River water bloom identification method, system, medium and device based on multi-source remote sensing

Abstract

The application discloses a kind of river water bloom identification method, system, medium and equipment based on multi-source remote sensing, it is related to water environment remote sensing monitoring field, including obtaining multi-source remote sensing dataset;To coast line vector line is segmented, determine coast line curvature value;Based on coast line curvature value, determine coast line complexity weight factor, and according to water body suspended substance spatial distribution characteristics, determine water body spectrum correction term;Based on coast line complexity weight factor and water body spectrum correction term, determine the MNDWI value after correction, based on the MNDWI value after correction, extract river water body boundary in remote sensing dataset, determine initial river water body vector layer;Extract width information and bending degree in initial river water body vector layer, determine river section type and complex coast line river section;Adjust the number of pixels of inward mask to eliminate interference pixels, determine pure river water body area;Based on FAI-Otsu algorithm, obtain the river water bloom identification result of pure river water body area.

Description

Technical Field

[0001] This invention relates to the field of water environment remote sensing monitoring, and in particular to a method, system, medium, and equipment for identifying algal blooms in rivers based on multi-source remote sensing. Background Technology

[0002] With the increasing prominence of eutrophication in water bodies, the frequency of algal blooms in rivers is increasing year by year, seriously damaging the balance of aquatic ecosystems and threatening the safety of drinking water sources and the ecological environment for human habitation.

[0003] Satellite remote sensing technology, with its advantages of macroscopic, rapid, and periodic observation, has become a core technical means for monitoring algal blooms. However, when traditional remote sensing extraction methods are applied to rivers, they face many specific technical bottlenecks. For example, traditional methods directly use the MNDWI-Otsu algorithm to segment water bodies in the target area. Due to the interference of riverbank curvature on water body segmentation, in river sections with complex and curved shorelines, the spectral differences between the river water and the background (vegetation, bare soil) are weakened, resulting in large deviations in riverbank segmentation and low sensitivity for water body identification in turbid rivers. Summary of the Invention

[0004] This invention provides a method, system, medium, and device for identifying river algal blooms based on multi-source remote sensing, to solve the aforementioned problems existing in the prior art, namely, how to accurately identify river algal blooms in the prior art. This invention provides a method for identifying river algal blooms based on multi-source remote sensing, which includes: Obtain multi-source remote sensing datasets for the target area; By segmenting the shoreline vector lines of the target area, the shoreline curvature value of each segment unit is determined; based on the shoreline curvature value, the shoreline complexity weighting factor is determined, and the water spectral correction term is determined according to the spatial distribution characteristics of suspended matter in the target area. Based on the shoreline complexity weighting factor and the water spectral correction term, the corrected MNDWI value is determined. Based on the corrected MNDWI value, the following is adopted: - The Otsu algorithm extracts the river water body boundaries from the remote sensing dataset to determine the initial river water body vector map layer; The width and curvature information of the initial river channel water vector layer are extracted. Based on the width information, the river segment is divided into multiple river segment types, and river segments with curvature greater than a preset threshold are marked as complex shoreline river segments. The river segment types include narrow segments, medium-width segments, and wide segments. For different river segment types and river segments with complex shorelines, the number of pixels in the inward mask is dynamically adjusted to eliminate interfering pixels and determine the pure river water area; based on the FAI-Otsu algorithm, the identification results of river algal blooms corresponding to the pure river water area are obtained.

[0005] Optionally, the acquisition of the initial river channel water vector map layer specifically includes: The ratio of the difference between the green light band reflectance and the shortwave infrared band reflectance to the sum of the green light band reflectance and the shortwave infrared band reflectance is used as the value of the original improved normalized difference water index (MNDWI). The method for obtaining this value is as follows: in, For green light band reflectivity, For shortwave infrared band reflectivity; By dividing the acquired shoreline vector line of the target area into equidistant segments, the original curvature value of each segment unit is determined. Based on the 95th quantile calibration of the original curvature values, global normalization is performed to determine the shoreline curvature value. Based on shoreline curvature value The shoreline complexity weighting factor W is determined as follows: Based on the original MNDWI value, the shoreline complexity weighting factor W, and the introduced water spectral correction term C, the following method is used to determine... : The weighted inter-class variance is determined using the weighted Otsu thresholding method. The threshold that maximizes the weighted inter-class variance is taken as the optimal segmentation threshold. The optimal segmentation threshold is then used to... The calculation results are used for threshold segmentation to extract the river water body boundary in the remote sensing dataset in order to determine the initial river water body vector map layer.

