Water surface area identification method and system
By using contrast extraction based on GLCM to determine the water surface area, the problem of drone positioning failure on calm water surfaces was solved, achieving efficient and robust water surface area recognition and improving recognition accuracy and generalization ability.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-26
- Publication Date
- 2026-03-13
AI Technical Summary
When drones fly over calm water surfaces, the lack of obvious textures and feature points on the water surface causes the visual system to fail in positioning, making it impossible to accurately determine the direction and distance of movement. This may lead to safety accidents such as loss of flight attitude control. Existing technologies for fusing water body indices and texture features suffer from poor algorithm generalization ability, efficiency bottlenecks, and insufficient robustness.
By using the contrast extracted based on GLCM and the total contrast at different distances as the judgment benchmark, the water surface area is identified, avoiding the fusion of water index and texture features. An integrated feature extraction mechanism is adopted, and the calculation process is simple and efficient.
It significantly improves the ability and accuracy of water surface area recognition under different geographical environments and imaging conditions, enhances the robustness of the algorithm, eliminates the influence of heterogeneous feature conflicts, avoids repeated parameter tuning, and achieves efficient water surface area recognition.
Smart Images
Figure CN121661542A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to a method and system for water surface area recognition. Background Technology
[0002] The core principle of a drone's visual positioning system is to calculate its own trajectory and position information by recognizing feature points such as patterns, textures, and color changes on the ground. However, when a drone flies over calm water (especially open water), the water surface lacks obvious textures and feature points, presenting a "uniform and indistinguishable" state. This causes the drone's visual system to be unable to obtain effective positioning references, leading to positioning failure. This manifests as the drone "getting lost," unable to accurately determine its own direction and distance of movement, and in severe cases, may lead to loss of flight control, crashes, and other safety accidents.
[0003] Existing technologies include techniques for extracting texture features from images based on the Gray-Level Co-occurrence Matrix (GLCM) and then weighted fusion with a water body index matrix to achieve water body identification, such as the technical solution in application number 2024118907074, entitled "A Rapid Water Body Extraction Method from Remote Sensing Images Based on Multi-Scale GLCM Feature Fusion". However, this method requires fusing water body indices with texture features, which leads to the following drawbacks:
[0004] First, in the above-mentioned schemes, the fusion weights of water index and texture features rely on manual parameter tuning based on experience, resulting in poor generalization ability of the algorithm. Furthermore, the parameters need to be readjusted after each scene change, impacting the workflow. Second, the remote sensing images and textures used in the above schemes are heterogeneous features, which are prone to response conflicts in complex areas such as shadows and turbid water bodies. This can actually amplify false positives and false negatives after fusion. Finally, the sequential superposition of water index and texture feature calculations in the above schemes creates a significant efficiency bottleneck, severely restricting the rapid processing capability of large-scale, high-resolution images. Therefore, the existing technologies mentioned above lack robustness and are difficult to meet the automation and universality requirements of practical applications. Summary of the Invention
[0005] The purpose of this invention is to provide a water surface area identification method and system, which can significantly improve the water surface area identification capability and accuracy under different geographical environments and imaging conditions by using the contrast extracted based on GLCM and further using the total contrast at different distances as the judgment benchmark, while enhancing the robustness of the algorithm to different water surface states.
[0006] To achieve the above objectives, the present invention provides the following technical solution:
[0007] On the one hand, a method for identifying water surface areas is provided, which includes the following steps:
[0008] Acquire scene grayscale images using imaging devices;
[0009] The grayscale image of the scene is divided into M. N image patches are obtained, and the GLCM features of each image patch are extracted. Based on the GLCM features of each image patch, the contrast of each image patch under different directions and distances is obtained.
[0010] Obtain the total contrast Contrast(d) for each image patch at each distance d. n );
[0011] Based on total contrast Contrast(d) n The region determination result (Score) for each image block is obtained.
[0012] Furthermore, the region determination result (Score) for each image patch is compared with the water surface contrast threshold (Contrast). s Compare the results; if the region determination result Score of the current image patch is less than Contrast... s If the region determination result of the current image patch is Score ≥ Contrast, then the image patch is considered to be a water surface region. s If the image block is not a water surface area, then it is considered to be a non-water surface area.
