the combination of real-time operational data from the ground further verifies and confirms suspected hazard areas, significantly improving the accuracy and efficiency of hazard identification.

[0093] In some embodiments, multispectral satellite imagery of the target disaster area is acquired; cloud-covered areas and boundary pixel coordinates of the cloud-covered areas are detected based on the multispectral satellite imagery; and the geographical location range of the cloud-covered areas is determined based on the boundary pixel coordinates of the cloud-covered areas, including:

[0094] Step S101: Acquire multispectral satellite images of the target disaster area and preprocess the multispectral satellite images.

[0095] The preprocessing includes removing salt-and-pepper noise using median filtering, eliminating random noise interference using Gaussian filtering, and histogram equalization processing, which enhances image contrast by redistributing pixel gray values, making the distinction between cloud and surface features more obvious.

[0096] Step S102: Obtain the band reflectance value of each pixel in the preprocessed multispectral satellite image, and mark the pixels with band reflectance values ​​greater than the preset reflectance threshold as potential cloud pixels, and merge multiple potential cloud pixels into a cloud pixel marker map.

[0097] Cloud detection is based on the difference in reflectivity of different ground features across visible and infrared bands. Clouds exhibit high reflectivity in the visible light band, typically exceeding 0.6, while their radiance is significantly lower than that of the ground surface in the thermal infrared band due to the lower temperature at the cloud top. The ratio of reflectivity of each pixel in the visible light band to that in the infrared band is calculated; pixels exceeding a preset threshold are identified as potential cloud layers.

[0098] Specifically, by setting a reasonable preset reflectivity threshold, pixels with reflectivity values ​​greater than this threshold are marked as potential cloud pixels. These potential cloud pixels are then arranged according to their respective positions and fused together to form a cloud pixel marker map, enabling accurate identification of cloud-occupied areas in subsequent operations.

[0099] Wherein, the band reflectivity value = (DN value × radiometric calibration coefficient - radiometric calibration offset) / solar irradiance.

[0100] Wherein, DN value is the original grayscale value of the satellite image, the radiometric calibration factor of the visible light band of Gaofen-6 is taken as 0.008, the offset is taken as -10, and the solar irradiance is taken as 1367W / m². 2 Set the threshold for summer typhoon cloud systems to 2.0 and the threshold for thin upper-level clouds in winter to 2.5.

[0101] Step S103: Perform morphological dilation and erosion operations on adjacent potential cloud pixels in the cloud pixel marker map to obtain the initial cloud occlusion area.

[0102] In this process, morphological dilation operations (using 3×3 or 5×5 structuring elements) are employed to fill small holes inside the cloud layer through local maximum operations, connect adjacent but discontinuous cloud layer pixels, and expand the range of the cloud layer area to compensate for the possible loss of cloud layer area due to inaccurate pixel labeling.

[0103] Subsequently, morphological erosion operations are used to refine the expanded cloud region, removing some noise and isolated pixels caused by the expansion operation, thereby obtaining a more accurate initial cloud occlusion region.

[0104] Step S104: Extract the boundary pixel coordinates of each point on the boundary of the initial cloud-covered area, perform coordinate transformation on the multiple boundary pixel coordinates, and obtain the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area.

[0105] The coordinate transformation is based on georeferenced information from satellite imagery. By converting pixel coordinates to geographic coordinates (latitude and longitude values), the actual location of the initial cloud-occupied area on the Earth's surface can be accurately determined. Specifically, the coordinate transformation matrix is ​​established based on the geocoding information of the satellite imagery, including the latitude and longitude coordinates of the top-left corner of the image, pixel resolution, and projection parameters. For each boundary pixel, the pixel row and column numbers are converted to projected coordinates using an affine transformation formula, and then inverse projection is performed according to the projection type to obtain the geographic coordinates.

[0106] For example, based on the spatial resolution and projection coordinate system information in the satellite image metadata, a transformation matrix between pixel coordinates and geographic coordinates is established. The affine transformation formula is x'=a1x+b1y+c1, y'=a2x+b2y+c2, where x and y are pixel coordinates, and x' and y' are projection coordinates. Specifically, a1=0.000278, b1=0, c1=118.3333, a2=0, b2=-0.000278, and c2=31.3333. The pixel coordinates of each point on the boundary of the cloud-obscured area are then extracted. The boundary pixel coordinates are transformed using a transformation matrix. Based on the solar altitude angle and azimuth angle parameters during satellite imaging, where the solar altitude angle θ = 60°, the azimuth angle α = 120°, the cloud height H = 2000m, the ground projection offset Δx = H × tan(90° - θ) = 2000 × tan30° ≈ 1154.7m, and Δy = H × tan(α - 90°) = 2000 × tan30° ≈ 1154.7m, the corrected boundary pixel ground coordinates = transformed coordinates + offset. The boundary pixel ground coordinates corresponding to each boundary pixel are calculated to obtain the latitude and longitude values ​​of the boundary points.

[0107] Step S105: Determine the minimum bounding rectangle of the initial cloud-covered area based on the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area, and determine the geographical location range of the cloud-covered area based on the minimum bounding rectangle.

[0108] The algorithm employs a minimum bounding rectangle algorithm to process the latitude and longitude values ​​of points on the boundary of the initial cloud-covered area. By calculating the minimum bounding rectangle of the polygon formed by these latitude and longitude values, the geographical location of the cloud-covered area is accurately defined. This algorithm ensures that the determined geographical location includes the entire cloud-covered area without excessive expansion, thus providing precise target area positioning for subsequent UAV inspection tasks. Specifically, by traversing all latitude and longitude points on the boundary, the minimum and maximum longitude, minimum and maximum latitude values ​​are found. The rectangle formed by these four values ​​is the minimum bounding rectangle, and the coordinates of its four vertices represent the geographical boundary of the cloud-covered area. This method of determining the geographical location has high accuracy and practicality, fully utilizing the advantages of geographic information from satellite imagery.

[0109] In some embodiments, terrain and weather environmental parameters of the cloud-covered area are obtained, environmental constraints are constructed, and multiple schedulable drone nests are selected from the candidate drone nest set based on the geographical location range of the cloud-covered area and the environmental constraints, resulting in a list of available drone nests, including:

[0110] Step S201: Based on the geographical location of the cloud-covered area, determine various terrain and weather environmental parameters within the cloud-covered area, and determine environmental constraints based on these parameters.

[0111] Based on the geographical location of the cloud-covered area, various terrain and weather environmental parameters, including elevation, slope, and wind speed, are retrieved from the ground.

