Hydropower station unmanned aerial vehicle inspection method based on AI intelligent analysis

Abstract

The invention relates to a hydropower station unmanned aerial vehicle inspection method based on AI intelligent analysis. The method is suitable for the dam safety field. According to the technical scheme, the method comprises the following steps: acquiring an overall surface image of a dam to be inspected; forming a dam defect probability distribution diagram on the whole image of the surface of the dam to be inspected based on the dam surface image in combination with historical data of dams of the same type; acquiring the number N of preset inspection detection points, and distributing the inspection detection points according to the probability of occurrence of defects in each block on the whole image of the surface of the dam to be inspected; randomly generating a corresponding number of inspection detection points in the range of each block based on the number of the inspection detection points allocated to each block; generating a normal vector perpendicular to the surface of the dam based on the inspection detection points in the blocks, and selecting two points on the normal vector as a close-shot inspection point and a long-shot inspection point respectively; and generating an unmanned aerial vehicle routing inspection path based on the close-shot routing inspection points and the long-shot routing inspection points corresponding to the surface of the routing inspection dam and the number of the routing inspection unmanned aerial vehicles.

Description

Technical Field

[0001] The present invention relates to a hydropower station drone inspection method based on AI intelligent analysis, which is applicable to the field of dam safety. Background Art

[0002] Traditional dam surface sinkhole detection mostly relies on manual work, which involves building observation platforms and operating observation instruments. Manual detection is highly subjective, inefficient, has many interference factors, and has high labor costs. At present, with the maturity of drone technology, drones have also been applied to dam inspections. Surface images of dams are obtained through drones, and defects on the dam surface, such as cracks and leakage, are identified based on the dam surface images.

[0003] The dam is huge, and the defects on the surface of the dam may be only at the millimeter level (such as cracks). Although the use of drones can effectively save manpower, it takes a long time because it requires collecting high-definition images of various parts of the dam surface at close range. Summary of the invention

[0004] The technical problem to be solved by the present invention is: in response to the above-mentioned problems, a hydropower station drone inspection method based on AI intelligent analysis is provided.

[0005] The technical solution adopted by the present invention is: a hydropower station drone inspection method based on AI intelligent analysis, comprising:

[0006] Obtain the overall surface image of the dam to be inspected;

[0007] Based on the dam surface image and the historical data of the same type of dams, a probability distribution map of dam defects is formed on the overall surface image of the dam to be inspected;

[0008] Obtain the preset number of inspection points N, and allocate the inspection points according to the probability of defects occurring in each block on the overall image of the dam surface to be inspected;

[0009] Based on the number of inspection points allocated to each block, a corresponding number of inspection points are randomly generated within each block;

[0010] Generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, where the distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1;

[0011] Based on the near-view inspection points and far-view inspection points corresponding to the inspected dam surface and the number of inspection drones, a drone inspection path is generated.

[0012] The method of forming a dam defect probability distribution map on the overall surface image of the dam to be inspected based on the dam surface image combined with historical data of dams of the same type includes:

[0013] Obtain historical data of dams of the same type as the dam to be inspected, which includes the locations where surface defects of the dam appeared;

[0014] The dam surface is divided into several blocks according to the preset block division rules, and the number of defects in each block in the historical data is counted;

[0015] Based on the number of defects occurring in each block, the probability of defects occurring in each block is determined, and the probability distribution map of dam defects is obtained.

[0016] The step of obtaining a preset number N of inspection points and allocating inspection points according to the probability of occurrence of defects in each block on the overall image of the surface of the dam to be inspected includes:

[0017]

[0018] Among them, n i is the number of inspection points assigned to the i-th block on the surface image of the dam to be inspected; P i is the probability of defects occurring in the i-th block on the surface image of the dam to be inspected, i = 1, 2, ... m, and m is the number of blocks divided on the surface image of the dam to be inspected.