[0013] Preferably, the imaging device is mounted on a drone.
[0014] Preferably, the scene grayscale image is divided into M... The process involves N image patches, extracting GLCM features from each patch, and obtaining the contrast of each patch at different directions and distances based on its GLCM features. The steps include:
[0015] The parameters of GLCM are defined as P(i,j|d,θ), where i and j are the gray values of the first and second pixels in the pixel pair of the current image block, respectively; d is the distance between the first and second pixels in the pixel pair; and θ is the direction.
[0016] Construct an initial GLCM and iterate through the pixel pairs in the current image patch to extract all GLCMs P(i,j|d,θ) with a specified distance d and direction θ.
[0017] Normalize all GLCMs of the current image patch to obtain the GLCM probability matrix P(i,j);
[0018] Calculate the contrast Contrast of the current image patch at different distances and directions based on the GLCM probability matrix P(i,j);
[0019] Repeat the above steps to obtain the contrast (Contrast) of each image patch at different distances and directions.
[0020] Preferably, the total contrast Contrast(d) of each image patch at each distance d is obtained. n The process includes the following steps:
[0021] Determine the main direction angle of the current image patch, main_angle;
[0022] Calculate the k-th distance d using the following formula. k Below, the angle difference Δθ between each direction θ and the main direction angle main_angle. k And obtain the k-th distance d k Below, the contrast weights for each direction:
[0023] Δθ k =min(|θ-main_angle|, 180-|θ-main_angle|)
[0024] Calculate the k-th distance d k The contrast-weighted cumulative value in all directions is used as the k-th distance d of the current image patch. k Total contrast (d) k ):
[0025]
[0026] Where i represents the k-th distance d k The i-th direction below; Contrast i For the k-th distance d k Below, the contrast in the i-th direction; weight i For the k-th distance d k Below, the contrast weight in the i-th direction;
[0027] Repeat the above steps to obtain the total contrast Contrast(d) of the current image patch at each distance d. n ), and k=1,2,...,n, and the total contrast Contrast(d) of each image patch at each distance d. n ).
[0028] Preferably, the k-th distance d k The following steps are involved in obtaining the contrast weight for each direction:
[0029] The original contrast weighting coefficient weight0 for each direction is calculated using the following formula:
[0030] weight0 = cos(Δθ) π / 180)
[0031] Furthermore, the original contrast weighting coefficients weight0 in all directions are normalized to obtain the k-th distance d. k Below, the contrast weight in each direction.
[0032] Preferably, based on total contrast ratio Contrast(d) n To obtain the region determination result (Score) for each image patch, the following steps are included:
[0033] Contrast(d) represents the total contrast of the same image patch at all distances d. n The weighted sum is then used to obtain the region determination result (Score) for the image patch.
[0034] Repeat the above steps until you obtain the region determination result (Score) for each image block.
[0035] Preferably, if a grayscale image of the scene below the drone is acquired in real time based on the imaging device on board the drone, then the distance d is calculated based on the drone's current flight altitude H and the ground sampling distance:
[0036]
[0037] Among them, L real For the preset physical scale.
[0038] Preferably, the water surface area identification method further includes:
[0039] The total number T of image patches in the water surface area is counted, and according to α=T / (M) N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as the water surface area.
[0040] Preferably, the water surface area threshold value ranges from [0.70, 0.99].
[0041] On the other hand, a water surface area identification system is also provided, which includes:
[0042] Imaging equipment used to acquire grayscale images of a scene in real time;
[0043] The image segmentation module is used to divide the scene grayscale image into M... N image patches;
[0044] The feature extraction module is used to extract GLCM features for each image patch;
[0045] The contrast calculation module obtains the contrast (Contrast) of each image patch at different directions and distances based on the GLCM features of each patch, and also obtains the total contrast (Contrast(d)) of each image patch at each distance d. n );
[0046] The image patch region determination module calculates the region determination result (Score) for each image patch and compares the Score of each image patch with the water surface contrast threshold (Contrast). s Compare the images and determine whether the current image patch is a water surface area based on the comparison results.
