Black and odorous water remote sensing recognition method based on spatiotemporal spectrum multi-dimensional feature information fusion

Abstract

The application provides a black and odorous water body remote sensing identification method based on spatiotemporal spectrum multi-dimensional feature information fusion, comprising the following steps: S1, acquiring a remote sensing image; S2, preprocessing the remote sensing image; S3, extracting the water body boundary of the remote sensing image; S4, generating a black and odorous water body index value of the water body pixel by pixel; S5, constructing a black and odorous water body sample; S6, forming long-time sequence reference spectrum information of the black and odorous water body; S7, forming water body image pixel-by-pixel spectrum time sequence information; S8, calculating the similarity of the long-time sequence reference spectrum information of the black and odorous water body and the water body image pixel-by-pixel spectrum time sequence information; S9, judging whether the water body belongs to the black and odorous water body; the application enriches the feature parameters of the black and odorous water body remote sensing identification by fusing the spectrum time sequence information, solves the misclassification and missing classification problems of the black and odorous water body caused by seasonal changes, and improves the accuracy of the remote sensing identification.

Description

Remote Sensing Identification Method for Black and Odorous Water Bodies Based on Spatiotemporal Spectral Multidimensional Feature Information Fusion Technical Field

[0001] This application relates to the field of black and odorous water body identification technology, specifically to a remote sensing identification method for black and odorous water bodies based on the fusion of spatiotemporal spectral multidimensional feature information. Background Technology

[0002] Remote sensing technology, with its advantages of wide coverage and high timeliness, can effectively compensate for the shortcomings of traditional methods. However, existing remote sensing technologies for identifying black and odorous water bodies have the following limitations: they mainly rely on spectral feature information and multi-dimensional feature information resulting from the fusion of spectral and spatial feature information to achieve remote sensing identification of black and odorous water bodies. Techniques such as single-band threshold segmentation, band ratio, band difference, and band interpolation can provide effective support for the identification of black and odorous water bodies to a certain extent. However, due to the significant seasonal variations in black and odorous water bodies, especially fluctuations in water quality, pollution levels, and biological activity, single-temporal remote sensing... Remote sensing images often fail to comprehensively and accurately identify black and odorous water bodies. Due to the significant differences in the black and odorous state of water bodies across different seasons and time periods, analysis based on single-phase images is prone to misclassification and omission. Furthermore, black and odorous water bodies are mostly relatively small in size, requiring high spatial resolution from remote sensing images. Currently, the image data provided by domestic high-spatial-resolution satellites such as Gaofen satellites are mainly concentrated in the red, green, blue visible light, and near-infrared bands. The limited spectral range of these bands makes it difficult to accurately implement atmospheric correction during remote sensing image processing, resulting in the inability to obtain true surface reflectance information of ground objects, thus affecting the final identification results.

[0003] In summary, the current technology for identifying black and odorous water bodies using remote sensing is not yet fully mature. Due to the significant seasonal changes and small size of black and odorous water bodies, there are still certain deficiencies in identification accuracy. Furthermore, the requirement for high spatial resolution remote sensing imagery makes the identification cost relatively high. Therefore, there is an urgent need for a new technical solution to address the above-mentioned technical problems. Summary of the Invention

[0004] This application provides a remote sensing identification method for black and odorous water bodies based on the fusion of spatiotemporal spectral multidimensional feature information, including the following steps:

[0005] S1: Acquire remote sensing images of the water body in the target area;

[0006] S2: Preprocess the remote sensing images;

[0007] S3: Extract water body boundaries based on preprocessed remote sensing images and output a binarized water body region mask;

[0008] S4: Within the generated water area mask, generate the black and odorous water body index value of the water body pixel by pixel according to the black and odorous water body index calculation model (BOI calculation model), and integrate the black and odorous water body index value of each pixel as a new band into the water area image generated by the water area mask.

[0009] S5: Collect black and odorous water body sample points to construct a black and odorous water body sample; the black and odorous water body sample points are divided into training sample points and validation sample points.

[0010] S6: Using the spectral time series feature extraction algorithm, the red band time series information, green band time series information, blue band time series information, near-infrared band time series information and BOI time series information of the pixel location of the training sample points are extracted respectively. Then, the different band time series information and BOI time series information of all training sample points are averaged to form long-term reference spectral information of black and odorous water bodies.