[0006] Optionally, obtaining the weighted inter-class variance specifically includes: The remote sensing dataset is divided into foreground and background regions, based on the probability of gray level i appearing in the dataset. The pixel ratio of the foreground region is determined using the following formula. and background pixel ratio : Based on the pixel ratio of the foreground region and background pixel ratio Combining probability The mean gray level of the foreground is determined using the following method. Compared with the average gray level of the background : By introducing spatial weighting factors Combined with the pixel ratio of the foreground area Background pixel ratio Mean gray level of foreground Compared with the average gray level of the background The weighted inter-class variance is determined as follows: : .

[0007] Optionally, the spatial weighting factor of the river channel water body boundary. The value ranges from 0.7 to 0.9.

[0008] Optionally, one, two, and two pixel inward masks are set for the narrow, medium, and wide sections, respectively, and a local mask compensation of 0.5 pixels is added for complex shoreline river sections to eliminate interfering pixels and determine the pure river water area.

[0009] Optionally, obtaining the identification results of the river algal bloom specifically includes: Based on the FAI-Otsu algorithm, the basic phytoplankton index (FAI) is determined. Then, by introducing a spectral adaptive correction factor, the basic FAI is corrected to determine the improved FAI. + The improved phytoplankton index is obtained using the following formula. : in, The basic FAI calculation formula is as follows: In the formula, K is the spectral adaptive correction factor; Chl−a is the chlorophyll a inversion value of the water body in the current image. The background value of chlorophyll a in the area to be tested; SS is the current concentration of suspended solids in the water body in the image. This represents the background value of suspended matter in the area to be tested. , , These are the surface reflectances in the red, near-infrared, and short-wave infrared bands, respectively. , , These are the center wavelengths of the red light, near-infrared, and short-wave infrared bands, respectively. The near-infrared band interpolated reflectance is obtained by linear interpolation between the red band and the short-wave infrared band. For each scene in a multi-scene remote sensing image, the temporal weighted Otsu thresholding method is used to adaptively segment the improved phytoplankton index (FAI+) to determine the FAI of each scene. + Initial threshold; Each FAI + The initial thresholds are sorted in ascending order, and their corresponding cumulative weights are determined; the FAI corresponding to the first time the cumulative weight reaches or exceeds a preset range is selected. + The initial threshold is used as a unified threshold for detecting algal blooms. ; FAI of each pixel in the image to be identified + Value and unified discrimination threshold When the pixel FAI is compared, + Value greater than or equal to the unified discrimination threshold If the condition is met, the pixel is determined to be an algal bloom pixel; otherwise, it is determined to be a non-algal bloom water body pixel.

[0010] Optionally, the multi-source remote sensing dataset specifically includes: Landsat−8 / 9 OLI imagery and Sentinel−2 MSI imagery.

[0011] This invention provides a river algal bloom identification system based on multi-source remote sensing, realizing the above-mentioned river algal bloom identification method based on multi-source remote sensing.

[0012] The present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the above-described method for identifying river algal blooms based on multi-source remote sensing.

[0013] The present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the above-mentioned method for identifying river algal blooms based on multi-source remote sensing.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a method for identifying river algal blooms based on multi-source remote sensing. This method introduces a shoreline complexity weighting factor and a water body spectral correction term to correct the MNDWI value, which can adaptively enhance the spectral signal of the water body in the meandering river, offset the spectral shift caused by sediment and algae, and reduce the extraction error of the water body boundary in complex shoreline sections. At the same time, this invention innovatively proposes an automated calculation and normalization method for shoreline curvature value Curv based on the GEE platform, which realizes the quantitative and accurate characterization of shoreline curvature and improves the accuracy of water body boundary extraction. Based on the extracted pure river water body area, the FAI-Otsu algorithm is used to obtain the identification results of river algal blooms corresponding to the pure river water body area, so as to improve the identification accuracy of river algal blooms. Attached Figure Description

[0015] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0016] Figure 1 A flowchart illustrating a method for identifying algal blooms in river channels based on multi-source remote sensing, provided as an embodiment of the present invention; Figure 2 A schematic diagram of a computer device for a river algal bloom identification method based on multi-source remote sensing provided in an embodiment of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0018] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0019] Example 1 Figure 1 This is a flowchart of a river algal bloom identification method based on multi-source remote sensing provided in an embodiment of the present invention, as shown below. Figure 1 As shown in the figure, this embodiment illustrates a method for identifying river algal blooms based on multi-source remote sensing, including: S1: Obtain the multi-source remote sensing dataset of the target area.