[0047] And, the overall region judgment module, which is used to count the total number T of image patches in the water surface region, and according to α=T / (M N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as the water surface area.
[0048] In summary, the present invention has the following advantages compared with the prior art:
[0049] This invention significantly improves the ability and accuracy of water surface area recognition under different geographical environments and imaging conditions by using the contrast extracted based on GLCM and further using the total contrast at different distances as the judgment benchmark. It also enhances the robustness of the algorithm to different water surface states (such as calm water, rippled water, turbid water, etc.). Since there is no need to fuse water indexes with texture features, it solves the problem of poor generalization caused by manually designed indices and subjectively set fusion weights. This significantly improves the ability and accuracy of water surface area recognition under different geographical environments and imaging conditions, fundamentally eliminates the adverse effects of heterogeneous feature conflicts on recognition accuracy, and eliminates the need for repeated parameter tuning, making the entire calculation process more efficient and simpler. Attached Figure Description
[0050] Figure 1 This is a flowchart illustrating the steps of the UAV-based water surface area identification method in this invention.
[0051] Figure 2 This is a grayscale image of the scene identified as a water surface area in this invention, along with its region determination result (Score).
[0052] Figure 3 This is a grayscale image of a scene identified as a non-water surface area in this invention, along with its region determination result (Score).
[0053] Figure 4 This is a schematic diagram of the water surface area identification system in this invention. Detailed Implementation
[0054] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.
[0055] Example 1
[0056] Based on this, such as Figure 1 As shown, this embodiment provides a method for water surface area identification based on unmanned aerial vehicles (UAVs), which includes the following steps:
[0057] S1. When the drone is flying, the scene grayscale image of the area below the drone is acquired in real time based on the imaging device on the drone, and the scene grayscale image is preprocessed.
[0058] In this embodiment, the imaging device may be a binocular camera on a drone; the image preprocessing includes one or more of the following: noise reduction and illumination correction, in order to improve image quality;
[0059] S2, Divide the scene grayscale image into M The process involves N image patches, extracting GLCM features from each patch, and obtaining the contrast of each patch at different directions and distances based on its GLCM features. The specific steps include:
[0060] S21. Define the parameters of GLCM (Gray Co-occurrence Matrix) as P(i,j|d,θ), where i and j are the gray values of the first and second pixels in the pixel pair of the current image block, respectively; d is the distance between the first and second pixels in the pixel pair; and θ is the direction, such as 0°, 45°, 90°, 135°, etc.
[0061] Preferredly, in this embodiment, the distance d is dynamically calculated based on the UAV's current flight altitude H and the ground sample distance (GSD):
[0062]
[0063] Among them, L real The preset physical scale, ranging from 50 to 100 meters, is used to describe the actual spatial scale of the texture features to be captured. If the flight altitude H is large, the water surface area occupies a smaller proportion in the image, and the value of d is small. Conversely, if the flight altitude H is small, the water surface area occupies a larger proportion in the image, and the value of d is large. This ensures that the physical scale of the texture features reflected by GLCM is consistent at different flight altitudes.
[0064] S22. Construct an initial GLCM and set the gray level (e.g., 8 bits). Then, traverse the pixel pairs in the current image block to extract all GLCM P(i,j|d,θ) at a specified distance d and direction θ.
[0065] For example, when a specified distance d=1 and directions θ are 0°, 45°, 90°, and 135°, four GLCMs of the current image patch can be obtained; thus, if there are m distances d and n directions θ, the current image patch has a total of m×n GLCMs.
[0066] S23. Normalize all GLCMs of the current image patch according to the following formula to obtain the GLCM probability matrix P(i,j):
[0067] P(i,j)=P(i,j|d,θ) / R
[0068] Where R is the total number of elements in all GLCMs, the GLCM probability matrix P(i,j) can be used to represent the joint probability of a pair of pixels with a specific gray value appearing at a specified direction and distance.
