Large-range water area time-space dynamic change monitoring method fusing remote sensing and GIS (Geographic Information System)
By integrating remote sensing and GIS methods, and utilizing hyperspectral remote sensing image data and graph cut algorithms to optimize water area segmentation, the problem of inaccurate water area edge identification caused by aquatic plant shading was solved, and accurate monitoring of dynamic changes in large-area water areas was achieved.
Patent Information
- Application Number
- CN202511584152.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-31
- Publication Date
- 2025-12-30
AI Technical Summary
Traditional water area segmentation methods struggle to accurately identify water area boundaries during periods of vigorous aquatic plant growth, leading to distorted water area monitoring results and affecting the accuracy of long-term dynamic water area monitoring.
By employing a method that integrates remote sensing and GIS, we acquire pixel grayscale differences, texture features, and normalized water vegetation index from hyperspectral remote sensing image data, construct feature vectors, and combine them with the energy function of the graph cut algorithm to segment water areas. We then optimize the segmentation algorithm to identify water area edges.
It improves the accuracy of monitoring the spatiotemporal dynamic changes of water bodies, avoids the problem of blurred boundaries caused by vegetation obstruction, and ensures accurate monitoring of dynamic changes in large areas of water bodies.
Smart Images
Figure CN121236624A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of image processing technology, specifically to a method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS. Background Technology
[0002] With the rapid development of remote sensing technology and geographic information systems (GIS), spatiotemporal dynamic monitoring of large-scale water areas has become an important means of water resource management, ecological protection and disaster early warning.
[0003] Current water area segmentation methods perform well in periods with sparse or no vegetation. However, during periods of vigorous growth of aquatic plants, the edges of water areas are largely covered by aquatic plants, resulting in a high degree of similarity between the water area edges and terrestrial vegetation. Traditional graph cut algorithms, which calculate the differences in color features of pixels, cannot effectively distinguish between aquatic and terrestrial vegetation areas, making it difficult to accurately identify the true edges of water areas. This leads to distorted water area information monitoring results and affects the accuracy of long-term dynamic monitoring of water areas. Summary of the Invention
[0004] In light of the above, it is necessary to provide a method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS. Compared with traditional methods for monitoring the spatiotemporal dynamic changes of large-scale water bodies, this method improves the accuracy of monitoring spatiotemporal dynamic changes by enhancing the accuracy of regional segmentation. The method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS in this application adopts the following technical solution: One embodiment of this application provides a method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS. The method includes the following steps: Real-time acquisition of hyperspectral remote sensing image data of the water area to be monitored; For each time point, the feature vector of each pixel and the center vector of each area label in the water body to be monitored are constructed by combining the gray level difference and texture feature difference between each pixel and its neighboring pixels in the hyperspectral remote sensing image data with the normalized water index and normalized vegetation index of each pixel. An energy function for a graph cut algorithm to segment the water area to be monitored is constructed. In the region term of the energy function, the penalty for assigning each pixel to each region label is obtained by comparing the feature vector of each pixel with the center vector of each region label. In the boundary term of the energy function, the penalty for assigning any two adjacent pixels to different region labels is obtained by comparing the feature vectors of any two adjacent pixels. The water area to be monitored is segmented at different times by using an energy function, and the spatiotemporal dynamic changes of the water area to be monitored are carried out by using the segmentation results at different times.
[0005] In one embodiment, the feature vector is composed of the gray value of each pixel, the texture feature value of each pixel, the normalized water index, and the normalized vegetation index. The texture feature values are obtained by the grayscale distribution of each pixel and its neighboring pixels.
[0006] In one embodiment, the process of obtaining the texture feature values is as follows: A gray-level co-occurrence matrix is constructed based on the gray values of each pixel and its preset nearest neighbor pixels. The texture feature value is the average of the entropy and contrast of the gray-level co-occurrence matrix corresponding to each pixel.
