Automatic acquisition method of reservoir area hydro-fluctuation belt rock crack image based on unmanned aerial vehicle technology

Abstract

The invention discloses a reservoir area hydro-fluctuation belt rock seam image automatic acquisition method based on an unmanned aerial vehicle technology, and the method comprises the steps: obtaining the geographical environment information and water level information of a reservoir area, and calculating the position distribution of a hydro-fluctuation belt in a geographical space; issuing an image acquisition instruction to the unmanned aerial vehicle through the unmanned aerial vehicle control platform based on the position distribution of the hydro-fluctuation belt; the unmanned aerial vehicle flies to a designated position based on the acquisition instruction and starts to acquire a rock image of the hydro-fluctuation belt; and analyzing the rock images to determine rock crack images in the rock images and summarizing the rock crack images to obtain a reservoir area hydro-fluctuation belt rock crack image set. According to the invention, the unmanned aerial vehicle can automatically collect the rock crack image, the workload and risk of manual on-site exploration are reduced, and compared with a traditional manual exploration method, the use of the unmanned aerial vehicle reduces the cost and improves the economic benefit.

Description

Technical Field

[0001] The invention belongs to the technical field of topographic survey, and in particular relates to an automatic acquisition method of rock crack images in a drawdown zone of a reservoir area based on unmanned aerial vehicle technology. Background Art

[0002] Due to the periodic water storage of the Three Gorges Dam reservoir, the water level in the reservoir area fluctuates greatly. In order to meet the needs of industrial and agricultural production and logistics transportation, the water level fluctuates periodically between 145m and 175m above sea level, and the cycle eventually forms the drawdown zone area. Since the Three Gorges Reservoir was filled with water, domestic and foreign experts and scholars have gradually begun to pay attention to the deformation and stability of the bank slope of the drawdown zone. The periodic water level changes make the drawdown zone area extremely active. Therefore, the continuous interaction between water and rock has intensified the loss and degradation process of the rock on the bank slope of the drawdown zone, causing local instability in the bank slope area of ​​the drawdown zone. In the process of the water level in the reservoir area, the rock mass in the bank slope of the drawdown zone has obvious damage and degradation. The cracks on the rock mass on the bank slope of the drawdown zone are the main areas of interaction between water and rock, and are an important part of the formation of the slip surface. The damage and degradation of the rock mass under the action of water level fluctuation is an irreversible process, especially the intermittent penetration into the crack area, which is also a process from the surface to the inside. Therefore, the identification and detection of rock cracks on the reservoir slope is particularly important, and it can also provide more appropriate and timely first-hand reference information for relevant geological maintenance departments in the reservoir area.

[0003] Affected by the periodic changes in the water level in the reservoir area and the limitations of current detection technology, detection data collection is often only targeted at areas above the 175m water level line. For the drawdown zone area where rock cracks are common, it is impossible to directly and effectively collect data, and then it is impossible to achieve qualitative and quantitative analysis of the deformation of the rock mass on the bank slope of the drawdown zone. In addition, the change mechanism of the drawdown zone area is complex and diverse, the change area is narrow and long, and the working environment is difficult. The slope of some drawdown zone banks is almost vertical. The traditional detection method that relies on auxiliary tools for manual visual inspection will have problems such as low efficiency, high cost and personnel safety, especially for some difficult-to-reach locations, there will be detection blind spots.

[0004] With the continuous development of UAV equipment, aerial photography and remote sensing technology, its data acquisition capabilities have received more and more attention. Due to its high efficiency, low cost, low risk and high accessibility, non-contact close-up photography technology based on UAVs has performed well in data collection on complex terrain. Using UAVs equipped with high-definition lenses to collect rock images in the drawdown zone of the Three Gorges Reservoir, rock crack information can be obtained without physical contact. This creates good prerequisites for crack recognition technology based on image processing. Using computer vision to detect and identify rock cracks in the drawdown zone has certain practical significance for the safe management of the reservoir bank slope, early detection of rock erosion and disaster warning. Summary of the invention

[0005] In view of this, the present invention proposes a method for automatically collecting rock fracture images in the drawdown zone of a reservoir area based on drone technology. The present invention can realize the automatic collection of rock fracture images by drones, reducing the workload and risks of manual on-site surveys. Compared with traditional manual survey methods, the use of drones reduces costs and improves economic benefits.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention provides a method for automatically collecting rock crack images in a drawdown zone of a reservoir area based on drone technology, comprising:

[0008] Obtain the geographical environment information and water level information of the reservoir area, and calculate the location distribution of the drawdown zone in geographical space;

[0009] Based on the location distribution of the drawdown zone, the UAV control platform sends image acquisition instructions to the UAV;

[0010] Based on the acquisition instructions, the UAV flies to the designated location and starts to collect rock images in the drawdown zone;

[0011] The rock images are analyzed to determine the existing rock fracture images and summarize them to obtain a rock fracture image set of the reservoir drawdown zone.