[0069] S24. Calculate the texture features of the current image patch at different distances and directions based on the GLCM probability matrix P(i,j), where the texture feature is the contrast. ;
[0070] Repeat steps S21-S24 above to obtain the contrast of each image block at different distances and directions;
[0071] S3. Obtain the total contrast (Contrast(d)) of each image patch at each distance d. n ), where n is the number of distances d, and it includes the following steps:
[0072] S31. Determine the main orientation angle (main_angle) of the current image patch through principal component analysis (PCA) or histogram of gradient orientation (HOG) to preliminarily estimate the dominant texture orientation;
[0073] S32. Calculate the k-th distance d according to the following formula. k Below, the angle difference Δθ between each direction θ and the main direction angle main_angle. k And based on the angle difference Δθ k Get the k-th distance d k Below, the contrast weights for each direction:
[0074] Where, Δθk =min(|θ-main_angle|, 180-|θ-main_angle|), for example, if the main direction angle main_angle is 28.5°, then when the direction θ is 0°, 45°, 90°, and 135°, the corresponding angle difference Δθ is 28.5°, 16.5°, 61.5°, and 73.5°, respectively.
[0075] Furthermore, the original contrast weighting coefficient weight0 for each direction is calculated according to the following formula:
[0076] weight0 = cos(Δθ) π / 180)
[0077] The original contrast weighting coefficients weight0 in all directions are normalized to obtain the k-th distance d. k Below, the contrast weight in each direction;
[0078] S33, Calculate the k-th distance d k The contrast-weighted cumulative value in all directions is used as the k-th distance d of the current image patch. k Total contrast (d) k ):
[0079]
[0080] Where i represents the k-th distance d k The i-th direction below; Contrast i For the k-th distance d k Below, the contrast in the i-th direction; weight i For the k-th distance d k Below, the contrast weight in the i-th direction;
[0081] S34. Repeat steps S32-S33 above to obtain the total contrast Contrast(d) of the current image block at each distance d. n ), and k = 1, 2, ..., n, where n and k are both positive integers greater than or equal to 2;
[0082] S35. Repeat steps S31-S34 to obtain the total contrast Contrast(d) of each image patch at each distance d. n );
[0083] S4. Contrast(d) of the total contrast of the same image patch at all distances d. n Perform a weighted summation (i.e., Sum(Contrast(d)) nTo obtain the region determination result Score for the image block, repeat this step until the region determination result Score for each image block is obtained;
[0084] And, S5, the region judgment result (Score) of each image block is compared with the water surface contrast threshold (Contrast). s Compare the results; if the region determination result Score of the current image patch is less than Contrast... s If the region determination result of the current image patch is Score ≥ Contrast, then the image patch is considered to be a water surface region. s If the image patch is not a water surface region, it is considered to be a non-water surface region, such as a shore or the boundary between the water surface and the shore; the water surface contrast threshold Contrast s It can be determined based on the different conditions of the water surface, such as whether the water is calm, has ripples, or is turbid.
[0085] And, count the total number T of image patches in the water surface area, and according to α=T / (M N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as a water surface area. The water surface area threshold ranges from [0.70, 0.99] (preferably [0.80, 0.90]), and its specific value can be set according to the actual detection accuracy requirements.
[0086] like Figure 2 As shown, the current scene grayscale image is divided into 5 (M) segments. 6(N) = 30 image blocks (the first row of values within each red box represents the region judgment result, Score, for that image block), and the region judgment result, Score, for each image block is calculated according to the steps in this embodiment, and the Contrast is taken. s =1.5, water surface area threshold =0.7, then image patches with a score <1.5 are water surface area image patches, and those with a score ≥1.5 are non-water surface area image patches. Finally, the total number of water surface area image patches is 22, so the proportion of water surface area image patches α = 22 / 30 = 0.73. At this time, the scene corresponding to the current scene grayscale image is considered to be a water surface area.
[0087] Similarly, such as Figure 3 As shown, the current scene grayscale image is also divided into 5 (M) segments. 6(N) = 30 images, and the region judgment result (Score) for each image block is calculated according to the steps in this embodiment, and the Contrast is taken. s=1.5, water surface area threshold =0.7, and the final count shows that there are a total of 13 water surface area image patches. Therefore, the proportion of water surface area image patches α = 13 / 30 = 0.43. At this point, the scene corresponding to the current scene grayscale image is considered to be a non-water surface area.
