Water resource information acquisition method based on remote sensing image

Abstract

The invention discloses a water resource information acquisition method based on a remote sensing image, which belongs to the technical field of image recognition, and specifically comprises the following steps: firstly, acquiring remote sensing image data of a target water area, and constructing a dynamic substrate spectrum template library by acquiring actually measured spectrum reflection curves of different substrate types in a water-free state; secondly, for each spatial position in the image, matching exposed substrate types of adjacent areas according to geographic coordinates, and extracting corresponding spectral features from a template library as basic reflection reference; establishing a light propagation path separation model, simulating a process that sunlight penetrates through a water layer and returns to a sensor through substrate reflection, and generating a substrate reflection contribution component in combination with basic reflection reference; and finally, deducting the contribution component from the spectral reflectivity to obtain a pure water optical response value, and outputting a water transparency parameter distribution diagram of each spatial position of the target water area according to a physical correlation model of the pure water optical response value and the water transparency. According to the invention, reliable remote sensing acquisition of shallow water transparency is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition, and in particular to a water resources information collection method based on remote sensing images. Background Art

[0002] Water resources are a core element in maintaining ecological security and human development. Obtaining key parameters such as water transparency is crucial for water environmental protection, water resource management, and ecological assessment. Remote sensing technology, leveraging its advantages in large-scale, simultaneous observations, has become a key means of collecting water resource information. Its ability to infer water transparency relies on accurately extracting the inherent water signal from spectral reflectance. However, in complex waters, such as shallow and nearshore areas, the spatial heterogeneity and dynamic changes in bottom sediment types significantly interfere with spectral signals, limiting the accuracy of traditional remote sensing methods.

[0003] Existing remote sensing inversion methods are mostly based on statistical or physical models of spectral reflectance and transparency, estimating parameters through band combinations or optical equations. To address sediment interference, some techniques employ preset thresholds or homogenization assumptions to separate signals, such as by dividing regions according to water depth and applying empirical corrections. Other studies have established sediment spectral libraries to match typical features, but these methods generally ignore the heterogeneity of sediments in spatial distribution, such as the interweaving of sand, mud, and gravel, and in temporal dimensions, such as the alternation of sediments caused by erosion and tides. They also fail to fully consider the complex influence of light coupled propagation at the water-sediment interface on the spectrum.

[0004] Shallow water bottoms, such as sandy and muddy, are affected by the sedimentary environment and water erosion, resulting in significant differences in particle composition, organic matter content, and surface state. This leads to complex variations in the bottom reflectance spectrum in different areas, with the reflectance of sandy bottoms significantly higher than that of muddy bottoms. These bottom reflection signals are superimposed on the water scattering and absorption signals to form a mixed spectrum. However, existing models assume that the optical properties of the bottom are uniform and stable, and fail to effectively characterize heterogeneity such as the staggered distribution of different bottom types and the gradient changes in coarse and fine particles within the same water area. As a result, it is impossible to accurately separate bottom reflections from inherent water signals during transparency inversion. For example, adjacent high-reflection sandy bottoms and low-reflection muddy bottoms may be misinterpreted by traditional models as differences in water turbidity, resulting in systematic deviations in the inversion results, affecting the reliability of shallow water quality assessment and ecological monitoring. Summary of the Invention

[0005] The purpose of this invention is to provide a method for collecting water resources information based on remote sensing images to solve the following technical problems:

[0006] Existing inversion models usually assume that the optical properties of the substrate are uniform and stable, ignoring the heterogeneity of the spatial distribution of the substrate, which leads to deviations in the inversion results of water transparency.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] A method for collecting water resources information based on remote sensing images, comprising the following steps:

[0009] Acquiring remote sensing image data of the target water area, wherein the remote sensing image data includes spectral reflectance corresponding to multiple spatial locations;

[0010] Constructing a dynamic substrate spectrum template library, wherein the dynamic substrate spectrum template library is established by collecting measured spectral reflectance curves of different substrate types in a water-free state, and each substrate type corresponds to an independent spectral feature data set;

[0011] For each spatial location in the remote sensing image, the exposed substrate type of the adjacent area is matched according to the geographic coordinates of the spatial location, and the spectral feature dataset of the corresponding substrate type is extracted from the dynamic substrate spectrum template library as the basic reflectance reference;

[0012] Establishing a light propagation path separation model, which simulates the complete path of sunlight passing through the water layer, reaching the bottom surface, and then returning to the sensor after secondary reflection, and generates a bottom reflection contribution component in combination with a basic reflection reference;

[0013] The bottom reflection contribution component is deducted from the spectral reflectance of the remote sensing image data to obtain the pure water optical response value. Based on the physical correlation model between the pure water optical response value and water transparency, a water transparency parameter distribution map of the spatial position of the target water area is output.

[0014] As a further solution of the present invention: the process of constructing a dynamic bottom sediment spectrum template library is:

[0015] Conduct field surveys in the target waters during the dry season and select shallow areas where the water level is below the natural threshold as exposed bottom sampling areas. Grid the exposed bottom sampling areas, collect physical samples of surface sediments from each grid cell, and record the geographic coordinates of the physical samples.