[0011] S7: Using the spectral temporal feature extraction algorithm, the red band temporal information, green band temporal information, blue band temporal information, near-infrared band temporal information and BOI temporal information are extracted pixel by pixel from the remote sensing image after the water body boundary extraction is completed, forming the pixel-by-pixel spectral temporal information of the water body image.

[0012] S8: Calculate the similarity between the long-term reference spectral information of black and odorous water bodies and the pixel-by-pixel spectral temporal information of water body images using the time-weighted dynamic time warping algorithm;

[0013] S9: Based on the similarity between the long-term reference spectral information of the black and odorous water body and the pixel-by-pixel spectral temporal information of the water body image, and combined with the threshold binary classification method, determine whether the water body belongs to the black and odorous water body.

[0014] As a preferred option, the remote sensing imagery in S1 is provided by the Sentinel-2 satellite system.

[0015] As a preferred embodiment, the preprocessing in S2 includes radiometric correction, atmospheric correction, and geometric correction of the remote sensing image.

[0016] As a preferred embodiment, in step S3, the processed remote sensing image is first processed by calculating the Normalized Difference Water Index (NDWI), and regions with NDWI ≥ the local water body NDWI threshold are extracted. Then, through morphological filtering and spatial rule constraints, abnormal erroneous extractions are eliminated, and small patches with an area smaller than the effective water body scale are removed. Finally, the water body boundary is extracted, and a binary water body region mask is output.

[0017] As a preferred embodiment, the ratio of training sample points to validation sample points in S5 is 7:3.

[0018] As a preferred embodiment, the calculation formula for extracting red band time-series information, green band time-series information, blue band time-series information, near-infrared band time-series information, and BOI time-series information at the pixel location of the training sample point is as follows:

[0019] in, Let b be the spectral value / index value of the k-th sample point at time t. Let be the red band spectral value of the k-th sample point at time t; Let be the green band spectral value of the k-th sample point at time t; Let be the blue band spectral value of the k-th sample point at time t; Let be the near-infrared spectral value of the k-th sample point at time t; Let t be the black and odorous water index value corresponding to the k-th sample point at time t; t represents the month, t=1, 2, ... , =12.

[0020] As a preferred approach, the formula for averaging the time series information of different bands and the BOI time series information of all training sample points is as follows:

[0021] , =12;

[0022] This represents the average value of the b-th type spectral value / index value of the k-th sample point over consecutive time intervals at time t; b represents the band or index type, including Red, Green, Blue, NIR, and the Black and Odorous Water Index (BOI); K represents the total number of training sample points.

[0023] As a preferred embodiment, the long-term time series reference spectral information of the black and odorous water body The formula is:

[0024] .

[0025] This represents the average value of the red band spectrum at time t; This represents the average value of the green band spectrum at time t; This represents the average value of the blue band spectrum at time t; This represents the average value of the near-infrared spectral values ​​at time t; This represents the average value of the black and odorous water body index at time t.

[0026] As a preferred embodiment, the calculation formula for extracting red band time-series information, green band time-series information, blue band time-series information, near-infrared band time-series information, and BOI time-series information pixel by pixel from the remote sensing image after water body boundary extraction using the spectral time-series feature extraction algorithm is as follows:

[0027] in, This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ), the value of the b-th type band or index at time t; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower red band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower green band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower blue band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower near-infrared band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time The index value of black and odorous water bodies.

[0028] As a preferred approach, the formula for calculating the pixel-by-pixel spectral temporal information Y of water body images is:

[0029] ,in, , =12.

[0030] As a preferred embodiment, step S8 specifically involves using long-term reference spectral information of black and odorous water bodies. Using the pixel-by-pixel spectral temporal information Y of the water body image as the target curve, the spectral temporal similarity distance between the target curve and the reference curve is calculated pixel-by-pixel and band-by-band based on the time-weighted dynamic time warping algorithm. Then, the similarity measurement results of the red band, green band, blue band, near-infrared wave and BOI are weighted and fused to generate a comprehensive spectral similarity index. This index is used as a new feature band and embedded into the processed water body remote sensing image data.

