A dam water seepage monitoring method based on remote sensing images

Abstract

The application discloses a dam water seepage monitoring method based on remote sensing images and belongs to the technical field of image processing. The application collects multiple remote sensing images of the same dam wall surface area at different time periods under sunlight conditions, and distinguishes background areas and contrast areas through clustering processing. Feature values of pixel water seepage are obtained according to color differences, and a feature map is generated. Overlapping contrast areas of the same water seepage target are found out, water seepage diffusion coincidence degrees and their mean values, and water seepage change degrees are calculated. The pixel water seepage feature map is enhanced by using the above indexes. Finally, the enhanced map is processed by a dam water seepage evaluation neural network to obtain a water seepage score, so that accurate monitoring and evaluation of dam water seepage conditions are realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to a dam seepage monitoring method based on remote sensing images. BACKGROUND

[0002] As the core facility of water conservancy projects, dams play an irreplaceable role in flood control, water supply, power generation and other fields. The safe and stable operation of dams is directly related to people's life and property safety and social and economic development. However, with the continuous development of water conservancy construction, many dams have increased in design height to meet multiple demands such as flood control and power generation, and some high dams even reach hundreds of meters. Such height makes manual visual observation extremely difficult, and the seepage condition of the upper part of the dam body is difficult to be discovered in time, greatly increasing the safety hazard.

[0003] Due to the influence of factors such as long-term water flow erosion, geological condition changes, and material aging, dam walls are prone to seepage problems. If not discovered and handled in time, seepage may gradually intensify, causing damage to the dam structure, reducing the strength, and even leading to dam collapse and other major safety accidents. Therefore, it is crucial to accurately and efficiently monitor the seepage condition of the dam.

[0004] Traditional dam seepage monitoring methods mostly rely on manual inspection or sensor deployment. Manual inspection not only consumes a lot of manpower and resources, but also is limited by the experience and field of view of the inspectors. Especially when facing high dams, it is difficult to cover all areas of the dam, and the seepage points at high altitudes of the dam body often become blind spots of monitoring, with a serious risk of missed detection. Although the sensor deployment method can achieve a certain degree of automatic monitoring, the installation and maintenance of sensors are costly and easily disturbed by environmental factors, and the monitoring range is also limited, making it difficult to meet the overall monitoring needs of large high dams.

[0005] The existing method can obtain macro information of the dam wall from remote sensing images, but how to highlight the seepage features from the remote sensing images and accurately identify the seepage condition is still a key problem currently faced. SUMMARY

[0006] In view of the above shortcomings in the prior art, the dam seepage monitoring method based on remote sensing images provided by the present application solves the problem that the seepage features in the remote sensing images are not prominent, resulting in inaccurate identification of the seepage condition.

[0007] In order to achieve the above-mentioned application purposes, the technical scheme adopted by the present application is as follows: a dam seepage monitoring method based on remote sensing images, comprising the following steps:

[0008] Under daylight conditions, multiple remote sensing images of the same dam wall area at different time periods are collected, each remote sensing image is subjected to clustering processing to obtain multiple connected regions, and the background region and the contrast region are distinguished;

[0009] According to the color difference between the contrast region and the background region, a pixel water seepage feature value of each pixel point in the contrast region is obtained, and a pixel water seepage feature map is obtained;

[0010] On the plurality of pixel water seepage feature maps, the contrast regions belonging to the same water seepage target with area overlap are found out;

[0011] The water seepage diffusion coincidence degree of each contrast region in the pixel water seepage feature map is obtained, and the water seepage diffusion coincidence degrees belonging to the same water seepage target are averaged;

[0012] According to the contrast regions belonging to the same water seepage target, the water seepage change degree of each water seepage target is obtained;

[0013] The water seepage diffusion coincidence degree average and the water seepage change degree are respectively used for enhancement processing on any one of the pixel water seepage feature maps, and a first pixel water seepage feature enhancement map and a second pixel water seepage feature enhancement map are obtained;

[0014] The dam water seepage evaluation neural network is used to process the first pixel water seepage feature enhancement map and the second pixel water seepage feature enhancement map, and a water seepage score is obtained.

