A method and device for identifying dam leakage areas based on color distance of remote sensing infrared thermal images from unmanned aerial vehicles

Abstract

A method and device for identifying dam leakage areas based on the color distance of unmanned aerial vehicle remote sensing infrared thermal images, the method comprising: (1) customizing the flight altitude, speed, path, and inclination angle of the airborne camera and other route planning parameters; (2) collecting images of the entire dam area; (3) performing distortion correction and noise reduction preprocessing operations on the initial frame remote sensing infrared thermal image; then converting the remote sensing infrared thermal image to the HSV (Hue, Saturation, Value) color space; (4) preliminarily segmenting the image into "hot spots" and "cold spots", that is, binary images; (5) automatically and intelligently judging the suspected leakage area based on the temperature field state of the target dam body; (6) extracting the temperature extreme values ​​in the binary image and calculating the extreme difference to evaluate the true value of the suspected leakage area. (7) For the binary image obtained by step (6), firstly use the image morphology processing algorithm to fill the holes and further reduce the noise; then use the area screening algorithm to adaptively eliminate the fragmented temperature areas in the binary image that do not fully match the actual temperature distribution, and obtain the "hot spots" or "cold spots" with continuous and smooth boundaries; (8) extract the subsequent remote sensing infrared thermal images frame by frame, repeat steps (3) to (7) until all images are adaptively processed; (9) use the heterogeneous image fusion algorithm based on wavelet transform to fuse the visible light image and the binary image frame by frame; then use the image stitching algorithm to stitch the fused images of each frame to construct a comprehensive image covering the entire area of ​​the target dam and locate the spatial position of the leakage area.

Description

Technical Field

[0001] The present invention relates to the technical field of dam prevention and control, and in particular to a method and device for identifying a dam leakage area based on color distance of an unmanned aerial vehicle remote sensing infrared thermal image. Background Art

[0002] Leakage is one of the most common diseases of dams. Failure to detect and deal with it in time may lead to serious consequences. Therefore, leakage detection of dams is crucial. However, the temporal and spatial randomness and concealment of leakage bring great challenges to detection. Usually, dam leakage is monitored by arranging pressure gauges and water measuring weirs, but this method is easily disturbed by surface runoff and rainfall, and it is difficult to accurately locate leakage. In order to improve the accuracy of leakage area identification results, geological radar method, parallel electrical method and isotope tracer method have received extensive attention and research. The geological radar method uses the reflection difference of electromagnetic waves in different media to reveal the spatial position and morphology of the leakage area, while the parallel electrical method distinguishes based on the conduction difference of current. Although both methods can improve the accuracy of leakage detection, they both rely on experts to interpret and analyze complex data results, which to a certain extent limits their wide application. In addition, the cumbersome sensor arrangement process does not match the needs of emergency detection. Similarly, the isotope tracer method also has the shortcomings of long result output cycle and difficult data interpretation. Therefore, the current flood season leakage detection work still relies on manual inspection, but it is difficult to achieve a quick and timely response. It can be seen that at this stage, there is still an urgent need for a simple, efficient and economical dam leakage detection method. Summary of the invention

[0003] The purpose of the present invention is to overcome the problems existing in the above-mentioned prior art and to provide a method and device for identifying dam leakage areas based on the color distance of unmanned aerial vehicle remote sensing infrared thermal images.

[0004] The invention not only realizes the instant detection and precise location of dam leakage, but also significantly improves the efficiency and cost-effectiveness of the detection process. The core of the invention lies in its adaptive nature, which eliminates the reliance on human intervention in traditional methods, while optimizing data processing and ensuring the high readability and user-friendliness of the output results.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] The first aspect of the present invention relates to a method for identifying leakage areas of dams based on color distance of infrared thermal images sensed by unmanned aerial vehicles, comprising the following steps:

[0007] (1) Taking into account the multi-dimensional factors such as the scale of the dam, the topographic features, the lighting environment, and the performance of key equipment such as the parameters of the drone and the onboard camera, the flight altitude, speed, path, and inclination angle of the onboard camera are customized.