[0031] As a preferred option, the comprehensive spectral similarity index The calculation formula is:

[0032] ;

[0033] ;

[0034] in, This represents the comprehensive spectral similarity index; This represents the spectral bands and spectral indices involved in the calculation. The spectral bands include the red band, green band, blue band, and near-infrared band, and the spectral indices include the BOI. The first one calculated based on the TWDTW algorithm A measure of spectral similarity across bands; The weights of each spectral band and spectral index reflect their relative importance in the identification of black and odorous water bodies. and These represent the long-term reference spectral information of black and odorous water bodies. The TWDTW algorithm is a time-weighted dynamic time warping algorithm that integrates pixel-by-pixel spectral temporal information of water bodies and water images.

[0035] As a preferred embodiment, step S9 specifically involves, based on the generated comprehensive spectral similarity index... A univariate decision tree classification model was constructed to determine whether water body pixels were black and odorous. The classification rules of the model were defined as follows:

[0036] in, Representing geographic coordinates The classification result of a pixel at a given location is encoded using binary codes: 0 indicates that the pixel is classified as a non-black and odorous water body, and 1 indicates that the pixel is classified as a black and odorous water body. The threshold h is the decision boundary for black and odorous water body classification, and its optimal value is determined through the validation sample set. By traversing the candidate thresholds with a fixed step size, the h that maximizes the classification accuracy is selected as the optimal threshold. The mathematical expression is as follows:

[0037] ;

[0038] in, Let i be the true label of the i-th validation sample; This represents the predicted label at a given classification threshold h, corresponding to ; This is an indicator function used to count the number of samples whose predicted results match the true labels.

[0039] This application enriches the feature parameters for remote sensing identification of black and odorous water bodies by fusing spectral time-series information, effectively solving the problems of misclassification and omission of black and odorous water bodies caused by seasonal changes and other characteristics, thereby improving the accuracy of remote sensing identification. Attached Figure Description

[0040] Figure 1 is a logic block diagram of this application; Detailed Implementation

[0041] The specific embodiments of the present invention will be described in detail below with reference to Figure 1. It should be noted that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0042] Example 1: This application provides a remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion, including the following steps:

[0043] S1: Acquire remote sensing images of the water body in the target area; the remote sensing images in S1 are provided by the Sentinel-2 satellite system; using Sentinel-2 remote sensing images to replace high spatial resolution images such as high-resolution satellites in the existing technology reduces the cost of remote sensing identification of black and odorous water bodies.

[0044] S2: Preprocess the remote sensing imagery; perform radiometric, atmospheric, and geometric corrections on the Sentinel-2 remote sensing imagery using ENVI 5.3 software;

[0045] S3: Extract water body boundaries based on preprocessed remote sensing images and output a binarized water body region mask;

[0046] Specifically, for the preprocessed remote sensing image, the Normalized Difference Water Index (NDWI) is first calculated, and regions with NDWI ≥ the local water body NDWI threshold are extracted. The specific threshold can be determined according to the local water body NDWI distribution pattern, such as a local water body NDWI threshold of 0.2. Subsequently, through morphological filtering and spatial rule constraints, abnormal erroneous extractions caused by shadows, bare soil, roads, cloud remnants, and low-reflectivity artificial features are removed, and small patches with an area smaller than the effective water body scale are removed. Finally, the water body boundary is extracted, and a binarized water body region mask is output.

[0047] S4: Within the generated water body area mask, generate the black and odorous water body index value (BOI value) of the water body pixel by pixel according to the black and odorous water body index calculation model (BOI calculation model), and integrate the black and odorous water body index value of each pixel as a new band into the water body area image generated by the water body area mask.

[0048] Specifically, based on the ENVI 5.3 platform, within the generated water area mask, the black and odorous water index value (BOI value) of the water body is generated pixel by pixel according to the BOI calculation model, and the result is integrated as a new band into the water area image generated by the water area mask. To ensure the comparability of multi-source band data in subsequent analysis, the Red band, Green band, Blue band, NIR band, and BOI are normalized to unify the dimensions and improve the consistency of spectral features and the robustness of the model.

[0049] S5: Collect black and odorous water body sample points and construct a black and odorous water body sample; the black and odorous water body sample points are divided into training sample points and validation sample points; the training sample points form a training set and the validation sample points form a validation set.