[0020] For example, leveraging the massive remote sensing datasets and cloud processing capabilities of the GEE platform, rapid acquisition, standardization processing, and extraction of water quality background parameters from multi-source data can be achieved. This addresses issues such as cloud contamination, spatial scale mismatch, and spectral incomparability among data from different sensors, providing parameter support for subsequent algorithm improvements. Specific operations may include: (1) Data acquisition: The Sentinel−2MSIL2A and Landsat−8 / 9OLIL2 preprocessing datasets built into the GEE platform are called. Based on the user-input boundary of the area to be measured and the monitoring time series parameters, the remote sensing images that meet the requirements are automatically selected without the need for manual download and import of data. (2) Cloud masking: Using the QA masking technology of the GEE platform, the cloud and cloud shadows in the image are automatically detected and removed, and only cloudless or low-cloud areas with less than 15% cloud cover are retained to reduce the interference of cloud and fog on the extraction of river algal bloom information. (3) Spatial resolution uniformity: The spatial resolution of Landsat−8 / 9OLI images can be uniformized to 10m by upsampling, which is consistent with the 10m resolution of Sentinel−2MSI images, thus solving the problem of spatial scale mismatch between different sensors and accurately depicting the fine boundary of algal blooms in the river channel. (4) Band matching: In the GEE platform, filter the blue, green, red, near-infrared and short-wave infrared bands that are common to Sentinel−2 and Landsat8 / 9 sensors or have similar spectral characteristics, and complete the one-to-one correspondence matching of bands to ensure the spectral comparability of multi-source remote sensing data. (5) Background parameter extraction: In GEE, the chlorophyll a background value Chl−a0 of the water body in the area to be measured is inverted by the three-band semi-analysis algorithm, and the suspended matter background value SS0 is inverted by the normalized suspended matter index combined with the empirical model, which provides the basis for the calculation of the spectral adaptive correction factor for the subsequent improved FAI algorithm.

[0021] S2: By segmenting the shoreline vector line of the target area, the shoreline curvature value of each segment unit is determined; based on the shoreline curvature value, the shoreline complexity weighting factor is determined, and the water spectral correction term is determined according to the spatial distribution characteristics of suspended matter in the target area.

[0022] For example, obtaining the shoreline curvature value (Curv) may include: converting the initial channel water vector surface features into shoreline vector line features and smoothing them by 10m; dividing the smoothed shoreline into 20m equidistant segments; and calculating the original curvature value of each segment. After 95% quantile calibration, global maximum-min normalization from 0 to 5 is performed to obtain Curv, which is then rasterized and assigned to the corresponding river channel pixels. The value of Curv ranges from 0 to 5, with larger values ​​for higher curvature.

[0023] S3: Based on the shoreline complexity weighting factor and the water spectral correction term, determine the corrected MNDWI value. Based on the corrected MNDWI value, use... The Otsu algorithm extracts the river water body boundaries from the remote sensing dataset to determine the initial river water body vector layer.

[0024] For example, this invention addresses the problem of large segmentation deviation in complex riverbanks caused by the traditional MNDWI-Otsu algorithm by introducing a shoreline complexity weighting factor and a water spectral correction term, thus improving it into an improved algorithm. The Otsu algorithm highlights the spectral characteristics of water bodies while improving the segmentation accuracy of complex shorelines, enabling precise and automated extraction of river water body boundaries within GEE. Specific operations include: For example, the improved MNDWI ( The calculation is expressed as: The formula for calculating MNDWI is as follows: in, For green light band reflectivity, denoted as shortwave infrared reflectance; W is the shoreline complexity weighting factor; Curv is the shoreline curvature value, ranging from 0 to 5, with higher curvature resulting in larger Curv values; C is the water body spectral correction term; the overall classification accuracy of water body identification is used as the benchmark for comparison. The closer the value is to 1, the purer the water body. Compared with the traditional MNDWI, the sensitivity of identifying river water bodies is improved by 15%.