[0112] Environmental constraints are determined by various terrain and weather parameters. For example, a maximum altitude threshold is set for drones to fly. When the altitude of a certain area within the cloud cover exceeds this threshold, drone flight will be restricted in that area. The slope range for safe take-off, landing and flight of drones is set based on the slope. If the slope of a certain area exceeds this range, the area is not suitable for drone operation. The maximum wind speed that drones can withstand is set based on wind speed data. When the wind speed exceeds this value, the flight safety of drones in that area will be threatened.

[0113] Step S202: Obtain the position coordinates of each first candidate UAV nest in the candidate UAV nest set that meets the environmental constraints, and for each first candidate UAV nest, determine the Euclidean distance from the position coordinates of the first candidate UAV nest to the center of the cloud-covered area.

[0114] The candidate drone nest set is a collection of all available drone nests covering the target disaster area. These nests may be located in different geographical locations and have different takeoff and landing conditions. When selecting the first candidate drone nests, the impact of environmental constraints on drone flight needs to be considered. From the candidate drone nest set, drone nests that meet the environmental constraints are selected. The elevation, slope, wind speed, and other parameters of these nests must meet the requirements for safe drone flight.

[0115] Then, for each first-candidate drone nest that meets the conditions, the Euclidean distance from its location coordinates to the center of the cloud-covered area is calculated. Euclidean distance is a method for measuring the straight-line distance between two points. By calculating it, the relative distance from each first-candidate drone nest to the target area can be obtained, which can quantify the relative positional relationship between each first-candidate drone nest and the cloud-covered area, providing a basis for subsequent screening of schedulable drone nests.

[0116] Step S203: If the Euclidean distance is less than the preset distance threshold, then the first candidate drone nest is determined as the second candidate drone nest, and the second candidate drone nest is added to the initial available nest list.

[0117] If the Euclidean distance is not less than a preset distance threshold, the first candidate drone nest is relatively far from the cloud cover area and is unsuitable as a scheduling nest for this inspection mission; therefore, it is not added to the initial list of available nests. Conversely, if the Euclidean distance is less than the preset distance threshold, it indicates that the first candidate drone nest is close to the cloud cover area and can reach the target area in a shorter time, meeting the scheduling requirements. Therefore, it is identified as the second candidate drone nest and added to the initial list of available nests. The preset distance threshold is set based on the actual needs of the inspection mission.

[0118] Step S204: Select the second candidate drone nest from the initial list of available nests. The second candidate drone nest has a range that meets the round-trip flight distance relative to the cloud cover area. The second candidate drone nest is the one that can be scheduled.

[0119] In one example, by combining standard drone range parameters such as 30,000m, candidate drone nests with a round-trip flight distance of less than 80% of the range are selected as schedulable drone nests. The round-trip flight distance is the sum of the straight-line distances calculated based on the position coordinates of each second candidate drone nest in the initial list of available nests and the boundary coordinates of the cloud-covered area.

[0120] Specifically, it is necessary to first determine the point on the boundary of the cloud-covered area that is farthest from the nest of each second candidate UAV, and then multiply the straight-line distance between this point and the nest coordinates by two to obtain the theoretical round-trip flight distance.

[0121] Subsequently, dynamic screening is performed based on the actual flight range parameters of the drones: when the theoretical round-trip flight distance is less than 80% of the drone's standard flight range, the drone nest is deemed eligible for scheduling. For example, if a drone has a flight range of 30,000m, its effective operating range is a round-trip distance of 24,000m. Only when the straight-line distance between the farthest point of the cloud-covered area and the drone nest is less than 12,000m will the drone nest be included in the final list of available drone nests. Through this dual verification mechanism, both the rationality of geographical distance and the sufficiency of energy for drones to perform inspection tasks are considered, thereby constructing a set of schedulable drone nests that meet the actual operational needs.

[0122] The 80% threshold for flight range is set based on safety redundancy considerations. In actual flight, the drone's range will be lower than the nominal value due to factors such as wind speed, load, and temperature. Reserving 20% ​​of the battery capacity ensures that the drone has enough power to return to its nest in case of emergencies.

[0123] For example, a drone with a nominal range of 30 kilometers has an actual usable flight distance limited to within 24 kilometers. If the distance from the drone's nest to the center of the obstructed area is 10 kilometers, a round trip would require 20 kilometers, thus meeting the requirements for safe flight.

[0124] Step S205: Sort the multiple schedulable drone nests according to their Euclidean distance to generate a list of available nests.

[0125] When sorting the selected schedulable drone nests, an ascending Euclidean distance sorting strategy is adopted.

[0126] In practice, the Euclidean distance from each schedulable drone nest to the center of the cloud-covered area is first extracted. Then, these distance values ​​are sorted from smallest to largest using a quicksort algorithm, ultimately generating a list of available drone nests containing nest numbers, location coordinates, and distance parameters. This list is presented according to spatial proximity priority, ensuring that the nests closest to the target area are scheduled first, thereby shortening the drone response time and improving inspection efficiency.

[0127] The sorted list structure contains three columns of data: the first column is the unique identifier of the drone nest, the second column is the latitude and longitude coordinates of the nest, and the third column is the Euclidean distance from the nest to the target area, in meters. This sorting method not only conforms to the principle of optimal path planning for drone scheduling, but also provides the system with an intuitive decision-making basis in the subsequent task allocation stage. When multiple nests simultaneously meet the scheduling conditions, the system can automatically select the nearest nest to execute the task, effectively reducing drone flight energy consumption and operation time.

[0128] In some embodiments, the method further includes:

[0129] Step S21: Obtain the current resolution parameters of the satellite image of the unobstructed area, and determine whether the current resolution parameters meet the preset compliance resolution conditions.

[0130] Among them, the resolution parameter of satellite imagery in unobstructed areas is a key indicator for measuring image quality, and its accuracy directly affects the accuracy of subsequent hazard identification.

[0131] First, the current resolution value is extracted from the satellite imagery metadata. This parameter is typically expressed in meters, representing the actual ground distance corresponding to a single pixel. Then, this value is compared to a preset compliance resolution threshold. For example, in a power transmission line inspection scenario, a compliance resolution of less than 0.5 meters is required to clearly identify minor defects such as broken insulators on towers. When the current resolution is detected to be better than the threshold, the image quality is deemed acceptable; if the resolution is lower than the threshold, the image enhancement process is triggered.