[0012] Preferably, obtaining geographical environment information and water level information of the reservoir area and calculating the location distribution of the drawdown zone in geographical space includes:

[0013] Scan the terrain environment near the reservoir with terrain scanning equipment in advance to obtain the terrain environment information near the reservoir;

[0014] For each terrain environment scanning point in the terrain environment information, determine its corresponding world coordinate information on the terrain scanning device, and establish a mapping relationship between the world coordinate information and the terrain environment information to obtain the geographical environment information;

[0015] Obtain the water level information of the reservoir area through the reservoir area data management platform;

[0016] Based on the water level information and geographical environment information, the three-dimensional model composed of the nearshore terrain environment is horizontally cut to obtain the drawdown zone model located above the water surface;

[0017] Based on the mapping relationship between the corresponding terrain environment information and the world coordinate information on the water-drawing zone model, the location distribution of the water-drawing zone in the geographic space is determined.

[0018] Preferably, sending an image acquisition instruction to the drone through the drone control platform based on the position distribution of the drawdown zone includes:

[0019] Determine the total number of drones that can be selected and activated within a preset range around the drawdown zone and the spatial coordinates of each drone;

[0020] The narrow ebb and flow zone is averagely segmented based on the total number of drones to obtain multiple image acquisition areas;

[0021] For any image acquisition area, determine the midpoint of the line connecting the two end points of the image acquisition area as the area coordinate;

[0022] Combine multiple drones with multiple regional coordinates in pairs to get a combination set.

[0023] Determine the straight-line distance between the spatial coordinates corresponding to each drone in the combination set and the unique regional coordinates corresponding to the drone, and sum up all straight-line distances to obtain the total distance;

[0024] Taking the shortest total distance as the optimization goal, the pairwise combination of multiple drones and multiple regional coordinates as the optimization item, and the unique correspondence between drones and regional coordinates as the boundary condition, the optimization calculation is performed by traversing all combinations to obtain the combination set corresponding to the optimal combination.

[0025] According to the combination set, the coordinates of the area corresponding to any UAV are determined as the flight destination, and an image acquisition task is issued to the UAV.

[0026] Preferably, the drone automatically flies to a designated location based on the collection instruction and starts collecting rock crack images of the drawdown zone, and the execution process includes:

[0027] Based on the acquired acquisition instructions, the aircraft can navigate and move to the designated working coordinate position, and perform automatic obstacle avoidance management during the flight process;

[0028] After reaching the working coordinate position, the relative position relationship between the drone and the water surface is detected by ultrasonic equipment, and the drone position is adjusted according to the relative position relationship and the predetermined position distribution of the drawdown zone in geographic space, so that the drone maintains a preset shooting distance from the water surface and the drawdown zone;

[0029] Use drones to take images of the ebb and flow zone, and upload the captured images to the drone control platform through the network equipment carried by the drone;

[0030] The UAV control platform inputs the uploaded image into a pre-trained neural network for rock identification, obtains the recognition result of the rock on the image, and performs image segmentation to obtain the rock image.

[0031] Preferably, after obtaining the rock recognition result on the image through the UAV control platform, the following operations are also performed:

[0032] The position of the rocks on the drawdown zone is located and marked according to the shooting position corresponding to the image, and a close-up shooting instruction is sent to the drone;

[0033] After receiving the close-up shooting command, the UAV takes a full-range close-up shot of the rock position based on the positioning and marking results to obtain a rock image corresponding to the rock at a specific location.

