Urban water body monitoring method and system based on multi-source remote sensing data

Abstract

The invention relates to the technical field of remote sensing monitoring and environment management, and discloses an urban water body monitoring method and system based on multi-source remote sensing data, and the method comprises the steps: collecting and correcting a multi-source remote sensing image, and obtaining a calibration data set; obtaining a standardized image set through band matching and noise suppression; edge features are extracted, spectral distribution is verified, and a water body boundary is identified; determining an initial water body distribution diagram through geometric correction and classification; analyzing narrow water body area features, and eliminating misjudgments through shadow difference comparison and boundary overlapping analysis; and calculating a change trend and marking a dynamic area, and finally performing superposition verification with an urban ground feature layer to generate a dynamic monitoring map of the narrow water body. According to the method, through multi-source data fusion and multi-level verification, the identification problem caused by shielding and spectrum confusion of urban narrow water bodies is effectively solved, and the monitoring accuracy and reliability are improved.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring and environmental management technology, and in particular to a method and system for monitoring urban water bodies based on multi-source remote sensing data. Background Technology

[0002] Urban water body monitoring, as an important component of environmental management and urban planning, plays an indispensable role in ensuring ecological balance and the quality of life for residents. With the acceleration of urbanization, water pollution and resource management issues are becoming increasingly prominent, making the accurate monitoring of urban water body distribution and changes an urgent problem to be solved.

[0003] In a current technology, water body identification typically employs a single remote sensing data source, relying on spectral features for water body extraction and combining threshold segmentation or classifiers for region division. For example, water body identification is performed based on Landsat or Sentinel series satellite imagery, using methods such as the normalized water index.

[0004] However, due to the scattered distribution of narrow rivers, ditches, and other water bodies in cities, and their frequent obstruction by tall buildings, roads, and other structures, traditional methods or single smart sensors struggle to effectively distinguish water bodies from shadows in terms of spectral resolution and spatial detail, leading to misjudgments and missed detections. Furthermore, different remote sensing sensors exhibit variations in spectral response and radiometric characteristics, making it easy to introduce systematic errors when directly fusing multi-source data, further impacting the accuracy and reliability of monitoring results. Therefore, existing technologies suffer from poor consistency in multi-source data fusion and difficulty in accurately identifying narrow water bodies. Summary of the Invention

[0005] This invention provides a method and system for urban water body monitoring based on multi-source remote sensing data, in order to solve the problems of poor consistency in multi-source data fusion and difficulty in accurately identifying narrow water bodies.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides a method for urban water body monitoring based on multi-source remote sensing data, comprising: Multi-source remote sensing images were acquired, and the acquired raw spectral data were corrected and radiometrically calibrated to unify data source compatibility and obtain a preliminary calibrated remote sensing dataset. Band matching adjustment and noise suppression are performed on the remote sensing dataset to achieve feature distribution equalization and obtain a standardized image set; The edge feature information of the standardized image set is extracted. If the edge intensity exceeds the preset first edge intensity threshold and is accompanied by the verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. Geometric correction is performed on the images in the boundary enhancement image set, and the water and non-water areas are preliminarily divided by the ground feature classification accuracy analysis to determine the initial water body distribution map; Extract the distribution characteristics of narrow water body regions from the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, then eliminate misjudgments through boundary overlap analysis to obtain the corrected water body distribution map. Calculate the change trend in the corrected water body distribution map. If the rate of change is higher than the preset rate of change threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained. The change monitoring distribution results are overlaid with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

[0007] Secondly, the present invention provides an urban water body monitoring system based on multi-source remote sensing data, comprising: The multi-source data acquisition and preprocessing module acquires multi-source remote sensing images and performs correction and radiometric calibration on the acquired raw spectral data to unify data source compatibility and obtain a preliminarily calibrated remote sensing dataset. The data standardization processing module performs band matching adjustment and noise suppression on the remote sensing dataset, completes feature distribution balancing, and obtains a standardized image set. The water body boundary enhancement and recognition module extracts edge feature information from the standardized image set. If the edge intensity exceeds a preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. The initial water body classification module performs geometric correction on the images in the boundary enhancement image set, and combines the land cover classification accuracy analysis to initially divide the water body and non-water body areas to determine the initial water body distribution map. The narrow water body fine correction module extracts the distribution characteristics of the narrow water body region in the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, the misjudgment is eliminated through boundary overlap analysis to obtain the corrected water body distribution map. The dynamic change detection module calculates the change trend in the corrected water body distribution map. If the change rate is higher than the preset change rate threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained. The monitoring results integration and output module overlays the change monitoring distribution results with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention effectively unifies the differences in spectral response, radiometric characteristics and spatial resolution between different sensors such as satellites and UAVs by adjusting the band matching of multi-source remote sensing data, correcting radiometric consistency and standardizing the format. It overcomes the technical difficulty of poor consistency in multi-source data fusion and provides a high-quality data foundation for subsequent accurate identification.

[0009] (2) This invention combines edge intensity detection with spectral feature distribution uniformity verification, first screening potential boundaries, then using spectral standards such as near-infrared bands for secondary confirmation, and through multiple verification mechanisms such as shadow difference analysis and boundary overlap analysis for narrow water bodies, significantly reduces the misjudgment and missed detection rate of narrow water bodies and ditches in the city.

[0010] (3) This invention constructs a complete technical chain from initial water body classification to dynamic change marking, identifies change trends through time series analysis, and spatially aligns and integrates the change results with a high-precision urban land cover layer. The resulting monitoring map not only accurately reflects the spatial distribution of water bodies, but also intuitively displays their dynamic change information, providing strong decision support for urban water environment management and planning. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the urban water body monitoring method based on multi-source remote sensing data provided in the first embodiment of the present invention; Figure 2 This is a schematic diagram of the urban water monitoring system based on multi-source remote sensing data provided in the second embodiment of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Reference Figure 1 The first embodiment of the present invention provides a method for monitoring urban water bodies based on multi-source remote sensing data, comprising the following steps: S11: Acquire multi-source remote sensing images and perform correction and radiometric calibration on the acquired raw spectral data to unify data source compatibility and obtain a preliminary calibrated remote sensing dataset. S12, perform band matching adjustment and noise suppression on the remote sensing dataset to achieve feature distribution balancing and obtain a standardized image set; S13, extract edge feature information from the standardized image set. If the edge intensity exceeds the preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. S14, geometric correction is performed on the images in the boundary enhancement image set, and the water and non-water areas are initially divided by combining the land cover classification accuracy analysis to determine the initial water body distribution map; S15, extract the distribution characteristics of narrow water body regions in the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, the misjudgment is eliminated through boundary overlap analysis to obtain the corrected water body distribution map. S16, calculate the change trend in the corrected water body distribution map. If the change rate is higher than the preset change rate threshold, mark it as a dynamically changing water body area and obtain the change monitoring distribution result. S17, the change monitoring distribution results are overlaid with the pre-established urban land cover layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

[0014] The various discrimination thresholds involved in each step of this method include, but are not limited to, edge intensity threshold, shadow difference threshold, rate of change threshold, and classification accuracy threshold. Their determination is based on a pre-constructed, rigorously quality-controlled historical sample library. The construction of this historical sample library involves collecting historical water body distribution data from multiple publicly available global geographic information databases such as Landsat Water Mask, JRC Global Surface Water, and OpenStreetMap, and combining this data with high-resolution commercial satellite imagery such as Google Earth and some field survey data for cross-validation and meticulous manual annotation. The sample library ultimately contains over 10,000 annotated sample units, covering various water bodies, including rivers, lakes, reservoirs, and ditches, and their surrounding shadows, vegetation, buildings, and other disturbing features, across different seasons, climate zones, and typical environments such as urban, suburban, and rural areas. Statistical analysis (primarily using percentile methods) of corresponding feature values ​​in this sample library, such as gradient magnitude, near-infrared reflectance, and area change rate, was conducted to determine the various preset thresholds mentioned in subsequent steps, ensuring the scientific validity and universality of the threshold settings.