[0008] (2) Place the drone in a flat and open area and start the drone. The drone automatically locates to the predetermined starting point and simultaneously collects remote sensing infrared thermal images and visible light images. Then, according to the preset route, it automatically flies to the subsequent designated sites and collects corresponding images at each site until the image collection task of the entire dam area is completed.

[0009] (3) Preprocessing operations such as distortion correction and noise reduction are performed on the initial frame remote sensing infrared thermal image. A pseudo-color processing algorithm is used to map the infrared thermal image from grayscale space to RGB (Red, Green, Blue) color space. The brightness V (Value) is set by the RGB extreme value, and the hue H (Hue) and saturation S (Saturation) are determined by the RGB extreme difference, and the remote sensing infrared thermal image is further converted to the HSV color space.

[0010] (4) For the HSV color space infrared thermal image obtained by step (3), the HSV color space hue channel, which is highly robust to environmental factors such as illumination, is used to calculate the HSV color histogram. The color distance is analyzed, and based on the color similarity criterion, the adaptive region growing algorithm is used to aggregate the color regions adjacent to the double peaks in the HSV color histogram, further shaping the HSV color histogram with a distinct double peak and valley structure. Subsequently, the valley hue value is used as the threshold to preliminarily segment the image into "hot spots" and "cold spots", i.e., a binary image.

[0011] (5) Based on the temperature field status of the target dam body, the suspected leakage area is automatically and intelligently judged: if the dam is in the heating stage, the "cold spot" is identified as the suspected leakage area; conversely, if the dam is in the heat release and cooling stage, the "hot spot" is identified as the suspected leakage area.

[0012] (6) For the binary image obtained by step (4), the temperature extremes in the image are extracted and the range is calculated, which is used as a basis for evaluating the authenticity of the suspected leakage area. For the misjudged foreground pixels, their labels are adjusted through the grayscale directional inversion algorithm so that these pixels are reclassified as background.

[0013] (7) For the binary image obtained by step (6), the image morphological processing algorithm is first used to fill the holes and further reduce the noise. Then, the area screening algorithm is used to adaptively eliminate the fragmented temperature areas in the binary image that do not fully match the actual temperature distribution, and obtain "hot spots" or "cold spots" with continuous and smooth boundaries.

[0014] (8) Extract subsequent remote sensing infrared thermal images frame by frame, and repeat steps (3) to (7) until all images are adaptively processed.

[0015] (9) A heterogeneous image fusion algorithm based on wavelet transform is used, with visible light images as the base layer and binary images as the surface layer, which are fused frame by frame to further eliminate possible misjudgments caused by factors such as slope water accumulation, vegetation itself and its shadows, and pit shadows. An image stitching algorithm is used to stitch the fused images of each frame to construct a comprehensive image covering the entire target dam area and locate the spatial position of the leakage area.

[0016] Furthermore, in step (1), considering that dam leakage usually leads to water seepage and pipe burst on the backwater slope, leakage detection should focus on the backwater slope of the dam. In addition, the infrared thermal camera lens needs to be set perpendicular to the dam to capture images, so as to reduce the temperature information reception error caused by path differences and ensure that the remote sensing infrared thermal image accurately reflects the actual temperature information of the dam.

[0017] Furthermore, in step (2), an equal-distance interval shooting strategy is adopted to ensure that all areas of the dam can be evenly covered. In particular, for dams with special areas (such as known leakage areas, areas with complex working conditions, etc.), an equal-time interval shooting strategy can be adopted: the drone's staged speed is preset, so that the image acquisition frequency is automatically refined during the flight process, ensuring that the temperature information of special areas is captured with optimal accuracy.

[0018] Furthermore, in step (6), the authenticity of the leakage area can be determined by comparing the temperature extreme difference with the temperature threshold. The temperature threshold can be determined by weighting based on historical exploration data, model data, experimental data, etc.

[0019] Furthermore, in step (7), the setting of the minimum screening area takes into account multi-dimensional factors such as the degree of leakage, dam size, and ground sampling distance, and is fine-tuned using a trial-and-error method.

[0020] The second aspect of the present invention relates to a device for identifying dam leakage areas based on the color distance of unmanned aerial vehicle remote sensing infrared thermal images, comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the method for identifying dam leakage areas based on the color distance of unmanned aerial vehicle remote sensing infrared thermal images of the present invention.