[0034] Preferably, the following steps are performed in the process of analyzing the rock images in detail to determine the rock fracture images existing therein and summarizing them to obtain the rock fracture image set of the reservoir drawdown zone:

[0035] The target pixel in the rock image is replaced by the median value of its surrounding pixels to achieve image noise reduction;

[0036] Linear contrast enhancement method is used to enhance the denoised rock image;

[0037] Screen the rock images, remove the images with high blur, and obtain clear images with obvious cracks;

[0038] The screened images are cropped to obtain multiple images of uniform size, and the cropped images are analyzed to determine the rock fracture images that exist therein and summarize them to obtain a rock fracture image set of the reservoir drawdown zone.

[0039] The present invention has achieved at least the following beneficial effects:

[0040] 1. The automated collection of rock fracture images by drones reduces the workload and risks of manual on-site surveys. Compared with traditional manual survey methods, the use of drones reduces costs and improves economic benefits.

[0041] 2. UAVs can quickly cover large areas, effectively improve the efficiency of data collection, complete image collection tasks in a short time, and provide real-time or near real-time data support for dynamically changing ebb and flow zones.

[0042] Other advantages, objectives and features of the present invention will be described in the following description and will be apparent to those skilled in the art to some extent, or those skilled in the art may be taught from the practice of the present invention. The objectives and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to make the purpose, technical solution and beneficial effects of the present invention clearer, the present invention provides the following drawings for illustration:

[0044] Figure 1It is a flowchart of the steps of a method for automatically collecting rock crack images of a drawdown zone in a reservoir area based on drone technology in an embodiment of the present invention;

[0045] Figure 2 A schematic diagram of the flight path of a drone when collecting rock surface images in an embodiment of the present invention;

[0046] Figure 3 This is a comparison diagram of Gaussian filtering denoising effects of different neighborhood windows in an embodiment of the present invention;

[0047] Figure 4 This is a comparison diagram of mean filtering denoising effects of different neighborhood windows in an embodiment of the present invention;

[0048] Figure 5 A comparison diagram of median filtering denoising effects of different neighborhood windows in an embodiment of the present invention;

[0049] Figure 6 A comparison diagram of different image enhancement effects in the embodiments of the present invention;

[0050] Figure 7 Schematic diagram of crack image samples in an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0052] The present invention provides an automatic acquisition method of rock crack images in the reservoir area based on drone technology, referring to Figure 1 ,include:

[0053] Obtain the geographical environment information and water level information of the reservoir area, and calculate the location distribution of the drawdown zone in geographical space;

[0054] Based on the location distribution of the drawdown zone, the UAV control platform sends image acquisition instructions to the UAV;

[0055] Based on the acquisition instructions, the UAV flies to the designated location and starts to collect rock images in the drawdown zone;

[0056] The rock images are analyzed to determine the existing rock fracture images and summarize them to obtain a rock fracture image set of the reservoir drawdown zone.

[0057] The working principle and beneficial effects of the above technical solution are as follows: by obtaining the geographical environment information of the reservoir area, the specific geographical location and environmental characteristics of the drawdown zone can be clarified, including terrain undulations, landform types, vegetation conditions, etc., which is crucial for the subsequent UAV flight route planning and image acquisition area selection. Water level information is a key factor in determining the location of the drawdown zone. By analyzing the changes in water levels, the location and range of the drawdown zone under different water level conditions can be determined, which is convenient for setting the flight altitude and shooting angle of the UAV. Combining geographical environment information and water level information, the precise location distribution of the drawdown zone in geographic space can be calculated. This step ensures that the UAV can fly to the correct area to collect rock fracture images. Accurate drawdown zone location distribution information can help the UAV quickly locate the target area, avoiding invalid searches and flights, thereby improving the efficiency and accuracy of image acquisition. Based on the location distribution of the drawdown zone, an image acquisition instruction is sent to the drone through the drone control platform. Based on the acquisition instruction, the drone flies to the specified location on its own and starts to collect rock images of the drawdown zone. Finally, the rock images are analyzed to determine the rock crack images existing therein and summarize them to obtain a rock crack image set of the drawdown zone in the reservoir area. The present invention reduces the workload and risk of manual on-site investigation by automatically collecting images through drones. The drone can quickly cover a large area and can effectively improve the efficiency of data collection. The drone can complete the image acquisition task in a short time and can provide real-time or near real-time data support for dynamically changing drawdown zone areas. Compared with traditional manual exploration methods, the use of drones reduces costs and improves economic benefits.