[0007] In one embodiment, the process of constructing the center vector of each region label is as follows: Thresholding is performed on the grayscale values of pixels in the hyperspectral remote sensing image data to obtain the first pixel threshold. and the second pixel threshold ; Threshold segmentation is performed on the texture feature values of pixels in hyperspectral remote sensing image data to obtain the first texture threshold. and the second texture threshold ; Thresholding was performed on the normalized water index and normalized vegetation index of pixels in the hyperspectral remote sensing image data to obtain the water index threshold. Vegetation index threshold ; The area labels include water areas, aquatic vegetation areas, terrestrial vegetation areas, and soil areas; Using the first pixel threshold, the second pixel threshold, the first texture threshold, the second texture threshold, the water index threshold, and the vegetation index threshold, a center vector is constructed for the water body region, the aquatic vegetation region, the terrestrial vegetation region, and the soil region.
[0008] In one embodiment, the construction of the center vectors for the water body region, aquatic vegetation region, terrestrial vegetation region, and soil region includes: The center vector of the water body region The central vector of the aquatic vegetation area The central vector of terrestrial vegetation regions , center vector of soil region They are represented as follows: , , , .
[0009] In one embodiment, obtaining the penalty for assigning each pixel to each region label includes: Calculate the squared difference between the feature vector of each pixel and the center vector of each region label for each corresponding positional component; take the square root of the weighted sum of all the squared differences between the feature vector of each pixel and the center vector of each region label as the penalty for assigning each pixel to each region label.
[0010] In one embodiment, during the calculation of the weighted sum of the squared results, the weights of the squared results are adjusted using the texture feature values and normalized vegetation index of each pixel.
[0011] In one embodiment, the method for adjusting the weights of the squared result using the texture feature values of each pixel and the normalized vegetation index is as follows: The texture feature value of each pixel and the normalized vegetation index are used as the exponent of an exponential function with the natural constant as the base. The reciprocal of the calculation result of the exponential function is used as the weight of the square result corresponding to the first component between each pixel and each region label; The calculation result of the exponential function is used as the weight of the squared result corresponding to the second component between each pixel and each region label.
[0012] In one embodiment, obtaining the penalty for assigning any two adjacent pixels to different region labels includes: Calculate the difference in feature vectors between any two adjacent pixels; The penalty for assigning any two adjacent pixels to different region labels is inversely proportional to the amount of difference.
[0013] In one embodiment, the penalty for assigning any two adjacent pixels to different region labels is represented by a weight function in the boundary term: ;in, Represents the weighting function; e represents the natural constant; , Let represent the feature vectors of the i-th and j-th pixels in the hyperspectral remote sensing image data, respectively. This represents the 2-norm operation; This represents the dispersion of grayscale values of all pixels in hyperspectral remote sensing image data.
[0014] This application has at least the following beneficial effects: This application constructs an energy function for a graph cut algorithm by comprehensively considering the color, texture, and spectral features of different regions in hyperspectral remote sensing data. This function can more comprehensively reflect the features of pixels. The optimized graph cut algorithm can more accurately segment water bodies, aquatic vegetation, terrestrial vegetation, and soil regions, avoiding the boundary blurring problem caused by vegetation occlusion. This allows for the identification of the true edge of water bodies and avoids the inaccurate segmentation problem caused by the reliance on a single feature in traditional graph cut methods. This ensures accurate monitoring of the dynamic changes of large areas of water at different times. Attached Figure Description
[0015] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 A flowchart illustrating the steps of the method for monitoring the spatiotemporal dynamic changes of large-scale water bodies that integrates remote sensing and GIS, as provided in this application. Figure 2 This is a schematic diagram of the process of dividing the water area to be monitored. Detailed Implementation
[0017] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.
[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".
[0019] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.
[0020] The following, in conjunction with the accompanying drawings, details the specific scheme of the method for monitoring the spatiotemporal dynamic changes of large-scale water bodies that integrates remote sensing and GIS, as provided in this application.