[0015] In step S11, multi-source remote sensing images are acquired, and the acquired raw spectral data are corrected and radiometrically calibrated to unify data source compatibility, resulting in a preliminarily calibrated remote sensing dataset, including: S1101 acquires multi-source remote sensing images through satellite and UAV platforms, corrects the acquired raw spectral data and unifies spectral response differences to obtain a corrected spectral dataset; S1102, Perform radiometric consistency adjustment on image data from different sources in the spectral dataset to determine a unified dataset after radiometric calibration; S1103, If the unified dataset has data compatibility deviations, then format standardization processing is performed to obtain a more compatible fused dataset; S1104, noise suppression and data integrity checks are performed on the fused dataset to obtain the final calibrated remote sensing dataset.

[0016] In step S1101, multi-source remote sensing images are acquired through satellite and UAV platforms, the acquired raw spectral data are corrected and the differences in spectral response are unified to obtain the corrected spectral dataset.

[0017] It should be noted that when acquiring multi-source remote sensing images with multiple time series and resolutions simultaneously through satellite platforms such as Landsat 8 OLI and UAV platforms such as those equipped with MicaSenseRedEdge-MX smart sensors, the raw spectral data collected by different sensors, including optical and infrared sensors, often have response differences. Spectral response correction adjusts the data to be unified to the same benchmark by comparing the spectral response functions of different sensors, thus obtaining a corrected spectral dataset.

[0018] In one embodiment, the spectral response of a satellite sensor is too high in the 500-600 nanometer range, while that of a drone sensor is too low. The two are adjusted based on a standard spectral curve to generate a corrected spectral dataset.

[0019] In step S1102, radiometric consistency adjustment is performed on image data from different sources in the spectral dataset to determine a unified dataset after radiometric calibration.

[0020] It should be noted that radiometric consistency adjustment is performed on image data from different sources in the spectral dataset to convert the original digital signals into physical radiance values, in units of watts per square meter per strata per micrometer, thereby determining a radiometrically calibrated unified dataset.

[0021] In one embodiment, the digital value range of a satellite image is 0-255, which can be converted into a radiance value through radiometric calibration, in watts per square meter per strata per micrometer, to ensure radiometric consistency of images from different sources.

[0022] In step S1103, if the unified dataset has data compatibility deviations, format standardization processing is performed to obtain a more compatible fused dataset.

[0023] It should be noted that if there are data compatibility discrepancies in the unified dataset after radiometric calibration, such as inconsistencies in spatial resolution or timestamps from different sources, format standardization will be performed. For spatial resolution unification, a bilinear interpolation resampling method is used to unify data of different resolutions to a 10-meter pixel size. This resolution is a baseline resolution determined based on the needs of urban water body monitoring and the accuracy requirements of multi-source data fusion. For temporal inconsistencies, a nearest-neighbor time-series matching method is used to align data using quarterly time windows.

[0024] In one embodiment, bilinear interpolation is used to uniformly resample satellite data with a resolution of 30 meters and UAV data with a resolution of 1 meter to a spatial resolution of 10 meters, and timestamps are aligned with quarters as time windows to generate a more compatible fused dataset.

[0025] In step S1104, noise suppression and data integrity checks are performed on the fused dataset to obtain the final calibrated remote sensing dataset.

[0026] It should be noted that noise suppression can be achieved by smoothing random noise points in an image through filtering techniques, while data integrity checks identify and locate missing or outlier data through statistical analysis.

[0027] In one embodiment, if some data in a certain area is missing due to cloud cover, the area is marked and a prompt to supplement the data is generated, ultimately resulting in a calibrated remote sensing dataset.

[0028] In step S12, the remote sensing dataset undergoes band matching adjustment and noise suppression to achieve feature distribution equalization, resulting in a standardized image set, including: S1201, Perform band alignment and consistency adjustment on the remote sensing dataset to obtain a data set after band matching; S1202, The random noise in the data set is filtered to obtain a noise-suppressed image group; S1203, If there is a deviation caused by sensor differences in the image group, it is corrected to determine the data group after sensor consistency adjustment; S1204, remove background noise and atmospheric interference from the data set to obtain a standardized image set with balanced features.

[0029] In step S1201, the remote sensing dataset is subjected to band alignment processing and consistency adjustment to obtain a data set after band matching.

[0030] It should be noted that the band alignment and consistency adjustment of the remote sensing dataset specifically involves using interpolation or remapping methods to unify the band range of each data source to a preset standard band, based on the spectral response characteristics of different sensors.

[0031] In one embodiment, when the visible light band of the satellite sensor is 400-500 nanometers and that of the UAV sensor is 450-550 nanometers, the data from both are aligned to the 400-550 nanometer band range using the method described above, thereby obtaining a band-matched data set.

[0032] In step S1202, the random noise in the dataset is filtered to obtain a noise-suppressed image group.

[0033] It should be noted that random noise often originates from electronic interference in the sensor itself or environmental factors. The random noise filtering of the data set is specifically performed using spatial domain filtering, which calculates new pixel values ​​based on the neighborhood window of the pixel to replace abnormal noise points.

[0034] In one embodiment, when there are isolated high-value points in the image caused by sensor jitter, a mean filtering method is used to correct the noise point by the average value of the pixels around it, thereby obtaining a noise-suppressed image group.

[0035] In step S1203, if there is a deviation caused by sensor differences in the image group, it is corrected to determine the data group after sensor consistency adjustment.

[0036] It should be noted that the systematic brightness or spectral deviations in the image group caused by sensor differences are corrected by comparing the distribution characteristics of data from different sensors and adjusting the brightness or spectral values ​​using a linear transformation method based on a standard reference.

[0037] In one embodiment, the brightness values ​​of a satellite image are generally too high, while those of a drone image are too low. The calibration tool will linearly adjust the brightness distribution of both images based on a standard reference value, ultimately generating a data set with sensor consistency adjustment.

[0038] In step S1204, background noise and atmospheric interference in the data set are removed to obtain a standardized image set with balanced features.