[0058] In a preferred embodiment, obtaining geographical environment information and water level information of the reservoir area and calculating the location distribution of the drawdown zone in geographical space includes:

[0059] Scan the terrain environment near the reservoir with terrain scanning equipment in advance to obtain the terrain environment information near the reservoir;

[0060] For each terrain environment scanning point in the terrain environment information, determine its corresponding world coordinate information on the terrain scanning device, and establish a mapping relationship between the world coordinate information and the terrain environment information to obtain the geographical environment information;

[0061] Obtain the water level information of the reservoir area through the reservoir area data management platform;

[0062] Based on the water level information and geographical environment information, the three-dimensional model composed of the nearshore terrain environment is horizontally cut to obtain the drawdown zone model located above the water surface;

[0063] Based on the mapping relationship between the corresponding terrain environment information and the world coordinate information on the water-drawing zone model, the location distribution of the water-drawing zone in the geographic space is determined.

[0064] The working principle and beneficial effects of the above technical solution are as follows: through terrain scanning and coordinate mapping, high-precision positioning of the drawdown zone can be achieved, ensuring that the drone can accurately reach the target area when collecting images. Using the water level information obtained by the reservoir data management platform, the drawdown zone model can be updated in real time to adapt to the impact of water level changes and improve the timeliness and accuracy of the data. Through the construction and cutting of the three-dimensional model, the spatial distribution of the drawdown zone can be intuitively displayed, which is convenient for researchers and managers to analyze and make decisions. Clear information on the location distribution of the drawdown zone can greatly improve the image acquisition efficiency of the drone and reduce unnecessary flights and time waste. Accurate drawdown zone location data provides important basic data for subsequent rock fracture analysis, geological stability assessment and environmental monitoring, supporting scientific research and decision-making. The automated data acquisition and processing process reduces manual intervention, reduces labor costs and improves work efficiency. By establishing a comprehensive management platform for geographic environment information and water level information, information sharing and utilization are promoted, which helps collaborative work among departments.

[0065] In a preferred embodiment, sending an image acquisition instruction to the drone through the drone control platform based on the position distribution of the drawdown zone includes:

[0066] Determine the total number of drones that can be selected and activated within a preset range around the drawdown zone and the spatial coordinates of each drone;

[0067] The narrow ebb and flow zone is averagely segmented based on the total number of drones to obtain multiple image acquisition areas;

[0068] For any image acquisition area, the midpoint of the line connecting the two endpoints of the image acquisition area is determined as the regional coordinate; in order to shorten the image acquisition path, the two endpoints of the image acquisition area are respectively the starting point or the ending point, and the midpoint of the line connecting the two endpoints of the image acquisition area is used as the regional coordinate. The drone placement point, the starting point, and the ending point together form a triangular path. The line connecting the regional coordinate and the drone placement point is the midline of the triangle, which can reflect the length of the drone's total working path.

[0069] Combine multiple drones with multiple regional coordinates in pairs to get a combination set.

[0070] Determine the straight-line distance between the spatial coordinates corresponding to each drone in the combination set and the unique regional coordinates corresponding to the drone, and sum up all straight-line distances to obtain the total distance;

[0071] Taking the shortest total distance as the optimization goal, the pairwise combination of multiple drones and multiple regional coordinates as the optimization item, and the unique correspondence between drones and regional coordinates as the boundary condition, the optimization calculation is performed by traversing all combinations to obtain the combination set corresponding to the optimal combination.

[0072] According to the combination set, the coordinates of the area corresponding to any UAV are determined as the flight destination, and an image acquisition task is issued to the UAV.

[0073] The working principle and beneficial effects of the above technical solution are as follows: by averaging segmentation of the narrow ebb and flow zone based on the total number of drones, it is convenient to allocate work content according to the number of enabled drones, thereby effectively improving the speed of collection work, averaging segmentation of the ebb and flow zone, ensuring that the entire area can be covered by drones, and avoiding collection blind spots. By optimizing the flight path of the drone, unnecessary flights are reduced, thereby reducing energy consumption and improving the endurance of the drone. By dynamically allocating drones to different collection areas, the flexibility of the task is improved, and the tasks of the drone can be quickly adjusted according to the actual situation of the drone's location. The optimization calculation ensures the reasonable allocation of drone resources and avoids the situation where some drones are overloaded while other drones are idle. Since the flight path and collection area of ​​the drone are optimized, the image collection time is shortened, so that data processing and analysis can be performed faster. Through intelligent optimization calculation and task allocation, this technical solution realizes the efficient, comprehensive, energy-saving and flexible execution of drone image collection tasks, which has a significant promoting effect on improving the quality and efficiency of rock crack image collection in the ebb and flow zone of the reservoir area.