[0015] In some preferred embodiments of the present application, the process of obtaining the pixel water seepage feature map comprises:

[0016] The average of each brightness value in the background region is obtained, and the background brightness is obtained;

[0017] The ratio of the background brightness to the brightness value of each pixel point in the contrast region is taken as the pixel brightness ratio;

[0018] The channel difference is obtained by subtracting the channel value of each pixel point in the contrast region from the background channel average;

[0019] After weighting and adding the three channel differences, a channel difference value is obtained;

[0020] The pixel water seepage feature value is obtained by multiplying the channel difference value of the same pixel point in the contrast region with the pixel brightness ratio, and the multiplication result is normalized, the pixel water seepage feature value of the pixel point in the background region is set to 0, and the pixel water seepage feature map is obtained.

[0021] In some preferred embodiments of the present application, the formula for obtaining the channel difference value is:

[0022] ,

[0023] Wherein, C i is the channel difference value of the i-th pixel point in the contrast region, △R i is the R channel difference of the i-th pixel point, △G i is the G channel difference of the i-th pixel point, and △Bi Let be the B channel difference of the i-th pixel.

[0024] In some preferred embodiments of the present invention, the process of obtaining the water diffusion fit includes:

[0025] Divide a contrast region of the pixel water seepage feature map into multiple image blocks of size N×M, where N and M are positive integers;

[0026] The average value of the pixel water seepage feature for each image block is taken;

[0027] The image block corresponding to the mean value of the maximum pixel water seepage feature located within the comparison area is selected as the center image block;

[0028] In the horizontal direction, starting from the central image block, the difference in the mean pixel water penetration feature between the left and right adjacent image blocks is calculated;

[0029] In the vertical direction, starting from the central image block, the difference in the mean pixel water penetration feature between the upper and lower adjacent image blocks is calculated.

[0030] The ratio of the number of differences greater than 0 to the total number of differences is taken as the degree of fit of seepage diffusion.

[0031] In some preferred embodiments of the present invention, in the direction to the left of the central image block, the average pixel water seepage feature of the right image block among the adjacent image blocks is subtracted from the average pixel water seepage feature of the left image block to obtain the difference.

[0032] To the right of the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the left image block from the average pixel water seepage feature of the right image block.

[0033] In the direction above the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the lower image block from the average pixel water seepage feature of the upper image block.

[0034] In the direction below the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the upper image block from the average pixel water seepage feature of the lower image block.

[0035] In some preferred embodiments of the present invention, the process of obtaining the degree of change in permeability for each permeation target includes:

[0036] Select the largest and smallest comparison areas from the comparison areas belonging to the same seepage target.

[0037] The maximum area difference is obtained by subtracting the largest contrast area from the smallest contrast area.

[0038] The maximum area difference is normalized to obtain the degree of change in seepage of the seepage target.

[0039] In some preferred embodiments of the present invention, the enhancement process includes:

[0040] The average water diffusion matching degree of the same water seepage target is used to enhance the contrast area of ​​the same water seepage target in the pixel water seepage feature map to obtain the first pixel water seepage feature enhancement map.

[0041] The comparison area of ​​the same seepage target in the pixel seepage feature map is enhanced by using the seepage variation degree of the same seepage target to obtain the second pixel seepage feature enhancement map.

[0042] In some preferred embodiments of the present invention, the formula for the enhancement process is as follows:

[0043] , ,

[0044] Among them, V z1，i V represents the i-th enhancement value in the contrast region of the same seepage target on the first pixel seepage feature enhancement map. i Let V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target on the pixel water seepage feature map, c be the mean water seepage diffusion matching degree of the same water seepage target, and V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target. z2，i ε represents the i-th enhancement value in the comparison area of ​​the same seepage target on the second pixel seepage feature enhancement map, and ε represents the seepage variation degree of the same seepage target.

[0045] In some preferred embodiments of the present invention, the dam seepage assessment neural network includes: a first parallel feature extraction unit, a second parallel feature extraction unit, a cross-branch gated interactive feature enhancement unit, a CNN unit, and a fully connected layer;

[0046] The input terminal of the first parallel feature extraction unit is used to input the water seepage feature enhancement map of the first pixel; the input terminal of the second parallel feature extraction unit is used to input the water seepage feature enhancement map of the second pixel.

[0047] The first input of the cross-branch gated interactive feature enhancement unit is connected to the output of the first parallel feature extraction unit, its second input is connected to the output of the second parallel feature extraction unit, and its output is connected to the input of the CNN unit; the input of the fully connected layer is connected to the output of the CNN unit, and its output serves as the output of the dam seepage assessment neural network.