[0016] Measure the continuous spectral reflectance curve of the physical samples in the visible to near-infrared band in a dry state; classify the physical samples into sandy substrate type, muddy substrate type, gravel substrate type, and mixed substrate type based on the sediment particle size distribution and organic matter content; perform cluster analysis on the spectral reflectance curves of substrate samples of the same type, and eliminate abnormal spectral reflectance curves with a dispersion exceeding the set threshold;

[0017] The weighted average value of the band of the retained spectral reflectance curve is calculated to form a standardized spectral reflectance set of the substrate type in the set band; the standardized spectral reflectance set is associated with the corresponding geographic coordinates and stored to form a spatially indexed dynamic substrate spectral template library.

[0018] As a further solution of the present invention: the process of matching the exposed substrate types of adjacent areas is:

[0019] A search buffer zone is established with the current spatial location to be processed as the center, wherein the radius of the search buffer zone is determined according to the historical maximum water receding range of the water area; all exposed bottom sediment sampling points located within the search buffer zone are retrieved from the dynamic bottom sediment spectrum template library;

[0020] Calculate the surface distance and elevation difference between the exposed bottom sampling point and the current spatial position, and exclude the exposed bottom sampling points whose elevation difference exceeds the terrain relief threshold; perform frequency statistics on the bottom types of the remaining exposed bottom sampling points, and select the bottom type with the highest frequency as the matching result; when there are multiple bottom types with the same frequency, select the bottom type of the exposed bottom sampling point with the smallest surface distance from the current spatial position;

[0021] If there is no exposed bottom sediment sampling point within the search buffer zone, the search will be extended to the typical bottom sediment distribution map of the basin to which the target water area belongs for analogical matching.

[0022] As a further solution of the present invention: the process of generating the bottom sediment reflection contribution component by the light propagation path separation model is as follows:

[0023] The incident angle of sunlight and the sensor observation angle are calculated based on the imaging time of the remote sensing image and the satellite orbit. The propagation process of the incident light penetrating the water layer to the bottom surface is simulated to calculate the first propagation attenuation. The propagation process of the light reflected by the bottom penetrating the water layer to the sensor is simulated to calculate the second propagation attenuation.

[0024] The first propagation attenuation and the second propagation attenuation are applied to the basic reflection reference value to generate a bottom sediment reflection contribution component. A water surface mirror reflection correction factor is introduced to dynamically adjust the bottom sediment reflection contribution component. The water surface mirror reflection correction factor is generated by a wind and wave disturbance model combined with real-time meteorological data.

[0025] As a further solution of the present invention: the calculation process of the propagation attenuation is:

[0026] The area in the target water area with a water depth greater than the optical depth is selected as the pure water reference area; the spectral reflectance of the pure water reference area in the remote sensing image is extracted as the pure water signal without bottom sediment interference;

[0027] Establishing a parameter fitting relationship between a pure water signal and a theoretical water absorption model, wherein the theoretical water absorption model includes the absorption bands of phytoplankton pigments, suspended particulate matter, and dissolved organic matter;

[0028] The parameters of the theoretical water absorption model are adjusted through a nonlinear optimization algorithm to minimize the root mean square error between the simulated spectral curve and the measured spectral curve. The optimized theoretical water absorption model parameters are used as the benchmark values ​​of the spectral absorption characteristics of the current water area.

[0029] As a further solution of the present invention: the specific selection process of the pure water reference area is:

[0030] The stability of the spectral reflectance of a selected pure water reference area is analyzed in a time series. When the deviation between the reflectance trend of a specific band in the reference area and the inversion results of the water transparency physical correlation model shown by multiple consecutive remote sensing images continues to increase, the reference area reselection mechanism is activated.

[0031] Based on the spatial distribution pattern of historical propagation attenuation, the optical stability index of each deep-water area in the target waters is calculated. The optical stability index reflects the correlation between water depth and the ability to suppress bottom reflection. The candidate area with the highest optical stability index is selected as the new pure water reference area, ensuring that the water depth in the candidate area is greater than that of the original reference area and the fluctuation amplitude of historical propagation attenuation is below the set tolerance threshold.

[0032] The spectral reflectance of the new pure water reference area is extracted, the theoretical water absorption model parameters are refitted, and the optimized theoretical water absorption model parameters are updated to the spectral absorption characteristic benchmark values ​​of the light propagation path separation model.

[0033] As a further solution of the present invention: the calculation process of the optical response value of the pure water body is:

[0034] Obtain the total spectral reflectance value of the remote sensing image in the set band, calculate the bottom sediment reflection contribution component at the current spatial position, and determine the direct reflection component of the water surface;

[0035] The pure water optical response value is obtained by deducting the bottom reflection contribution component and the water surface direct reflection component from the total spectral reflectance value;

[0036] When the optical response value of pure water is lower than the preset noise threshold, the attenuation gradient of the adjacent spatial position is used for spatial interpolation compensation, and the atmospheric scattering effect correction is performed on the optical response value of pure water after spatial interpolation compensation.