[0019] The method of generating a drone inspection path based on the near-view inspection points and the far-view inspection points corresponding to the inspection dam surface and the number of inspection drones includes:

[0020] According to the number of inspection drones and the distribution of inspection points on the dam surface, the dam surface is divided into inspection areas corresponding to the number of inspection drones;

[0021] Based on the near-field inspection points and far-field inspection points in each inspection area, the drone inspection path is planned for the corresponding inspection drone.

[0022] The method of planning a drone inspection path for a corresponding inspection drone based on the near-view inspection points and the far-view inspection points in each inspection area includes:

[0023] The inspection path of the UAV is divided into a long-range inspection path to be executed first and a short-range inspection path to be executed later. The long-range inspection path is obtained based on the planning of long-range inspection points in the inspection area, and the short-range inspection path is obtained based on the planning of short-range inspection points in the inspection area.

[0024] Based on the dam surface images collected by the inspection drone on the long-range inspection path, the defects in the dam surface images are identified through the trained defect recognition model;

[0025] After the defect is identified, the position of the defect on the overall image of the dam surface to be inspected is determined based on the corresponding position of the inspection drone on the long-range inspection path during image acquisition and the corresponding position of the defect on the image, and an inspection point is added at the position, and corresponding short-range inspection points are added based on the newly added inspection points;

[0026] Replan the close-range inspection route based on the newly added close-range inspection points.

[0027] A hydropower station drone inspection device based on AI intelligent analysis, comprising:

[0028] The first module is used to obtain the overall surface image of the dam to be inspected;

[0029] The second module is used to form a probability distribution map of dam defects on the overall surface image of the dam to be inspected based on the dam surface image combined with historical data of dams of the same type;

[0030] The third module is used to obtain the preset number of inspection points N and allocate the inspection points according to the probability of defects appearing in each block on the overall image of the dam surface to be inspected;

[0031] The fourth module is used to randomly generate a corresponding number of inspection points within each block based on the number of inspection points allocated to each block;

[0032] The fifth module is used to generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, respectively, where the distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1;

[0033] The sixth module is used to generate a drone inspection route based on the near-view inspection points and far-view inspection points corresponding to the inspection dam surface and the number of inspection drones.

[0034] A storage medium stores a computer program that can be executed by a processor, and when the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis are implemented.

[0035] A computer device comprises a memory and a processor, wherein the memory stores a computer program executable by the processor, and when the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis are implemented.

[0036] A hydropower station drone inspection system based on AI intelligent analysis, including a drone hangar and at least one inspection drone;

[0037] The drone hangar includes a controller, a wireless communication module, a drone storage warehouse, and a charging module for charging the drone;

[0038] The controller has a memory and a processor, and the memory stores a computer program that can be executed by the processor. When the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis are implemented;

[0039] The controller is connected to the inspection drone via the wireless communication module, sends the inspection drone a corresponding inspection path, and obtains the dam surface image collected by the drone.

[0040] The beneficial effects of the present invention are as follows: the present invention allocates inspection points to each block on the dam surface based on the probability distribution map of dam defects, and based on the number of inspection points in each block, randomly generates a corresponding number of inspection points within each block, thereby achieving key inspections on blocks with a high probability of defect occurrence, improving inspection efficiency, and because the positions of the inspection points are randomly generated during each inspection, the entire dam surface area can be basically covered after multiple inspection tasks.

[0041] The present invention generates near-view inspection points and far-view inspection points based on inspection detection points. A wide-range image of the dam surface is acquired through the far-view inspection points, thereby collecting more dam surface information; a small-range but clearer image of the dam surface is acquired through the near-view inspection points, thereby facilitating the identification of small defects on the dam surface.