[0074] In a specific implementation process, the present invention uses a DJI M300 RTK drone equipped with a Zenmuse P1 gimbal camera as an image acquisition device.

[0075] The M300 RTK DJI drone has the following advantages:

[0076] It is easy to carry, assemble and disassemble, can ensure stable hovering when shooting close up, and its endurance can meet the needs of a single mission;

[0077] Equipped with an automatic obstacle avoidance system based on high-precision sensors to ensure safety during close-flight;

[0078] It has real-time laser ranging function, which can measure the distance to the subject while taking images;

[0079] Possess certain water resistance, wind resistance and earthquake resistance;

[0080] It can be equipped with the Zenmuse P1 gimbal camera that can shoot high-definition and high-resolution images. The technical parameters of the M300 RTK drone are as follows:

[0081]

[0082] Zenmuse P1 gimbal and camera technical parameters are as follows:

[0083]

[0084]

[0085] When using a drone to photograph the rock surface, first locate the rock area of ​​the bank slope in the drawdown zone of the study section and scan the surrounding flight environment, and use automatic obstacle avoidance technology to ensure that the acquisition equipment can shoot under a certain safety. Considering that the acquisition task of the entire rock surface requires comprehensive and detailed image acquisition from the bottom of the rock surface to the top of the rock surface. Based on the acquisition requirements, the present invention adopts a "bow"-shaped flight path to take close shots of the determined rock surface, referring to Figure 2 , if the drone has any safety or low battery problems, it will automatically return. In addition, set appropriate relative altitude and absolute altitude. When performing the collection task, maintain a collection distance of 5m from the collection surface through the drone controller, and take comprehensive photos according to the preset flight path. In case of narrow or dangerous areas, a request message should be sent when necessary to request manual shooting to ensure that no important rock surfaces are missed and safety is guaranteed. Secondly, ensure that the camera lens is parallel to the collection surface, and the overlap rate of the drone's heading and lateral direction is set to 50% during collection. In case of special circumstances, appropriate adjustments need to be made to ensure the quality of the collected images, so as to ensure that the collection task covers the entire study section until the shooting is completed.

[0086] During the implementation of the present invention, the resolution of each image is 8192×5460. High-resolution images help to accurately identify and extract the geometric representation of rock fractures, and serve as a basic data set for sample production, providing rich visual data for the construction of subsequent deep learning models. When the drone performs a collection task, it automatically saves the high-precision three-dimensional coordinate information of the collection location to ensure that each image matches its exact location in the actual terrain, and can be further three-dimensionally modeled as needed.

[0087] In a preferred embodiment, the drone automatically flies to a designated location based on a collection instruction to start collecting rock crack images of the drawdown zone, and the execution process includes:

[0088] Based on the acquired acquisition instructions, the aircraft can navigate and move to the designated working coordinate position, and perform automatic obstacle avoidance management during the flight process;

[0089] After reaching the working coordinate position, the relative position relationship between the drone and the water surface is detected by ultrasonic equipment, and the drone position is adjusted according to the relative position relationship and the predetermined position distribution of the drawdown zone in geographic space, so that the drone maintains a preset shooting distance from the water surface and the drawdown zone;

[0090] Use drones to take images of the ebb and flow zone, and upload the captured images to the drone control platform through the network equipment carried by the drone;

[0091] The UAV control platform inputs the uploaded image into a pre-trained neural network for rock identification, obtains the recognition result of the rock on the image, and performs image segmentation to obtain the rock image.