[0039] It should be noted that after acquiring the data set with sensor consistency adjustments, data quality will be further optimized. Background noise and atmospheric interference, such as haze or water vapor, can affect the feature representation of remote sensing images. By analyzing the spectral characteristics of the images, these interfering factors are identified and removed. The FLAASH atmospheric correction model is used to automatically generate atmospheric parameters based on image acquisition time, geographic location information, and the MODTRAN standard atmospheric database. The mid-latitude summer atmospheric profile and rural aerosol model are the system default configurations. The optical thickness of water vapor and aerosols is calculated based on the latitude and longitude information of the image acquisition time and location. The signal attenuation caused by water vapor, aerosols, and other factors is quantified and compensated based on the spectral characteristics of the images. The final remote sensing image set has undergone radiometric calibration and atmospheric correction, converting its original radiance values ​​into surface reflectance, thereby eliminating spectral deviations caused by sensor differences and atmospheric conditions. This provides a standardized and consistent data foundation for subsequent water body identification based on reflectance thresholds.

[0040] In one embodiment, when the image is darkened overall due to fog, the surface reflectance is inverted and adjusted using atmospheric parameters automatically generated by the FLAASH model based on the MODTRAN atmospheric database, including mid-latitude summer atmospheric profiles, rural aerosol models, and 40km visibility parameters, to enhance the effective signal strength and thereby obtain a set of remote sensing images with balanced features and standardized.

[0041] In step S13, edge feature information is extracted from the standardized image set. If the edge intensity exceeds a preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is determined to be a potential water body boundary region, resulting in a boundary enhancement image set, including: S1301, Perform feature extraction on the standardized image set to obtain an edge feature image set with prominent edge features; S1302, compare the edge intensity of the edge feature image set, and if it exceeds the preset first edge intensity threshold, it is judged as a potential boundary region, and a high edge intensity region image set is obtained. S1303, detect the spectral feature distribution of the high edge intensity region image set and verify the distribution balance. If it meets the preset specific land cover type distribution standard, it is determined as a candidate water body boundary region, and a spectrally balanced region image set is obtained. S1304, perform detail optimization on the candidate water body boundary regions in the spectrally balanced region image set to obtain the final boundary-enhanced image set.

[0042] In step S1301, feature extraction is performed on the standardized image set to obtain an edge feature image set with prominent edge features.

[0043] It should be noted that edge feature extraction of the remote sensing image set mainly involves identifying regions with significant grayscale or color changes to extract boundary information. This is achieved by calculating pixel value gradient changes and employing the Canny edge detection operator. First, Gaussian filtering is applied to smooth image noise. Then, the image gradient is calculated (typically using the Sobel operator), followed by non-maximum suppression. Finally, double thresholding is used to detect and connect edges. To adapt to remote sensing data with different spatial resolutions and noise levels, the standard deviation parameter of the Gaussian filter is adjusted. (Unit: pixels) Adjustments need to be made based on the ground plane pixel size; generally, the larger the ground plane pixel size (the lower the resolution), the better for effectively smoothing noise. The value should be relatively increased; the smaller the pixel ground size (the higher the resolution), the more details are retained. The value should be relatively reduced; for example, through experiments, for an image with a spatial resolution of 10 meters, the value can be set to... =1.0; For images with a spatial resolution of 30 meters, the setting can be adjusted. =1.5; The high and low thresholds of the Canny operator are adaptively set based on the statistical distribution of the gradient magnitude image. For example, the high threshold is set to the 70th percentile of the gradient magnitude distribution, and the low threshold is set to the 35th percentile. The Canny operator, with these parameters set, identifies and marks such regions, thereby obtaining an image set of edge features with prominent edge characteristics.

[0044] In step S1302, the edge intensity of the edge feature image set is compared. If it exceeds a preset first edge intensity threshold, it is determined to be a potential boundary region, and a high edge intensity region image set is obtained.

[0045] It should be noted that the preset edge strength threshold is based on statistical analysis of the gradient magnitude (range 0-255) of 8-bit remote sensing images in the sample image library. The gradient magnitude of each pixel is calculated using the Canny operator, and the calculation formula is as follows:

[0046] in, This represents the edge strength of the pixel, which is calculated using the Sobel operator to determine the gradient magnitude of each pixel in the image. and These are the horizontal and vertical gradients calculated using the Sobel operator, respectively. By statistically analyzing the gradient magnitude distribution of all known true boundary pixels in the sample library, the 85th percentile was selected as the threshold benchmark. After experimental verification, the threshold was determined to be 50 (in units of gradient magnitude). This method ensures that most true boundaries are preserved while effectively suppressing unstructured noise. The gradient magnitude of each pixel in the edge feature image set is calculated as the edge intensity. Pixels whose edge intensity exceeds a preset first edge intensity threshold are identified as potential boundary regions.

[0047] In one embodiment, based on statistical analysis of 8-bit remote sensing images (gradient magnitude range 0-255), the distribution of gradient magnitudes of all known true boundary pixels is calculated, and the 85th percentile is selected as the threshold benchmark. The first edge intensity threshold is set to 50 (gradient magnitude unit). When the gradient magnitude of a certain edge region reaches 60, the region is identified as a potential boundary and extracted, thereby obtaining a high edge intensity region image set.

[0048] In step S1303, the spectral feature distribution of the high edge intensity region image set is detected and the distribution balance is verified. If it meets the preset specific land cover type distribution standard, it is determined as a candidate water body boundary region, and a spectrally balanced region image set is obtained.

[0049] It should be noted that the verification of the balanced distribution of spectral features specifically refers to determining whether the reflectance of the target area in a specific band (such as the near-infrared band) meets the preset spectral feature standard for water bodies. The preset distribution standard for specific land cover types is established by analyzing the spectral feature data of historical water body area samples. Specifically, it uses standardized multi-source remote sensing data, including the reflectance (dimensionless, range 0-1) distribution of water body samples in the near-infrared band from Landsat 8 OLI and Sentinel-2 MSI, and selects the 90th percentile of 0.15 as the critical value to distinguish water bodies from other land cover types. This critical value setting takes into account the characteristics of different sensors and is based on surface reflectance data after atmospheric correction, effectively adapting to multi-source data conditions. The spectral feature distribution of the high-edge-intensity area image set is detected and its balance verified. Reflectance data of each area in different bands are extracted, and the spectral curve shape and numerical distribution are analyzed to determine whether the area meets the characteristic distribution standard for specific land cover types.

[0050] In one embodiment, based on statistical analysis of the near-infrared reflectance of water samples from multiple regions and multiple sensors, and after standardization, a near-infrared reflectance value below 0.15 is selected as the water body identification standard, using the lower 90th quantile. When the near-infrared reflectance value of a high-edge-intensity region is 0.12, and its spectral distribution conforms to the characteristics of a water body, this region is identified as a candidate water body boundary region and included in the spectrally balanced region image set.

[0051] In step S1304, the candidate water body boundary regions in the spectrally equalized region image set are optimized in detail to obtain the final boundary enhancement image set.