[0021] This application provides an embodiment of a method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by fusing remote sensing and GIS. Specifically, the method is described below. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps: Step 1: Acquire hyperspectral remote sensing image data of the water area to be monitored in real time.
[0022] Real-time hyperspectral remote sensing imagery and GIS (Geographic Information System) data of the water area to be monitored are acquired using UAV remote sensing and satellite remote sensing technologies. The hyperspectral remote sensing imagery data is used to monitor changes in the characteristics of the monitored water area at different times, including vegetation features on the water surface, water edges, within the water body, and on land at the water edges. The GIS data is vector data of the monitored water area, containing its geographical location information.
[0023] After acquiring hyperspectral remote sensing image data of the water area to be monitored, radiometric correction is performed on the hyperspectral remote sensing image data to eliminate radiometric distortion caused by factors such as illumination conditions. Subsequently, geometric correction is performed on the radiometrically corrected hyperspectral remote sensing image data to correct image distortion and positional shifts caused by changes in UAV flight attitude, terrain undulations, or lens distortion, ensuring accurate matching of the hyperspectral remote sensing image data with geographic coordinates and guaranteeing spatial consistency of multi-temporal and multi-source data. Both radiometric and geometric correction are well-known techniques and will not be elaborated upon in this application.
[0024] For GIS data, preprocessing operations including coordinate system unification, attribute cleaning, and topology repair are performed. Specifically, vector data is projected onto the WGS84 coordinate system to ensure spatial consistency; attribute fields are standardized, such as unifying units, completing missing information, and removing duplicate records; and topology checks are performed to repair overlapping or unclosed polygons, ensuring boundary closure. The specific processes for these preprocessing steps are well-known techniques and will not be elaborated upon in this application.
[0025] Step 2: For each time point, the graph cut algorithm is optimized by combining the color information, texture information and spectral features of the pixels, and then the optimized graph cut algorithm is used to segment the water area to be monitored.
[0026] During long-term monitoring of water bodies, aquatic plants at the water's edge, as seen in hyperspectral remote sensing imagery, can interfere with edge detection. While the area of the monitored water body can be detected normally during periods with sparse or no aquatic plants, traditional segmentation algorithms struggle to accurately detect water edges during periods of abundant aquatic vegetation. This can lead to misidentification as densely vegetated soil areas, affecting the identification of the true water body and causing errors during long-term dynamic monitoring.
[0027] To address the aforementioned issues, this application optimizes the graph cut algorithm by integrating color information, texture information, and spectral features. The optimized algorithm is then used to meticulously segment water, aquatic vegetation, terrestrial vegetation, and soil regions within hyperspectral remote sensing image data. Traditional graph cut algorithms typically construct energy functions based on pixel grayscale differences. However, in complex scenes, the grayscale differences of some pixels are not significant, leading to unclear boundaries and blurred segmentation during image segmentation. This results in inaccurate edge recognition of real water bodies, hindering accurate long-term dynamic monitoring of large water areas.
[0028] By comprehensively considering the color, texture, and spectral features of different regions in hyperspectral remote sensing image data, an energy function for graph cut algorithm is constructed, thereby more accurately calculating the edge weights of the graph and avoiding the problem of blurred segmentation edges in different regions caused by the reliance on pixel grayscale differences in traditional graph cut algorithms.
[0029] Step 2.1: By combining the grayscale differences and texture feature differences between each pixel and its neighboring pixels in the hyperspectral remote sensing image data with the normalized water index and normalized vegetation index of each pixel, feature vectors of each pixel are constructed.
[0030] In the visible light spectrum, the color characteristics of different land features are significantly affected by their reflectivity. Pure water appears dark due to its high reflectivity in the blue-green band and strong absorption in the red band. Aquatic and terrestrial plants, rich in chlorophyll in their leaves, absorb red light strongly and reflect near-infrared light very highly, appearing as bright green or dark green in images. Soil appears yellow or brownish-yellow, with moist soil appearing darker overall due to water absorption. Color characteristics have significant limitations in complex scenes because aquatic plants and terrestrial vegetation have similar colors, making them difficult to distinguish using only RGB values, which can easily lead to misjudgment.