[0037] As a further solution of the present invention: the construction process of the physical association model is:

[0038] Deploy a fixed monitoring buoy array in the target waters to synchronously obtain field measurements of water transparency at different spatial locations, and extract the pure water optical response values ​​at the corresponding spatial locations at the time of remote sensing image acquisition;

[0039] Analyzing the statistical correlation between the specific blue-green band ratio of the pure water optical response value and the field measurement value of the water transparency; constructing a piecewise linear regression equation, wherein the piecewise linear regression equation adopts a first slope linear relationship when the pure water optical response value is in a low value range, and adopts a second slope linear relationship when the pure water optical response value is in a high value range;

[0040] The calibration sub-areas were divided into hydrogeological units, and the parameters of the piecewise linear regression equation were optimized for each calibration sub-area.

[0041] As a further solution of the present invention, it also includes obtaining multiple periods of remote sensing image data of the target water area in a continuous time series, independently generating a water transparency parameter distribution map for each period of remote sensing image data, and establishing a spatiotemporal change analysis matrix to detect sudden changes in water transparency at the same spatial location;

[0042] When the duration of the mutation point exceeds the set period and the change amplitude exceeds the natural fluctuation threshold, the sediment spectrum template library update process is triggered;

[0043] The process of updating the substrate spectrum template library includes re-collecting exposed substrate samples at a spatial location, measuring the spectral reflectance curves of the new exposed substrate samples and updating the dynamic substrate spectrum template library data, and using the updated dynamic substrate spectrum template library to recalculate the water transparency parameters of the historical period to generate a consistency-corrected time series dataset.

[0044] Beneficial effects of the present invention:

[0045] The present invention constructs a dynamic bottom sediment spectrum template library, combines it with the exposed bottom type matching in adjacent areas, and accurately obtains the bottom sediment spectrum characteristic benchmark, thereby solving the problem of inaccurate reflection reference caused by the spatiotemporal heterogeneity of bottom sediment types; through the light propagation path separation model, the complete path of light penetrating the water layer, bottom sediment reflection, and returning to the sensor is simulated, and the bottom sediment reflection contribution component is accurately generated, effectively separating the bottom sediment reflection and the inherent optical signal of the water body, overcoming the defect of the existing model that assumes the bottom sediment is uniform and cannot quantify the bottom sediment contribution; by deducting the bottom sediment reflection contribution component from the spectral reflectance, the pure water body optical response value is obtained, and the transparency is inverted by combining the physical correlation model, which significantly reduces the inversion deviation caused by bottom sediment interference; at the same time, through time series analysis and the dynamic update mechanism of the template library, it can adapt to the changes of bottom sediment type with hydrological conditions, ensure the spatial consistency and temporal stability of water transparency parameters in long-term monitoring, and ultimately achieve high-precision and high-reliability remote sensing acquisition of water transparency in shallow water areas. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The present invention will be further described below with reference to the accompanying drawings.

[0047] Figure 1 It is a schematic flow diagram of the present invention. DETAILED DESCRIPTION

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] See also Figure 1 As shown, the present invention is a water resources information collection method based on remote sensing images, comprising the following steps:

[0050] First, remote sensing image data of the target waters is obtained. This type of data covers multiple bands from visible light to near-infrared, and records the spectral reflectance information of multiple spatial locations in the target waters at different wavelengths.

[0051] When constructing a dynamic substrate spectral template library, we selected shallow, exposed substrate areas where the water level falls below the natural threshold during the dry season as sampling areas. Surface sediment samples were collected in a gridded manner, and their geographic coordinates were recorded. The continuous spectral reflectance of the samples in their dry state was measured, and the samples were classified into sandy and muddy types based on particle size distribution and organic matter content. Cluster analysis was used to remove anomalous spectra, and the weighted average of the retained spectra was calculated to form a standardized spectral set and store it in association with geographic coordinates. This dynamic template library, with spatial indexing, was constructed to adapt to the spatiotemporal changes in the substrate.

[0052] For each spatial location in the image, a search buffer is set based on the historical maximum water receding range and its coordinates. Sample points with exposed sediment within the buffer are retrieved. After eliminating outliers with large elevation differences, the frequency of sediment types at the remaining sampling points is counted, and the type with the highest frequency or closest proximity is selected as the matching result. If no data is available in the buffer, a typical sediment distribution map of the watershed is used for matching, and the spectral characteristics of the corresponding sediment type are obtained as a basic reflectance reference.

[0053] A light propagation path separation model is established. The solar incidence and sensor observation angles are determined according to the imaging time and satellite orbit parameters. The process of light penetrating the water layer to reach the bottom surface, penetrating the water layer again after reflection and returning to the sensor is simulated. The contribution of bottom reflection to the spectrum is quantified by combining the basic reflection reference to generate the bottom reflection contribution component.

[0054] The bottom reflectance contribution is deducted from the spectral reflectance to obtain the pure water optical response value. Fixed monitoring points are deployed in the target waters to simultaneously obtain the measured transparency values ​​and the pure water optical response values ​​at the corresponding locations. A physical correlation model is established between the two, and the optical response values ​​are converted into transparency parameters. Ultimately, a transparency parameter distribution map is output for each spatial location in the target waters.