[0042] In the present invention, the inspection drone first executes a long-range inspection path, identifies defects in the image through the dam image collected by the drone on the long-range inspection path, and adds a close-range inspection point after the defect is identified, thereby obtaining a clear image of the defective area to more accurately judge the dam defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of an embodiment. DETAILED DESCRIPTION

[0044] This embodiment is a hydropower station drone inspection method based on AI intelligent analysis, which specifically includes the following steps:

[0045] S1. Obtaining an overall surface image of the dam to be inspected, wherein the overall surface image includes the entire upstream surface or the entire downstream surface of the dam.

[0046] S2. Based on the dam surface image and the historical data of the same type of dams, a dam defect probability distribution map is formed on the overall surface image of the dam to be inspected.

[0047] S21. Obtain historical data of dams of the same type as the dam to be inspected. If the dam to be inspected is an earth-rock dam, obtain historical data of existing earth-rock dams. The historical data includes locations where surface defects (cracks, leakage, etc.) of the dam appear.

[0048] S22. Divide the dam surface into several blocks according to a preset block division rule (e.g., divide the dam surface into m blocks and the relative position of each block on the dam surface adopts a unified position distribution rule), and count the number of defects in each block in the historical data.

[0049] S23. Based on the number of defects in each block, determine the probability of defects in each block, and obtain a probability distribution diagram of dam defects. If there are 1000 samples in the historical data, and the number of defects in the ith block is 10 (e.g., 1 defect in the ith block of 5 samples, 2 defects in the ith block of 1 sample, and 3 defects in the ith block of 1 sample), the probability of defects in the ith block is 10 / 1000=1%.

[0050] S3. Obtain a preset number N of inspection points, and allocate the number of inspection points according to the probability of occurrence of defects in each block on the overall image of the dam surface to be inspected.

[0051] In this embodiment, the number of inspection points N is determined in an approximate numerical range according to the area of ​​the dam surface, and is appropriately adjusted according to the actual situation before the inspection. For example, if the number of inspection points in the previous inspections was relatively small, the number of inspection points can be appropriately increased in this inspection. If a natural disaster such as a flood or earthquake that affects the safety of the dam has occurred before this inspection, a large number of inspection points will need to be increased.

[0052] In this example, the number of inspection points in each block is determined according to the following formula:

[0053]

[0054] Among them, n i is the number of inspection points assigned to the i-th block on the surface image of the dam to be inspected; P i is the probability of defects occurring in the i-th block on the surface image of the dam to be inspected, i = 1, 2, ... m, and m is the number of blocks divided on the surface image of the dam to be inspected.

[0055] S4. Based on the number of inspection points allocated to each block, a corresponding number of inspection points are randomly generated within each block.

[0056] S5. Generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, respectively. The distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1, and H1 and H2 are determined by the user based on the performance of the inspection drone camera.

[0057] S6. Generate a drone inspection path based on the near-view inspection points and far-view inspection points corresponding to the inspection dam surface and the number of inspection drones.

[0058] S61. Divide the dam surface into multiple inspection areas according to the number of inspection drones. The number of inspection areas corresponds to the number of inspection drones. The scope of each inspection area is determined according to the distribution of inspection detection points on the dam surface.

[0059] S62. Based on the near-view inspection points and far-view inspection points in each inspection area, a drone inspection path is planned for the corresponding inspection drone. The drone inspection path is divided into a far-view inspection path that is executed first and a near-view inspection path that is executed later. The far-view inspection path is planned based on the far-view inspection points in the inspection area, and the near-view inspection path is planned based on the near-view inspection points in the inspection area.

[0060] S621, based on the long-term inspection points in the inspection area, taking the shortest overall path as the planning goal, planning a long-term inspection path;

[0061] S622, obtaining a dam surface image collected by the inspection drone on a long-range inspection path, processing the dam surface image through a trained defect recognition model, and identifying defects on the dam surface image;

[0062] S623, after the defect is identified, based on the corresponding position of the inspection drone on the long-range inspection path during image acquisition and the corresponding position of the defect on the image, determine the approximate position of the defect on the overall image of the dam surface to be inspected, add an inspection point at the position, and add corresponding short-range inspection points based on the newly added inspection points;

[0063] S624: Based on the near-view inspection points in the inspection area, a near-view inspection path is planned, and the near-view inspection path needs to be different from the far-view inspection path.