[0088] Therefore, in this embodiment, only the contrast extracted based on GLCM is used as an indicator, and the total contrast at different distances is used as a judgment benchmark to enhance the robustness of the algorithm to different water surface states (such as calm water, ripples, turbid water, etc.). This can significantly improve the recognition accuracy of difficult areas such as shadows and turbid water. Moreover, there is no need to fuse the water index with texture features. Therefore, it can solve the problem of poor generalization caused by manually designed indices and subjectively set fusion weights, and significantly improve the water surface area recognition ability and recognition accuracy under different geographical environments and imaging conditions.
[0089] Meanwhile, the integrated feature extraction mechanism in this embodiment can fundamentally eliminate conflicts between heterogeneous features (such as remote sensing images and textures in the prior art), and it does not require repeated parameter tuning. The entire calculation process is more efficient and simple, and can provide a reliable technical solution for realizing large-scale, high-precision automated water surface area identification and extraction.
[0090] Example 2:
[0091] This embodiment provides a water surface area identification system based on unmanned aerial vehicles (UAVs), such as Figure 4 As shown, it includes:
[0092] Drone 1;
[0093] Imaging device 2 is mounted on the drone 1, and when the drone is flying, it acquires the scene grayscale image of the area below the drone in real time based on the imaging device mounted on the drone.
[0094] Image segmentation module 3 is used to divide the scene grayscale image into M... N image patches;
[0095] Feature extraction module 4 is used to extract GLCM features for each image patch, and its process is the same as step S2.
[0096] The contrast calculation module 5 obtains the contrast (Contrast) of each image block at different directions and distances based on the GLCM features of each image block, and obtains the total contrast (Contrast(d)) of each image block at each distance d. n The process is the same as steps S2-S3;
[0097] Image patch region determination module 6 is used to calculate the region determination result Score for each image patch, and to compare the region determination result Score of each image patch with the water surface contrast threshold Contrast. s A comparison is performed, and based on the comparison result, it is determined whether the current image patch is a water surface area. Specifically, determining whether the current image patch is a water surface area based on the comparison result includes: if the region determination result of the current image patch, Score < Contrast... s If the region determination result of the current image patch is Score ≥ Contrast, then the image patch is considered to be a water surface region. s If so, the image patch is considered to be a non-water surface area;
[0098] The overall region judgment module 7 is used to count the total number T of image blocks in the water surface area, and according to α=T / (M N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as the water surface area.
[0099] In summary, this invention significantly improves the ability and accuracy of water surface area recognition under different geographical environments and imaging conditions by using the total contrast at different distances as the judgment benchmark, based solely on the contrast extracted by GLCM. At the same time, it enhances the robustness of the algorithm to different water surface states (such as calm water, rippled water, turbid water, etc.).
[0100] Meanwhile, this invention does not require the fusion of water body index and texture features, thus solving the problem of poor generalization caused by manually designed indices and subjectively set fusion weights. It can significantly improve the ability and accuracy of water surface area recognition under different geographical environments and imaging conditions, and can fundamentally eliminate the adverse effects of heterogeneous feature conflicts on recognition accuracy. Moreover, it does not require repeated parameter tuning, and the entire calculation process is more efficient and simple.
[0101] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method for identifying water surface areas, characterized in that, Includes the following steps: Acquire scene grayscale images using imaging devices; The grayscale image of the scene is divided into M. N image patches are obtained, and the GLCM features of each image patch are extracted. Based on the GLCM features of each image patch, the contrast of each image patch under different directions and distances is obtained. Obtain the total contrast Contrast(d) for each image patch at each distance d. n ); Based on total contrast Contrast(d) n The region determination result (Score) for each image block is obtained. Furthermore, the region determination result (Score) for each image patch is compared with the water surface contrast threshold (Contrast). s Compare the results; if the region determination result Score of the current image patch is less than Contrast... s If the region determination result of the current image patch is Score ≥ Contrast, then the image patch is considered to be a water surface region. s If the image block is not a water surface area, then it is considered to be a non-water surface area.
2. The water surface area identification method as described in claim 1, characterized in that, The imaging device is mounted on a drone.