[0025] For example, weighted Otsu thresholding adaptive segmentation may specifically include: The calculation results are used as input, and the optimal segmentation threshold is solved using a weighted Otsu's algorithm, introducing a spatial weight factor. This approach prioritizes pixels in the center of the river channel over those at the boundaries, avoiding interference from boundary noise on threshold segmentation. It also maximizes the weighted inter-class variance between water bodies (foreground) and non-water bodies (background). The core Otsu weighted calculation formula is as follows: In the formula, As a spatial weighting factor, the central pixel of the river channel =1.0, boundary cell =0.7~0.9, automatically calculated by the GEE platform through distance transformation; The remaining parameters are defined the same as in the traditional Otsu method: t is the segmentation threshold, and L is the total number of gray levels in the image. Let i be the probability of gray level i appearing. , These represent the pixel percentages of the foreground and background after segmentation based on threshold t. , These are the average gray values ​​of the foreground and background after segmentation based on threshold t, respectively. For inter-class variance, select the one that makes The maximum value of t is used as the optimal segmentation threshold; For example, the threshold that maximizes the variance between weighted classes is selected as the optimal segmentation threshold. The automatic extraction of the river water body boundary is realized in GEE, and the initial river water body vector map layer is obtained. The average deviation between the extracted boundary and the actual shoreline is less than 1 pixel, which is more than 50% lower than the deviation of traditional algorithms.

[0026] S4: Extract the width and curvature information from the initial river water vector map layer. Based on the width information, divide the river segment into multiple river segment types and mark the river segment with a curvature greater than a preset threshold as a complex shoreline river segment. The river segment types include narrow segment, medium-width segment, and wide segment.

[0027] For example, the initial river water vector layer obtained in the GEE platform is subjected to width quantization and curvature analysis. The curvature is the ratio of the actual length of the shoreline to the straight distance between the two ends of the river segment (the ratio is ≥1). The distance perpendicular to the shoreline is calculated by the surface-to-line tool, and the river is divided into narrow, medium-wide, and wide segments. The river segment with a curvature >1.2 is marked as a complex shoreline river segment by the curvature calculation tool.

[0028] S5: For different river section types and river sections with complex shorelines, dynamically adjust the number of pixels of the inward mask to eliminate interfering pixels and determine the pure river water area; based on the FAI-Otsu algorithm, obtain the identification results of river algal blooms corresponding to the pure river water area.

[0029] For example, addressing the shortcomings of traditional fixed masking techniques, this invention designs a dedicated adaptive multi-scale inward masking technology. This technology dynamically adjusts masking parameters based on river width and curvature, eliminating near-shore interference while preventing the loss of effective water body and mitigating the impact of near-shore boundary effects on subsequent algal bloom identification. 0.5-pixel masking compensation is achieved at the sub-pixel level through raster resampling and distance-weighted interpolation on the GEE platform. Specific operations are as follows: (1) Adaptive mask parameter settings: a 1-pixel inward mask is applied to narrow sections, and a 2-pixel inward mask is applied to medium and wide sections; a local mask compensation of 0.5 pixels is added to complex shoreline river sections, and sub-pixel level accurate masking is achieved through GEE raster operation; (2) Interference removal and pure area output: Based on the above adaptive parameters, the buffer analysis function is called to perform inward masking, automatically removing pixels at the water body boundary that are disturbed by shoreline vegetation, suspended sediment, human buildings, etc., to obtain a pure river water body to be tested without near-shore background interference. The interference removal rate is over 99%, and the effective water body retention rate is over 98%.

[0030] For example, this invention addresses the problems of poor spectral adaptability and low threshold stability in the traditional FAI-Otsu algorithm by introducing a spectral adaptive correction factor and a seasonal weighting factor, thus improving it into a FAI algorithm. + The time-weighted Otsu algorithm (an improved FAI-Otsu algorithm), combined with time-series weight quantile optimization, achieves adaptive and highly stable selection of algal bloom thresholds, eliminates false alarms due to low biomass noise, and improves the sensitivity of identification during peak algal bloom periods. FAI + After normalization from 0 to 100, the specific operation is as follows: (1) Improved FAI (FAI) + Calculation: Based on the obtained pure river water area, according to the formula: ; in, Where K is the spectral adaptive correction factor; , These are the real-time inversion of chlorophyll a and suspended solids concentration, respectively. , This represents the regional water quality background value. , Reflectance of red, near-infrared, and short-wave infrared bands and their corresponding center wavelengths; For linear interpolation of near-infrared reflectance, Based on the basic phytoplankton index, This invention improves the phytoplankton index.