[0031] Based on the above analysis, a color space conversion method is used for each time point to obtain the color features of different regions in the hyperspectral remote sensing image data. Specifically, the hyperspectral remote sensing image data is converted to the CIE XYZ color space using a color space conversion method. This process involves integrating the spectral reflectance with the CIE color matching function to obtain the values of the X, Y, and Z components. Then, the obtained XYZ values are further converted to the sRGB standard color space to obtain an RGB image. The specific processes of integrating the spectral reflectance with the CIE color matching function and converting the obtained XYZ values to the sRGB standard color space are well-known techniques and will not be elaborated upon in this application. Since the RGB image is obtained by converting hyperspectral remote sensing image data, there is a one-to-one correspondence between the RGB image and the pixels in the hyperspectral remote sensing image data. The RGB image is then grayscaled, and the grayscale value is used as a simplified representation of the color. The grayscaled processing of the RGB image is also a well-known technique and will not be elaborated upon in this application.
[0032] Thresholding is performed on the grayscale values of pixels in the hyperspectral remote sensing image data to obtain the first pixel threshold. and the second pixel threshold The gray value of the i-th pixel in the hyperspectral remote sensing image data is denoted as... ,like The i-th pixel may belong to the water area. The i-th pixel may belong to a vegetation area. The i-th pixel may belong to the soil region.
[0033] In this embodiment, an image multi-threshold segmentation algorithm based on Otsu's method is used to obtain the first pixel threshold and the second pixel threshold. The image multi-threshold segmentation algorithm based on Otsu's method is a well-known technology and will not be described in detail in this application. Based on the ability to obtain the first pixel threshold and the second pixel threshold, the implementer may use other existing feasible technologies, and this application does not impose any special restrictions.
[0034] Furthermore, texture features are a key basis for distinguishing water bodies, vegetation, and soil, and their spatial distribution characteristics can be quantified using a gray-level co-occurrence matrix (GLCM). Calm water surfaces are uniform and smooth, while soil features exhibit smooth or fine-grained textures, resulting in low entropy and weak contrast in the GLCM. Aquatic plant cover areas, such as lotus leaves, appear as continuous circular patches in hyperspectral remote sensing imagery, showing significant variations in texture features but without forming dense, coarse textures, resulting in moderate entropy and contrast in the GLCM. In terrestrial vegetation, such as forests with their interlaced canopies and densely packed grasslands, the resulting coarse textures exhibit high entropy and high contrast in the GLCM.
[0035] Furthermore, the texture features of the grayscale image are quantized using a gray-level co-occurrence matrix. Specifically, taking the i-th pixel in the hyperspectral remote sensing image data as an example, the gray-level co-occurrence matrix is obtained based on the gray values of the pixels within a preset square window centered on the i-th pixel, and the entropy of the gray-level co-occurrence matrix is calculated. and contrast Among them, the gray-level co-occurrence matrix is a well-known technique, which will not be described in detail in this application. The entropy... and contrast The mean value is used as the texture feature value of the i-th pixel. .
[0036] Threshold segmentation is performed on the texture feature values of pixels in hyperspectral remote sensing image data to obtain the first texture threshold. and the second texture threshold ;like The i-th pixel may belong to a water area or a soil area. The i-th pixel may belong to the aquatic vegetation area. The i-th pixel may belong to a terrestrial vegetation area.
[0037] In this embodiment, the side length of the preset square window is 3. The side length of the preset square window is preset by the user and can be set by the implementer according to the actual situation. This application does not impose any special restrictions.
[0038] In this embodiment, an image multi-threshold segmentation algorithm based on Otsu's method is used to obtain the first texture threshold and the second texture threshold. The image multi-threshold segmentation algorithm based on Otsu's method is a well-known technology and will not be described in detail in this application. Based on the ability to obtain the first texture threshold and the second texture threshold, the implementer may use other existing feasible technologies, and this application does not impose any special restrictions.