[0048] In some preferred embodiments of the present invention, the cross-branch gating interaction feature enhancement unit includes: a first gating module, a second gating module, a multiplier M1, a multiplier M2, and an adder A1;

[0049] The input terminal of the first gating module is connected to the first input terminal of multiplier M1 and serves as the first input terminal of the cross-branch gating interactive feature enhancement unit; the input terminal of the second gating module is connected to the first input terminal of multiplier M2 and serves as the second input terminal of the cross-branch gating interactive feature enhancement unit; the second input terminal of multiplier M1 is connected to the output terminal of the first gating module, and its output terminal is connected to the first input terminal of adder A1; the second input terminal of multiplier M2 is connected to the output terminal of the second gating module, and its output terminal is connected to the second input terminal of adder A1; the output terminal of adder A1 serves as the output terminal of the cross-branch gating interactive feature enhancement unit.

[0050] The beneficial effects of this invention are as follows:

[0051] 1. This invention performs clustering processing on remote sensing images of the same dam wall area at different time periods, which can accurately distinguish between background and contrast areas. Then, by combining the color difference between the contrast and background areas, pixel seepage feature values ​​are obtained, highlighting the seepage characteristics of potential seepage areas, making the seepage characteristics significant, and improving the accuracy of seepage identification.

[0052] 2. This invention identifies comparative regions belonging to the same seepage target in multiple pixel seepage feature images, and enhances the pixel seepage feature images by combining the mean of seepage diffusion consistency and seepage variability. This highlights the characteristics of the seepage target, further enhances the distinction between the seepage area and other areas, and reduces the influence of interference factors. Then, a dam seepage assessment neural network is used to process the first pixel seepage feature enhancement image and the second pixel seepage feature enhancement image. Combining the enhanced seepage features in the two images improves the accuracy of seepage identification. Attached Figure Description

[0053] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0054] Figure 1 A flowchart of a dam seepage monitoring method based on remote sensing images provided in an embodiment of the present invention;

[0055] Figure 2 A schematic diagram illustrating an embodiment of the present invention, showing an image region divided into multiple N×M size blocks;

[0056] Figure 3 This is a schematic diagram of the structure of a neural network for assessing dam seepage, provided in an embodiment of the present invention.

[0057] Figure 4 This is a schematic diagram of the structure of a cross-branch gating interaction feature enhancement unit provided in an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of the structure of a first gate control module and a second gate control module provided in an embodiment of the present invention;

[0059] Figure 6 This is a schematic diagram of the structure of a first parallel feature extraction unit and a second parallel feature extraction unit provided in an embodiment of the present invention. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0061] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0062] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0063] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship commonly used when the product of this invention is in use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention. In addition, the terms "first," "second," and "third," etc., are only used to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0064] Furthermore, terms such as "horizontal," "vertical," and "sag" do not imply that components must be absolutely horizontal or suspended, but rather that they can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal relative to "vertical," and does not mean that the structure must be completely horizontal, but can be slightly tilted.

[0065] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0066] The following detailed description of some embodiments of the present invention is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0067] Example 1

[0068] like Figure 1 As shown, a method for monitoring dam seepage based on remote sensing images includes the following steps:

[0069] Under sunlight conditions, multiple remote sensing images of the same dam wall area were collected at different times. Each remote sensing image was clustered to obtain multiple connected regions, and the background region and contrast region were distinguished.

[0070] Based on the color difference between the contrast area and the background area, the pixel water seepage feature value of each pixel in the contrast area is obtained to obtain the pixel water seepage feature map;

[0071] On multiple pixel-based water seepage feature maps, identify the contrasting regions that have overlapping areas and belong to the same water seepage target;

[0072] For each comparison region in the pixel water seepage feature map, obtain the water seepage diffusion matching degree, and take the average value of the water seepage diffusion matching degrees belonging to the same water seepage target.

[0073] Based on the comparison areas belonging to the same seepage target, obtain the degree of seepage variation for each seepage target;

[0074] The mean value of water diffusion matching degree and the water diffusion change degree are used to enhance the water seepage feature map of any pixel to obtain the first pixel water seepage feature enhancement map and the second pixel water seepage feature enhancement map.

[0075] A dam seepage assessment neural network is used to process the first pixel seepage feature enhancement image and the second pixel seepage feature enhancement image to obtain a seepage score.