[0031] The FAI (Fluorescence Indication) is calculated from the background water quality parameters of the area to be measured and the chlorophyll a and suspended matter inversion values ​​of the current image. + The value is only related to algal abundance, eliminating interference from water quality background; FAI + The higher the value, the higher the algal abundance in the water body, and the correlation with chlorophyll a concentration increases to over 0.9; (2) Temporally weighted Otsu initial threshold segmentation: The temporally weighted Otsu thresholding method is used for FAI segmentation. + The calculation results are subjected to adaptive threshold segmentation, and a seasonal weighting factor Wt is introduced to weight the inter-class variance of a single scene image. Wt=1.0 during the summer peak algal bloom period, Wt=0.8 during the winter low biomass period, and Wt=0.9 during spring and autumn. After time-series weighting of the inter-class variance, the FAI under different time series and environmental conditions is obtained. + The initial threshold, compared to the seasonal accuracy of the threshold, shows that the seasonal accuracy of the threshold is improved by more than 20% compared to the traditional Otsu.

[0032] (3) Determining a unified discrimination threshold by optimizing the temporal weight quantile: FAI for multiple images + The initial threshold is assigned a temporal weight consistent with Wt, which is consistent with the inter-class variance weighting of single-scene images. After statistical analysis, the 20th percentile of the weighted threshold is selected as the unified discrimination threshold for algal blooms. This effectively eliminates the false alarm phenomenon of algal blooms caused by low biomass and high noise in winter, while improving the sensitivity of low-concentration algal blooms during the high incidence of algal blooms in summer. The false alarm rate is reduced by more than 15%, and the false negative rate is reduced by more than 10%.

[0033] For example, quantitative classification of algal bloom intensity: based on FAI normalized from 0 to 100 in multiple images. + The 20%, 40%, and 60% weighted quantiles of the threshold are used as cutoff values ​​to classify river algal blooms into four levels, achieving quantitative grading of bloom intensity. The specific standards are as follows: Algal bloom-free: FAI + <20th weighted quantile; Mild algal bloom: 20th weighted quantile ≤ FAI + <40th weighted quantile; Moderate algal bloom: 40th weighted quantile ≤ FAI + <60th weighted quantile; severe algal bloom: FAI + ≥60% weighted quantile.

[0034] For example, the confusion matrix method combined with the multi-dimensional error discrimination method can be used to verify the accuracy of the extraction results. The unified accuracy qualification standard is: overall classification accuracy ≥ 95%, Kappa coefficient ≥ 0.9, weighted Kappa coefficient ≥ 0.9, and visual interpretation boundary overlap ≥ 90%. Using the visual interpretation results of river algal blooms as the real reference data, the classification error discrimination and boundary extraction error discrimination are completed in the GEE platform, without any local operation required. (1) Classification error judgment: Sample points are randomly arranged in the test area according to the principle of stratified random sampling, and a binary classification confusion matrix is ​​constructed. The producer accuracy, user accuracy, overall classification accuracy and Kappa coefficient are calculated. At the same time, the omission error = 1 − producer accuracy and misclassification error = 1 − user accuracy are used to quantify the omission and misclassification problems in algal bloom extraction. The omission error and misclassification error are the error indicators of water body / non-water body binary classification. For the four-level algal bloom intensity classification results, the weighted Kappa coefficient is calculated to evaluate the classification accuracy in multi-level scenarios and make up for the neglect of level deviation by the traditional Kappa coefficient. (2) Boundary extraction error judgment: In GEE, the river water body boundary, algal bloom boundary and the true boundary interpreted by visual interpretation are rasterized to a spatial resolution of 10m through code commands. The raster cells are assigned values ​​of 1 = boundary cell and 0 = non-boundary cell. The number of overlapping cells, the total number of true boundary cells and the number of cells with a deviation of 1 are counted by the GEE raster operation tool. The overlap degree of the visually interpreted boundary is automatically calculated. The basic formula is: For rivers with winding or complex shorelines, a modified formula is used: If the accuracy requirement is not met, readjust the threshold or optimize the mask parameters in GEE until the accuracy requirement is met.