[0039] Furthermore, spectral characteristics are a core basis for distinguishing ground features, especially with significant differences in the near-infrared (NIR) and short-wave infrared (SWIR) bands. Pure water bodies have extremely low reflectance in the NIR band and exhibit a high Normalized Difference Water Index (NDWI). Aquatic plants and terrestrial vegetation, due to the abrupt increase in reflectance in the NIR band caused by their leaf cell structure, exhibit a high Normalized Difference Vegetation Index (NDVI). NDWI is effective in identifying water bodies; NDVI identifies vegetation but cannot distinguish between aquatic and terrestrial categories. However, subtle differences exist between aquatic and terrestrial vegetation areas. The gaps between aquatic plants are affected by water reflection, influencing their NDVI values, and NDWI can be detected. In terrestrial vegetation areas, continuous vegetation zones exhibit high NDVI but no NDWI, while soil areas show extremely low levels of both NDVI and NDWI.
[0040] Based on the above analysis, if a certain area in hyperspectral remote sensing image data has a coarse texture and high NDVI, it may be a terrestrial plant area. If a certain area has a fine but not smooth texture and high NDVI and slight NDWI, it may be an aquatic vegetation area.
[0041] Furthermore, by analyzing the reflectance of each pixel in the hyperspectral remote sensing image data in the near-infrared, red, and green bands, the normalized water index and normalized vegetation index of each pixel are obtained, expressed as follows: ; In the formula, , Representing the i-th pixel Normalized water index and normalized vegetation index; This represents the average reflectance of the i-th pixel across all wavelengths in the green light band. This represents the average reflectance of the i-th pixel across all wavelengths in the near-infrared band. This represents the average reflectance of the i-th pixel across all wavelengths in the red light band.
[0042] Thresholding was performed on the normalized water index and normalized vegetation index of pixels in the hyperspectral remote sensing image data to obtain the water index threshold. Vegetation index threshold ;when , When, the i-th pixel may belong to a vegetation area; when , At that time, the i-th pixel may belong to a non-vegetated area.
[0043] In this embodiment, the Otsu threshold segmentation algorithm is used to obtain the water index threshold and the vegetation index threshold respectively. The Otsu threshold segmentation algorithm is a well-known technology and will not be described in detail in this application. Based on the ability to obtain the water index threshold and the vegetation index threshold, the implementer may use other existing feasible technologies, and this application does not impose any special restrictions.
[0044] Furthermore, by fusing the multidimensional features of each pixel in the hyperspectral remote sensing image data, the feature vector of each pixel is obtained, specifically as follows: ;in, This represents the feature vector of the i-th pixel in the hyperspectral remote sensing image data; , , , This represents the grayscale value, texture feature value, normalized vegetation index, and normalized water index of the i-th pixel in the hyperspectral remote sensing image data.
[0045] Step 2.2: Construct the center vector of each region label; construct the energy function of the graph cut algorithm for segmenting the water area to be monitored. In the region term of the energy function, by comparing the feature vector of each pixel with the center vector of each region label, obtain the penalty for assigning each pixel to each region label; in the boundary term of the energy function, by comparing the feature vectors of any two adjacent pixels, obtain the penalty for assigning any two adjacent pixels to different region labels; segment the water area to be monitored at each time step using the energy function.
[0046] The graph cut algorithm maps pixels in an image to a graph, uses an energy function to determine the weights of edges in the graph, and constructs a weighted graph. Then, the maximum flow algorithm is used to find the maximum flow and minimum cut of the weighted graph. The weighted graph is partitioned according to the minimum cut, and finally, the partitioning result of the weighted graph is mapped to the image segmentation result, completing the region segmentation of the image.