[0076] In this embodiment, the clustering process includes: taking any unclassified pixel as the initial cluster point, including pixels with similar pixel values ​​in the neighborhood of the cluster point into the same region, using the newly included pixels as new cluster points, and repeating the expansion until no more pixels can be added to the region, forming a connected region, and continuing to classify the remaining unclassified pixels until the remote sensing image of the dam wall is classified, resulting in multiple connected regions.

[0077] The detailed clustering process includes the following steps:

[0078] A1. Take any pixel that does not have a connected region as a cluster point, obtain the similarity of pixel value between each other pixel in the neighborhood of the cluster point and the cluster point, and when the similarity is greater than the similarity threshold, classify the corresponding pixels in the neighborhood and the cluster point into a connected region.

[0079] A2. Take any pixel in the connected region that is not a cluster point as a new cluster point, obtain the similarity of pixel value between each pixel in the neighborhood of the new cluster point and the other pixels in the neighborhood that are not connected regions. When the similarity is greater than the similarity threshold, include the corresponding pixels in the neighborhood into the connected region.

[0080] A3. Repeat A2 until the connected region can no longer be expanded, at which point the connected region partitioning is complete.

[0081] A4. Repeat A1 to extract the next connected region until the remote sensing image of the dam wall is completely divided.

[0082] In this embodiment, the largest connected region is used as the background region, and other connected regions are used as the comparison regions.

[0083] In this embodiment, the process of obtaining the pixel water seepage feature map includes:

[0084] The average value of each brightness value in the background area is calculated to obtain the background brightness.

[0085] The ratio of the background brightness to the brightness value of each pixel in the contrast area is used as the pixel brightness ratio;

[0086] The channel difference is obtained by subtracting the channel value of each pixel in the contrast region from the mean of the background channel.

[0087] The channel difference values ​​are obtained by weighting the three channel differences and then summing them.

[0088] The pixel water penetration feature value of each pixel in the comparison area is obtained by multiplying the channel difference value of the same pixel in the comparison area with the pixel brightness ratio and normalizing the multiplication result. The pixel water penetration feature value of the pixels in the background area is set to 0 to obtain the pixel water penetration feature map.

[0089] In this embodiment, the formula for normalizing the multiplication result is: sz = z / z max Where sz is the pixel water seepage feature value, z is the product of any pixel, and z max This is the result of the maximum multiplication.

[0090] Due to the moisture covering the surface of the seepage area, the reflectivity is reduced, and the overall brightness is lower than that of the background. Therefore, the ratio of "background brightness / pixel brightness" will be greater than 1 (the larger the pixel brightness ratio, the more significant the darkening feature).

[0091] Moisture absorbs visible light (especially blue light) more strongly, resulting in significantly lower RGB channel values ​​(especially blue channel) in the water-soaked area compared to the dry background area.

[0092] This invention uses the product operation of "channel difference value × pixel brightness ratio" to synergistically amplify the two features of "color shift" and "brightness reduction" caused by moisture, so that the pixel water seepage feature value in the water seepage area is significantly higher than that in the non-water seepage area (such as dry stains and textures). It utilizes the law of the effect of moisture on light to make the feature value positively correlated with the degree of water seepage.

[0093] In this embodiment, the formula for obtaining the channel difference value is: , where C i To compare the channel difference value of the i-th pixel in the region, △R i Let ΔG be the R channel difference of the i-th pixel. i Let △B be the G channel difference of the i-th pixel. i Let be the B channel difference of the i-th pixel.

[0094] In this invention, the weight of the R channel difference is assigned as 0.2, the weight of the G channel difference as 0.3, and the weight of the B channel difference as 0.5. Moisture absorbs visible light of different wavelengths to varying degrees, with a relatively stronger absorption of blue light (corresponding to the B channel). When calculating the channel difference value, assigning the highest weight of 0.5 to the B channel difference, a weight of 0.3 to the G channel difference, and a weight of 0.2 to the R channel difference better highlights the difference between the seepage area and the background on the B channel, thereby accurately capturing the optical feature changes caused by the presence of moisture in the seepage area and improving the accuracy of seepage area identification.

[0095] In this embodiment, identifying the comparison area for the same seepage target includes the following process:

[0096] Let any contrast region in any pixel water seepage feature map R be denoted as contrast region A;

[0097] Obtain the intersection of the comparison region A with each comparison region in the remaining pixel water seepage feature map K at the pixel point. When the intersection is not empty, and there is only one comparison region in each pixel water seepage feature map K with a non-empty intersection with the comparison region A, the comparison region in the pixel water seepage feature map K is recorded as comparison region B. If there are multiple comparison regions, the comparison region corresponding to the largest intersection in the pixel water seepage feature map K is recorded as comparison region B.