[0064] The planning method of the close-range inspection path in this embodiment includes: planning point by point with the close-range inspection point corresponding to the end of the long-range inspection path as the starting point, calculating the distance between the starting point and the remaining inspection points (excluding the planned close-range inspection points and the close-range inspection points corresponding to the long-range inspection points connected to the long-range inspection points corresponding to the starting point on the long-range inspection path), taking the point with the shortest distance as the connection point connected to the starting point, and calculating the connection point of the new starting point again with this connection point as the starting point, until all close-range inspection points are planned.

[0065] S625. Obtain the dam surface image collected by the inspection drone on the close-range inspection path, process the dam surface image through the trained defect recognition model, and identify defects on the dam surface image.

[0066] This embodiment also provides a hydropower station drone inspection device based on AI intelligent analysis, including:

[0067] The first module is used to obtain the overall surface image of the dam to be inspected;

[0068] The second module is used to form a probability distribution map of dam defects on the overall surface image of the dam to be inspected based on the dam surface image combined with historical data of dams of the same type;

[0069] The third module is used to obtain the preset number of inspection points N and allocate the inspection points according to the probability of defects appearing in each block on the overall image of the dam surface to be inspected;

[0070] The fourth module is used to randomly generate a corresponding number of inspection points within each block based on the number of inspection points allocated to each block;

[0071] The fifth module is used to generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, respectively, where the distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1;

[0072] The sixth module is used to generate a drone inspection route based on the near-view inspection points and far-view inspection points corresponding to the inspection dam surface and the number of inspection drones.

[0073] This embodiment also provides a storage medium on which a computer program that can be executed by a processor is stored. When the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis in this example are implemented.

[0074] This embodiment also provides a computer device having a memory and a processor, wherein the memory stores a computer program that can be executed by the processor, and when the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis in this example are implemented.

[0075] This embodiment also provides a hydropower station drone inspection system based on AI intelligent analysis, including a drone hangar and at least one inspection drone, wherein the drone hangar includes a controller, a wireless communication module, a drone storage warehouse and a charging module for charging the drone.

[0076] In this example, the controller has a memory and a processor. The memory stores a computer program that can be executed by the processor. When the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis in this example are implemented.

[0077] In this embodiment, the controller is connected to the inspection drone via the wireless communication module, sends the inspection drone a corresponding inspection path, and obtains the dam surface image collected by the drone.

Claims

1. A hydropower station drone inspection method based on AI intelligent analysis, characterized in that: include: Obtain the overall surface image of the dam to be inspected; Based on the dam surface image and the historical data of the same type of dams, a probability distribution map of dam defects is formed on the overall surface image of the dam to be inspected; Obtain the preset number of inspection points N, and allocate the inspection points according to the probability of defects occurring in each block on the overall image of the dam surface to be inspected; Based on the number of inspection points allocated to each block, a corresponding number of inspection points are randomly generated within each block; Generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, where the distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1; Based on the near-view inspection points and far-view inspection points corresponding to the inspected dam surface and the number of inspection drones, a drone inspection path is generated.

2. According to claim 1, a hydropower station drone inspection method based on AI intelligent analysis is characterized in that: The method of forming a dam defect probability distribution map on the overall surface image of the dam to be inspected based on the dam surface image combined with historical data of dams of the same type includes: Obtain historical data of dams of the same type as the dam to be inspected, which includes the locations where surface defects of the dam appeared; The dam surface is divided into several blocks according to the preset block division rules, and the number of defects in each block in the historical data is counted; Based on the number of defects occurring in each block, the probability of defects occurring in each block is determined, and the probability distribution map of dam defects is obtained.