[0047] Traditional graph cut algorithms typically construct energy functions using pixel grayscale differences. However, in complex scenes, the grayscale differences of some pixels are not significant, leading to unclear boundaries and blurred segmentation when segmenting regions of the image. This, in turn, results in inaccurate edge recognition of real water bodies, making it difficult to accurately monitor the long-term dynamic changes of large water areas. To address this issue, this paper analyzes the color, texture, and spectral features of pixels in different regions of hyperspectral remote sensing image data. By integrating multiple features of pixels instead of their grayscale values, an energy function is constructed. This allows for more accurate determination of edge weights when building a weighted graph, comprehensively quantifying the differences between pixels.
[0048] The energy function is constructed by including a domain term and a boundary term, and is expressed as follows: ;in, Represents the energy function; This is a region item, representing region segmentation information. These are boundary items, representing boundary segmentation information. It is the scaling factor of the energy function, which is set to 1 in this embodiment. Implementers can set it according to the actual situation. This application does not impose any special restrictions.
[0049] First, we define region labels. In this embodiment, there are four region labels: water body region, aquatic vegetation region, terrestrial vegetation region, and soil region. ,in, Represents a water body area. Represents aquatic vegetation areas. Represents terrestrial vegetation areas. This represents a soil region. For each region label, a center vector is set. center vector The expression is: ;in, , , , These represent the area labels. The first, second, third, and fourth components of the center vector; the center vector of the water body region. The central vector of the aquatic vegetation area The central vector of terrestrial vegetation regions , center vector of soil region Specifically, they are expressed as follows: ; ; ; ; in, , , , These represent the area labels. The first, second, third, and fourth components in the center vector; , , , These represent the area labels. The first, second, third, and fourth components in the center vector; , , , These represent the area labels. The first, second, third, and fourth components in the center vector; , , , These represent the area labels. The first, second, third, and fourth components in the center vector; Indicates the threshold value of the first pixel; Indicates the threshold value for the second pixel; Indicates the first texture threshold; Indicates the second texture threshold; Indicates the water index threshold; This represents the vegetation index threshold.
[0050] The region term measures the cost of a pixel belonging to a certain region label. It is jointly modeled based on the color, texture, and spectral features of pixels in hyperspectral remote sensing image data, and is represented as follows: Where P is the set of all pixels in the hyperspectral remote sensing image data. This represents assigning the i-th pixel in the hyperspectral remote sensing image data to a region label. The penalty is determined by comparing the feature vector of the i-th pixel in the hyperspectral remote sensing image data with the region label. Obtaining the center vector, Represented as: ;in, This indicates the adjustment weight, specifically: In vegetated areas, aquatic and terrestrial vegetation are difficult to distinguish accurately in color space, and the values of texture and spectral characteristics tend to be medium or high. Higher values can reduce the role of color features in region labeling and increase the role of texture and spectral features, thus improving the accuracy of separating aquatic and terrestrial vegetation regions. For water and soil regions, texture and spectral feature values are lower, and they are easily distinguishable by color features. This will decrease, thereby increasing the role of color features in classifying water and soil areas.
[0051] The boundary term measures the cost of assigning different region labels to two adjacent pixels, and is expressed as: Where N represents the set of all adjacent pixel pairs in the hyperspectral remote sensing image data; , Let represent the i-th pixel and the j-th pixel in the hyperspectral remote sensing image data, respectively. The weighting function represents the penalty for assigning adjacent ith and j pixels in hyperspectral remote sensing image data to different region labels. The greater the difference between the feature vectors of the ith and j pixels, the greater the probability that the ith and j pixels will be assigned to different region labels. This represents the maximum penalty value for the boundary term, when... hour, ,when hour, ; , These represent the region labels assigned to the i-th and j-th pixels, respectively.
[0052] Where e represents the natural constant; , Let represent the feature vectors of the i-th and j-th pixels in the hyperspectral remote sensing image data, respectively. This represents the 2-norm operation; This represents the dispersion of grayscale values of all pixels in hyperspectral remote sensing image data. By comprehensively considering the color, texture, and spectral features of pixels, it is possible to better distinguish between pixels and avoid the problem of blurred boundaries in graph cut results caused by constructing graph edge weights solely based on pixel variations in color space.