[0098] Comparison area A and each comparison area B belong to the same seepage target. Repeat this process to obtain the comparison areas that belong to each seepage target.

[0099] Comparison areas belonging to the same seepage target meet the following conditions: Where G1 is the set of pixel coordinates of the comparison region belonging to the seepage target in the pixel seepage feature map of the first time period, and G2 is the set of pixel coordinates of the comparison region belonging to the seepage target in the pixel seepage feature map of the second time period. t Let G be the set of pixel coordinates in the contrast region of the water seepage feature map belonging to the water seepage target in time period t. T Let be the set of pixel coordinates in the comparison region of the water seepage feature map belonging to the water seepage target in time period T, and ∩ represents the intersection. It is an empty set.

[0100] There are cases where the conditions are not met. The contrast area does not need to be considered, that is, its pixel water seepage feature value is set to 0.

[0101] In this embodiment, the process of obtaining the consistency of water diffusion includes:

[0102] Divide a contrast region of the pixel water seepage feature map into multiple image blocks of size N×M, where N and M are positive integers;

[0103] The average value of the pixel water seepage feature for each image block is taken;

[0104] The image block corresponding to the mean value of the maximum pixel water seepage feature located within the comparison area is selected as the center image block;

[0105] In the horizontal direction, starting from the central image block, the difference in the mean pixel water penetration feature between the left and right adjacent image blocks is calculated;

[0106] In the vertical direction, starting from the central image block, the difference in the mean pixel water penetration feature between the upper and lower adjacent image blocks is calculated.

[0107] The ratio of the number of differences greater than 0 to the total number of differences is taken as the degree of fit of seepage diffusion.

[0108] In this embodiment, the size of the image block is set to 5×5, that is, it is vertically divided by a width of 5 pixels and horizontally divided by a width of 5 pixels. There are cases where the size of the edge image block is less than 5×5, and only the pixel seepage feature value existing in the edge image block is considered.

[0109] In this embodiment, in the direction to the left of the central image block, the average pixel water seepage feature of the right image block among the adjacent image blocks is subtracted from the average pixel water seepage feature of the left image block to obtain the difference.

[0110] To the right of the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the left image block from the average pixel water seepage feature of the right image block.

[0111] In the direction above the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the lower image block from the average pixel water seepage feature of the upper image block.

[0112] In the direction below the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the upper image block from the average pixel water seepage feature of the lower image block.

[0113] Dam seepage occurs gradually from the seepage point (central area) outwards, with the strongest seepage characteristics (such as humidity and optical properties) in the central area, gradually weakening towards the periphery.

[0114] This invention selects the image block corresponding to the average value of the maximum pixel seepage feature as the central image block, which can accurately locate the potential core area of ​​leakage. When calculating the difference of the feature average value, the left side is calculated by "right minus left", the right side by "left minus right", the upper side by "lower minus upper", and the lower side by "upper minus lower". This difference calculation method is used to determine whether the feature of the outer image block is weaker than that of the adjacent block in the central direction: if the difference is greater than 0, it means that from the center to that direction, the feature shows a decreasing trend of "strong in the center and weak in the periphery", which can effectively capture the diffusion characteristics of real seepage.

[0115] like Figure 2 As shown, the area where the red arrow points upward in the center image block O is the upper direction, the area where the red arrow points downward in the center image block O is the lower direction, the area where the red arrow points to the left in the center image block O is the left direction, and the area where the red arrow points to the right in the center image block O is the right direction. In the lower direction, H1 and H2 are a pair of adjacent image blocks, with H1 being the upper image block and H2 being the lower image block.

[0116] In this embodiment, the process of obtaining the degree of seepage change for each seepage target includes:

[0117] Select the largest and smallest comparison areas from the comparison areas belonging to the same seepage target.

[0118] The maximum area difference is obtained by subtracting the largest contrast area from the smallest contrast area.

[0119] The maximum area difference is normalized to obtain the degree of change in seepage of the seepage target.