[0092] The working principle and beneficial effects of the above technical solution are as follows: when the drone is performing the shooting task, through precise position detection and adjustment, the drone can maintain the best shooting distance, thereby obtaining high-definition rock crack images. Real-time image uploading and processing capabilities enable data collection and analysis to be carried out in parallel, improving work efficiency. The use of deep learning models for image recognition and segmentation greatly improves the automation and accuracy of lithology identification. The application of the automatic obstacle avoidance system improves the flight safety of drones in complex environments and reduces the risk of accidents. Through automated flight and acquisition processes, the utilization of human resources is optimized and the reliance on manual operations is reduced. High-quality rock crack images and accurate lithology identification results provide valuable data support for geological research and analysis. This technical solution realizes the efficient collection and analysis of rock crack images in the drawdown zone through automated flight control, precise position detection, real-time data processing and intelligent image recognition, which has important application value for geological monitoring and research.

[0093] In a preferred embodiment, after obtaining the rock recognition result on the image through the UAV control platform, the following operations are performed:

[0094] The position of the rocks on the drawdown zone is located and marked according to the shooting position corresponding to the image, and a close-up shooting instruction is sent to the drone;

[0095] After receiving the close-up shooting command, the UAV takes a full-range close-up shot of the rock position based on the positioning and marking results to obtain a rock image corresponding to the rock at a specific location.

[0096] The working principle and beneficial effects of the above technical solution are as follows: by shooting at close range, clearer and more detailed rock images can be obtained, which helps to improve the accuracy and reliability of subsequent analysis. High-resolution images can better show the microstructure and crack characteristics of rocks, which are of great value for geological analysis and evaluation. Accurate rock location data and detailed image information can provide more comprehensive data support for geological engineers and researchers, helping them make more accurate decisions. Automated rock positioning and close-range shooting command sending reduce the need for manual operation and improve work efficiency. The close-range shooting of drones reduces the need for personnel to work in dangerous or difficult-to-reach areas and reduces operational risks. This technical solution supports dynamic monitoring of rocks on the drawdown zone, and can collect and update images regularly or as needed to monitor rock changes and potential geological disasters. This technical solution realizes accurate monitoring and detailed analysis of rocks on the drawdown zone through automated rock positioning, labeling and close-range shooting processes, which has important application value in geological research, disaster prevention and environmental monitoring.

[0097] In a preferred embodiment, the following steps are performed in the process of analyzing rock images to determine the rock fracture images existing therein and summarizing them to obtain a rock fracture image set of the reservoir drawdown zone:

[0098] The target pixel in the rock image is replaced by the median value of its surrounding pixels to achieve image noise reduction;

[0099] Linear contrast enhancement method is used to enhance the denoised rock image;

[0100] Screen the rock images, remove the images with high blur, and obtain clear images with obvious cracks;

[0101] Crop the filtered images to get multiple images of uniform size, refer to Figure 7 , and analyze the cropped images to determine the existing rock fracture images and summarize them to obtain the rock fracture image set of the reservoir drawdown zone.

[0102] During the process of drone image acquisition, there may be strong winds or rugged terrain, and it is inevitable to encounter the problem of flight vibration. In this specific environment, it may also be affected by electromagnetic interference such as high-voltage power lines and base station signals, thereby interfering with the stability and image quality of image transmission, thereby reducing the accuracy of crack image extraction. Therefore, it is necessary to denoise the collected images. The present invention analyzes the denoising effect of the collected images with added noise through different image denoising techniques. Currently, the commonly used image filtering and denoising methods include Gaussian filtering, mean filtering and median filtering. By comparing the experimental analysis and evaluation of the denoising performance and effect of the above-mentioned commonly used image denoising methods in rock images, the most suitable method for rock crack image denoising is determined.

[0103] As a denoising technology for smoothing images, Gaussian filtering is dedicated to reducing Gaussian noise and volume noise in images. The value of each point in the image is replaced by the weighted average of the surrounding points, which can effectively remove image noise while maintaining the edge characteristics of the image. Gaussian filters of different sizes have different smoothing effects on the image. The value represents the value of the Gaussian filter kernel in the X direction. This value is usually related to the size of the neighborhood window because it determines the distribution of weights. During the implementation of the present invention, three neighborhood windows of different sizes of 33, 55 and 77 are used to process the captured images. When set to 0, the OpenCV package in Python will automatically calculate the appropriate value based on the size of the kernel, refer to Figure 3 The results show that with the increase of the domain window size, the image noise of the Gaussian filter denoising method is reduced or even eliminated. However, with the increase of the neighborhood window, the value of also increases. In the process of gradually eliminating noise, it is inevitable to blur the image and lose the detailed information of the image, which will reduce the pixel-level extraction accuracy of rock cracks.