[0053] In this embodiment, the dispersion is variance. As other implementation methods, based on the ability to measure the unevenness of gray value distribution, implementers may use other existing technologies, such as standard deviation, coefficient of variation, etc. This application does not impose any special restrictions.
[0054] Furthermore, the edge weights of the graph are determined by the energy function, thereby enabling the construction of a weighted graph.
[0055] Furthermore, the Ford-Fulkerson algorithm is used to find the maximum flow and minimum cut of the weighted graph. The weighted graph is then partitioned based on the minimum cut, and the partitioned weighted graph is mapped to the segmentation results of the hyperspectral remote sensing image data. The process of using the Ford-Fulkerson algorithm to find the maximum flow and minimum cut of the weighted graph, and then partitioning the weighted graph, is a well-known technique and will not be elaborated upon here. A schematic diagram of the segmentation process of the monitored water area is shown below. Figure 2 As shown.
[0056] Step 3: Monitor the spatiotemporal dynamic changes of the water area to be monitored by using the segmentation results at different times.
[0057] Mask data of the actual water body is extracted from hyperspectral remote sensing image data of segmented water body areas and aquatic vegetation areas; the water body mask data at different times are combined with GIS data for geospatial alignment, and the change in water area at different times is calculated by spatial overlay. The process of extracting the water body mask data is a known technique and will not be described further in this application; spatial overlay is also a known technique and will not be described further in this application.
[0058] Furthermore, a pixel-level edge displacement vector method is used to quantify the direction and amplitude of water boundary displacement, and a heat map of the spatiotemporal changes of the water body is generated, thereby realizing the spatiotemporal dynamic monitoring of large-area water bodies. The edge displacement vector method is a known technique and will not be described further in this application.
[0059] In summary, this application constructs an energy function for the graph cut algorithm by comprehensively considering the color, texture, and spectral features of different regions in hyperspectral remote sensing data. This function can more comprehensively reflect the features of pixels. Based on the optimized graph cut algorithm, water bodies, aquatic vegetation, terrestrial vegetation, and soil regions can be segmented more accurately, avoiding the boundary blurring problem caused by vegetation occlusion. This allows for the identification of the true edge of water areas and avoids the inaccurate segmentation problem caused by the reliance on a single feature in traditional graph cut methods. This ensures accurate monitoring of the dynamic changes of large-area water areas at different times.
[0060] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0061] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.
Claims
1. A method for monitoring large-scale water body spatio-temporal dynamic change by fusing remote sensing and GIS, characterized in that, The method comprises the following steps: Real-time acquisition of hyperspectral remote sensing image data of the water area to be monitored; For each time, the feature vector of each pixel point in the hyperspectral remote sensing image data and the center vector of each region label in the water area to be monitored are constructed by comparing the gray difference and texture feature difference between each pixel point and its adjacent pixel points, and combining the normalized water index and the normalized vegetation index of each pixel point; An energy function of a graph cut algorithm for segmenting the water area to be monitored is constructed, in the region term of the energy function, the penalty of assigning each pixel point to each region label is obtained by comparing the feature vector of each pixel point and the center vector of each region label; in the boundary term of the energy function, the penalty of assigning any two adjacent pixel points to different region labels is obtained by comparing the feature vectors of the two adjacent pixel points. The water area to be monitored is segmented at each time by the energy function, and the spatio-temporal dynamic change of the water area to be monitored is monitored by the segmentation results at different times.
2. The method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS as described in claim 1, characterized in that, The feature vector is composed of the gray value of each pixel point, the texture feature value of each pixel point, the normalized water index and the normalized vegetation index. The texture feature value is obtained by the gray distribution of each pixel point and its adjacent pixel points.