[0021] The third aspect of the present invention relates to a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the method of identifying dam leakage areas based on color distance of unmanned aerial vehicle remote sensing infrared thermal images of the present invention.

[0022] The embodiments of the present invention have the following advantages:

[0023] (1) For each frame of remote sensing infrared thermal image, the HSV color space is used to eliminate the influence of environmental factors such as light, and the segmentation threshold of each frame of remote sensing infrared thermal image is adaptively calculated through the HSV color histogram to identify "hot spots" or "cold spots", forming an automatic and intelligent method for identifying dam leakage areas without human intervention.

[0024] (2) Taking into account the temperature distribution characteristics of the single-frame infrared thermal image of the dam and the temperature characteristics of the leakage area, the image threshold segmentation algorithm is used to identify the "hot spots" or "cold spots". Subsequently, image morphological processing, area screening and other algorithms are used for detail processing to complete the identification of the dam leakage area, forming a dam leakage area identification method that does not require training samples, has strong readability and is highly user-friendly. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 Flow chart of the main implementation steps of this embodiment. DETAILED DESCRIPTION

[0026] The specific implementation of the present invention will be further described below through specific examples in conjunction with the accompanying drawings.

[0027] Example 1

[0028] A method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images from unmanned aerial vehicles comprises the following steps:

[0029] (1) Taking into account the multi-dimensional factors such as the scale of the dam, the topographic features, the lighting environment, and the key equipment performance such as the parameters of the drone and the parameters of the airborne camera, the route planning parameters such as the altitude, speed, path, and the inclination angle of the airborne camera are customized to ensure the efficiency of the detection process and the accuracy of the detection results. The relevant calculation formula for the route planning parameters is:

[0030]

[0031] Where W is the ground image width, H is the flight altitude, and θ is the camera field of view;

[0032]

[0033]

[0034] Among them, GSD is the ground sampling distance, f is the focal length of the camera lens, a is the size of the pixel, and h is c With w c are the height and width of the sensor, h x With w x are the height and width of effective pixels respectively;

[0035] In particular, considering that dam leakage usually causes seepage or piping on the backwater slope, the focus of drone detection should be on the backwater slope of the dam. In order to accurately capture temperature information, it is necessary to consult the target dam structure drawings and related materials before the operation to ensure that the onboard camera lens can shoot perpendicular to the dam. The use of oblique photography technology can effectively minimize the impact of path differences on the reception of temperature information, ensuring that the remote sensing infrared thermal images obtained can truly and accurately reflect the actual temperature field conditions of the dam.

[0036] (2) Carefully assess local climate conditions to prevent the drone from encountering adverse weather such as strong winds and heavy rain during operation. At the same time, conduct a comprehensive inspection of the environment around the dam to eliminate flight obstacles such as power lines, trees, birds, and protruding rocks on the mountain to ensure the safety of the equipment during the mission. Subsequently, conduct detailed functional tests and status checks on the drone body and its key components, including but not limited to precise calibration of the gyroscope and compass, confirmation of sufficient battery power, firm fastening of the wings, and evaluation of the working status of the camera and sensors to ensure the stability and reliability of the equipment during the mission.

[0037] After ensuring that the surrounding environment and equipment conditions meet the operational requirements, place the drone in a flat and open area and start the drone. After reaching the starting measurement point, the drone will simultaneously capture infrared thermal imaging and visible light imaging dual-modal image data. The drone will then automatically navigate to each scheduled inspection site based on the planned preset route and perform systematic image acquisition to ensure the continuity and consistency of image acquisition. After completing image acquisition at all designated sites, the drone will return to the starting point along the optimal return path.

[0038] (3) In view of the inherent distortion characteristics of infrared thermal imaging cameras, the use of advanced distortion correction technology is a necessary preliminary operation, the purpose of which is to correct image distortion and ensure the geometric accuracy of the image, so as to facilitate accurate comparison with visible light images and effectively eliminate false positive recognition results. In addition, considering that remote sensing infrared thermal images are easily affected by various factors such as camera resolution, jitter during drone flight, water vapor and dust in the environment, and changes in light intensity, filtering technologies such as median filtering and Gaussian filtering are used to remove noise in the image and optimize image contrast.