[0052] It should be noted that the detailed optimization is achieved through local contrast enhancement and edge sharpening. The contrast-limited adaptive histogram equalization method is used to improve the contrast between the boundary and the surrounding ground features, and a guided filter is used for edge selective sharpening to enhance linear features and suppress noise.

[0053] In one embodiment, for a candidate region whose spectral characteristics match those of a water body but whose boundaries are blurred, the clarity of the boundary lines of the region is significantly improved after applying the local contrast enhancement and edge sharpening processing, and the transition between the region and the adjacent ground features is more distinct, ultimately forming a remote sensing image dataset with enhanced boundary details.

[0054] In step S14, geometric correction is performed on the images in the standardized image set, and the water and non-water areas are initially divided based on the accuracy analysis of land cover classification to determine the initial water body distribution map, including: S1401, Perform position calibration on the images in the standardized image set to obtain the first image set after calibration; S1402, perform deformation correction on the first image set to obtain the corrected second image set; S1403, classify the surface features of the second image set into categories. If the classification result meets the preset classification accuracy threshold, it is judged as a preliminary water body or non-water body area, and the third image set after classification is determined. S1404, delineate the boundaries of the water areas in the third image set and integrate the classification results to obtain an initial water distribution map.

[0055] In step S1401, the images in the standardized image set are positionally calibrated to obtain the first calibrated image set.

[0056] It should be noted that the location calibration uses stable and reliable ground feature points in the image as control points. Based on the coordinates of these control points, a polynomial transformation is used to uniformly register images from different sources or at different times to the same geographic coordinate system. The selection criteria for control points are that they are clearly identifiable and have stable positions in multiple images, such as permanent ground features like road intersections and building corners. The parameters of the polynomial transformation are solved using the least squares method to minimize the registration error of the control points. The calibration accuracy is controlled by the root mean square error to ensure that the spatial offset of the registered image meets the preset tolerance requirements.

[0057] In one embodiment, for two remote sensing images with a spatial offset of about 5 kilometers due to differences in sensor angles, five road intersections are selected as control points, and coordinate correction is performed using a quadratic polynomial transformation model to reduce the spatial offset to within 0.1 kilometers, thereby obtaining a calibrated first image set.

[0058] In step S1402, the first image set is deformed and corrected to obtain the corrected second image set.

[0059] It should be noted that the deformation correction is based on the position calibration of the first image set. It further combines a digital elevation model with sensor imaging geometric parameters, constructing a rigorous rational function model to precisely correct the image geometric deformation caused by terrain undulations and sensor attitude. This process involves accurately mapping image pixel coordinates to the corresponding three-dimensional coordinates of ground points and using a resampling method to generate a deformation-reduced image.

[0060] In one embodiment, for remote sensing images in which mountainous areas appear stretched and deformed in the image due to high mountain terrain, the rational function model is applied to perform geometric fine correction by inputting the digital elevation model data of the area and the orbit and attitude parameters of the sensor, so that the mountain outline accurately matches the spatial position of the real terrain, thereby obtaining a corrected second image set.

[0061] In step S1403, the surface features of the second image set are classified. If the classification result meets the preset classification accuracy threshold, it is determined to be a preliminary water body or non-water body area, and the third image set after classification is determined.

[0062] It should be noted that the preset classification accuracy threshold is the overall classification accuracy calculated based on the confusion matrix of the classification results of the validation sample set. This threshold is determined statistically through historical classification experimental data and is usually set to no less than 90% to ensure reliable accuracy of the classification results. When classifying the surface features of the second image set, a random forest supervised classification method is adopted, using eight spectral texture features, including near-infrared reflectance, NDVI, texture variance (based on gray-level co-occurrence matrix), blue band reflectance, green band reflectance, red band reflectance, NDWI (normalized water index), and texture contrast. The training samples come from 5000 historical surface data annotation points (including 2000 water samples and 3000 non-water samples). The number of decision trees is set to 100. By analyzing the spectral features and texture information of the images, pixels are divided into different categories, such as water, vegetation, or bare land.

[0063] In one embodiment, a random forest classifier is trained using eight features, including near-infrared reflectance, NDVI, and texture variance. The accuracy is determined statistically from historical classification experimental data and is typically set to be no less than 90%, i.e., a preset classification accuracy threshold of 90%. The preset water body classification threshold requires a near-infrared reflectance value below 0.2. If a pixel in a certain area has a near-infrared reflectance value of 0.15 and its texture feature value is below the set threshold, it meets the characteristics of a water body and is marked as such. When the classification results are verified, the overall classification accuracy reaches 92%, exceeding the preset 90% accuracy threshold. Therefore, the classification result is adopted, and a third image set after classification is generated. The spectral classification threshold used to distinguish water bodies from other land features is determined by analyzing the statistical characteristics of reflectance in various bands of a large number of typical water and non-water body samples. Specifically, it involves statistically analyzing the reflectance distribution of the two types of samples in the near-infrared band and selecting the critical value that achieves the best classification effect.

[0064] In step S1404, the water areas of the third image set are delineated and the classification results are integrated to obtain an initial water distribution map.

[0065] It should be noted that the boundary delineation involves identifying all pixels classified as water bodies, connecting adjacent water body pixels into continuous regions based on the eight-neighbor connectivity principle, and extracting the outer contour of each connected region as the water body boundary. During the integration process, area-based spatial filtering is applied, and a minimum water body patch area threshold is set to eliminate isolated pixel blocks generated by classification noise. This area threshold is determined based on statistical analysis of historical water body distribution data. In remote sensing images with a spatial resolution of 30 meters, areas of no less than 9 pixels (equivalent to an actual surface area of ​​8100 square meters) are typically retained as valid water bodies.

[0066] In one embodiment, for areas where the classification results show scattered water patches, adjacent water pixels are connected to form a complete polygonal boundary. At the same time, isolated classification results with an area of ​​less than 9 pixels are filtered out in the image with a spatial distribution rate of 30 meters, and finally a continuous and complete initial water distribution map is formed.

[0067] In step S15, the distribution characteristics of narrow water body regions in the initial water body distribution map are extracted. If the spectral value of the narrow water body region differs from the surrounding shadows by less than a preset shadow difference threshold, misjudgments are eliminated through boundary overlap analysis to obtain a corrected water body distribution map, including: S1501, extract features from the narrow water body regions of the initial water body distribution map to obtain preliminary regional classification information and determine the first distribution dataset; S1502, compare the spectral values ​​of the narrow water body area in the first distribution dataset with the surrounding shadows. If the values ​​are lower than the preset shadow difference threshold, they are judged as shadow interference, and a second distribution dataset is obtained. S1503, perform boundary overlap analysis between the narrow water body area and the surrounding area in the second distribution dataset. If the degree of overlap is higher than the preset boundary overlap threshold, remove the falsely judged part and obtain the third distribution dataset. S1504, adjust the boundaries of the water body regions in the third distribution dataset, and re-divide the narrow water body regions according to the preset classification criteria to obtain the corrected water body distribution map.