[0104] Mean filtering is a simple and practical linear filtering algorithm that removes noise by replacing the pixel value with the average value of all pixel values ​​in its neighborhood. This method is suitable for removing random noise, but it may also reduce the sharpness of the image. Its calculation process can be regarded as a step-by-step convolution of the image pixels to achieve the purpose of denoising. Similarly, different image denoising effects can be obtained by taking different values ​​of the neighborhood window size. In the implementation process of the present invention, three different sizes of domain windows, 3×3, 5×5 and 7×7, are used to filter and denoise the collected image. Figure 4 ,The results show that as the size of the neighborhood window increases, ,the noise of the UAV image can be effectively reduced and eliminated.,Denoising also causes image distortion, which will increase the difficulty of identifying cracks and reduce the ,extraction efficiency to a certain extent.

[0105] Median filtering is widely used in the field of image denoising and is a classic nonlinear filtering algorithm. Denoising is achieved by replacing the target pixel with the median of its surrounding pixel values. In the denoising experiment implemented in the present invention, the denoising effect of the median filtering method under different neighborhood window sizes was summarized and analyzed. Specifically, denoising was performed by using neighborhood windows of three different sizes: 3×3, 5×5, and 7×7. The choice of window size is crucial to the denoising effect and the retention of image details. The goal of this part of the experiment is to evaluate the impact of different window sizes on the denoising effect and find the optimal window size setting. Figure 5 The results show that the denoising effect of the 5×5 window is particularly significant, especially for dense noise points. However, it blurs the details of the image to a certain extent. The median filter of the 7×7 window suppresses noise very effectively, but the larger window size leads to the loss of image details and image distortion, especially for cracks that occupy a small area in the rock image. The 3×3 neighborhood window retains the details and clarity of the image as much as possible while denoising.

[0106] In summary, the use of median filtering in different sizes of neighborhood windows shows a trade-off between denoising effect and detail preservation. Although a larger neighborhood window size can remove noise more effectively, a 3×3 neighborhood window can achieve good denoising effect while retaining image details, and is more suitable for rock fracture identification that requires fine extraction.

[0107] After the rock images collected by the drone are subjected to median filtering and denoising, in order to further improve the accuracy and efficiency of rock crack detection, the contrast between the cracks and the surrounding background is made more obvious by applying image enhancement technology. Currently, the commonly used image enhancement algorithms include histogram equalization, contrast-limited adaptive histogram equalization (CLAHE) and linear contrast enhancement. In the implementation process of the present invention, the above three methods are used to perform image enhancement processing on the denoised images, referring to Figure 6 The results show that histogram equalization can enhance the image details by changing the image contrast, making the overall image clearer and more visible, but it over-enhances the background noise or irrelevant surface texture in the rock image containing cracks. Contrast-limited adaptive histogram equalization is suitable for local image enhancement, and histogram equalization is applied in a smaller area while reducing noise amplification. However, over-enhancing irrelevant rock surfaces will reduce the accuracy of crack detection, and too many elements will lead to decreased efficiency. By using simple and practical linear contrast enhancement to achieve a saturation effect on some areas of the rock image, the distinction between cracks and background is higher, which meets the research requirements of this paper. In summary, this paper uses the linear contrast enhancement method to enhance the acquired image after denoising.

[0108] The working principle and beneficial effects of the above technical solution are as follows: the present invention performs filtering, denoising and image enhancement preprocessing on the rock images collected by the drone, compares and analyzes the preprocessing results of different methods on the collected images, and finally selects the median filter and linear contrast enhancement methods to preprocess the collected images. Subsequently, the original rock crack image data set that meets the requirements is obtained through image cropping and screening. Subsequently, the image annotation software can be used to construct the crack segmentation model data set, and the data set can be expanded by using the data enhancement method, so as to obtain rock crack image samples that are convenient for directional neural network training.

[0109] Finally, it should be noted that the above preferred embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail through the above preferred embodiments, those skilled in the art should understand that various changes can be made in form and details without departing from the scope defined by the claims of the present invention.

Claims

1. A method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology, characterized in that: include: Obtain the geographical environment information and water level information of the reservoir area, and calculate the location distribution of the drawdown zone in geographical space; Based on the location distribution of the drawdown zone, the UAV control platform sends image acquisition instructions to the UAV; Based on the acquisition instructions, the UAV flies to the designated location and starts to collect rock images in the drawdown zone; The rock images are analyzed to determine the existing rock fracture images and summarize them to obtain a rock fracture image set of the reservoir drawdown zone.