3. According to the hydropower station drone inspection method based on AI intelligent analysis according to claim 1, it is characterized in that: The step of obtaining a preset number N of inspection points and allocating inspection points according to the probability of occurrence of defects in each block on the overall image of the surface of the dam to be inspected includes: Among them, n i is the number of inspection points assigned to the i-th block on the surface image of the dam to be inspected; P i is the probability of defects occurring in the i-th block on the surface image of the dam to be inspected, i = 1, 2, ... m, and m is the number of blocks divided on the surface image of the dam to be inspected.

4. According to the hydropower station drone inspection method based on AI intelligent analysis according to claim 1, it is characterized in that: The method of generating a drone inspection path based on the near-view inspection points and the far-view inspection points corresponding to the inspection dam surface and the number of inspection drones includes: According to the number of inspection drones and the distribution of inspection points on the dam surface, the dam surface is divided into inspection areas corresponding to the number of inspection drones; Based on the near-field inspection points and far-field inspection points in each inspection area, the drone inspection path is planned for the corresponding inspection drone.

5. According to claim 4, a hydropower station drone inspection method based on AI intelligent analysis is characterized in that: The method of planning a drone inspection path for a corresponding inspection drone based on the near-view inspection points and the far-view inspection points in each inspection area includes: The inspection path of the UAV is divided into a long-range inspection path to be executed first and a short-range inspection path to be executed later. The long-range inspection path is obtained based on the planning of long-range inspection points in the inspection area, and the short-range inspection path is obtained based on the planning of short-range inspection points in the inspection area. Based on the dam surface images collected by the inspection drone on the long-range inspection path, the defects in the dam surface images are identified through the trained defect recognition model; After the defect is identified, the position of the defect on the overall image of the dam surface to be inspected is determined based on the corresponding position of the inspection drone on the long-range inspection path during image acquisition and the corresponding position of the defect on the image, and an inspection point is added at the position, and corresponding short-range inspection points are added based on the newly added inspection points; Replan the close-range inspection route based on the newly added close-range inspection points.

6. A hydropower station drone inspection device based on AI intelligent analysis, characterized in that: include: The first module is used to obtain the overall surface image of the dam to be inspected; The second module is used to form a probability distribution map of dam defects on the overall surface image of the dam to be inspected based on the dam surface image combined with historical data of dams of the same type; The third module is used to obtain the preset number of inspection points N and allocate the inspection points according to the probability of defects appearing in each block on the overall image of the dam surface to be inspected; The fourth module is used to randomly generate a corresponding number of inspection points within each block based on the number of inspection points allocated to each block; The fifth module is used to generate a normal vector perpendicular to the dam surface based on the inspection points in the block, and select two points on the normal vector as the near-view inspection point and the far-view inspection point, respectively, where the distance between the near-view inspection point and the dam surface is H1, and the distance between the far-view inspection point and the dam surface is H2, H2>H1; The sixth module is used to generate a drone inspection route based on the near-view inspection points and far-view inspection points corresponding to the inspection dam surface and the number of inspection drones.

7. A storage medium having stored thereon a computer program executable by a processor, characterized in that: When the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis as described in any one of claims 1 to 5 are implemented.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program executable by the processor, wherein: When the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis as described in any one of claims 1 to 5 are implemented.

9. A hydropower station drone inspection system based on AI intelligent analysis, characterized by: Includes a drone hangar and at least one inspection drone; The drone hangar includes a controller, a wireless communication module, a drone storage warehouse, and a charging module for charging the drone; The controller has a memory and a processor, and the memory stores a computer program that can be executed by the processor, and when the computer program is executed, the steps of the hydropower station drone inspection method based on AI intelligent analysis according to any one of claims 1 to 5 are implemented; The controller is connected to the inspection drone via the wireless communication module, sends the inspection drone a corresponding inspection path, and obtains the dam surface image collected by the drone.

Images