[0068] In step S1501, features are extracted from the narrow water body regions of the initial water body distribution map to obtain preliminary regional classification information and determine the first distribution dataset.

[0069] It should be noted that the feature extraction is achieved by calculating the gray-level mean and texture variance of pixels within each narrow water body region. The gray-level mean is based on the DN value (digital quantization value, range 0-255) of the 8-bit remote sensing image, reflecting the color depth characteristics of the region; the texture variance is calculated based on the contrast characteristics of the gray-level co-occurrence matrix to characterize the smoothness of the region by assessing the dispersion of pixel values. The feature determination criteria for water body regions are determined based on statistical analysis of the gray-level mean and texture variance of 3000 historical water body samples. The 85th percentile of the gray-level mean data (45) is taken as the gray-level threshold, and the 90th percentile of the texture variance data (20) is taken as the texture threshold. Regions that simultaneously meet the criteria of a gray-level mean below 45 and a texture variance less than 20 are initially classified as water bodies.

[0070] In one embodiment, for a narrow water body region with a width of about 3 pixels at the edge of a lake, its grayscale mean is calculated to be 38 and its texture variance is 15. Both indicators meet the preset water body feature standards, so the region is initially classified as a water body region, forming the first distribution dataset.

[0071] In step S1502, the spectral values ​​of the narrow water body region in the first distribution dataset are compared with the surrounding shadows. If the spectral values ​​are lower than the preset shadow difference threshold, it is determined to be shadow interference, and a second distribution dataset is obtained.

[0072] It should be noted that the preset shadow difference threshold was determined by statistically analyzing the reflectance difference distribution between typical water bodies and shadow areas in the near-infrared band. Specifically, based on the atmospherically corrected reflectance values ​​in multi-source remote sensing data, the spectral differences between water bodies and shadows in a large number of independent samples were calculated, and the 5th percentile of all difference values ​​was selected as the base threshold. To enhance the applicability of the threshold in different scenarios, an adaptive adjustment mechanism based on urban and rural regional characteristics was established. Through statistical analysis of samples from different regional types, a stricter threshold of 0.03 was adopted in densely built-up urban areas, a threshold of 0.05 was adopted in suburban areas with high vegetation coverage, and a threshold of 0.07 was adopted in open rural areas. This adjustment mechanism was established based on the research results of Jensen et al. (2015) on the differences in spectral characteristics of shadows in urban and rural areas and was verified by localized samples. When comparing the spectral values ​​of the narrow water body area in the first distribution dataset with the surrounding shadows, the average reflectance values ​​in the near-infrared band of the narrow water body area to be judged and its surrounding shadow area were extracted respectively, and the absolute difference between the two was calculated as a difference measure to determine whether shadow interference exists.

[0073] In one embodiment, based on statistical analysis of the reflectance differences between historical water bodies and shadow samples in the near-infrared band, the 5th percentile of the difference value is selected as the base threshold, and adjusted according to the characteristics of urban and rural areas. A narrow water body located in an urban area has a near-infrared reflectance value of 0.18, while the reflectance value of its surrounding shadow area is 0.16, with a difference of 0.02. This difference value is lower than the applicable shadow difference threshold of 0.03 for urban areas, indicating that the spectral characteristics of this narrow water body area are highly similar to those of the shadow area. Therefore, it is determined to be shadow interference rather than a real water body, and is removed from the water body classification results to generate a second distribution dataset.

[0074] In step S1503, the narrow water body region in the second distribution dataset is subjected to boundary overlap analysis with the surrounding region. If the degree of overlap is higher than the preset boundary overlap threshold, the falsely judged part is removed, and the third distribution dataset is obtained.

[0075] It should be noted that the preset boundary overlap threshold was determined by analyzing sample data from 200 sets of known classification results. Statistical analysis revealed that when the Jaccard coefficient exceeds 0.6, the probability of the area being misclassified as a water body exceeds 95%. Therefore, this value was set as the judgment threshold. The boundary overlap analysis is achieved by calculating the Jaccard similarity coefficient between the boundary of the narrow water body area and the boundaries of surrounding land features. This coefficient is defined as the ratio of the length of the overlapping boundary to the total boundary length.

[0076] In one embodiment, the Jaccard similarity coefficient between the boundary of a narrow water body and the boundary of the surrounding vegetation area reaches 0.8, which is significantly higher than the preset threshold of 0.6. This indicates that the area is very likely a vegetation area that has been misidentified as a water body. Therefore, it is removed from the water body distribution data to form a third distribution dataset.

[0077] In step S1504, the boundaries of the water body regions in the third distribution dataset are adjusted, and the narrow water body regions are re-divided according to the preset classification criteria to obtain the corrected water body distribution map.

[0078] It should be noted that the boundary adjustment first employs a morphological opening operation-based method (erosion followed by dilation, using 3×3 circular structuring elements) to remove fine burrs while preserving the main boundary morphology. Subsequently, a buffer with a width of 5 pixels is established for analysis. This buffer width is determined based on the width statistics (75th percentile) of 200 urban waterway samples. Narrow water body regions located at the boundary between water and non-water bodies undergo boundary smoothing. The reclassification process is based on a preset multi-feature classification standard, which is determined through statistical analysis of the spectral and textural features of historically accurately classified water body samples. This standard requires that the region simultaneously meet the requirements of a normalized water index greater than 0.3 and a near-infrared reflectance less than 0.2.

[0079] In one embodiment, for a narrow water body region with jagged boundaries, morphological opening operations are first applied to remove isolated pixels, and then the boundary contour is smoothed through analysis using a buffer with a width of 5 pixels. The normalized water index of the region is then calculated to be 0.35 and the near-infrared reflectance is 0.18, which meets the preset water body classification criteria. Therefore, it is confirmed as a valid water body region, and finally a corrected water body distribution map with smooth boundaries and accurate classification is generated.

[0080] In step S16, the change trend in the corrected water body distribution map is calculated. If the rate of change is higher than a preset rate of change threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained, including: S1601, Obtain the distribution information within a continuous time period in the corrected water body distribution map, and determine the preliminary time series dataset; S1602, perform difference calculation on the water body areas in each time period of the time series dataset. If the rate of change is higher than the preset rate of change threshold, it is marked as a potential dynamic change area, and the labeled change dataset is obtained. S1603, perform boundary confirmation on the potential dynamic change regions in the change dataset, integrate the regions by combining spatial distribution information, and obtain the final set of dynamic change regions; S1604, the corrected water body distribution map is adjusted using the set of dynamically changing regions to generate a monitoring distribution map containing dynamic change markers, and the final change monitoring distribution result is determined.

[0081] In step S1601, the distribution information within a continuous time period in the corrected water body distribution map is obtained to determine the preliminary time series dataset.

[0082] It should be noted that the selection criteria for the continuous time period are based on the remote sensing image data acquisition cycle and hydrological change characteristics, typically collecting water distribution maps for at least 12 consecutive periods on a monthly basis. To address the issue of missing data, cloud mask generation technology is used to identify and mark cloud-covered areas, and images with cloud coverage exceeding 50% are removed. Simultaneously, linear temporal interpolation methods are used to fill in the missing water distribution data for each period, ensuring the continuity of the time series. By performing spatial overlay analysis on these chronologically arranged water distribution maps, the water area for each period is calculated and its changes are recorded, forming a preliminary time series dataset.