3. The method according to claim 2, wherein the method comprises the following steps: 1) collecting remote sensing data of the large-scale water area; 2) collecting GIS data of the large-scale water area; 3) fusing the remote sensing data and the GIS data; 4) monitoring the spatio-temporal dynamic changes of the large-scale water area. The process of obtaining the texture feature value is as follows: According to the gray values of each pixel point and its preset adjacent pixel points, a gray level co-occurrence matrix is constructed, and the texture feature value is the mean value of the entropy and contrast of the gray level co-occurrence matrix corresponding to each pixel point.
4. The method according to claim 2, wherein the method comprises the steps of: The process of constructing the center vector of each region label is as follows: Thresholding is performed on the grayscale values of pixels in the hyperspectral remote sensing image data to obtain the first pixel threshold. and the second pixel threshold ; Threshold segmentation is performed on the texture feature values of pixels in hyperspectral remote sensing image data to obtain the first texture threshold. and the second texture threshold ; The normalized water index and the normalized vegetation index of pixel points in the hyperspectral remote sensing image data are respectively subjected to threshold segmentation to obtain water index threshold values and vegetation index threshold values ; The region label includes water area, aquatic vegetation area, terrestrial vegetation area and soil area. The center vectors of the water area, aquatic vegetation area, terrestrial vegetation area and soil area are constructed by the first pixel threshold, the second pixel threshold, the first texture threshold, the second texture threshold, the water index threshold and the vegetation index threshold.
5. The method according to claim 4, wherein the method comprises the steps of: The center vectors of the water area, aquatic vegetation area, terrestrial vegetation area and soil area are constructed by the first pixel threshold, the second pixel threshold, the first texture threshold, the second texture threshold, the water index threshold and the vegetation index threshold. a center vector of the water body area a center vector of the aquatic vegetation area a center vector of the terrestrial vegetation area a center vector of the soil area are respectively represented as: , , , .
6. The method according to claim 1, wherein the method comprises the steps of: The process of constructing the center vectors of the water area, aquatic vegetation area, terrestrial vegetation area and soil area comprises: The process of obtaining the penalty of assigning each pixel point to each region label comprises:
7. The method according to claim 6, wherein the method comprises the steps of: The square result of the difference value of each same position component between the feature vector of each pixel point and the center vector of each region label is calculated; the square root of the weighted sum of all the square results between the feature vector of each pixel point and the center vector of each region label is taken as the penalty of assigning each pixel point to each region label.
8. The method according to claim 7, wherein the method comprises the steps of: In the process of calculating the weighted sum of the square results, the weight of the square result is adjusted by the texture feature value and the normalized vegetation index of each pixel point. The method of adjusting the weight of the square result by the texture feature value and the normalized vegetation index of each pixel point is as follows: The texture feature value and the normalized vegetation index of each pixel point are taken as the exponent of an exponential function with a natural constant as the base; The reciprocal of the calculation result of the exponential function is taken as the weight of the square result corresponding to the first component between each pixel point and each region label; The calculation result of the exponential function is taken as the weight of the square result corresponding to the second component between each pixel point and each region label.
9. The method for monitoring the spatiotemporal dynamic changes of large-scale water bodies by integrating remote sensing and GIS as described in claim 1, characterized in that, The penalty for assigning the arbitrary two adjacent pixel points to different region labels includes: calculating a difference amount of feature vectors of the arbitrary two adjacent pixel points; the penalty for assigning the arbitrary two adjacent pixel points to different region labels is inversely proportional to the difference amount.
10. The method according to claim 9, wherein the method comprises the steps of: The penalty for assigning the arbitrary two adjacent pixel points to different region labels is represented by a weight function in a boundary term: ; wherein, represents a weight function; e represents a natural constant; , respectively represent the feature vector of the i-th pixel point and the j-th pixel point in the hyperspectral remote sensing image data; represents the 2-norm operation; represents the dispersion of the gray values of all pixel points in the hyperspectral remote sensing image data.