[0132] Step S22: If it is determined that the current resolution parameter meets the preset compliance resolution conditions, then the current resolution parameter is used as the compliance resolution parameter of the drone's camera.

[0133] Step S23: If it is determined that the current resolution parameter does not meet the preset compliance resolution conditions, then the backup resolution parameter that meets the preset compliance resolution conditions from the preset backup resolution parameter list shall be used as the compliance resolution parameter of the drone's shooting camera.

[0134] The backup resolution parameter list is a pre-built set of resolution options based on different satellite image sources and mission requirements. This list typically contains multiple validated resolution parameters, such as gradient values ​​of 0.3 meters, 0.5 meters, and 0.8 meters, each corresponding to the optimal imaging effect in a specific scenario. When the main image resolution is insufficient, the system automatically traverses the backup list and uses a fast comparison algorithm to select the resolution parameter closest to and better than the compliance threshold. For example, in a mountainous inspection mission, if the main image resolution is only 1.2 meters, and the backup list contains two optional parameters, 0.5 meters and 0.8 meters, the system will prioritize the 0.5-meter parameter to ensure the visibility of tower details. This dynamic parameter adjustment mechanism ensures a minimum image quality while avoiding the storage and transmission burden caused by excessively pursuing high resolution.

[0135] For example, environmental constraints impose resolution requirements on multiple dimensions. Wind speed constraints primarily consider the drone's stable shooting capability under different wind speed conditions. When the wind speed exceeds 15 m / s, the drone needs to lower its flight altitude to maintain stability, correspondingly requiring higher image resolution to ensure the shooting range. Elevation constraints reflect the limitation of drone flight altitude caused by terrain undulations. For every 1000 meters increase in altitude, air density decreases by approximately 10%, reducing the drone's lift and necessitating adjustments to shooting parameters. Slope constraints affect the drone's takeoff, landing, and hovering stability; steep terrain requires finer resolution to identify safe flight paths. Based on these constraints, corresponding resolution requirement ranges are set, such as a resolution of 2-5 meters corresponding to wind speeds of 5-10 m / s, and a resolution of 5-10 meters corresponding to wind speeds of 10-15 m / s. Compliance is determined by judging whether the current satellite image resolution falls within the intersection of all constraint ranges.

[0136] When a parameter is deemed non-compliant, the backup resolution parameter list is generated based on the actual needs of power grid inspection. The list includes multiple resolution levels such as 0.5 meters, 1 meter, 2 meters, 5 meters, 10 meters, 15 meters, and 30 meters, covering different application scenarios from fine-grained inspections to large-scale monitoring.

[0137] Taking wind speed fit as an example, if the wind speed constraint requires a resolution of 5 meters and the spare parameter is 3 meters, then the absolute value of the difference is 2 meters. All differences are mapped to the interval between 0 and 1 using a linear normalization formula. Specifically, the formula is: Fit = 1 - Difference / Maximum Difference Range. The smaller the difference, the closer the fit is to 1, indicating a higher degree of matching.

[0138] In some embodiments, based on the geographical range of the cloud-covered area, an idle drone closest to the cloud-covered area is selected as the inspection drone from an available drone nest list. Combined with preset compliant resolution parameters of the camera in the inspection drone, a flight trajectory of the inspection drone covering the cloud-covered area is generated, including:

[0139] Step S301: Based on the geographical range of the cloud-covered area, select the idle drone closest to the cloud-covered area from the list of available drone nests as the inspection drone.

[0140] When selecting inspection drones, the system first extracts real-time status information for all available drone nests from the list of available nests, including key indicators such as whether the drone is idle, its battery level, and its maintenance status. A status filtering algorithm then excludes drones that are currently performing tasks, have insufficient battery power, or require maintenance, retaining only the set of available drones in standby mode.

[0141] Subsequently, the Euclidean distance from the nest of each drone in the set to the center of the cloud-covered area is calculated. A fast nearest neighbor search algorithm is used to determine the nest with the closest spatial distance, and the drones configured in that nest are selected as the main entities for this inspection task. This spatial proximity-based selection mechanism can shorten the drone response time and ensure equipment availability through the nest-drone binding relationship, providing a reliable platform for subsequent trajectory planning.

[0142] Step S302: Determine the image coverage area of ​​a single image captured by the inspection drone based on the preset compliant resolution parameters, focal length, and sensor size of the camera in the inspection drone.

[0143] The calculation of the image coverage area requires comprehensive consideration of camera parameters and flight altitude. First, the shooting altitude is calculated using the camera focal length f, sensor size (referring to the sensor diagonal length d), and compliant resolution parameter R, i.e.:

[0144] H=R×f / d

[0145] Then, calculate the camera field of view angle θ using the following formula:

[0146] θ = 2 × arctan(d / (2f)).

[0147] For example, when using a 24mm focal length camera with a 36mm×24mm full-frame sensor, the camera's field of view is approximately 84 degrees.

[0148] Then, the camera field of view θ is defined as the ground coverage width:

[0149] W = 2 × H × tan(θ / 2)

[0150] For example, at a flight altitude of 500 meters, a single image from this camera can cover an area approximately 840 meters wide. The longitudinal coverage length is calculated using the sensor's vertical dimensions and resolution, ultimately forming a rectangular coverage area. This calculation method ensures a precise match between image coverage and resolution requirements, providing fundamental data for subsequent trajectory planning.

[0151] Step S303: Determine the distribution of shooting points covering the cloud-covered area based on the image coverage area and the preset shooting point interval.

[0152] The system employs a specific algorithm to determine the number of flight paths and lateral overlap rates. These rates, typically set to 60%-80%, take into account the image stitching requirements, ensuring sufficient common areas for feature matching between adjacent images. The flight path spacing is calculated as D = W × (1 - lateral overlap rate), and the shooting point spacing is calculated as I = L × (1 - flight path overlap rate), where L is the flight path coverage length for a single UAV shot. These parameters determine the number of flight paths required to cover the entire obscured area and the number of shooting points on each flight path, resulting in a regular grid layout.

[0153] By combining the image coverage area with the number of flight paths and shooting points calculated based on the shooting point intervals, the precise location of each shooting point can be further determined.

[0154] Specifically, using a vertex in the cloud-covered area as a reference point, the starting point of each flight path is determined sequentially according to a preset flight path direction (usually due east or due north) and flight path spacing D. On each flight path, shooting points are marked at equal intervals starting from the starting point, according to the shooting point interval I, until the entire covered area is covered. This grid-like layout ensures the uniformity and comprehensiveness of the shooting point distribution, avoiding missed or repeated shots. When determining the shooting point locations, the impact of terrain undulations on the drone's flight altitude must also be considered; by adjusting the flight altitude in real time, the accuracy of the shooting point locations is ensured.