[0083] In one embodiment, monthly water distribution maps of a lake area over the past 12 months were selected for time series analysis. Monthly data for month 3 was excluded due to severe cloud cover, and supplementary data for month 3 was generated using linear interpolation based on data from months 2 and 4. Spatial overlay revealed that the water area in month 6 decreased by 12% compared to month 5. This change exceeded the 8% normal fluctuation threshold determined based on data from the same period over the past 5 years. Therefore, this significant change was recorded in the preliminary time series dataset to provide a basis for subsequent trend analysis. The normal fluctuation threshold used to determine significant area changes was determined by analyzing historical water area fluctuations in the region; changes exceeding the 95th percentile of the historical normal fluctuation range were considered significant changes.

[0084] In step S1602, the differences of water bodies in each time period in the time series dataset are calculated. If the rate of change is higher than the preset rate of change threshold, it is marked as a potential dynamic change area, and the labeled change dataset is obtained.

[0085] It should be noted that the preset rate of change threshold is determined based on statistical analysis of monthly water level monitoring data and water area change sequences extracted from remote sensing images of the region over the past 10 years (2013-2023). A sliding window method is used to calculate the rate of change of water area in each month relative to the previous month, forming a rate of change sequence. The upper limit of its 95% confidence interval, 8%, is calculated as the basic threshold based on this sequence. To adapt to the changing characteristics of different water bodies, an adaptive threshold mechanism based on water body type is further established. A threshold of 6% is set for lakes and reservoirs with gradual changes, a threshold of 10% is set for rivers with seasonal changes, and a threshold of 15% is set for temporary water bodies susceptible to rainfall. Differences in water body areas within each time period in the time series dataset are calculated by comparing the water body distribution maps of adjacent time periods, thus calculating the area change rate of each local water body area.

[0086] In one embodiment, by comparing the lake water distribution maps for two consecutive months, the water area change rate of a certain marginal sub-region was calculated to be 15%. The applicable change rate threshold for this lake region is 6%, and the actual change of 15% is significantly higher than this threshold, indicating that the change in this region is significantly beyond the normal range. Therefore, it is marked as a potential dynamic change region and included in the labeled change dataset.

[0087] In step S1603, the boundaries of the potential dynamic change regions in the change dataset are confirmed, and the regions are integrated in combination with spatial distribution information to obtain the final set of dynamic change regions.

[0088] It should be noted that the boundary confirmation process first divides the change areas into three types based on the continuous temporal water area change rate: expansion type (change rate greater than +5%), shrinkage type (change rate less than -5%), and stable type (change rate between ±5%). Then, it performs spatial overlay analysis on potential dynamic change areas and land use classification maps from the same period. The Jaccard similarity coefficient is used to quantify the degree of overlap with known land cover boundaries, eliminating spurious changes caused by classification errors. Regional integration is based on the principle of consistency in change types, employing the Euclidean distance clustering algorithm. Based on statistical analysis of the spatial distribution characteristics of historical urban water body change areas, 50 meters is set as the typical spatial clustering distance threshold. This value corresponds to the average interval between adjacent water body patches in the urban environment. For change areas of the same type, adjacent areas with boundary distances within 50 meters and consistent change characteristics are merged into the same change unit. For overlapping areas where expansion and shrinkage coexist, buffer overlap analysis is used to identify conflict areas, and the dominant change type is determined based on the majority principle and spectral continuity.

[0089] In one embodiment, two adjacent change regions were identified as belonging to the expansion type (change rate +8%) and the shrinkage type (change rate -12%), respectively. First, spatial overlay analysis confirmed that the boundary overlap rate between a potential change region and the surrounding farmland area was as high as 75%, confirming that this part was a classification confusion rather than a real water body change, and therefore it was removed. At the same time, based on the 50-meter spatial aggregation distance threshold, three adjacent patches belonging to the shrinkage type, 30-45 meters apart and with continuous spectral characteristics were merged into a complete dynamic change region. For the existing expansion-shrinkage conflict region, by analyzing the change trend of its near-infrared band reflectance (continuous decrease), it was finally classified as a shrinkage region, and finally a set of dynamic change regions with accurate spatial boundaries and consistent change characteristics was formed.

[0090] In step S1604, the corrected water body distribution map is adjusted using the set of dynamically changing regions to generate a monitoring distribution map containing dynamic change markers, and the final change monitoring distribution result is determined.

[0091] It should be noted that the adjustment is achieved through spatial overlay analysis, which merges the dynamically changing areas with the corrected water distribution map. The changing areas are marked using a specific color coding system; for example, expanding areas are represented by dark blue, and shrinking areas by red. The size of the marker symbol is determined based on the ratio of the changing area to the total water area.

[0092] In one embodiment, three areas of water shrinkage were identified in a lake area, with the changed areas accounting for 1.5%, 3.2%, and 4.1% of the total water area, respectively. According to preset standards, the latter two changed areas were marked with large red triangles, and the water boundary lines on the distribution map were updated. Finally, a monitoring distribution map containing information on the type, degree, and spatial location of the changes was generated, forming a complete change monitoring distribution result.

[0093] In step S17, the change monitoring distribution results are overlaid with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained, including: S1701, The change monitoring distribution results are fused with the pre-established urban land cover layer, and spatial alignment is performed for narrow water bodies to obtain a preliminary overlay layer; S1702, the distribution characteristics of the narrow water body area are analyzed through the preliminary overlay layer and spatial boundary information is obtained to obtain the updated layer; S1703, compare the boundary features with the actual ground feature information according to the updated layer. If the deviation exceeds the preset boundary deviation threshold, the boundary features are corrected to obtain a corrected layer that meets the monitoring requirements. S1704, The dynamic change information of the narrow water body area is integrated with the urban land feature information through the corrected layer to generate the final dynamic monitoring map of the narrow water body in the city.

[0094] In step S1701, the change monitoring distribution results are fused with the pre-established urban land cover layer, and spatial alignment is performed for narrow water bodies to obtain a preliminary overlay layer.

[0095] It should be noted that the fusion process employs a registration method based on control points, selecting stable and reliable feature points in the urban feature layer, such as the centers of road intersections and building corners, as benchmarks to optimize the coordinate transformation parameters of the water body distribution map. For spatial alignment of narrow water bodies, based on the matching relationship analysis between urban feature positioning accuracy and remote sensing image spatial resolution, a maximum permissible offset threshold of 2 pixels (2 meters) is set in remote sensing images with a 1-meter spatial resolution. Digital elevation model (DEM) data is then introduced to correct registration errors caused by terrain undulations. This threshold setting comprehensively considers both the accuracy requirements of urban monitoring and the practical feasibility of multi-source data fusion.