[0055] In a preferred embodiment of the present invention, the process of constructing a dynamic sediment spectrum template library is as follows:

[0056] Field surveys are conducted during the dry season of the target waters, when water levels drop below natural thresholds and the bottom of shallow areas is exposed, providing ideal conditions for directly collecting bottom samples uncovered by the water. Representative shallow areas are selected as exposed bottom sampling areas, covering the main types of bottom sediments likely to occur in the target waters. For example, muddy bottoms are likely to form where rivers enter lakes, sandy bottoms are more common in lakeside shallows, and gravelly and mixed bottoms may occur in river mouths. The sampling area is gridded, with grid density determined based on the spatial variability of bottom sediment types. A coarser grid is used in areas with flat terrain and relatively uniform bottom sediment composition, while a denser grid is used in areas with complex terrain and a staggered distribution of bottom sediment types. This ensures that the collected samples fully reflect the spatial heterogeneity of the bottom sediments in the target waters.

[0057] When collecting physical samples of surface sediments from each grid cell, uniform sampling specifications are followed: Non-disturbance sampling tools are used to obtain surface sediments at a depth of 0-10 cm to prevent the incorporation of deeper sediments that could affect sample representativeness; the geographic coordinates of the samples are simultaneously recorded, and high-precision positioning equipment is used to ensure sub-meter accuracy, providing a basis for subsequently accurately correlating spectral data with spatial locations. The obtained physical samples are naturally air-dried in a laboratory environment to remove moisture interference with spectral measurements. Spectral measurement equipment is then used to perform continuous spectral reflectance scans of the dried samples from the visible to near-infrared bands, typically covering a wavelength range of 400-2500 nanometers. This allows the differences in reflectance characteristics of different substrates across the entire spectral range to be captured. For example, sandy substrates have higher reflectance in the visible band, while muddy substrates have lower reflectance in the near-infrared band.

[0058] The sample classification process is based on the physical and chemical properties of the sediments: first, the ratio of sand, silt, and clay in the sediments is determined through particle size analysis. Combined with the detection of organic matter content, the samples are measured using methods such as loss on ignition to classify the samples into sandy, muddy, gravelly, and mixed substrate types. Sandy substrates are defined as those with a sand content of more than 50%, muddy substrates as those with a clay content of more than 30%, gravelly substrates as those with a gravel content of more than 40%, and mixed substrates as those with a balanced distribution of multiple particle components. A cluster analysis algorithm is used to clean the spectral reflectance curves of the same type of substrate samples. By calculating the Euclidean distance or correlation coefficient between the spectral curves, anomalous spectral curves with a dispersion significantly higher than the group mean are identified and eliminated. These anomalous data may be caused by contamination during the sampling process or temporary errors in the measuring equipment. Eliminating them can improve the reliability of the spectral template.

[0059] The retained valid spectral curves are statistically processed to calculate the weighted average of each band. The weighting factors are set according to the band response characteristics of the remote sensing sensor, for example, enhancing sensitivity to the blue, green, red, and near-infrared bands commonly used in remote sensing imagery. This results in a standardized set of spectral reflectances for each substrate type in the specified bands. Finally, the standardized spectral data is associated with the corresponding geographic coordinates and stored to construct a dynamic, spatially indexed substrate spectral template library. This template library allows users to quickly retrieve substrate spectral characteristics of adjacent areas by geographic coordinates and can be dynamically updated through subsequent supplementary field sampling data, ensuring that the substrate spectral information remains synchronized with changes in substrate types caused by natural processes such as erosion and deposition in the target waters.

[0060] In another preferred embodiment of the present invention, the process of matching the exposed substrate types of adjacent areas is as follows:

[0061] For each spatial location to be processed in the remote sensing image, a search buffer is established with the geographic coordinates of that location as the center. The buffer radius is determined based on the historical maximum water withdrawal range of the target water area. This range is obtained by analyzing years of water level monitoring data for the target water area or using a watershed hydrological model to simulate water level changes under different hydrological conditions. This ensures that the buffer covers areas of exposed sediments that may be exposed during the dry season and avoids missing sediment samples due to water level fluctuations. All exposed sediment sampling points that fall within the search buffer are retrieved from the dynamic sediment spectral template library. These sampling points contain historically collected sediment types and their corresponding spectral characteristics.

[0062] In order to eliminate the interference of terrain factors on the matching of substrate types, it is necessary to calculate the surface distance and elevation difference between each exposed substrate sampling point and the current spatial position. The surface distance is calculated using the plane Euclidean distance, and the elevation difference is determined by comparing the altitude difference between the two. A terrain undulation threshold is set, such as 5 meters, and sampling points with elevation differences exceeding the threshold are eliminated, because significant terrain differences may lead to systematic changes in substrate types. For example, gravel substrates are mostly distributed in high-altitude areas, while muddy substrates are easily deposited in low-lying areas at low altitudes. There are obvious differences in optical properties between the two. If they are not eliminated, it will lead to deviations in matching results. The frequency of the substrate types of the remaining valid sampling points is counted, and the substrate type with the highest frequency is preferentially selected as the matching result of the current spatial position. This strategy reflects the principle of dominance of substrate types in spatial distribution, and believes that the dominant substrate type in the same area has a dominant influence on the substrate characteristics of the current position.