[0050] Specifically, based on the definition of urban black and odorous water bodies in the "Guidelines for the Treatment of Urban Black and Odorous Water Bodies", a total of M black and odorous water body sample points were collected and divided into training set and validation set in a 7:3 ratio: 70%, i.e., K=0.7M, were used as training sample points for model training and initial threshold selection; the remaining 30%, i.e., N=0.3M, were used as independent validation sample points to evaluate classification performance and determine the optimal decision boundary.

[0051] S6: Using a spectral temporal feature extraction algorithm, red, green, blue, near-infrared, and BOI temporal information is extracted from the pixel locations of the training sample points. Then, the temporal information of different bands and BOI temporal information of all training sample points are averaged to form long-term reference spectral information of black and odorous water bodies. ;

[0052] in, Let b be the spectral value / index value of the k-th sample point at time t. Let be the red band spectral value of the k-th sample point at time t; Let be the green band spectral value of the k-th sample point at time t; Let be the blue band spectral value of the k-th sample point at time t; Let be the near-infrared spectral value of the k-th sample point at time t; Let t be the black and odorous water index value corresponding to the k-th sample point at time t; t represents the month, t=1, 2, ... , =12.

[0053] , =12;

[0054] in, This represents the average value of the b-th type spectral value / index value of the k-th sample point over consecutive time intervals at time t; b represents the band or index type, including Red, Green, Blue, NIR, and the Black and Odorous Water Index (BOI); and K represents the total number of training sample points.

[0055] ;

[0056] This represents the average value of the red band spectrum at time t; This represents the average value of the green band spectrum at time t; This represents the average value of the blue band spectrum at time t; This represents the average value of the near-infrared spectral values ​​at time t; This represents the average value of the black and odorous water body index at time t; the above extracts the time series information of all bands month by month; It is an array containing the mean time series information of the red band, the mean time series information of the green band, the mean time series information of the blue band, the mean time series information of the near-infrared band, and the mean time series information of the BOI.

[0057] S7: Using a spectral temporal feature extraction algorithm, the temporal information of the red band, green band, blue band, near-infrared band, and BOI is extracted pixel by pixel from the Sentinel-2 remote sensing image after water body boundary extraction. This forms complete five-dimensional spectral temporal information, namely, pixel-by-pixel spectral temporal information Y of the water body image;

[0058] in, This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ), the value of the b-th type band or index at time t; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower red band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower green band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower blue band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower near-infrared band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time The index value of black and odorous water bodies.

[0059] ;

[0060] in, , =12.

[0061] S8: Calculate the similarity between the long-term reference spectral information of black and odorous water bodies and the pixel-by-pixel spectral temporal information of water body images using the time-weighted dynamic time warping algorithm;

[0062] Specifically: using long-term reference spectral information of black and odorous water bodies Using the pixel-by-pixel spectral temporal information Y of the water body image as the target curve, the spectral temporal similarity distance between the target curve and the reference curve is calculated pixel-by-pixel and band-by-band based on the Time-Weighted Dynamic Time Warping (TWDTW) algorithm. Subsequently, the similarity measurement results of the red band, green band, blue band, near-infrared band, and BOI are weighted and fused. The weights are set according to the contribution of each band to the identification of black and odorous water bodies; here, a uniform value of 0.2 is used to ensure the weight sums equal to 1, thus generating a comprehensive spectral similarity index. This index, as a newly added feature band, is embedded in the water area image to support high-precision identification of black and odorous water bodies.

[0063] ;

[0064] ;

[0065] in, This represents the comprehensive spectral similarity index; The spectral bands and spectral indices used in the calculation include red, green, blue, and near-infrared bands, and the spectral index is the Black and Odorous Water Index (BOI). The first one calculated based on the TWDTW algorithm A measure of spectral similarity across bands. This refers to the time-weighted dynamic time warping algorithm; The weights of each spectral band and spectral index are assigned to reflect their relative importance in the identification of black and odorous water bodies. In this implementation... = ; and These represent the long-term reference spectral information of black and odorous water bodies. And pixel-by-pixel spectral temporal information Y of water body image.