2. The method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology according to claim 1 is characterized in that: The obtaining of geographical environment information and water level information of the reservoir area and calculating the geographical spatial distribution of the drawdown zone includes: Scan the terrain environment near the reservoir with terrain scanning equipment in advance to obtain the terrain environment information near the reservoir; For each terrain environment scanning point in the terrain environment information, determine its corresponding world coordinate information on the terrain scanning device, and establish a mapping relationship between the world coordinate information and the terrain environment information to obtain the geographical environment information; Obtain the water level information of the reservoir area through the reservoir area data management platform; Based on the water level information and geographical environment information, the three-dimensional model composed of the nearshore terrain environment is horizontally cut to obtain the drawdown zone model located above the water surface; Based on the mapping relationship between the corresponding terrain environment information and the world coordinate information on the water-drawing zone model, the location distribution of the water-drawing zone in the geographic space is determined.

3. The method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology according to claim 1 is characterized in that: The sending of image acquisition instructions to the drone through the drone control platform based on the position distribution of the water-falling zone includes: Determine the total number of drones that can be selected and activated within a preset range around the drawdown zone and the spatial coordinates of each drone; The narrow ebb and flow zone is averagely segmented based on the total number of drones to obtain multiple image acquisition areas; For any image acquisition area, the midpoint of the line connecting the two end points of the image acquisition area is determined as the area coordinate; Combine multiple drones with multiple regional coordinates in pairs to get a combination set. Determine the straight-line distance between the spatial coordinates corresponding to each drone in the combination set and the unique regional coordinates corresponding to the drone, and sum up all straight-line distances to obtain the total distance; Taking the shortest total distance as the optimization goal, the pairwise combination of multiple drones and multiple regional coordinates as the optimization item, and the unique correspondence between drones and regional coordinates as the boundary condition, the optimization calculation is performed by traversing all combinations to obtain the combination set corresponding to the optimal combination. According to the combination set, the coordinates of the area corresponding to any UAV are determined as the flight destination, and an image acquisition task is issued to the UAV.

4. The method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology according to claim 1 is characterized in that: Based on the acquisition instruction, the UAV flies to the designated location and starts to collect rock crack images of the drawdown zone. The execution process includes: Based on the acquired acquisition instructions, the aircraft can navigate and move to the designated working coordinate position, and perform automatic obstacle avoidance management during the flight process; After reaching the working coordinate position, the relative position relationship between the drone and the water surface is detected by ultrasonic equipment, and the drone position is adjusted according to the relative position relationship and the predetermined position distribution of the drawdown zone in geographic space, so that the drone maintains a preset shooting distance from the water surface and the drawdown zone; Use drones to take images of the ebb and flow zone, and upload the captured images to the drone control platform through the network equipment carried by the drone; The UAV control platform inputs the uploaded image into a pre-trained neural network for rock identification, obtains the recognition result of the rock on the image, and performs image segmentation to obtain the rock image.

5. The method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology according to claim 4 is characterized in that: After obtaining the rock recognition results on the image through the UAV control platform, the following operations are performed: The position of the rocks on the drawdown zone is located and marked according to the shooting position corresponding to the image, and a close-up shooting instruction is sent to the drone; After receiving the close-up shooting command, the UAV takes a full-range close-up shot of the rock position based on the positioning and marking results to obtain a rock image corresponding to the rock at a specific location.

6. The method for automatically collecting rock crack images in the drawdown zone of a reservoir area based on drone technology according to claim 1 is characterized in that: The following steps are performed in the process of analyzing the rock images to determine the existing rock fracture images and summarizing them to obtain the rock fracture image set of the reservoir drawdown zone: The target pixel in the rock image is replaced by the median value of its surrounding pixels to achieve image noise reduction; Linear contrast enhancement method is used to enhance the denoised rock image; Screen the rock images, remove the images with high blur, and obtain clear images with obvious cracks; The screened images are cropped to obtain multiple images of uniform size, and the cropped images are analyzed to determine the rock fracture images that exist therein and summarize them to obtain a rock fracture image set of the reservoir drawdown zone.

Images