[0063] When there are multiple substrate types with the same frequency, a distance-first strategy is adopted: the substrate type of the sampling point with the shortest surface distance from the current spatial position is selected. It is believed that substrate types within a short distance have higher spatial correlation, which can reduce matching ambiguities caused by the staggered distribution of substrate types. For example, in an area where sandy and muddy substrates are staggered, the type of the sampling point closest to the current position is more likely to reflect the actual substrate conditions at that location. If there are no exposed substrate sampling points in the search buffer, such as areas where monitoring is being carried out for the first time or special areas after extreme hydrological events, the matching range is extended to the typical substrate distribution map of the basin to which the target water area belongs. This map is constructed based on historical monitoring data from basin geological surveys and expert knowledge. It records in detail the dominant substrate types and their distribution patterns in different sub-basins. Through analogical analysis, the most likely substrate type is determined by considering factors such as the similarity of geological causes and the consistency of water transport paths, ensuring that the matching process can still be carried out effectively in the absence of data.

[0064] Throughout the matching process, high-precision geographic coordinate positioning provides the foundation for spatial analysis. Dynamic adjustment of the buffer zone adapts to changing hydrological conditions. Topographic filtering eliminates interference from irrelevant data. A multi-level matching strategy addresses the identification of sediment types at scales from local to basin-wide. These technical approaches work together to form a sediment type identification method that adapts to complex terrain and hydrological conditions. This method leverages the accuracy of field sampling data and incorporates basin-scale sediment distribution patterns. It effectively addresses the matching challenges presented by the spatial heterogeneity of sediment types in shallow waters and lays a solid foundation for the subsequent precise separation of sediment reflection signals.

[0065] In another preferred embodiment of the present invention, the process of generating the bottom sediment reflection contribution component by the light propagation path separation model is as follows:

[0066] To generate the sediment reflectance contribution, the solar incident angle and sensor observation angle are first determined based on the remote sensing image's imaging time and satellite orbit parameters. These angular parameters, calculated using satellite platform ephemeris data and surface positioning information, are used to construct a geometric model of light propagation. Next, the propagation of incident light from the atmosphere into the water column, through the water layer, and onto the sediment surface is simulated. During this process, the light is absorbed and scattered by components in the water column, such as phytoplankton pigments, suspended particulate matter, and dissolved organic matter, resulting in energy attenuation. This attenuation is referred to as the first propagation attenuation. The propagation of light reflected from the sediment surface back through the water layer back to the sensor is then simulated. Here, the light is also attenuated by the aforementioned components in the water column, resulting in the second propagation attenuation. The simulation of these two attenuation processes requires consideration of the inherent optical properties of water: namely, the differences in the absorption and scattering coefficients of light of different wavelengths in water, which affect the degree of energy loss in the water layer.

[0067] The first propagation attenuation and the second propagation attenuation are applied to the basic reflection reference value obtained in the early stage, that is, the standardized spectral reflectance of the corresponding bottom type, and the energy loss of light in the water layer is calculated layer by layer to generate the contribution component of the bottom reflection signal to the sensor receiving spectrum. It is worth noting that the water surface state has a significant impact on the propagation of light, so the water surface mirror reflection correction factor is introduced to dynamically adjust the bottom reflection contribution component. This correction factor is generated by combining a wind and wave disturbance model with real-time meteorological data. The wind and wave disturbance model simulates the water surface roughness based on parameters such as wind speed and direction. The real-time meteorological data comes from ground meteorological stations or satellite inversion products, which are used to evaluate the changes in the intensity of the water surface mirror reflection to ensure that the interference of mirror reflection on the bottom signal can be accurately corrected under different hydrodynamic conditions, such as calm water surface and undulating waves.

[0068] In a preferred embodiment of the present invention, the calculation process of the propagation attenuation is as follows:

[0069] To accurately capture the spectral absorption characteristics of water, we first select an area within the target water body with a depth greater than the optical depth (OD) as a pure water reference zone. The OOD is the depth at which light energy decays to 1% of the surface layer. Beyond this depth, the bottom sediment reflection signal is negligible. Therefore, the selected reference zone is guaranteed to be free of interference from bottom sediment reflections. The spectral reflectance of the pure water reference zone is extracted from remote sensing imagery as a pure water signal free of bottom sediment interference. This signal, which contains only the scattering and absorption characteristics of the water itself, truly reflects the inherent optical properties of the water.

[0070] A parameter fitting relationship between the pure water signal and the theoretical water absorption model is established. The theoretical water absorption model takes into account the absorption characteristics of components such as phytoplankton pigments, suspended particulate matter, and dissolved organic matter in different bands. These components correspond to specific absorption spectral ranges. For example, phytoplankton pigments mainly absorb the blue light band, suspended particulate matter absorbs the green light band strongly, and dissolved organic matter absorbs significantly in the ultraviolet band. The model parameters are adjusted through a nonlinear optimization algorithm to minimize the root mean square error between the simulated spectral curve and the measured pure water signal spectral curve. This process is achieved through iterative calculation. After each parameter adjustment, the difference between the simulated value and the measured value is compared until the error converges to the preset accuracy range. Finally, the optimized model parameters are used as the baseline value of the spectral absorption characteristics of the current water area for the subsequent calculation of the propagation attenuation to ensure that the model can accurately characterize the optical absorption characteristics of the current water body.