[0066] S9: Based on the similarity between the long-term reference spectral information of black and odorous water bodies and the pixel-by-pixel spectral temporal information of water body images, and combined with the threshold binary classification method, determine whether the water body belongs to black and odorous water bodies;

[0067] Specifically, based on the Comprehensive Spectral Similarity Index (SSI) generated by S8, a univariate decision tree classification model is constructed to determine whether a water body pixel is a black and odorous water body; the classification rule is defined as follows:

[0068] ;

[0069] in, Representing geographic coordinates The classification result of a pixel at a given location is binary encoded: 0 indicates that the pixel is classified as a non-black and odorous water body, and 1 indicates that the pixel is classified as a black and odorous water body. The threshold h is the decision boundary for black and odorous water body classification, and its optimal value is determined through the validation sample set. By using a fixed step size, such as 0.1, the candidate thresholds are traversed, and the h that maximizes the classification accuracy is selected as the optimal threshold. The mathematical expression is as follows:

[0070] ;

[0071] in, Let i be the true label of the i-th validation sample; This represents the predicted label at a given classification threshold h, corresponding to ; This is an indicator function used to count the number of samples whose predicted results match the true labels.

[0072] The computing device required for this invention is a PC computer, and the operating system must be Windows 7 / 10.

[0073] The core of this invention is: based on the fusion of spatiotemporal spectral multidimensional feature information, and through the time-weighted dynamic time warping algorithm and decision tree classification algorithm, accurate identification of black and odorous water bodies can be achieved; the fusion of spatiotemporal spectral multidimensional features can effectively capture the long-term spectral dynamic change information of black and odorous water bodies, reducing the omission of black and odorous water bodies due to seasonal variation characteristics; in addition, the time-weighted dynamic time warping algorithm is an algorithm used to measure the similarity between two time series. By measuring the similarity of time series curves in different bands, it enriches the feature parameters for remote sensing identification of black and odorous water bodies and improves the accuracy of remote sensing identification of black and odorous water bodies.

[0074] The preferred embodiments of this application have been described in detail above with reference to the accompanying drawings. However, this application is not limited to the specific details of the above embodiments. Within the scope of the technical concept of this application, various simple modifications can be made to the technical solution of this application, and these simple modifications all fall within the protection scope of this application.

[0075] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable way without contradiction. In order to avoid unnecessary repetition, the various possible combinations in this application will not be described separately.

[0076] Furthermore, various different implementations of this application can be combined in any way, as long as they do not violate the spirit of this application, and such combinations should also be regarded as the content disclosed in this application.

Claims

1. A remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion, characterized in that, The process includes the following steps: S1: Acquire remote sensing images of the target area's water body; S2: Preprocess the remote sensing images; S3: Extract the water body boundaries based on the preprocessed remote sensing images and output a binarized water body region mask. S4: Within the generated water body area mask, generate the black and odorous water body index value pixel by pixel according to the black and odorous water body index calculation model, and integrate the black and odorous water body index value of each pixel as a new band into the water body area image generated by the water body area mask; S5: Collect black and odorous water body sample points to construct black and odorous water body samples; the black and odorous water body sample points are divided into training sample points and validation sample points; S6: Using the spectral time series feature extraction algorithm, the red band time series information, green band time series information, blue band time series information, near-infrared band time series information and BOI time series information of the pixel location of the training sample points are extracted respectively. Then, the time series information of different bands and BOI time series information of all training sample points are averaged to form long-term reference spectral information of black and odorous water bodies. S7: Using a spectral temporal feature extraction algorithm, extract the red band temporal information, green band temporal information, blue band temporal information, near-infrared band temporal information, and BOI temporal information pixel by pixel from the remote sensing image after water body boundary extraction: to form the pixel-by-pixel spectral temporal information of the water body image. S8: Calculate the similarity between the long-term reference spectral information of black and odorous water bodies and the pixel-by-pixel spectral temporal information of water body images using the time-weighted dynamic time warping algorithm; S9: Based on the similarity between the long-term reference spectral information of the black and odorous water body and the pixel-by-pixel spectral temporal information of the water body image, and combined with the threshold binary classification method, determine whether the water body belongs to the black and odorous water body.

2. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 1, characterized in that, In step S3, the processed remote sensing image is first processed by calculating the Normalized Difference Water Index (NDWI), and regions with NDWI ≥ the local water body NDWI threshold are extracted. Then, through morphological filtering and spatial rule constraints, abnormal erroneous extractions are eliminated, and small patches with an area smaller than the effective water body scale are removed. Finally, the water body boundary is extracted, and a binary water body region mask is output.

3. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 1, characterized in that, The calculation formula for extracting red band time-series information, green band time-series information, blue band time-series information, near-infrared band time-series information, and BOI time-series information at the pixel location of the training sample point is as follows: in, Let b be the spectral value / index value of the k-th sample point at time t. Let be the red band spectral value of the k-th sample point at time t; Let be the green band spectral value of the k-th sample point at time t; Let be the blue band spectral value of the k-th sample point at time t; Let be the near-infrared spectral value of the k-th sample point at time t; Let t be the black and odorous water index value corresponding to the k-th sample point at time t; t represents the month, t=1, 2, ... ， =12。 4. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 3, characterized in that, The formula for averaging the time series information of different bands and the BOI time series information of all training sample points is as follows: ， =12; where, This represents the average value of the b-th type spectral value / index value of the k-th sample point over consecutive time intervals at time t; b represents the band or index type, including Red, Green, Blue, NIR, and the Black and Odorous Water Index (BOI); K represents the total number of training sample points.

5. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 4, characterized in that, The long-term reference spectral information of the black and odorous water body The formula is: ； This represents the average value of the red band spectrum at time t; This represents the average value of the green band spectrum at time t; This represents the average value of the blue band spectrum at time t; This represents the average value of the near-infrared spectral values ​​at time t; This represents the average value of the black and odorous water body index at time t.

6. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 1, characterized in that, The calculation formula for extracting red band time-series information, green band time-series information, blue band time-series information, near-infrared band time-series information, and BOI time-series information pixel by pixel from the remote sensing image after water body boundary extraction using the spectral time-series feature extraction algorithm is as follows: in, This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ), the value of the b-th type band or index at time t; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower red band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower green band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Lower blue band spectral values; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time Spectral values ​​in the lower near-infrared band; This indicates the latitude and longitude of the pixel center within the masked area of ​​the water body. ),time The index value of black and odorous water bodies.

7. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 6, characterized in that, The formula for calculating the pixel-by-pixel spectral temporal information Y of water body image is: ;in, ， =12。 8. The remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 1, characterized in that, Specifically, S8 refers to using long-term reference spectral information of black and odorous water bodies. Using the pixel-by-pixel spectral temporal information Y of the water body image as the target curve, the spectral temporal similarity distance between the target curve and the reference curve is calculated pixel by pixel and band by band based on the time-weighted dynamic time warping algorithm. Then, the similarity measurement results of the red band, green band, blue band, near-infrared wave and BOI are weighted and fused to generate a comprehensive spectral similarity index.

9. A remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion as described in claim 8, characterized in that, Comprehensive spectral similarity index The calculation formula is: ； in, This represents the comprehensive spectral similarity index; This represents the spectral bands and spectral indices involved in the calculation. The spectral bands include the red band, green band, blue band, and near-infrared band, and the spectral indices include the BOI. The first one calculated based on the TWDTW algorithm A measure of spectral similarity across bands; The weights for each spectral band and spectral index; and These represent the long-term reference spectral information of black and odorous water bodies. The TWDTW algorithm is a time-weighted dynamic time warping algorithm that integrates pixel-by-pixel spectral temporal information of water bodies and water images.

10. A remote sensing identification method for black and odorous water bodies based on spatiotemporal spectral multidimensional feature information fusion according to claim 9, characterized in that, Specifically, S9 involves, based on the generated comprehensive spectral similarity index... A univariate decision tree classification model was constructed to determine whether water body pixels were black and odorous. The classification rules of the model were defined as follows: in, Representing geographic coordinates The classification result of a pixel at a given location is encoded using binary codes: 0 indicates that the pixel is classified as a non-black and odorous water body, and 1 indicates that the pixel is classified as a black and odorous water body. The threshold h is the decision boundary for black and odorous water body classification, and its optimal value is determined through the validation sample set. By traversing the candidate thresholds with a fixed step size, the h that maximizes the classification accuracy is selected as the optimal threshold. The mathematical expression is as follows: ； in, Let i be the true label of the i-th validation sample; This represents the predicted label at a given classification threshold h, corresponding to ； This is an indicator function.

Images