3. The water surface area identification method as described in claim 1, characterized in that, The grayscale image of the scene is divided into M. The process involves N image patches, extracting GLCM features from each patch, and obtaining the contrast of each patch at different directions and distances based on its GLCM features. The steps include: The parameters of GLCM are defined as P(i,j|d,θ), where i and j are the gray values of the first and second pixels in the pixel pair of the current image block, respectively; d is the distance between the first and second pixels in the pixel pair; and θ is the direction. Construct an initial GLCM and iterate through the pixel pairs in the current image patch to extract all GLCMs P(i,j|d,θ) with a specified distance d and direction θ. Normalize all GLCMs of the current image patch to obtain the GLCM probability matrix P(i,j); Calculate the contrast Contrast of the current image patch at different distances and directions based on the GLCM probability matrix P(i,j); Repeat the above steps to obtain the contrast (Contrast) of each image patch at different distances and directions.
4. The water surface area identification method as described in claim 3, characterized in that, Obtain the total contrast Contrast(d) for each image patch at each distance d. n The process includes the following steps: Determine the main direction angle of the current image patch, main_angle; Calculate the k-th distance d using the following formula. k Below, the angle difference Δθ between each direction θ and the main direction angle main_angle. k And obtain the k-th distance d k Below, the contrast weights for each direction: Dth k =min(|θ-main_angle|, 180-|θ-main_angle|); Calculate the k-th distance d k The contrast-weighted cumulative value in all directions is used as the k-th distance d of the current image patch. k Total contrast (d) k ): ； Where i represents the k-th distance d k The i-th direction below; Contrast i For the k-th distance d k Below, the contrast in the i-th direction; weight i For the k-th distance d k Below, the contrast weight in the i-th direction; Repeat the above steps to obtain the total contrast Contrast(d) of the current image patch at each distance d. n ), and k=1,2,...,n, and the total contrast Contrast(d) of each image patch at each distance d. n ).
5. The water surface area identification method as described in claim 4, characterized in that, The k-th distance d k The following steps are involved in obtaining the contrast weight for each direction: The original contrast weighting coefficient weight0 for each direction is calculated using the following formula: weight0= cos(Δθ (π / 180) Furthermore, the original contrast weighting coefficients weight0 in all directions are normalized to obtain the k-th distance d. k Below, the contrast weight in each direction.
6. The water surface area identification method as described in claim 4, characterized in that, Based on total contrast Contrast(d) n To obtain the region determination result (Score) for each image patch, the following steps are included: Contrast(d) represents the total contrast of the same image patch at all distances d. n The weighted sum is then used to obtain the region determination result (Score) for the image patch. Repeat the above steps until you obtain the region determination result (Score) for each image block.
7. The water surface area identification method as described in claim 3, characterized in that, If a grayscale image of the scene below the drone is acquired in real time using an imaging device onboard the drone, then the distance d is calculated based on the drone's current flight altitude H and the ground sampling distance: ； Among them, L real For the preset physical scale.
8. The water surface area identification method as described in claim 1, characterized in that, The water surface area identification method also includes: The total number T of image patches in the water surface area is counted, and according to α=T / (M) N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as the water surface area.
9. The water surface area identification method as described in claim 8, characterized in that, The threshold value for the water surface area is in the range of [0.70, 0.99].
10. A water surface area identification system, characterized in that, include: Imaging equipment used to acquire grayscale images of a scene in real time; The image segmentation module is used to divide the scene grayscale image into M... N image patches; The feature extraction module is used to extract GLCM features for each image patch; The contrast calculation module obtains the contrast (Contrast) of each image patch at different directions and distances based on the GLCM features of each patch, and also obtains the total contrast (Contrast(d)) of each image patch at each distance d. n ); The image patch region determination module calculates the region determination result (Score) for each image patch and compares the Score of each image patch with the water surface contrast threshold (Contrast). s Compare the images and determine whether the current image patch is a water surface area based on the comparison results. And, the overall region judgment module, which is used to count the total number T of image patches in the water surface region, and according to α=T / (M N) Calculate the proportion α of all water surface area image blocks. If α ≥ water surface area threshold, then the scene corresponding to the current scene grayscale image is considered as the water surface area.