[0071] In another preferred embodiment of the present invention, the specific selection process of the pure water reference area is as follows:

[0072] To ensure the long-term validity of the spectral absorption characteristic benchmark values, a dynamic management mechanism for pure water reference zones was designed. First, the stability of the spectral reflectance of selected reference zones was analyzed over time series. Specifically, the reflectance trends of specific reference zone bands, such as the blue and green bands, were compared across multiple consecutive remote sensing imagery periods with the inversion results of a physical correlation model for water transparency. When the deviation between the two values ​​continues to increase, indicating possible changes in the optical properties of the reference zone, such as changes in water composition or exposure to exposed sediments, the reference zone reselection mechanism is activated.

[0073] The reselection process is based on the spatial distribution pattern of historical propagation attenuation and calculates the optical stability index of each deep-water area in the target waters. This index comprehensively considers the correlation between water depth and the ability to suppress bottom reflection. The deeper the water, the smaller the impact of bottom reflection, and the higher the optical stability. During the specific calculation, the fluctuation amplitude of the propagation attenuation in each area is statistically analyzed in combination with historical data. The smaller the fluctuation, the higher the stability of the area. The candidate area with the highest optical stability index is selected as the new pure water reference area. Two key conditions must be met: first, the water depth of the candidate area is greater than the original reference area to ensure that the bottom reflection signal is effectively suppressed; second, the fluctuation amplitude of the historical propagation attenuation is lower than the set tolerance threshold to ensure the stability of the optical characteristics of the water body and avoid calculation deviations caused by fluctuations in the regional optical environment.

[0074] After identifying the new reference area, its spectral reflectance is extracted and the parameters of the theoretical water absorption model are refitted. The parameters are then recalibrated using a nonlinear optimization algorithm. The optimized model parameters are then updated into the light propagation path separation model as the new spectral absorption benchmark. This dynamic update mechanism ensures that the model can adapt to changes in the optical properties of the target waters due to factors such as seasonal changes and current disturbances, maintaining the accuracy of propagation attenuation calculations and ensuring long-term reliable calculation of the sediment reflectance contribution.

[0075] In another preferred embodiment of the present invention, the calculation process of the optical response value of pure water is:

[0076] To calculate the optical response of pure water, the total spectral reflectance value for a specified band is first extracted from the remote sensing image. This total value comprises multiple components, including the water's own optical signal, the contribution from the bottom sediment reflection, and direct surface reflection. Based on a previously constructed light propagation path separation model, the bottom sediment reflection contribution at the current spatial location is calculated. Simultaneously, the direct surface reflection component is determined by analyzing the relationship between the solar altitude and the water surface roughness. This signal originates from the specular reflection of sunlight on the water surface and is unrelated to the inherent optical properties of the water, requiring separate analysis.

[0077] The bottom reflection contribution component and the direct reflection component of the water surface are deducted from the total spectral reflectance value in sequence to obtain the pure water optical response value that only reflects the scattering and absorption characteristics of the water body. When the pure water optical response value of some spatial locations is lower than the preset noise threshold, it may be caused by sensor noise or extreme water conditions. At this time, the attenuation gradient of the adjacent area is used for spatial interpolation compensation. Based on the surrounding valid data points, the reasonable value of the target position is calculated according to the distance weight to ensure the continuity of the spatial data. After compensation, the pure water optical response value is corrected for the atmospheric scattering effect to eliminate the scattering interference of atmospheric molecules and aerosols on light, and finally a pure signal that can be directly used for transparency inversion is obtained.

[0078] In another preferred embodiment of the present invention, the process of constructing the physical association model is:

[0079] To construct a physical correlation model between pure water optical response values ​​and water transparency, an array of fixed monitoring buoys was deployed in the target waters, covering varying water depths, bottom types, and pollution gradients to ensure data representativeness. The buoys simultaneously recorded field transparency measurements, acquired using standard transparency measurement tools, and accurately labeled the geographic coordinates of the measurement locations. The pure water optical response values ​​at these locations at the time of remote sensing image acquisition were simultaneously extracted, forming a matching dataset of optical signals and transparency.

[0080] The statistical correlation between the ratio of specific blue-green bands of the optical response values ​​of pure water and the transparency measurements in the dataset was analyzed. The blue band is sensitive to suspended particulate matter in the water, while the green band is significantly affected by chlorophyll. The ratio of the two can comprehensively reflect the clarity of the water. Based on this correlation, a piecewise linear regression equation was constructed. When the optical response values ​​of pure water are in the low range, corresponding to high turbidity, a linear relationship with the first slope is used. When the optical response values ​​are in the high range, corresponding to high transparency, a linear relationship with the second slope is used to accommodate the nonlinear relationship between the optical signal and the actual value in different transparency ranges.