[0039] Considering that the grayscale nature of infrared thermal images limits their ability to intuitively display areas of abnormal temperature, a pseudo-color processing algorithm is used to convert grayscale images to RGB color space through color mapping, using different RGB colors to represent different temperatures, and indirectly present the temperature field. However, environmental factors such as light may cause problems such as color distortion and color shift in RGB color space images, resulting in a wide and flat transition band in the RGB color histogram, posing certain challenges to the accurate determination of grayscale thresholds. Unlike the RGB color space, the hue channel in the HSV color space mainly represents the color type and can better eliminate the influence of environmental factors such as light. Therefore, the infrared thermal image is further converted from the RGB color space to the HSV color space to optimize color presentation and promote color feature separation. The calculation steps for converting the RGB color space to the HSV color space are:

[0040] First, normalize the three color components of the RGB color space:

[0041]

[0042] Secondly, the calculation formulas for the three components of the HSV color space are:

[0043] max=max(R,G,B)

[0044] min=min(R,G,B)

[0045]

[0046]

[0047] V = max;

[0048] (4) Analyze the hue channel of the HSV color space infrared thermal image obtained by step (3), and calculate and generate an HSV color histogram based on the hue channel. Use peak detection technology to identify the double peaks of the HSV color histogram, namely the "hot spots" and "cold spots" in the image. Then analyze the color distance, and based on the color similarity criterion, use the adaptive region growing algorithm to aggregate the color areas adjacent to the double peaks in the HSV color histogram, and further shape the HSV color histogram with a distinct double peak and valley structure. Select the valley hue value as the threshold, and preliminarily segment the image into "hot spots" and "cold spots", that is, a binary image.

[0049] (5) Automatically and intelligently judge the suspected leakage area based on the temperature field state of the target dam body: Since the specific heat capacity of soil is much smaller than that of water (the former is about 0.84×10 3 J / kg, the latter is about 4.2×10 3J / kg), the leakage area with high water content will show a temperature field that is significantly different from the non-leakage area. In short, during the heating stage, the temperature of the leakage area rises slowly, showing low temperature characteristics, and the "cold spots" in the image are intelligently identified as suspected leakage areas; conversely, if the dam is in the heat release and cooling stage, the "hot spots" are automatically identified as suspected leakage areas.

[0050] (6) For the binary image obtained by step (4), calculate the extreme value of the temperature of its pixel point, that is, obtain the highest temperature value (hottest point) T in the image. max With the lowest temperature (coldest point) T min Then, the difference between these two temperature extremes is calculated as ΔT = T max -T min , and after weighted analysis based on historical exploration data, model data, experimental data, etc. of the target dam, the temperature threshold T is determined threshold .

[0051] According to the comparison result of the temperature extreme difference and the temperature threshold, the identification strategy is dynamically adjusted to evaluate the authenticity of the suspected leakage area in step (5). Specifically, if ΔT>T threshold , the suspected leakage area is judged to be real; otherwise, these areas are automatically identified as misjudged. For the misjudged foreground pixels, the grayscale directional inversion algorithm is used to adjust their labels to ensure that these pixels are reclassified as background.

[0052] (7) For the binary image obtained by step (6), the image morphological processing algorithm is first used to optimize the binary image: In view of the complexity of the geometric structure of the dam and the diversity of the environmental background, the binary image often faces problems such as noise pollution, regional clutter and edge discontinuity, which affect the accuracy and reliability of the interpretation of the results. The image morphological processing algorithm is an image analysis algorithm based on mathematical morphology theory, which has significant advantages in extracting the main features of the image. Among them, dilation is the process of merging all background points that are in contact with the target object into the object, which can be used to fill holes and smooth the uneven concave parts on the image boundary. In contrast, the erosion operation can eliminate bright areas smaller than the structural elements, effectively removing isolated noise points and uneven convex parts on the boundary. The calculation formulas of the two are:

[0053]

[0054] (AΘB)(x,y)=min{A(x+x',y+y')-B(x',y')|(x',y')∈B}

[0055] Among them, A is the input image, B is the structural element, is the dilation operation, Θ is the erosion operation, (x, y) is the pixel position in the image, and (x', y') is the pixel position in the structure element B;

[0056] The closed operation processing, that is, the operation strategy of expansion first and then corrosion, can be expressed as:

[0057]

[0058] By using this formula to operate on binary images, the contact surface of the "hot spot" or "cold spot" and its surrounding background points can be expanded by dilation operation, effectively filling the tiny gaps caused by factors such as image resolution limitation or uneven local moisture distribution. Subsequently, the region boundary is refined through corrosion operation, the non-target background information that may be introduced by the dilation operation is removed, the definition of the region boundary is optimized, and the accuracy and reliability of the leakage area identification results are improved.