[0096] In one embodiment, a 5-meter-wide canal on the edge of a city has a spatial offset of 3 pixels in a 1-meter resolution change monitoring distribution map. By selecting 4 road intersections in the surrounding area as control points and combining the digital elevation model data of the area for terrain correction, the offset is reduced to less than 1 pixel, so that the canal boundary is precisely aligned with the road boundary in the urban feature layer, forming a preliminary overlay layer with consistent spatial position.

[0097] In step S1702, the distribution characteristics of the narrow water body area are analyzed through the preliminary overlay layer, and spatial boundary information is obtained from it to obtain the updated layer.

[0098] It should be noted that the distribution characteristics of the narrow water body area are analyzed through the preliminary overlay layer to obtain spatial boundary information, extract the continuous contour of the narrow water body area, and calculate its geometric feature parameters, including maximum width, perimeter, and aspect ratio. The maximum width threshold (5 meters) used to define a "narrow water body" is determined based on statistical analysis of specification measurement data of typical urban waterways and drainage facilities. The algorithm can accurately identify areas with widths within the threshold and exhibiting significant linear characteristics.

[0099] In one embodiment, for an urban waterway, its continuous outline is identified, and the maximum width of the water body is calculated to be 2 meters, its length to be 500 meters, and its aspect ratio to be 250:1, meeting the definition criteria for a narrow water body. This precise spatial boundary information is recorded and used to generate an updated layer with complete geometric features.

[0100] In step S1703, the boundary features are compared with the actual ground features based on the updated layer. If the deviation exceeds a preset boundary deviation threshold, the boundary features are corrected to obtain a corrected layer that meets the monitoring requirements.

[0101] It should be noted that the preset boundary deviation threshold is determined based on statistical analysis of the positioning accuracy of urban features. Specifically, it is determined by measuring the coordinate deviation distribution of known control points and using the 95th percentile value of 2 meters as the criterion. The boundary features are compared with the actual feature information based on the updated layer. During the comparison process, the deviation is quantified by calculating the average Euclidean distance between the water body boundary line in the updated layer and the reference position of the actual feature.

[0102] In one embodiment, the boundary line of a narrow water body area shows an average deviation of 5 meters compared with actual measurement data, exceeding the preset threshold of 2 meters. In this case, a spline function-based geometric correction method is used, with surrounding roads and buildings as spatial references, to perform elastic registration of the water body boundary, correcting the deviation to within 1.5 meters, and finally generating a corrected layer that meets the accuracy requirements.

[0103] In step S1704, the dynamic change information of the narrow water body area is integrated with the urban land feature information through the corrected layer to generate the final dynamic monitoring map of the narrow urban water body.

[0104] It should be noted that the integration process employs a spatial overlay analysis method to fuse dynamic change information with urban feature layers. The degree of change is visually represented through a color coding system; for example, areas with a reduction of more than 10% are marked in orange, and areas with a reduction of more than 20% are marked in red. Simultaneously, the layer accuracy level is calculated based on the sampling verification results. A stratified random sampling strategy is used, with sample weights set according to different change types (expansion, shrinkage, stability) and regional characteristics (urban areas, suburbs). A total of 50 sample points are selected, and cross-validation is performed using field verification and the Landsat Water Mask public dataset. The accuracy is expressed as the percentage of correctly identified points out of the total sample points.

[0105] In one embodiment, a city's waterway area shrank by 22% in the past six months. The system marked this area as a red change zone and overlaid it with urban landmark information such as road networks and building outlines. Thirty feature points were selected using a stratified random sampling method (15 from densely populated urban areas and 15 from open suburban areas). Cross-validation was performed using field surveys and the JRC Global Surface Water dataset. Twenty-eight of these points matched the actual situation, resulting in a layer accuracy level of 93.3%. This ultimately generated a dynamic monitoring map of narrow urban water bodies that met the required monitoring accuracy.

[0106] In summary, this invention discloses a method for urban water body monitoring based on multi-source remote sensing data. The method includes acquiring and correcting multi-source remote sensing images to obtain a preliminary calibration dataset; performing band matching and noise suppression to obtain a standardized image set; extracting edge features and combining them with spectral verification to identify water body boundaries, obtaining a boundary enhancement dataset; determining an initial water body distribution map through geometric correction and land cover classification; extracting features of narrow water body areas, eliminating misclassifications through shadow difference comparison and boundary overlap analysis to obtain a corrected distribution map; calculating change trends and marking dynamic areas to obtain change monitoring results; and finally, overlaying and verifying the results with urban land cover layers to generate a dynamic monitoring map of narrow urban water bodies. This method solves the problem of identifying narrow water bodies through multi-source data fusion and multi-level verification, improving monitoring accuracy and reliability.

[0107] Reference Figure 2 The second embodiment of the present invention provides an urban water monitoring system based on multi-source remote sensing data, comprising: The multi-source data acquisition and preprocessing module acquires multi-source remote sensing images and performs correction and radiometric calibration on the acquired raw spectral data to unify data source compatibility and obtain a preliminarily calibrated remote sensing dataset. The data standardization processing module performs band matching adjustment and noise suppression on the remote sensing dataset, completes feature distribution balancing, and obtains a standardized image set. The water body boundary enhancement and recognition module extracts edge feature information from the standardized image set. If the edge intensity exceeds a preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. The initial water body classification module performs geometric correction on the images in the boundary enhancement image set, and combines the land cover classification accuracy analysis to initially divide the water body and non-water body areas to determine the initial water body distribution map. The narrow water body fine correction module extracts the distribution characteristics of the narrow water body region in the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, the misjudgment is eliminated through boundary overlap analysis to obtain the corrected water body distribution map. The dynamic change detection module calculates the change trend in the corrected water body distribution map. If the change rate is higher than the preset change rate threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained. The monitoring results integration and output module overlays the change monitoring distribution results with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

[0108] It should be noted that the urban water monitoring system based on multi-source remote sensing data provided in this embodiment of the invention is used to execute all the process steps of the urban water monitoring method based on multi-source remote sensing data in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0109] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an urban water body monitoring program based on multi-source remote sensing data. When the processor executes the computer program, it implements the steps described in the various embodiments of the urban water body monitoring method based on multi-source remote sensing data, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above system embodiments, such as the narrow water body fine correction module.

[0110] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0111] The electronic device may be a desktop computer, laptop, handheld computer, or smart tablet, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0112] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting all parts of the electronic device via various interfaces and lines.