[0035] Example 2 This embodiment takes a typical northern river channel with varying width and complex shoreline as the research object. Based on the method of this invention, it conducts accurate remote sensing extraction of algal blooms from 2021 to 2025 on the GEE cloud platform to verify the accuracy improvement effect of the improved algorithm. The specific implementation steps may include: (1) Data acquisition and standardization preprocessing: The data acquisition and preprocessing module was called in the GEE platform. 150 Sentinel-2MSIL2A images, 30 Landsat-8OLIL2 images, and 10 Landsat-9OLIL2 images from 2021 to 2025 were selected within the area to be measured. The module automatically completed cloud masking, 10m resolution unification, and blue-green-red near-infrared and short-wave infrared band matching. The chlorophyll a background value Chl−a0=10μg / L was obtained by inverting the three-band semi-analysis algorithm. The background value of suspended matter SS0=50mg / L was obtained by inverting the normalized suspended matter index combined with the empirical model. (2) Adaptive extraction of water body boundaries: The water body boundary extraction module automatically performs 10m smoothing, 20m equidistant segmentation and curvature calculation on the shoreline of the test area. After 95% quantile calibration and 0~5 normalization, the shoreline curvature value Curv distribution of the test area is obtained as 0~4.8. Based on this, the shoreline complexity weight factor W=1.0~1.96 is calculated. At the same time, according to the spatial distribution characteristics of suspended matter in the water body of the test area, the water body spectral correction term C=0.02~0.04 is automatically assigned to complete MNDWI. + Band operations and weighted Otsu thresholding were used to obtain the initial river channel water vector layer. Experimental results show that the average deviation between the water body boundary extracted by this method and the actual shoreline is 0.8 pixels, which is about 62% lower than the traditional MNDWI-Otsu algorithm (average deviation of 2.1 pixels), perfectly matching the characteristic of traditional algorithms in the background technology where the deviation exceeds 2 pixels.

[0036] (3) Near-shore interference suppression: The near-shore interference suppression module performs width quantization and curvature analysis on the initial river water vector layer, automatically dividing the river area to be measured into narrow segments (width ≤ 5 x 10m pixels, accounting for 30%), medium-wide segments (5 pixels < width ≤ 10 pixels, accounting for 50%), and wide segments (width > 10 pixels, accounting for 20%), and marking complex shoreline river segments with curvature > 0.8 (accounting for 25%). Adaptive multi-scale masking is dynamically applied to different river segments: narrow segments are masked with 1 pixel inward, medium-wide and wide segments are masked with 2 pixels inward, and complex shoreline river segments are compensated with 0.5 pixels local masking through raster resampling and distance weighted interpolation. After processing, the nearshore interference pixel removal rate in the test area reached 99.5%, and the effective water body retention rate reached 98%. In contrast, the effective water body retention rate of the traditional fixed 2-pixel mask method in the test area was only 75%. This method improves the effective water body retention rate by about 31% compared with the traditional method, and completely solves the problems of loss of effective water body in narrow sections and incomplete removal of interference in wide sections by the traditional mask.

[0037] (4) Algal bloom identification and threshold determination: Based on the water quality background parameters of the area to be tested and the chlorophyll a and suspended solids concentration retrieved in real time from the image, the spectral adaptive correction factor K = 0.95~1.15 is automatically calculated through the algal bloom identification and threshold determination module to complete the 0~100 normalized FAI. + Band calculations were performed; simultaneously, seasonal weighting factors Wt were assigned based on the monitoring time series (Wt=1.0 for the high incidence of algal blooms in summer, Wt=0.9 for spring and autumn, and Wt=0.8 for the low biomass period in winter). The time-weighted Otsu algorithm was used to complete the initial threshold segmentation. Then, the initial thresholds from multiple images were statistically weighted according to time series, and the 20th percentile after weighting was selected as the unified algal bloom discrimination threshold. Field tests verified that, under this method, the false alarm rate of winter algal blooms in the tested area decreased from 20% with the traditional FAI-Otsu algorithm to 4.5%, and the missed detection rate of low-concentration algal blooms in summer decreased from 15% with the traditional algorithm to 2.8%, significantly improving the accuracy and stability of algal bloom identification.

[0038] (5) Algal bloom intensity classification and multi-dimensional cloud-based accuracy verification: The intensity classification and multi-dimensional accuracy verification module is based on FAI normalized from 0 to 100. +The 20%, 40%, and 60% weighted quantiles of the threshold values ​​were used to classify algal blooms in the test area into four levels: none, mild, moderate, and severe. The correlation between the classification results and the measured chlorophyll a concentration in the field reached 0.92, achieving accurate quantitative characterization of bloom intensity. To verify the extraction accuracy, 3000 sample points (1500 water / non-water samples and 1500 bloom intensity classification samples) were deployed in the test area according to the stratified random sampling principle. An automatic confusion matrix was constructed to complete multi-dimensional accuracy verification. The results showed that the overall classification accuracy reached 96.5%, the Kappa coefficient was 0.932, and the weighted Kappa coefficient was 0.938. The visual interpretation boundary overlap of the conventional river section was 92%, and the boundary overlap of the curved and complex shoreline river section reached 97% after correction, with minimal omission error. The overall classification accuracy is 3.5%, and the misclassification error is 2.8%. All indicators far exceed the accuracy standards set by this invention (overall classification accuracy ≥ 95%, Kappa coefficient ≥ 0.9, weighted Kappa coefficient ≥ 0.9, visual interpretation boundary overlap ≥ 90%). Compared with the traditional algorithm combination (traditional MNDWI-Otsu + fixed mask + traditional FAI-Otsu), this method improves the overall classification accuracy by 11.5%, the Kappa coefficient by 0.111, and the extraction accuracy achieves a leapfrog improvement.