[0120] This invention acquires multiple remote sensing images of the same dam wall area at different times under sunlight conditions. Therefore, dam seepage is a dynamic process that evolves over time. Under sunlight, the seepage area changes in size due to continuous water infiltration and evaporation at different times (e.g., morning, noon, and evening). By selecting the largest and smallest contrasting areas of the same seepage target from multiple time-series remote sensing images, the extreme differences in the spatial range of the seepage area over time can be captured. This difference directly corresponds to the activity level of seepage; the larger the area difference, the more significant the development or regression of seepage over time, further demonstrating the characteristic that "seepage is a dynamic process, and its area changes over time."

[0121] In this embodiment, the formula for normalizing the maximum area difference is: rE=E / (E+1), where rE is the degree of change of the seepage target and E is the maximum area difference.

[0122] In this embodiment, the enhancement process includes:

[0123] The average water diffusion matching degree of the same water seepage target is used to enhance the contrast area of ​​the same water seepage target in the pixel water seepage feature map to obtain the first pixel water seepage feature enhancement map.

[0124] The comparison area of ​​the same seepage target in the pixel seepage feature map is enhanced by using the seepage variation degree of the same seepage target to obtain the second pixel seepage feature enhancement map.

[0125] The contrast areas in the first pixel water seepage feature enhancement map were all enhanced, and the contrast areas in the second pixel water seepage feature enhancement map were all enhanced.

[0126] In this embodiment, the formula for the enhancement process is: , , where V z1，i V represents the i-th enhancement value in the contrast region of the same seepage target on the first pixel seepage feature enhancement map. i Let V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target on the pixel water seepage feature map, c be the mean water seepage diffusion matching degree of the same water seepage target, and V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target. z2，i ε represents the i-th enhancement value in the comparison area of ​​the same seepage target on the second pixel seepage feature enhancement map, and ε represents the seepage variation degree of the same seepage target.

[0127] This invention utilizes the average consistency of seepage diffusion and the degree of seepage variation of the same seepage target to enhance corresponding contrasting regions in the pixel seepage feature map. The consistency of seepage diffusion reflects the degree of fit of the diffusion characteristics of the seepage area, while the degree of seepage variation reflects the magnitude of change in the seepage area. Through enhancement processing, this invention makes the features of the seepage area significantly different from other areas, making the seepage characteristics more prominent and improving the accuracy of seepage monitoring.

[0128] like Figure 3 As shown, the dam seepage assessment neural network includes: a first parallel feature extraction unit, a second parallel feature extraction unit, a cross-branch gated interactive feature enhancement unit, a CNN unit, and a fully connected layer;

[0129] The input terminal of the first parallel feature extraction unit is used to input the water seepage feature enhancement map of the first pixel; the input terminal of the second parallel feature extraction unit is used to input the water seepage feature enhancement map of the second pixel.

[0130] The first input of the cross-branch gated interactive feature enhancement unit is connected to the output of the first parallel feature extraction unit, its second input is connected to the output of the second parallel feature extraction unit, and its output is connected to the input of the CNN unit; the input of the fully connected layer is connected to the output of the CNN unit, and its output serves as the output of the dam seepage assessment neural network.

[0131] This invention extracts features from the first pixel water seepage feature enhancement map and the second pixel water seepage feature enhancement map through a first parallel feature extraction unit and a second parallel feature extraction unit, respectively. Then, the two parallel extracted features are interactively processed by a cross-branch gating interactive feature enhancement unit. The gating mechanism adaptively adjusts the weights of features from different branches, so that the fused features can more accurately reflect the essential features of the water seepage target and improve the expressive power of the features. Finally, CNN units are used to extract features, and a fully connected layer is used to output the water seepage score.

[0132] like Figure 4 As shown, the cross-branch gating interaction feature enhancement unit includes: a first gating module, a second gating module, a multiplier M1, a multiplier M2, and an adder A1;

[0133] The input terminal of the first gating module is connected to the first input terminal of multiplier M1 and serves as the first input terminal of the cross-branch gating interactive feature enhancement unit; the input terminal of the second gating module is connected to the first input terminal of multiplier M2 and serves as the second input terminal of the cross-branch gating interactive feature enhancement unit; the second input terminal of multiplier M1 is connected to the output terminal of the first gating module, and its output terminal is connected to the first input terminal of adder A1; the second input terminal of multiplier M2 is connected to the output terminal of the second gating module, and its output terminal is connected to the second input terminal of adder A1; the output terminal of adder A1 serves as the output terminal of the cross-branch gating interactive feature enhancement unit.