[0155] Step S304: Determine the gridded waypoint coordinate matrix of the area covered by cloud cover based on the distribution of shooting points.

[0156] The gridded waypoint coordinate matrix was constructed using a rasterization method. With the southwest corner of the obscured area as the origin, waypoints were evenly distributed in the east-west and north-south directions according to the calculated flight path spacing and shooting point intervals. Each waypoint recorded its three-dimensional coordinate information, including longitude, latitude, and flight altitude.

[0157] For example, for a cloud-covered area of ​​2 square kilometers, if a single image covers 200 meters × 150 meters and an overlap rate of 70% is set, then the flight path spacing is 60 meters and the shooting point spacing is 45 meters. Approximately 34 flight paths need to be set up, with approximately 45 shooting points on each flight path, for a total of approximately 1,530 waypoints.

[0158] Step S305: Based on the gridded waypoint coordinate matrix, connect the positions of each shooting point according to the preset flight order to form a flight path sequence. Adjust the flight altitude of each shooting point position according to environmental constraints to generate a flight trajectory that includes the flight path sequence and flight altitude.

[0159] The locations of each shooting point are called waypoints. The generation of the flight trajectory not only needs to connect all waypoints, but also needs to take into account the flight characteristics of the UAV. A "bow" shaped flight path is adopted. After the UAV flies to the end of one route, it moves laterally to the next route and then flies in the opposite direction. This method reduces the number of turns and improves flight efficiency.

[0160] At the same time, the flight altitude of each shooting point is adjusted according to environmental constraints. The flight altitude must be higher than the elevation value of the obstructed area.

[0161] In some embodiments, the method further includes:

[0162] Step S31: Based on the flight path sequence, extract feature points within the image coverage area of ​​each shooting point, count the number of feature points per unit area within the image coverage area, and obtain the feature point density value of the image coverage area.

[0163] Among them, feature points are corner points and edge points within the image coverage area, which can be identified using Harris corner detection or SIFT feature detection methods.

[0164] The unit area is usually measured in square meters. The image coverage area is divided into several small regions of equal size. Then, the number of feature points in each small region is counted. Finally, the average number of feature points in all small regions is calculated. This average value is the feature point density value of the image coverage area. The feature point density value reflects the richness of detail in the image coverage area. The richer the detail information in the area, the more effective information it can provide for tasks such as hazard identification.

[0165] Step S32: For each image coverage area, compare the feature point density value with the preset density threshold.

[0166] A density threshold is set to distinguish between sparse and dense regions. Specifically, based on the minimum requirements of the image stitching algorithm, the density threshold is set to 8 feature points per square meter; regions with a density lower than this value are marked as sparse regions.

[0167] Step S33: If the feature point density value of the image coverage area is lower than the preset density threshold, a new shooting point is added at the midpoint of the adjacent flight path in the image coverage area, and the flight height of the new shooting point is interpolated based on the flight height of the two adjacent shooting points to obtain the flight height of the new shooting point.

[0168] If the feature point density value of the image coverage area is lower than a preset density threshold, a new shooting point is added at the midpoint of the adjacent flight path in the image coverage area. This increases the density of the image coverage, ensuring that detailed information within the area is fully captured. After adding a new shooting point, interpolation processing is required based on the flight altitudes of its two adjacent shooting points to determine a reasonable flight altitude for the new shooting point. This interpolation processing can employ linear interpolation or other suitable interpolation methods to ensure that the flight altitude of the new shooting point meets both the drone's flight safety requirements and the image acquisition quality standards. Through this adjustment, even areas with originally low feature point density can achieve sufficient image detail richness by adding shooting points and optimizing flight altitudes, providing reliable data support for subsequent hazard identification work.

[0169] For example, the flight altitude of the supplementary waypoint is determined by linear interpolation, where the linear interpolation formula is: H=H1×d2 / (d1+d2)+H2×d1 / (d1+d2). For example, if the adjacent waypoints are H1=400m, H2=450m, and the distances are d1=200m and d2=300m, then H=400×300 / 500+450×200 / 500=240+180=420m.

[0170] Step S34: Update the flight path sequence based on the new shooting point and the flight altitude of the new shooting point to obtain a new flight trajectory.

[0171] After obtaining new shooting points and their flight altitudes, they need to be integrated into the existing flight path sequence. Specifically, the insertion point in the flight path sequence is determined based on the adjacent flight path positions of the new shooting point. For example, if the new shooting point is located between the i-th and i+1-th flight paths, it is inserted between the shooting point sequences corresponding to these two flight paths. Simultaneously, the shooting point numbers of subsequent flight paths are adjusted to ensure the continuity and consistency of the entire flight path sequence. Subsequently, based on the flight altitude of the new shooting point, the flight altitude information of its corresponding flight segment is updated, forming a complete flight path sequence containing the new shooting point and its flight altitude. Finally, by reconnecting all waypoints, an updated flight trajectory is generated, ensuring that the UAV can fly along the optimized path, thereby improving the comprehensiveness and accuracy of image acquisition.

[0172] In one example, the nearest neighbor algorithm is used. Starting from the starting waypoint, each time the nearest unvisited waypoint is selected, all waypoints are connected, duplicate waypoints with a distance less than a preset threshold of 50m are merged, the flight order is adjusted to minimize the turning angle of the drone, i.e., the turning angle is ≤30°, and the optimized flight trajectory is determined.

[0173] In some embodiments, the process of collecting images of cloud-obscured areas along a flight path using an inspection drone and transmitting these images back to the ground includes:

[0174] Step S401: Collect images of cloud-obscured areas along the flight path using the inspection drone, and transmit the images of cloud-obscured areas back to the ground via a wireless transmission link.

[0175] The selection of wireless transmission links must comprehensively consider transmission distance, bandwidth requirements, and environmental interference factors. In mountainous or complex terrain conditions, low-frequency communication technologies with strong anti-interference capabilities should be prioritized. For example, a digital image transmission system using the 5.8GHz band can achieve a transmission bandwidth of 20Mbps, supporting real-time transmission of 4K resolution images and ensuring stable signal transmission. For large-scale inspection tasks, relay drones can be deployed to form an aerial communication network, extending coverage through multi-hop transmission.