[0039] The above are one or more embodiments of the river algal bloom identification method based on multi-source remote sensing provided in this specification. Based on the same idea, this specification also provides a corresponding river algal bloom identification system based on multi-source remote sensing to implement the above-mentioned river algal bloom identification method based on multi-source remote sensing.

[0040] For example, FAI can be based on 0-100 normalization. + The system automatically performs four-level quantitative classification of algal bloom intensity using 20%, 40%, and 60% weighted quantiles, outputting a visual representation of algal bloom distribution and intensity classification in GEE. The system supports cloud export and visualization of the results. Simultaneously, it incorporates a built-in accuracy verification model combining a confusion matrix and multi-dimensional error discrimination. This model automatically deploys sample points, constructs a binary confusion matrix, calculates various accuracy evaluation indicators, automatically derives missed / misclassified errors, calculates weighted Kappa coefficients, and automates the calculation of boundary rasterization and visual interpretation boundary overlap, generating a full-dimensional accuracy verification report. If the accuracy does not meet the standards, a feedback mechanism is automatically triggered, returning the system to the water body boundary extraction module for users to readjust parameters, ensuring the reliability of the extraction results.

[0041] The improved algorithm model library of this system includes algorithm logic and encapsulated functions for shoreline curvature calculation, weighted Otsu segmentation, adaptive masking, spectral adaptive correction, and temporal weighted quantile optimization, which can be directly called and executed in the GEE platform; the parameter self-learning library can automatically optimize algorithm parameters such as weight factors, correction terms, and mask parameters in the GEE platform according to the topography, hydrology, and water quality characteristics of different river channels, without the need for repeated manual debugging.

[0042] Specific limitations regarding the multi-source remote sensing-based river algal bloom identification system can be found in the limitations of the multi-source remote sensing-based river algal bloom identification method mentioned above, and will not be repeated here. Each module in the aforementioned multi-source remote sensing-based river algal bloom identification system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0043] The present invention also provides a computer-readable storage medium storing a computer program that can be used to execute the above-described method for identifying river algal blooms based on multi-source remote sensing.

[0044] The present invention also provides Figure 2 The schematic diagram of the computer device shown is as follows: Figure 2 As shown, at the hardware level, the computer device includes a processor, an internal bus, a network interface, memory, and non-volatile memory, and may also include other hardware required for business operations. The processor reads the corresponding computer program from the non-volatile memory into memory and then runs it to implement the river algal bloom identification method based on multi-source remote sensing provided in the above embodiments.

[0045] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. Any references to memory, storage, databases, or other media used in the embodiments provided by this invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0046] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this invention.

Claims

1. A method for identifying river algal blooms based on multi-source remote sensing, characterized in that, include: Obtain multi-source remote sensing datasets for the target area; By segmenting the obtained shoreline vector lines of the target area, the shoreline curvature value of each segment unit is determined; Based on the shoreline curvature value, the shoreline complexity weighting factor is determined, and the water body spectral correction term is determined according to the spatial distribution characteristics of suspended matter in the target area. Based on the shoreline complexity weighting factor and the water spectral correction term, the corrected MNDWI value is determined. Based on the corrected MNDWI value, the following is adopted: The Otsu algorithm extracts the river water body boundaries from the remote sensing dataset to determine the initial river water body vector map layer. The width and curvature information of the initial river channel water vector layer are extracted. Based on the width information, the river segment is divided into multiple river segment types, and river segments with curvature greater than a preset threshold are marked as complex shoreline river segments. The river segment types include narrow segments, medium-width segments, and wide segments. For different river segment types and river segments with complex shorelines, the number of pixels in the inward mask is dynamically adjusted to eliminate interfering pixels and determine the pure river water area; based on the FAI-Otsu algorithm, the identification results of river algal blooms corresponding to the pure river water area are obtained.

2. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 1, characterized in that, The acquisition of the initial river channel water vector map layer specifically includes: The ratio of the difference between the green light band reflectance and the shortwave infrared band reflectance to the sum of the green light band reflectance and the shortwave infrared band reflectance is used as the value of the original improved normalized difference water index (MNDWI). The method for obtaining this value is as follows: in, For green light band reflectivity, For shortwave infrared band reflectivity; By dividing the acquired shoreline vector line of the target area into equidistant segments, the original curvature value of each segment unit is determined. Based on the 95th quantile calibration of the original curvature values, global normalization is performed to determine the shoreline curvature value. Based on shoreline curvature value The shoreline complexity weighting factor W is determined as follows: Based on the original MNDWI value, the shoreline complexity weighting factor W, and the introduced water spectral correction term C, the following method is used to determine... : The weighted inter-class variance is determined using the weighted Otsu thresholding method. The threshold that maximizes the weighted inter-class variance is taken as the optimal segmentation threshold. The optimal segmentation threshold is then used to... The calculation results are used for threshold segmentation to extract the river water body boundary in the remote sensing dataset in order to determine the initial river water body vector map layer.

3. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 2, characterized in that, The acquisition of the weighted inter-class variance specifically includes: The remote sensing dataset is divided into foreground and background regions, based on the probability of gray level i appearing in the dataset. The pixel ratio of the foreground region is determined using the following formula. and background pixel ratio : Based on the pixel ratio of the foreground region and background pixel ratio Combining probability The mean gray level of the foreground is determined using the following method. Compared with the average gray level of the background : By introducing spatial weighting factors Combined with the pixel ratio of the foreground area Background pixel ratio Mean gray level of foreground Compared with the average gray level of the background The weighted inter-class variance is determined as follows: : 。 4. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 3, characterized in that, The spatial weighting factor of the river water body boundary The value ranges from 0.7 to 0.

9.

5. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 1, characterized in that, One, two, and two pixels of inward masking are set for the narrow, medium, and wide sections, respectively, and a local masking compensation of 0.5 pixels is added for the complex shoreline river section to eliminate interfering pixels and determine the pure river water area.

6. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 1, characterized in that, The acquisition of the identification results of the river algal bloom specifically includes: Based on the FAI-Otsu algorithm, the basic phytoplankton index (FAI) is determined. Then, by introducing a spectral adaptive correction factor, the basic FAI is corrected to determine the improved FAI. + The improved phytoplankton index is obtained using the following formula. : in, The basic FAI calculation formula is as follows: In the formula, K is the spectral adaptive correction factor; Chl−a is the chlorophyll a inversion value of the water body in the current image. The background value of chlorophyll a in the area to be tested; SS is the current concentration of suspended solids in the water body in the image. This represents the background value of suspended matter in the area to be tested. , , These are the surface reflectances in the red, near-infrared, and short-wave infrared bands, respectively. , , These are the center wavelengths of the red light, near-infrared, and short-wave infrared bands, respectively. The near-infrared band interpolated reflectance is obtained by linear interpolation between the red band and the short-wave infrared band. For each scene in a multi-scene remote sensing image, the temporal weighted Otsu thresholding method is used to adaptively segment the improved phytoplankton index (FAI+) to determine the FAI of each scene. + Initial threshold; Each FAI + The initial thresholds are sorted in ascending order, and their corresponding cumulative weights are determined; the FAI corresponding to the first time the cumulative weight reaches or exceeds a preset range is selected. + The initial threshold is used as a unified threshold for detecting algal blooms. ; FAI of each pixel in the image to be identified + Value and unified discrimination threshold When the pixel FAI is compared, + Value greater than or equal to the unified discrimination threshold If the condition is met, the pixel is determined to be an algal bloom pixel; otherwise, it is determined to be a non-algal bloom water body pixel.

7. The method for identifying river algal blooms based on multi-source remote sensing as described in claim 1, characterized in that, The multi-source remote sensing dataset specifically includes: Landsat−8 / 9 OLI imagery and Sentinel−2 MSI imagery.

8. A river algal bloom identification system based on multi-source remote sensing, characterized in that, Implement the river algal bloom identification method based on multi-source remote sensing as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the river algal bloom identification method based on multi-source remote sensing as described in any one of claims 1-7.

10. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the river algal bloom identification method based on any one of claims 1-7.

Images