[0081] The target water area is divided into multiple calibration sub-regions based on hydrogeological units, such as mountain river sections, plain lakes, and estuary deltas. Each sub-region has different water composition and sediment backgrounds. The slope and intercept parameters of the piecewise linear regression equation are optimized for each sub-region. By comparing the deviation between the model inversion values ​​and field measurements, continuous adjustments are made to ensure that the inversion results in each sub-region are consistent with the local water characteristics.

[0082] In another preferred embodiment of the present invention, in order to achieve long-term dynamic monitoring, the invention also includes acquiring multiple periods of remote sensing image data of a continuous time series of the target water area, performing the aforementioned steps independently for each period of imagery, and generating a distribution map of water transparency parameters for the corresponding time period. Based on these distribution maps, a spatiotemporal change analysis matrix is ​​established, and by comparing the transparency values ​​of the same spatial location at different time periods, the presence of a mutation point is detected. A mutation point refers to an area where transparency fluctuates abnormally within a short period of time, which may be caused by changes in bottom sediment type, sudden pollution, or hydrological events.

[0083] When a mutation point persists for longer than a set period and its magnitude exceeds the natural fluctuation threshold, it indicates a systematic change in the sediment or water properties, not a random disturbance. This triggers an update of the sediment spectral template library. During the update, exposed sediment samples are recollected from the area where the mutation point is located, prioritizing exposed areas during the dry season. Spectral reflectance curves of the new samples are measured, replacing the old data at the corresponding location in the template library. If the overall sediment type in the area changes, the sampling range is expanded to cover the newly emerged sediment type.

[0084] After the template library is updated, water transparency parameters for the historical period are recalculated, using the updated sediment spectra as a benchmark. This corrects errors in past data due to biased sediment assumptions, generating a consistent time series dataset. This mechanism ensures the comparability of long-term monitoring data, avoids invalidation of historical data due to sediment changes, and provides a reliable basis for analyzing water resource evolution trends.

[0085] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A water resources information collection method based on remote sensing images, characterized in that: The following steps are involved: Acquiring remote sensing image data of the target water area, wherein the remote sensing image data includes spectral reflectance corresponding to multiple spatial locations; Constructing a dynamic substrate spectrum template library, wherein the dynamic substrate spectrum template library is established by collecting measured spectral reflectance curves of different substrate types in a water-free state, and each substrate type corresponds to an independent spectral feature data set; For each spatial location in the remote sensing image, the exposed substrate type of the adjacent area is matched according to the geographic coordinates of the spatial location, and the spectral feature dataset of the corresponding substrate type is extracted from the dynamic substrate spectrum template library as the basic reflectance reference; Establishing a light propagation path separation model, which simulates the complete path of sunlight passing through the water layer, reaching the bottom surface, and then returning to the sensor after secondary reflection, and generates a bottom reflection contribution component in combination with a basic reflection reference; The bottom reflection contribution component is deducted from the spectral reflectance of the remote sensing image data to obtain the pure water optical response value. Based on the physical correlation model between the pure water optical response value and water transparency, a water transparency parameter distribution map of the spatial position of the target water area is output.

2. The water resources information collection method based on remote sensing images according to claim 1 is characterized in that: The process of constructing the dynamic sediment spectrum template library is as follows: Conduct field surveys in the target waters during the dry season and select shallow areas where the water level is below the natural threshold as exposed bottom sampling areas. Grid the exposed bottom sampling areas, collect physical samples of surface sediments from each grid cell, and record the geographic coordinates of the physical samples. Measure the continuous spectral reflectance curve of the physical samples in the visible to near-infrared band in a dry state; classify the physical samples into sandy substrate type, muddy substrate type, gravel substrate type, and mixed substrate type based on the sediment particle size distribution and organic matter content; perform cluster analysis on the spectral reflectance curves of substrate samples of the same type, and eliminate abnormal spectral reflectance curves with a dispersion exceeding the set threshold; The weighted average value of the band of the retained spectral reflectance curve is calculated to form a standardized spectral reflectance set of the substrate type in the set band; the standardized spectral reflectance set is associated with the corresponding geographic coordinates and stored to form a spatially indexed dynamic substrate spectral template library.

3. The method for collecting water resources information based on remote sensing images according to claim 1, characterized in that: The process of matching the exposed substrate types of adjacent areas is as follows: A search buffer zone is established with the current spatial location to be processed as the center, wherein the radius of the search buffer zone is determined according to the historical maximum water receding range of the water area; all exposed bottom sediment sampling points located within the search buffer zone are retrieved from the dynamic bottom sediment spectrum template library; Calculate the surface distance and elevation difference between the exposed bottom sampling point and the current spatial position, and exclude the exposed bottom sampling points whose elevation difference exceeds the terrain relief threshold; perform frequency statistics on the bottom types of the remaining exposed bottom sampling points, and select the bottom type with the highest frequency as the matching result; when there are multiple bottom types with the same frequency, select the bottom type of the exposed bottom sampling point with the smallest surface distance from the current spatial position; If there is no exposed bottom sediment sampling point within the search buffer zone, the search will be extended to the typical bottom sediment distribution map of the basin to which the target water area belongs for analogical matching.