[0176] Step S402: The received images of the cloud-obscured area are preprocessed by the ground terminal, and the preprocessed images are sorted in time sequence according to the shooting timestamp to obtain the image time sequence.

[0177] The preprocessing includes noise filtering of the image using a Gaussian filtering algorithm. Gaussian filtering uses a two-dimensional Gaussian kernel function to perform convolution operations on the image. The kernel size is determined based on the image resolution; for a 4000×3000 pixel image, a 5×5 kernel matrix is ​​used. The standard deviation parameter σ is typically set between 1.0 and 1.5, and this value determines the smoothness of the filter. Smaller σ values ​​retain more details but have limited noise suppression, while larger σ values ​​remove noise better but may lose edge information. In power grid inspection scenarios, it is necessary to preserve the edge features of equipment such as towers and insulators; therefore, the σ value should not be too large. Gaussian filtering can effectively remove thermal noise from the image sensor, pulse interference during transmission, and noise caused by atmospheric scattering, thus improving image quality.

[0178] Meanwhile, the preprocessed images are sorted chronologically according to the shooting timestamps, and the images are arranged into an image time sequence in chronological order to facilitate subsequent analysis and processing.

[0179] Step S403: Obtain the position information of each image in the image time sequence, and determine the projected area of ​​the region covered by all images in the image time sequence based on the position information of each image.

[0180] The location information of the images can be obtained through the GPS module or differential positioning system carried by the drone. These positioning devices can record the latitude and longitude coordinates of each image in real time.

[0181] After obtaining the location information of each image, the projection transformation function of the geographic information system is used to transform the image coverage area from a geographic coordinate system (such as WGS84) to a planar coordinate system (such as Gauss-Krüger projection), and then the coverage area of ​​each image on the plane is calculated.

[0182] Step S404: Calculate the ratio of the projected area to the area of ​​the cloud-covered region to obtain the coverage rate.

[0183] The coverage calculation involves merging the projection areas of all images using Boolean union operations to form a comprehensive coverage area, and then calculating the ratio of this comprehensive coverage area to the total area of ​​the cloud-covered area.

[0184] Step S405: If the coverage rate is greater than or equal to the preset coverage rate threshold, it is marked as fully covered, and the image of the cloud-covered area is output as a complete image.

[0185] The preset coverage threshold is typically set to 95% or higher to ensure that the vast majority of the obscured area is covered by the image. If the coverage rate is greater than or equal to the preset coverage threshold, it is marked as complete coverage, and the image of the cloud-obscured area is output as a complete image.

[0186] Step S406: If the coverage rate is less than the preset coverage rate threshold, add shooting points in the flight trajectory, update the flight trajectory, and based on the updated flight trajectory, re-execute the process of collecting images of the cloud-covered area along the flight trajectory by the inspection drone, and transmit the images of the cloud-covered area back to the ground, until the coverage rate is greater than or equal to the preset coverage rate threshold, and output the latest image of the cloud-covered area as a complete image.

[0187] If the coverage rate is less than a preset coverage threshold, it is necessary to analyze the areas with insufficient coverage in the current flight trajectory and optimize the trajectory by adding shooting points in the uncovered or insufficiently covered areas. Specifically, based on the geographical location and range of the uncovered areas, and in conjunction with the original flight path layout, new shooting points are reasonably inserted between adjacent flight paths. The location of the new shooting points must ensure that their image coverage can fill the blank areas of the original trajectory, while avoiding excessive overlap with existing shooting points. After inserting new shooting points, the flight altitude between adjacent shooting points needs to be recalculated. Linear interpolation or other suitable interpolation methods are used to determine the flight altitude of the new shooting points to ensure a smooth transition and safety in flight altitude. Subsequently, the flight path sequence is updated to integrate the new shooting points into the original trajectory, and the shooting point numbers and flight altitude information of subsequent flight paths are adjusted to form a complete flight path sequence containing the new shooting points. Finally, by reconnecting all waypoints, an updated flight trajectory is generated, enabling the UAV to fly along the optimized path, repeating the image acquisition and transmission steps until the coverage rate reaches or exceeds the preset threshold, and outputting the latest complete image.

[0188] In some embodiments, satellite images of the unobstructed area of ​​the target disaster area are acquired, and the images of the cloud-obstructed area and the satellite images of the unobstructed area are stitched together to obtain a stitched and fused image, including:

[0189] Step S501: Based on multispectral satellite imagery and cloud-covered areas, determine the unobstructed satellite imagery of the target disaster area.

[0190] Among them, the areas in the multispectral satellite imagery other than those obscured by clouds are the unobstructed satellite images of the target disaster area.

[0191] Step S502: Extract feature points from the satellite unobstructed area image and the cloud-obstructed area image, and determine the matching feature point pairs between the satellite unobstructed area image and the cloud-obstructed area image through a feature point matching algorithm.

[0192] The feature points are the boundary contour coordinates of the satellite unobstructed area image and the cloud-obstructed area image. The feature points are extracted using SIFT (Scale Invariant Feature Transform) or SURF (Speeded Up Robust Features) algorithms. These algorithms can detect key points and calculate their feature descriptors at different scales.

[0193] For satellite-unobstructed area imagery and cloud-obstructed area imagery, salient features such as corner points and edge points are extracted to generate two sets of feature points. The feature point matching algorithm employs RANSAC (Random Sample Consensus) combined with a nearest neighbor matching strategy. Initial matching point pairs are selected by calculating the Euclidean distance of the feature descriptors, and then RANSAC is used to remove mismatched points, retaining matching pairs with strong geometric consistency. For example, a distance threshold of 0.8 times the feature descriptor dimension is set, and the optimal matching set is obtained through iterative optimization to ensure accurate spatial alignment between satellite imagery and UAV imagery. The quantity and quality of matched feature point pairs directly affect the stitching accuracy; typically, at least 50 matching point pairs are required, and they must be evenly distributed to support subsequent geometric correction and fusion processing.

[0194] Step S503: Based on the matching feature point pairs, calculate the transformation parameters between the satellite unobstructed area image and the cloud-obstructed area image. The transformation parameters are used to describe the geometric transformation relationship between the satellite unobstructed area image and the cloud-obstructed area image.

[0195] Specifically, based on matching feature point pairs, the least squares method is used to fit the optimal transformation parameters, constructing mapping parameters from the satellite image coordinate system to the UAV image coordinate system. The transformation parameters include translation, rotation, and scaling factors.