[0134] This invention uses a first gating module and a second gating module to adaptively generate weights based on the characteristics of the input features. The weights are then multiplied bitwise with the input feature map. The two weighted feature maps are then added bitwise by adder A1. This allows the proportion of features from different sources in the fusion result to be dynamically adjusted based on the numerical value of the weights at corresponding positions in the feature maps.

[0135] like Figure 5 As shown, both the first and second gating modules include: a first convolutional layer, a Batch Normalization (BN) layer, a ReLU layer, a second convolutional layer, and a Sigmoid layer. The gating module generates a weight map with the same size as the input feature map through the process of "1×1 convolution → BN → ReLU → 1×1 convolution → Sigmoid".

[0136] like Figure 6 As shown, both the first parallel feature extraction unit and the second parallel feature extraction unit include: a third convolutional layer, a fourth convolutional layer, a fifth convolutional layer, and a Concat layer. The kernel size of the third convolutional layer is 3×3, the kernel size of the fourth convolutional layer is 5×5, and the kernel size of the fifth convolutional layer is 7×7.

[0137] The seepage score ranges from 0 to 100, with the corresponding situations as follows:

[0138] 0-30 points: There is no obvious seepage in the dam, and the dam body is in a normal and stable state.

[0139] 31-60 points: There is slight seepage in the dam, which will not pose a serious threat to the safety of the dam body at present, but monitoring needs to be strengthened.

[0140] 61-85 points: The dam is experiencing significant seepage. If not addressed promptly, it may gradually worsen and affect the dam structure.

[0141] 86-100 points: The dam is experiencing severe seepage, which poses a significant threat to the safety of the dam structure and requires immediate action.

[0142] This invention uses clustering processing on remote sensing images of the same dam wall area at different time periods to accurately distinguish between background and contrast areas. By combining the color differences between the contrast and background areas, pixel seepage feature values ​​are obtained, highlighting the seepage characteristics of potential seepage areas, making the seepage features significant, and improving the accuracy of seepage identification.

[0143] This invention identifies comparative regions belonging to the same seepage target in multiple pixel seepage feature images. It then enhances the pixel seepage feature images by combining the mean of seepage diffusion consistency and seepage variability, which highlights the characteristics of the seepage target and further enhances the distinction between the seepage area and other areas, reducing the influence of interference factors. The first and second pixel seepage feature enhancement images are then processed using a dam seepage assessment neural network. By combining the enhanced seepage features from the two images, the accuracy of seepage identification is improved.

[0144] In the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in the present invention based on the specific circumstances.

[0145] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0146] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring dam seepage based on remote sensing images, characterized in that, Includes the following steps: Under sunlight conditions, multiple remote sensing images of the same dam wall area were collected at different times. Each remote sensing image was clustered to obtain multiple connected regions, and the background region and contrast region were distinguished. Based on the color difference between the contrast area and the background area, the pixel water seepage feature value of each pixel in the contrast area is obtained to obtain the pixel water seepage feature map; On multiple pixel-based water seepage feature maps, identify the contrasting regions that have overlapping areas and belong to the same water seepage target; For each comparison region in the pixel water seepage feature map, the water seepage diffusion matching degree is obtained, and the average value of the water seepage diffusion matching degrees belonging to the same water seepage target is taken. The process of obtaining the water seepage diffusion matching degree includes: dividing a comparison region of the pixel water seepage feature map into multiple image blocks of size N×M, where N and M are positive integers; taking the average value of the pixel water seepage feature value of each image block; selecting the image block corresponding to the largest average value of the pixel water seepage feature value within the comparison region as the center image block; horizontally, starting from the center image block, calculating the difference in the average value of the pixel water seepage feature value between the left and right adjacent image blocks; vertically, starting from the center image block, calculating the difference in the average value of the pixel water seepage feature value between the upper and lower adjacent image blocks; and taking the ratio of the number of differences greater than 0 to the total number of differences as the water seepage diffusion matching degree. Based on the comparison areas belonging to the same seepage target, obtain the degree of seepage variation for each seepage target; The mean value of water diffusion matching degree and the water diffusion change degree are used to enhance the water seepage feature map of any pixel to obtain the first pixel water seepage feature enhancement map and the second pixel water seepage feature enhancement map. A dam seepage assessment neural network is used to process the first pixel seepage feature enhancement image and the second pixel seepage feature enhancement image to obtain a seepage score.

2. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, The process of obtaining the pixel water seepage feature map includes: The average value of each brightness value in the background area is calculated to obtain the background brightness. The ratio of the background brightness to the brightness value of each pixel in the contrast area is used as the pixel brightness ratio; The channel difference is obtained by subtracting the channel value of each pixel in the contrast region from the mean of the background channel. The channel difference values ​​are obtained by weighting the three channel differences and then summing them. The pixel water penetration feature value is obtained by multiplying the channel difference value of the same pixel in the comparison area with the pixel brightness ratio and normalizing the multiplication result. The pixel water penetration feature value of the pixels in the background area is set to 0 to obtain the pixel water penetration feature map.

3. The method for monitoring dam seepage based on remote sensing images according to claim 2, characterized in that, The formula for obtaining the channel difference value is: ， Among them, C i To compare the channel difference value of the i-th pixel in the region, △R i Let ΔG be the R channel difference of the i-th pixel. i Let △B be the G channel difference of the i-th pixel. i Let be the B channel difference of the i-th pixel.

4. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, To the left of the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the right image block from the average pixel water seepage feature of the left image block. To the right of the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the left image block from the average pixel water seepage feature of the right image block. In the direction above the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the lower image block from the average pixel water seepage feature of the upper image block. In the direction below the central image block, the difference is obtained by subtracting the average pixel water seepage feature of the upper image block from the average pixel water seepage feature of the lower image block.

5. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, The process of obtaining the degree of change in permeability for each permeation target includes: Select the largest and smallest comparison areas from the comparison areas belonging to the same seepage target. The maximum area difference is obtained by subtracting the largest contrast area from the smallest contrast area. The maximum area difference is normalized to obtain the degree of change in seepage of the seepage target.

6. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, The enhancement process includes: The average water diffusion matching degree of the same water seepage target is used to enhance the contrast area of ​​the same water seepage target in the pixel water seepage feature map to obtain the first pixel water seepage feature enhancement map. The comparison area of ​​the same seepage target in the pixel seepage feature map is enhanced by using the seepage variation degree of the same seepage target to obtain the second pixel seepage feature enhancement map.

7. The method for monitoring dam seepage based on remote sensing images according to claim 6, characterized in that, The formula for enhancement processing is: ， ， Among them, V z1，i V represents the i-th enhancement value in the contrast region of the same seepage target on the first pixel seepage feature enhancement map. i Let V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target on the pixel water seepage feature map, c be the mean water seepage diffusion matching degree of the same water seepage target, and V be the water seepage feature value of the i-th pixel in the comparison region of the same water seepage target. z2，i ε represents the i-th enhancement value in the comparison area of ​​the same seepage target on the second pixel seepage feature enhancement map, and ε represents the seepage variation degree of the same seepage target.

8. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, The neural network for dam seepage assessment includes: a first parallel feature extraction unit, a second parallel feature extraction unit, a cross-branch gated interactive feature enhancement unit, a CNN unit, and a fully connected layer; The input terminal of the first parallel feature extraction unit is used to input the water seepage feature enhancement map of the first pixel; the input terminal of the second parallel feature extraction unit is used to input the water seepage feature enhancement map of the second pixel. The first input of the cross-branch gated interactive feature enhancement unit is connected to the output of the first parallel feature extraction unit, its second input is connected to the output of the second parallel feature extraction unit, and its output is connected to the input of the CNN unit; the input of the fully connected layer is connected to the output of the CNN unit, and its output serves as the output of the dam seepage assessment neural network.

9. The method for monitoring dam seepage based on remote sensing images according to claim 1, characterized in that, The cross-branch gating interaction feature enhancement unit includes: a first gating module, a second gating module, a multiplier M1, a multiplier M2, and an adder A1; The input terminal of the first gating module is connected to the first input terminal of multiplier M1 and serves as the first input terminal of the cross-branch gating interactive feature enhancement unit; the input terminal of the second gating module is connected to the first input terminal of multiplier M2 and serves as the second input terminal of the cross-branch gating interactive feature enhancement unit; the second input terminal of multiplier M1 is connected to the output terminal of the first gating module, and its output terminal is connected to the first input terminal of adder A1; the second input terminal of multiplier M2 is connected to the output terminal of the second gating module, and its output terminal is connected to the second input terminal of adder A1; the output terminal of adder A1 serves as the output terminal of the cross-branch gating interactive feature enhancement unit.

Images