[0059] Then, an area screening algorithm is used to optimize the binary image: Although image morphological processing can significantly optimize phenomena such as regional clutter and undesirable edges, there may still be fragmented temperature areas in the binary image that do not fully match the actual leakage situation. Area screening, as a high-precision image area screening technology, can not only effectively remove irrelevant tiny details and interference in the image and screen out fragmented areas, but also ensure that under different detection environments and conditions, the present invention can accurately identify and retain key image areas related to leakage. Specifically, by setting a minimum screening area, all connected areas in the binary image are evaluated, and only areas with an area larger than the set screening area are retained, while fragmented areas with an area smaller than the screening area are regarded as background noise and eliminated. The process can be expressed as:

[0060] A={A i |A i >Min area}

[0061] Among them, A is the set of regions after screening, A i is the area of ​​the i-th connected region in the binary image, Min area is the minimum screening area;

[0062] In particular, the setting of the minimum screening area is not static, but needs to be comprehensively considered in combination with multi-dimensional factors such as the degree of dam leakage, dam scale, ground sampling distance, etc. In order to achieve the optimal screening effect, the trial and error method is further used for fine adjustment to dynamically determine the most appropriate minimum screening area.

[0063] (8) Extract subsequent remote sensing infrared thermal images frame by frame, and repeat steps (3) to (7) until all remote sensing images are adaptively processed.

[0064] (9) The use of a heterogeneous image fusion algorithm based on wavelet transform can significantly improve recognition accuracy. Visible light images can be used to eliminate misjudgments caused by factors such as slope water accumulation, vegetation itself and its shadows, and pit shadows, thereby reducing false positive or false negative leakage detection results. By performing primary wavelet transform on the visible light image and the binary image respectively, each frame of the image will be decomposed into a low-frequency sub-band LL1 and three high-frequency sub-bands HL1, LH1, and HH1. Subsequently, a secondary wavelet transform is performed on the low-frequency sub-band LL1, which is further refined into LL2, HL2, LH2, and HH2 sub-bands. This process can be extended to three or more levels of wavelet transform to extract more detailed image features. In the heterogeneous image fusion stage, by integrating and merging the corresponding sub-bands (for example, LL of the visible light image and the binary image), the image is decomposed into a low-frequency sub-band LL1 and three high-frequency sub-bands HL1, LH1, and HH1. i The geomorphic information of the dam is integrated into the leakage segmentation result, thereby effectively eliminating the interference of external environmental factors through cross-validation with the visible light image.

[0065] Through multi-resolution fusion technology, the edges of each frame of the fused image are smoothed to optimize the transition at the seam and reduce visual differences. Subsequently, the fused frames are spliced ​​into a comprehensive image covering the entire area of ​​the target dam in order of spatial position. On the global map, the graphic annotation tool is used to automatically represent the leakage area information identified in the previous step to intuitively locate the approximate location of the leakage area.

[0066] This embodiment can use the HSV color space to eliminate the influence of environmental factors such as illumination based on a series of image data measured by drones. Based on the temperature distribution characteristics of the infrared thermal image of a single-frame earth-rock dam, the temperature characteristic law of the leakage area is used to automatically and intelligently generate image segmentation thresholds through the HSV color histogram to identify "hot spots" or "cold spots", and use image morphological processing, area screening and other algorithms for detail processing to complete the identification of earth-rock dam leakage areas. A heterogeneous image fusion algorithm based on wavelet transform is used to further eliminate misjudgments that may be caused by environmental factors, and an image stitching algorithm is used to locate the spatial position of the leakage area.