[0196] For example, select three pairs of corresponding points: (100,200)→(150,250), (300,400)→(350,450), (500,600)→(550,650), and solve the system of equations to obtain the translation t. x =50、t y =50, rotate θ=0°, scale s x =1、s y =1, perform geometric correction on the UAV imagery to align its boundaries with the satellite imagery boundaries, obtaining a preliminary registered image pair. For the preliminary registered image pair, use the SIFT algorithm to extract feature points and their descriptors in the overlapping areas. Select corresponding point pairs through descriptor matching. A reliable match is defined as a ratio of nearest neighbor to second nearest neighbor distance ≤0.7 (e.g., nearest neighbor distance 0.5, second nearest neighbor distance 1.0, ratio 0.5 < 0.7). After removing mismatched points, calculate the precise transform parameters.

[0197] Step S504: Perform geometric transformation on the image of the cloud-covered area using transformation parameters to align the image of the cloud-covered area with the image of the satellite uncovered area in spatial position, and then stitch and fuse the aligned image of the cloud-covered area and the image of the satellite uncovered area to obtain the initial stitched and fused image.

[0198] Specifically, by using the calculated transformation parameters, affine or projection transformations are performed on the image of the cloud-covered area to precisely align it with the image of the uncovered satellite area in spatial position.

[0199] Step S505: Eliminate the stitching seams in the initially stitched and fused image, and unify the pixel resolution of the initially stitched and fused image to obtain the stitched and fused image.

[0200] The process employs a gradient blending method to eliminate stitching seams. This gradient blending technique achieves seamless stitching by constructing a weighted mask in the overlapping areas. Distance transformation is used to calculate the shortest distance from each pixel to the image boundary; this normalized distance value is then used as the blending weight.

[0201] Specifically, by constructing a weighted gradient template, transition areas are set on both sides of the stitching seam, and the template weight changes linearly with the distance from the stitching seam. For example, the width of the transition area is set to 50 pixels, the weight at the stitching seam is 0, and the weight at the edge of the transition area is 1. The images on both sides are weighted and averaged. This gradient transition can effectively eliminate the stitching seam, avoid obvious brightness or color jumps, and make the pixel values ​​transition smoothly.

[0202] Meanwhile, a bicubic interpolation algorithm is used to unify the pixel resolution of the stitched and fused image. The grayscale value of the new pixel is calculated based on the target resolution, and interpolation is performed based on the grayscale values ​​of the surrounding 16 known pixels to ensure consistent overall image resolution. The result is a seamlessly stitched and uniformly resolution fused image, providing a high-quality data foundation for subsequent hazard identification.

[0203] In some embodiments, by using a preset power grid hazard feature template, suspected power grid hazard areas are selected from the stitched and fused images. Combined with real-time operational data of these suspected power grid hazard areas obtained from the ground, the actual location of the hazard areas is identified, including:

[0204] Step S601: Divide the stitched and fused image into sliding windows to obtain sub-image regions under multiple windows.

[0205] The sliding window scanning employs a multi-scale detection strategy. The window size is set according to the typical size of different hazard types; for example, tree obstruction hazards typically use windows of 50×50 to 200×200 pixels, while hanging foreign objects use windows of 30×30 to 100×100 pixels. The window slides across the image row by row and column by column with a set step size, typically 1 / 4 to 1 / 2 of the window size, ensuring sufficient overlap between adjacent windows and avoiding missing hazards at boundaries. The sliding window starts from the upper left corner of the image and moves sequentially along the row and column directions, moving a step size each time, until it covers the entire image area, generating a sub-image sequence covering the entire map.

[0206] Step S602: For each sub-image region, determine the similarity between the sub-image region and the preset power grid hidden danger feature template.

[0207] The preset power grid hazard feature template includes morphological features such as foreign objects in conductors, vegetation intrusion, and equipment corrosion. At each window location, features such as color histogram, edge direction histogram, and local binary pattern are extracted and matched with the features of the corresponding hazard type in the preset power grid hazard feature template. Similarity calculation uses a normalized cross-correlation coefficient.

[0208] The normalized cross-correlation coefficient is obtained by calculating the dot product of the window feature vector and the template feature vector, and then dividing by the product of their moduli. Its value ranges from -1 to 1, with a value closer to 1 indicating higher similarity. For example, for a wire foreign object template, the RGB color values ​​of all pixels within the window are extracted to construct a color histogram. The pixel distribution ratio of each color channel is statistically analyzed and compared with the typical color histogram of the wire foreign object in the template to calculate the normalized cross-correlation coefficient. Simultaneously, the gradient direction information of the window edges is extracted to construct an edge direction histogram, which is then matched with the edge direction features in the template to further verify similarity. By comprehensively considering the matching results of multi-dimensional features such as color and edges, the similarity score between the sub-image region and the preset template is determined.

[0209] Step S603: If the similarity exceeds the preset similarity threshold, the sub-image region is taken as a suspected power grid hazard region, and the location coordinates of the suspected power grid hazard region are extracted.

[0210] The similarity threshold is set to 0.7. When the similarity exceeds 0.7, the sub-image area of ​​the window is marked as a suspected power grid hazard area, and the location coordinates of the suspected power grid hazard area are extracted.

[0211] Step S604: Based on the location coordinates of the suspected power grid hazard area, obtain the real-time operation data of the suspected power grid hazard area from the ground.

[0212] Specifically, based on the location coordinates of suspected power grid hazard areas, real-time operational data for the corresponding areas is retrieved from the ground. This real-time operational data covers multi-dimensional information such as voltage, current, temperature, and equipment status.

[0213] Step S605: Compare the real-time operating data with the preset normal operating data range. If the real-time operating data does not meet the preset normal operating data range, then determine the location coordinates of the suspected power grid hazard area as the actual power grid hazard area location.

[0214] The process involves comparing the retrieved real-time operational data with preset normal operating data ranges one by one. For example, for voltage data, if the value exceeds the preset normal voltage fluctuation range; for current data, if the value is not within the preset reasonable current range; for temperature data, if the value exceeds the upper limit of the temperature range during normal equipment operation, etc. If any one or more real-time operational data points do not meet the preset normal operating data range, the area corresponding to the location coordinates of the suspected power grid hazard area is determined to be the actual location of the power grid hazard area. Simultaneously, to ensure the accuracy of the determination results, a certain fault tolerance mechanism can be set. For example, only when the real-time operational data at multiple consecutive time points does not meet the preset range is the actual location of the power grid hazard area finally determined, avoiding misjudgments due to accidental data fluctuations.