[0113] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0114] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0115] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the system embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0116] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for monitoring urban water bodies based on multi-source remote sensing data, characterized in that, include: Multi-source remote sensing images were acquired, and the acquired raw spectral data were corrected and radiometrically calibrated to unify data source compatibility and obtain a preliminary calibrated remote sensing dataset. Band matching adjustment and noise suppression are performed on the remote sensing dataset to achieve feature distribution equalization and obtain a standardized image set; The edge feature information of the standardized image set is extracted. If the edge intensity exceeds the preset first edge intensity threshold and is accompanied by the verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. Geometric correction is performed on the images in the boundary enhancement image set, and the water and non-water areas are preliminarily divided by the ground feature classification accuracy analysis to determine the initial water body distribution map; Extract the distribution characteristics of narrow water body regions from the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, then eliminate misjudgments through boundary overlap analysis to obtain the corrected water body distribution map. Calculate the change trend in the corrected water body distribution map. If the rate of change is higher than the preset rate of change threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained. The change monitoring distribution results are overlaid with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

2. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, Multi-source remote sensing images were acquired, and the acquired raw spectral data were corrected and radiometrically calibrated to ensure data source compatibility, resulting in a preliminary calibrated remote sensing dataset, including: Multi-source remote sensing images are acquired through satellite and UAV platforms. The raw spectral data are then corrected and the differences in spectral response are unified to obtain a corrected spectral dataset. Radiometric consistency adjustment is performed on image data from different sources in the spectral dataset to determine a unified dataset after radiometric calibration. If the unified dataset has data compatibility deviations, format standardization processing is performed to obtain a more compatible fused dataset. The fused dataset is subjected to noise suppression and data integrity checks to obtain the final calibrated remote sensing dataset.

3. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, The remote sensing dataset is subjected to band matching adjustment and noise suppression to achieve feature distribution equalization, resulting in a standardized image set, including: The remote sensing dataset is subjected to band alignment and consistency adjustment to obtain a band-matched dataset; The random noise in the dataset is filtered to obtain a noise-suppressed image set; If there are deviations caused by sensor differences in the image group, they are corrected to determine the data group after sensor consistency adjustment; Background noise and atmospheric interference in the data set are removed to obtain a standardized image set with balanced features.

4. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, Edge feature information is extracted from the standardized image set. If the edge intensity exceeds a preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is determined to be a potential water body boundary region, resulting in a boundary-enhanced image set, including: Feature extraction is performed on the standardized image set to obtain an image set with prominent edge features; The edge intensity of the edge feature image set is compared. If it exceeds a preset first edge intensity threshold, it is judged as a potential boundary region, and a high edge intensity region image set is obtained. The spectral feature distribution of the high edge intensity region image set is detected and the distribution balance is verified. If it meets the preset distribution standard of a specific land cover type, it is determined as a candidate water body boundary region, and a spectrally balanced region image set is obtained. The candidate water body boundary regions in the spectrally balanced region image set are optimized for detail to obtain a boundary-enhanced image set.

5. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, Geometric correction is performed on the images in the boundary enhancement image set. Combined with the accuracy analysis of land cover classification, water and non-water body regions are initially divided to determine the initial water body distribution map, including: Position calibration is performed on the images in the boundary enhancement image set to obtain a first calibrated image set; The first image set is deformed and corrected to obtain the corrected second image set; The surface features of the second image set are classified. If the classification result meets the preset classification accuracy threshold, it is judged as a preliminary water body or non-water body area, and the third image set after classification is determined. The water areas in the third image set are delineated, and the classification results are integrated to obtain an initial water distribution map.

6. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, Extract the distribution characteristics of narrow water body regions from the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than a preset shadow difference threshold, then false judgments are eliminated through boundary overlap analysis to obtain a corrected water body distribution map, including: Feature extraction is performed on the narrow water body regions of the initial water body distribution map to obtain preliminary regional classification information and determine the first distribution dataset; The spectral values ​​of the narrow water body region in the first distribution dataset are compared with the surrounding shadows. If the difference is lower than the preset shadow difference threshold, it is judged as shadow interference, and the second distribution dataset is obtained. Perform boundary overlap analysis between the narrow water body region and the surrounding region in the second distribution dataset. If the degree of overlap is higher than the preset boundary overlap threshold, remove the falsely judged part and obtain the third distribution dataset. The boundaries of the water body regions in the third distribution dataset are adjusted, and the narrow water body regions are re-divided according to the preset classification criteria to obtain the corrected water body distribution map.

7. The urban water body monitoring method based on multi-source remote sensing data according to claim 1, characterized in that, Calculate the change trend in the corrected water body distribution map. If the rate of change is higher than a preset rate of change threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained, including: Obtain the distribution information of the corrected water body distribution map within a continuous time period to determine the preliminary time series dataset; Differences are calculated for water bodies in each time period of the time series dataset. If the rate of change is higher than a preset rate of change threshold, it is marked as a potential dynamic change area, thus obtaining a labeled change dataset. Boundary confirmation is performed on the potential dynamic change regions in the change dataset, and regional integration is performed in combination with spatial distribution information to obtain the final set of dynamic change regions; The corrected water body distribution map is adjusted by the set of dynamically changing regions to generate a monitoring distribution map containing dynamic change markers, and the final change monitoring distribution results are determined.

8. The urban water body monitoring method based on multi-source remote sensing data according to claim 7, characterized in that, The changes in the monitored distribution results are overlaid with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, a final dynamic monitoring map of narrow urban water bodies is obtained, including: The change monitoring distribution results are fused with a pre-established urban feature layer, and spatial alignment is performed for narrow water bodies to obtain a preliminary overlay layer. The distribution characteristics of the narrow water body area are analyzed by the initial overlay layer, and spatial boundary information is obtained from it to obtain the updated layer; The boundary features are compared with the actual ground features based on the updated layer. If the deviation exceeds the preset boundary deviation threshold, the boundary features are corrected to obtain a corrected layer that meets the monitoring requirements. The corrected layer integrates the dynamic change information of the narrow water body area with urban landform information to generate the final dynamic monitoring map of the narrow urban water body.

9. A city water body monitoring system based on multi-source remote sensing data, characterized in that, include: The multi-source data acquisition and preprocessing module acquires multi-source remote sensing images and performs correction and radiometric calibration on the acquired raw spectral data to unify data source compatibility and obtain a preliminarily calibrated remote sensing dataset. The data standardization processing module performs band matching adjustment and noise suppression on the remote sensing dataset, completes feature distribution balancing, and obtains a standardized image set. The water body boundary enhancement and recognition module extracts edge feature information from the standardized image set. If the edge intensity exceeds a preset first edge intensity threshold and is accompanied by verification of balanced spectral feature distribution, it is judged as a potential water body boundary region, and a boundary enhancement image set is obtained. The initial water body classification module performs geometric correction on the images in the boundary enhancement image set, and combines the land cover classification accuracy analysis to initially divide the water body and non-water body areas to determine the initial water body distribution map. The narrow water body fine correction module extracts the distribution characteristics of the narrow water body region in the initial water body distribution map. If the difference between the spectral value of the narrow water body region and the surrounding shadow is lower than the preset shadow difference threshold, the misjudgment is eliminated through boundary overlap analysis to obtain the corrected water body distribution map. The dynamic change detection module calculates the change trend in the corrected water body distribution map. If the change rate is higher than the preset change rate threshold, it is marked as a dynamically changing water body area, and the change monitoring distribution results are obtained. The monitoring results integration and output module overlays the change monitoring distribution results with a pre-established urban feature layer. If the accuracy of the overall monitoring is verified, the final dynamic monitoring map of narrow urban water bodies is obtained.

Images