[0067] Example 2

[0068] The present embodiment relates to a device for identifying leakage areas of dams based on the color distance of remote sensing infrared thermal images of unmanned aerial vehicles (UAVs), comprising a memory and one or more processors, wherein the memory stores executable codes, and when the one or more processors execute the executable codes, they are used to implement the method for identifying leakage areas of dams based on the color distance of remote sensing infrared thermal images of unmanned aerial vehicles (UAVs) of embodiment 1.

[0069] Example 3

[0070] The present embodiment relates to a computer-readable storage medium having a program stored thereon. When the program is executed by a processor, the method for identifying dam leakage areas based on color distance of infrared thermal images sensed by unmanned aerial vehicle remote sensing of embodiment 1 is implemented.

[0071] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention.

Claims

1. A method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images from unmanned aerial vehicles, comprising the following steps: (1) Taking into account the multi-dimensional factors of dam scale, topographic features, and lighting environment, as well as the key equipment performance of drone parameters and airborne camera parameters, the flight altitude, speed, path, and airborne camera inclination angle are customized to set the route planning parameters; (2) The drone is placed in a flat and open area and started. The drone automatically locates to the predetermined starting point and simultaneously collects remote sensing infrared thermal images and visible light images. Then, according to the preset route, the drone automatically flies to the subsequent designated sites and collects corresponding images at each site until the image collection task of the entire dam area is completed. (3) Perform distortion correction and noise reduction preprocessing operations on the initial frame remote sensing infrared thermal image; use a pseudo-color processing algorithm to map the infrared thermal image from grayscale space to RGB (Red, Green, Blue) color space; set the brightness V (Value) by the RGB extreme value, and use the RGB extreme difference to determine the hue H (Hue) and saturation S (Saturation), and further convert the remote sensing infrared thermal image to the HSV color space; (4) For the HSV color space infrared thermal image obtained by step (3), the HSV color space hue channel with strong robustness to lighting environment factors is used to calculate the HSV color histogram; the color distance is analyzed, and according to the color similarity criterion, the adaptive region growing algorithm is used to aggregate the color regions adjacent to the double peaks in the HSV color histogram, and further shape the HSV color histogram with a distinct double peak and valley structure; then, the valley hue value is used as the threshold to preliminarily segment the image into "hot spots" and "cold spots", that is, a binary image; (5) Based on the temperature field state of the target dam body, the suspected leakage area is automatically and intelligently judged: if the dam is in the heating stage, the "cold spot" is identified as the suspected leakage area; conversely, if the dam is in the heat release and cooling stage, the "hot spot" is identified as the suspected leakage area; (6) For the binary image obtained by processing in step (4), the temperature extremes in the image are extracted and the range is calculated, and the authenticity of the suspected leakage area is evaluated based on this; for the misjudged foreground pixels, their labels are adjusted through the grayscale directional inversion algorithm so that these pixels are reclassified as background; (7) For the binary image obtained by step (6), firstly, an image morphological processing algorithm is used to fill the holes and further reduce the noise points; then, an area screening algorithm is used to adaptively eliminate the fragmented temperature areas in the binary image that do not fully match the actual temperature distribution, and obtain "hot spots" or "cold spots" with continuous and smooth boundaries; (8) extracting subsequent remote sensing infrared thermal images frame by frame, and repeating steps (3) to (7) until all images are adaptively processed; (9) A heterogeneous image fusion algorithm based on wavelet transform is used, with the visible light image as the base layer and the binary image as the surface layer. The images are fused frame by frame to further eliminate the misjudgment caused by factors such as slope water accumulation, vegetation itself and its shadows, and pit shadows. An image stitching algorithm is used to stitch the fused images of each frame to construct a comprehensive image covering the entire target dam and locate the spatial position of the leakage area.

2. The method for identifying dam leakage areas based on color distance of UAV remote sensing infrared thermal images as claimed in claim 1, characterized in that: In step (1), considering that dam leakage usually leads to water seepage and pipe burst on the back slope, leakage detection should focus on the back slope of the dam. In addition, the infrared thermal camera lens needs to be set perpendicular to the dam to capture images in order to reduce the temperature information reception error caused by path differences and ensure that the remote sensing infrared thermal image accurately reflects the actual temperature information of the dam.

3. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles as claimed in claim 1, characterized in that: In step (2), an equidistant interval shooting strategy is adopted to ensure that all areas of the dam can be evenly covered; in particular, for dams with special areas, such as known leakage areas and areas with complex working conditions, an equidistant interval shooting strategy is adopted: the drone's staged speed is preset to automatically refine the image acquisition frequency during the flight process, ensuring that the temperature information of special areas is captured with optimal accuracy.

4. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles as claimed in claim 1, characterized in that: The calculation steps for converting the RGB color space to the HSV color space in step (3) are: First, normalize the three color components of the RGB color space: Secondly, the calculation formulas for the three components of the HSV color space are: max=max(R,G,B) min=min(R,G,B) V=max.

5. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles as claimed in claim 1, characterized in that: In step (6), the authenticity of the leakage area is determined by comparing the temperature extreme difference with the temperature threshold; the temperature threshold is determined based on the weighted historical exploration data, model data, and experimental data.

6. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles as claimed in claim 1, characterized in that: Step (7) specifically includes: for the binary image obtained by step (6), firstly, optimizing the binary image by using an image morphological processing algorithm: dilation is the process of merging all background points in contact with the target object into the object, filling holes and smoothing the uneven concave parts on the image boundary; on the contrary, erosion operation eliminates bright areas smaller than the structural element, effectively removing isolated noise points and uneven convex parts on the boundary; the calculation formulas of the two are respectively: (AΘB)(x,y)=min{A(x+x',y+y')-B(x',y')|(x',y')∈B} Among them, A is the input image, B is the structural element, is the dilation operation, Θ is the erosion operation, (x, y) is the pixel position in the image, and (x', y') is the pixel position in the structure element B; The closed operation processing, that is, the operation strategy of expansion first and then corrosion, is expressed as: This formula is used to operate the binary image, that is, the contact surface of the "hot spot" or "cold spot" and its surrounding background points is first expanded by means of the dilation operation, effectively filling the tiny gaps caused by factors such as image resolution limitation or uneven local moisture distribution; then, the region boundary is refined by the corrosion operation, the non-target background information that may be introduced by the dilation operation is removed, the definition of the region boundary is optimized, and the accuracy and reliability of the leakage area recognition result are improved; Then, an area screening algorithm is used to optimize the binary image: a minimum screening area is set, all connected areas in the binary image are evaluated, and only areas with an area larger than the set screening area are retained, while fragment areas with an area smaller than the screening area are regarded as background noise and removed; the process is expressed as: A = {A} i |A i >Min area } Among them, A is the set of regions after screening, A i is the area of ​​the i-th connected region in the binary image, Min area is the minimum screening area.

7. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles according to claim 6, characterized in that: The setting of the minimum screening area in step (7) takes into account multiple factors including the degree of leakage, dam size, and ground sampling distance, and is fine-tuned using the trial-and-error method.

8. The method for identifying dam leakage areas based on color distance of remote sensing infrared thermal images of unmanned aerial vehicles as claimed in claim 1, characterized in that: Step (9) By performing primary wavelet transform on the visible light image and the binary image respectively, each frame image will be decomposed into a low-frequency sub-band LL1 and three high-frequency sub-bands HL1, LH1, and HH1; then, a secondary wavelet transform is performed on the low-frequency sub-band LL1, and it is further refined into LL2, HL2, LH2, and HH2 sub-bands; this process is extended to three or more levels of wavelet transform to extract more detailed image features; in the heterogeneous image fusion stage, by integrating and merging the corresponding sub-bands, the geomorphological information of the dam is integrated into the leakage segmentation result, thereby effectively eliminating the interference of external environmental factors through cross-validation with the visible light image.

9. A device for identifying dam leakage areas based on color distance of remote sensing infrared thermal images from drones, characterized in that: It includes a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, it is used to implement the method for identifying dam leakage areas based on color distance of unmanned aerial vehicle remote sensing infrared thermal images as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that: A program is stored thereon, and when the program is executed by a processor, the method for identifying dam leakage areas based on color distance of unmanned aerial vehicle remote sensing infrared thermal images as described in any one of claims 1 to 8 is implemented.

Images