[0215] Based on the same inventive concept, this application also provides a power grid hazard identification system for implementing the power grid hazard identification method described above.

[0216] The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more power grid hazard identification system embodiments provided below can be found in the limitations of the power grid hazard identification method described above, and will not be repeated here.

[0217] like Figure 3 As shown in the figure, this application provides a power grid hazard identification system, including:

[0218] The obscured area determination module 100 is used to acquire multispectral satellite images of the target disaster area, detect cloud-obscured areas and boundary pixel coordinates of the cloud-obscured areas based on the multispectral satellite images, and determine the geographical location range of the cloud-obscured areas based on the boundary pixel coordinates of the cloud-obscured areas.

[0219] Available UAV nest determination module 200 is used to obtain terrain and weather environmental parameters of the cloud-covered area, construct environmental constraints, and select multiple schedulable UAV nests from the candidate UAV nest set based on the geographical location range of the cloud-covered area and the environmental constraints, to obtain a list of available UAV nests.

[0220] The flight trajectory generation module 300 is used to select the idle drone closest to the cloud-covered area as the inspection drone based on the geographical range of the cloud-covered area through the list of available drone nests, and generate the flight trajectory of the inspection drone covering the cloud-covered area by combining the preset compliant resolution parameters of the camera in the inspection drone.

[0221] The image acquisition and transmission module 400 is used to acquire images of cloud-obscured areas along the flight path of the inspection drone and transmit the images of the cloud-obscured areas back to the ground.

[0222] The image stitching and fusion module 500 is used to acquire satellite images of the unobstructed area of ​​the target disaster area, and to stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain the stitched and fused image.

[0223] The hazard area identification module 600 is used to filter out suspected power grid hazard areas from the stitched and fused images using a preset power grid hazard feature template, and to identify the actual location of the power grid hazard area by combining it with real-time operational data of the suspected power grid hazard area obtained from the ground.

[0224] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the system described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0225] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0226] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0227] In the several embodiments provided by this invention, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0228] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for identifying potential hazards in power grids, characterized in that, include: Acquire multispectral satellite imagery of the target disaster area, detect cloud-covered areas and boundary pixel coordinates of the cloud-covered areas based on the multispectral satellite imagery, and determine the geographical location range of the cloud-covered areas based on the boundary pixel coordinates of the cloud-covered areas. Obtain the terrain and weather environment parameters of the cloud-covered area, construct environmental constraints, and based on the geographical location range of the cloud-covered area and the environmental constraints, select multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list. Based on the geographical location of the cloud-covered area, the idle drones closest to the cloud-covered area are selected as inspection drones from the list of available drone nests. Combined with the preset compliant resolution parameters of the camera in the inspection drone, the flight trajectory of the inspection drone covering the cloud-covered area is generated. The inspection drone collects images of the cloud-obscured area along the flight path and transmits the images of the cloud-obscured area back to the ground. Acquire satellite images of the unobstructed area of ​​the target disaster area, and stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain a stitched and fused image; By using a preset power grid hazard feature template, suspected power grid hazard areas are selected from the stitched and fused images. Combined with the real-time operation data of the suspected power grid hazard areas obtained from the ground terminal, the actual location of the power grid hazard areas is identified.

2. The power grid hazard identification method according to claim 1, characterized in that, The process of acquiring multispectral satellite imagery of the target disaster area, detecting cloud-covered areas and their boundary pixel coordinates based on the multispectral satellite imagery, and determining the geographical location range of the cloud-covered areas based on the boundary pixel coordinates of the cloud-covered areas includes: Acquire multispectral satellite images of the target disaster area and preprocess the multispectral satellite images; The band reflectance values ​​of each pixel in the preprocessed multispectral satellite image are obtained, and the pixels with band reflectance values ​​greater than a preset reflectance threshold are marked as potential cloud pixels. Multiple potential cloud pixels are then merged into a cloud pixel marker map. Morphological dilation and erosion operations are performed on adjacent potential cloud pixels in the cloud pixel marker map to obtain the initial cloud occlusion region. Extract the boundary pixel coordinates of each point on the boundary of the initial cloud-covered area, perform coordinate transformation on the multiple boundary pixel coordinates, and obtain the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area. Based on the latitude and longitude values ​​of each point on the boundary of the initial cloud-covered area, the minimum bounding rectangle of the initial cloud-covered area is determined, and the geographical location range of the cloud-covered area is determined based on the minimum bounding rectangle.

3. The power grid hazard identification method according to claim 1, characterized in that, The process involves obtaining terrain and weather environmental parameters of the cloud-covered area, constructing environmental constraints, and, based on the geographical location of the cloud-covered area and the environmental constraints, selecting multiple schedulable drone nests from the candidate drone nest set to obtain an available nest list, including: Based on the geographical location of the cloud-covered area, various terrain and weather environmental parameters within the cloud-covered area are determined, and environmental constraints are determined based on these various terrain and weather environmental parameters. Obtain the position coordinates of each first candidate drone nest in the candidate drone nest set that satisfies the environmental constraints, and for each first candidate drone nest, determine the Euclidean distance from the position coordinates of the first candidate drone nest to the center of the cloud-covered area. If the Euclidean distance is less than a preset distance threshold, the first candidate drone nest is determined to be the second candidate drone nest, and the second candidate drone nest is added to the initial available nest list. Select second candidate drone nests from the initial list of available nests that have sufficient range for the round-trip flight distance relative to the cloud cover area; these will be the schedulable drone nests. The multiple schedulable UAV nests are sorted according to their Euclidean distance to generate the list of available nests.

4. The power grid hazard identification method according to claim 1, characterized in that, Also includes: Obtain the current resolution parameters of the satellite's unobstructed area image, and determine whether the current resolution parameters meet the preset compliance resolution conditions; If it is determined that the current resolution parameter meets the preset compliance resolution condition, then the current resolution parameter is used as the compliance resolution parameter of the drone's camera. If it is determined that the current resolution parameter does not meet the preset compliance resolution condition, then the backup resolution parameter that meets the preset compliance resolution condition from the preset backup resolution parameter list is used as the compliance resolution parameter of the drone's camera.