4. The water resources information collection method based on remote sensing images according to claim 1 is characterized in that: The process of generating the sediment reflection contribution component by the light propagation path separation model is as follows: The incident angle of sunlight and the sensor observation angle are calculated based on the imaging time of the remote sensing image and the satellite orbit. The propagation process of the incident light penetrating the water layer to the bottom surface is simulated to calculate the first propagation attenuation. The propagation process of the light reflected by the bottom penetrating the water layer to the sensor is simulated to calculate the second propagation attenuation. The first propagation attenuation and the second propagation attenuation are applied to the basic reflection reference value to generate a bottom sediment reflection contribution component. A water surface mirror reflection correction factor is introduced to dynamically adjust the bottom sediment reflection contribution component. The water surface mirror reflection correction factor is generated by a wind and wave disturbance model combined with real-time meteorological data.

5. The method for collecting water resources information based on remote sensing images according to claim 4 is characterized in that: The calculation process of propagation attenuation is: The area in the target water area with a water depth greater than the optical depth is selected as the pure water reference area; the spectral reflectance of the pure water reference area in the remote sensing image is extracted as the pure water signal without bottom sediment interference; Establishing a parameter fitting relationship between a pure water signal and a theoretical water absorption model, wherein the theoretical water absorption model includes the absorption bands of phytoplankton pigments, suspended particulate matter, and dissolved organic matter; The parameters of the theoretical water absorption model are adjusted through a nonlinear optimization algorithm to minimize the root mean square error between the simulated spectral curve and the measured spectral curve. The optimized theoretical water absorption model parameters are used as the benchmark values ​​of the spectral absorption characteristics of the current water area.

6. The water resources information collection method based on remote sensing images according to claim 5 is characterized in that: The specific selection process of the pure water reference area is as follows: The stability of the spectral reflectance of a selected pure water reference area is analyzed in a time series. When the deviation between the reflectance trend of a specific band in the reference area and the inversion results of the water transparency physical correlation model shown by multiple consecutive remote sensing images continues to increase, the reference area reselection mechanism is activated. Based on the spatial distribution pattern of historical propagation attenuation, the optical stability index of each deep-water area in the target waters is calculated. The optical stability index reflects the correlation between water depth and the ability to suppress bottom reflection. The candidate area with the highest optical stability index is selected as the new pure water reference area, ensuring that the water depth in the candidate area is greater than that of the original reference area and the fluctuation amplitude of historical propagation attenuation is below the set tolerance threshold. The spectral reflectance of the new pure water reference area is extracted, the theoretical water absorption model parameters are refitted, and the optimized theoretical water absorption model parameters are updated to the spectral absorption characteristic benchmark values ​​of the light propagation path separation model.

7. The method for collecting water resources information based on remote sensing images according to claim 1, characterized in that: The calculation process of the optical response value of pure water is: Obtain the total spectral reflectance value of the remote sensing image in the set band, calculate the bottom sediment reflection contribution component at the current spatial position, and determine the direct reflection component of the water surface; The pure water optical response value is obtained by deducting the bottom reflection contribution component and the water surface direct reflection component from the total spectral reflectance value; When the optical response value of pure water is lower than the preset noise threshold, the attenuation gradient of the adjacent spatial position is used for spatial interpolation compensation, and the atmospheric scattering effect correction is performed on the optical response value of pure water after spatial interpolation compensation.

8. The method for collecting water resources information based on remote sensing images according to claim 1, characterized in that: The construction process of the physical association model is as follows: Deploy a fixed monitoring buoy array in the target waters to synchronously obtain field measurements of water transparency at different spatial locations, and extract the pure water optical response values ​​at the corresponding spatial locations at the time of remote sensing image acquisition; Analyzing the statistical correlation between the specific blue-green band ratio of the pure water optical response value and the field measurement value of the water transparency; constructing a piecewise linear regression equation, wherein the piecewise linear regression equation adopts a first slope linear relationship when the pure water optical response value is in a low value range, and adopts a second slope linear relationship when the pure water optical response value is in a high value range; The calibration sub-areas were divided into hydrogeological units, and the parameters of the piecewise linear regression equation were optimized for each calibration sub-area.

9. The method for collecting water resources information based on remote sensing images according to claim 1, characterized in that: It also includes obtaining multiple periods of remote sensing image data of the target water area in a continuous time series, independently generating a water transparency parameter distribution map for each period of remote sensing image data, and establishing a spatiotemporal change analysis matrix to detect sudden changes in water transparency at the same spatial location; When the duration of the mutation point exceeds the set period and the change amplitude exceeds the natural fluctuation threshold, the sediment spectrum template library update process is triggered; The process of updating the substrate spectrum template library includes re-collecting exposed substrate samples at a spatial location, measuring the spectral reflectance curves of the new exposed substrate samples and updating the dynamic substrate spectrum template library data, and using the updated dynamic substrate spectrum template library to recalculate the water transparency parameters of the historical period to generate a consistency-corrected time series dataset.

Images