5. The power grid hazard identification method according to claim 1 or 4, characterized in that, The process involves selecting the nearest available drone to the cloud-covered area from the list of available drone nests based on the geographical location of the cloud-covered area, and generating a flight trajectory for the inspection drone covering the cloud-covered area by combining the preset compliant resolution parameters of the camera in the inspection drone. This includes: Based on the geographical location of the cloud-covered area, the available drones closest to the cloud-covered area are selected as inspection drones from the list of available drone nests. Based on the preset compliant resolution parameters, focal length, and sensor size of the camera in the inspection drone, the image coverage area of ​​a single image captured by the inspection drone is determined. Based on the image coverage area, the distribution of shooting points covering the cloud-covered area is determined by combining the preset shooting point intervals. Based on the distribution of the shooting points, a gridded waypoint coordinate matrix covering the cloud-obscured area is determined; Based on the gridded waypoint coordinate matrix, the positions of each shooting point are connected in a preset flight order to form a flight path sequence. The flight altitude of each shooting point is adjusted according to environmental constraints to generate a flight trajectory that includes the flight path sequence and flight altitude.

6. The power grid hazard identification method according to claim 5, characterized in that, Also includes: Based on the flight path sequence, feature points within the image coverage area of ​​each shooting point are extracted, and the number of feature points per unit area within the image coverage area is counted to obtain the feature point density value of the image coverage area. For each image coverage area, the feature point density value is compared with a preset density threshold. If the feature point density value of the image coverage area is lower than the preset density threshold, a new shooting point is added at the midpoint of the adjacent flight path in the image coverage area, and the flight height of the new shooting point is interpolated based on the flight height of the two adjacent shooting points to obtain the flight height of the new shooting point. Based on the new shooting point and its flight altitude, the flight path sequence is updated to obtain a new flight trajectory.

7. The power grid hazard identification method according to claim 1, characterized in that, The step of collecting images of the cloud-obscured area along the flight path using the inspection drone and transmitting the images of the cloud-obscured area back to the ground includes: The inspection drone collects images of the cloud-obscured area along the flight path and transmits the images of the cloud-obscured area back to the ground via a wireless transmission link. The image of the cloud-obscured area received by the ground terminal is preprocessed, and the preprocessed image is sorted in time sequence according to the shooting timestamp to obtain the image time sequence. The location information of each image in the image time sequence is obtained, and the projected area of ​​the region covered by all images in the image time sequence is determined based on the location information of each image. The coverage rate is obtained by calculating the ratio of the projected area to the area of ​​the cloud-covered region. If the coverage rate is greater than or equal to a preset coverage rate threshold, it is marked as complete coverage, and the image of the cloud-covered area is output as a complete image. If the coverage rate is less than a preset coverage rate threshold, then the number of shooting points in the flight trajectory is increased, the flight trajectory is updated, and based on the updated flight trajectory, the process of collecting images of the cloud-obscured area by the inspection drone along the flight trajectory is repeated, and the images of the cloud-obscured area are transmitted back to the ground, until the coverage rate is greater than or equal to the preset coverage rate threshold, and the latest image of the cloud-obscured area is output as a complete image.

8. The power grid hazard identification method according to claim 1, characterized in that, The process of acquiring satellite images of the unobstructed area of ​​the target disaster area, and then stitching and fusing the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain a stitched and fused image includes: Based on the multispectral satellite imagery and the cloud-covered areas, determine the unobstructed satellite imagery of the target disaster area; Feature points are extracted from the satellite unobstructed area image and the cloud-obstructed area image, and matching feature point pairs between the satellite unobstructed area image and the cloud-obstructed area image are determined by a feature point matching algorithm; Based on the matching feature point pairs, the transformation parameters between the satellite unobstructed area image and the cloud-obstructed area image are calculated. The transformation parameters are used to describe the geometric transformation relationship between the satellite unobstructed area image and the cloud-obstructed area image. Geometric transformation is performed on the image of the cloud-covered area using transformation parameters to align the image of the cloud-covered area with the image of the satellite uncovered area in spatial position. The aligned image of the cloud-covered area and the image of the satellite uncovered area are then stitched together to obtain an initial stitched and fused image. The stitching seams of the initially stitched and fused image are eliminated, and the pixel resolution of the initially stitched and fused image is unified to obtain the stitched and fused image.

9. The power grid hazard identification method according to claim 1, characterized in that, The process involves using a preset power grid hazard feature template to filter out suspected power grid hazard areas from the stitched and fused image, and combining this with real-time operational data of the suspected power grid hazard areas acquired from the ground terminal to identify the actual location of the hazard areas. This includes: The stitched and fused image is divided into multiple sub-image regions under a sliding window. For each of the sub-image regions, determine the similarity between the sub-image region and the preset power grid hidden danger feature template; If the similarity exceeds a preset similarity threshold, the sub-image region is designated as a suspected power grid hazard region, and the location coordinates of the suspected power grid hazard region are extracted. Based on the location coordinates of the suspected power grid hazard area, real-time operational data of the suspected power grid hazard area is obtained from the ground terminal; The real-time operating data is compared with a preset normal operating data range. If the real-time operating data does not meet the preset normal operating data range, the location coordinates of the suspected power grid hazard area are determined to be the location of the actual power grid hazard area.

10. A power grid hazard identification system, characterized in that, include: The obscured area determination module is used to acquire multispectral satellite images of the target disaster area, detect cloud-obscured areas and boundary pixel coordinates of the cloud-obscured areas based on the multispectral satellite images, and determine the geographical location range of the cloud-obscured areas based on the boundary pixel coordinates of the cloud-obscured areas. The available drone nest determination module is used to obtain the terrain and weather environment parameters of the cloud-covered area, construct environmental constraints, and select multiple schedulable drone nests from the candidate drone nest set based on the geographical location range of the cloud-covered area and the environmental constraints, thereby obtaining an available drone nest list. The flight trajectory generation module is used to select the idle drone closest to the cloud-covered area as the inspection drone based on the geographical location range of the cloud-covered area through the list of available drone nests, and generate the flight trajectory of the inspection drone covering the cloud-covered area by combining the preset compliance resolution parameters of the camera in the inspection drone. The image acquisition and transmission module is used to acquire images of the cloud-obscured area along the flight path of the inspection drone, and transmit the images of the cloud-obscured area back to the ground. The image stitching and fusion module is used to acquire satellite images of the unobstructed area of ​​the target disaster area, and to stitch and fuse the images of the cloud-obstructed area and the satellite images of the unobstructed area to obtain the stitched and fused image. The hidden danger area identification module is used to filter out suspected power grid hidden danger areas from the stitched and fused images using a preset power grid hidden danger feature template, and to identify the actual location of the power grid hidden danger area by combining the real-time operation data of the suspected power grid hidden danger area obtained by the ground terminal.

Images