A method and device for identifying high-risk areas of red tide based on a water color satellite

Abstract

The application discloses a red tide high-risk area distinguishing method and device based on a water color satellite, and relates to the field of satellite remote sensing image processing.The method is as follows: S1, acquiring satellite remote sensing images and performing pretreatment; S2, using a plurality of red tide distinguishing algorithms to distinguish red tide to obtain algorithm distinguishing results; S3, obtaining a red tide certainty index RTCI and an algorithm distinguishing result matrix corresponding to each value of the RTCI according to the algorithm distinguishing results; S4, acquiring red tide measured data, comparing the measured data with the algorithm distinguishing result matrix to obtain a confusion matrix corresponding to each value of the RTCI; S5, evaluating each confusion matrix by using an F-measure algorithm to obtain FM values corresponding to each value of the RTCI, and taking the RTCI value with the highest FM value as a judgment threshold of a red tide high-risk area; and S6, determining a range formed by pixels with an RTCI value greater than or equal to the judgment threshold as the red tide high-risk area.The application constructs a red tide certainty index based on a plurality of red tide distinguishing algorithms, and reduces the uncertainty of a single algorithm in red tide monitoring.

Description

Technical Field

[0001] This invention relates to the field of satellite remote sensing image processing, and in particular to a method and apparatus for identifying high-risk areas for red tide occurrence based on water color satellites. Background Technology

[0002] Red tides have become one of the major marine disasters in my country. Utilizing water color remote sensing technology for effective, real-time, and accurate red tide monitoring is crucial for scientifically understanding red tides, responding to red tide disasters, and avoiding their various negative impacts. When a red tide occurs, the red tide water body exhibits significant changes in reflectance spectrum compared to the surrounding non-red tide water body. Therefore, algorithms based on the spectral differences of the red tide water body can distinguish it from non-red tide water bodies. However, due to the complex bio-optical characteristics of nearshore waters and the influence of the algorithms themselves, the location and range identified by each red tide algorithm vary to some extent. These differences can cause some confusion in operational applications, as users are unsure which algorithm provides more accurate and reliable results. Summary of the Invention

[0003] To address the above problems, this invention proposes a method and device for identifying high-risk areas of red tides based on ocean color satellites. After downloading and preprocessing ocean color data, multiple red tide identification algorithms are constructed using remote sensing reflectance. Then, a red tide certainty index is built based on these algorithms. The value of the red tide certainty index is evaluated using measured red tide data. When the red tide certainty index is greater than a certain threshold, it indicates that the identified location has the highest certainty of experiencing a red tide and is a high-risk area for red tide occurrence. This reduces the uncertainty of using a single algorithm in red tide monitoring.

[0004] On the one hand, a method for identifying high-risk areas for red tides based on ocean color satellites, with the following specific steps:

[0005] S1. Acquire satellite remote sensing images of multiple time phases during the red tide, preprocess the acquired satellite remote sensing images, and obtain the remote sensing reflectance of each band of the satellite remote sensing images.

[0006] S2, Based on the remote sensing reflectance, several red tide discrimination algorithms are used to distinguish red tides, and the algorithm discrimination results of whether each pixel of the satellite remote sensing image is a red tide point are obtained respectively;

[0007] S3. Obtain the red tide determinism index based on the discrimination results of each algorithm, and construct the algorithm discrimination result matrix corresponding to the red tide determinism index based on the algorithm discrimination results of all pixels in the satellite remote sensing image.

[0008] S4. Obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix.

[0009] The comparison results include: both the algorithm's determination result and the measured red tide data indicate a red tide point; the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point; the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point; and both the algorithm's determination result and the measured red tide data indicate a non-red tide point. The total number of cases where both the algorithm's determination result and the measured red tide data indicate a red tide point is TP; the total number of cases where the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point is FP; the total number of cases where the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point is FN; and the total number of cases where both the algorithm's determination result and the measured red tide data indicate a non-red tide point is TN. The TP, FP, FN, and TN corresponding to a red tide determinism index value constitute the confusion matrix corresponding to that value.

[0010] S5. The confusion matrix is ​​evaluated based on the F-measure algorithm to obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index. The red tide certainty index with the highest FM value is used as the judgment threshold for high-risk areas of red tide occurrence.

[0011] S6 defines the area consisting of all pixels whose red tide certainty index values ​​are greater than or equal to the judgment threshold as a high-risk area for red tide.

[0012] Preferably, in S1, the preprocessing is projection transformation and Mosaic stitching.

[0013] Preferably, in S2, the plurality of red tide discrimination algorithms include the red tide index method, the red tide detection index method, the improved fluorescence height method, and the normalized differential chlorophyll index method;

[0014] The results of the algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point using the red tide index method are as follows:

[0015]

[0016] Where RI represents the red tide index calculated for the current pixel; R rs (560) represents the remote sensing reflectance of the current pixel in the 560nm band; R rs (443) represents the remote sensing reflectance of the current pixel in the 443nm band; R rs (490) represents the remote sensing reflectance of the current pixel in the 490nm band;

[0017] RI>4.0 was used as the criterion for judging the occurrence of red tide in each pixel;

[0018] The results of the algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point using the red tide detection index method are as follows:

[0019]

[0020] Where RDI represents the calculated value of the red tide detection index for the current pixel; R rs (λ1) represents the remote sensing reflectance of the current pixel in the λ1 band, where λ1 = 665 nm; R rs (λ2) represents the remote sensing reflectance of the current pixel in the λ2 band, where λ2 = 560 nm; R rs (λ3) represents the remote sensing reflectance of the current pixel in the λ3 band, where λ3 = 753 nm;

[0021] RDI>0.16 was used as the criterion for judging the occurrence of red tide in each pixel;

[0022] The results of the improved fluorescence height method algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point are as follows:

[0023]

[0024] Where MFLH represents the calculated value of the current pixel using the improved fluorescence height method; nL w (681) represents the water radiance of the current pixel in the 681nm band; nL w (665) represents the water-free radiance of the current pixel in the 665nm band; nL w (754) represents the water-free radiance of the current pixel in the 754nm band; the water-free radiance nL in the λ band. W (λ)=R rs (λ)×F0,R rs (λ) represents the remote sensing reflectance of the λ band, and F0 is the solar irradiance outside the atmosphere;

[0025] Using MFLH>0.4mW cm -2 μm -1 sr -1 And b bp / MFLH<0.2mW cm -2 μm -1 sr -1 As the criteria for judging the occurrence of red tides in each pixel, b bp The backscattering coefficient;

[0026] The algorithm results for determining whether each pixel in a satellite remote sensing image is a red tide point using the normalized differential chlorophyll index method are as follows:

[0027]

[0028] Wherein, NDCI represents the calculated value of the current pixel using the normalized differential chlorophyll index method; R rs (708) represents the remote sensing reflectance of the current pixel in the 708nm band; R rs(665) represents the remote sensing reflectance of the current pixel in the 665nm band;

[0029] NDCI>0 is used as the criterion for judging the occurrence of red tide in each pixel.

[0030] Preferably, in S3, the process of obtaining the red tide determinism index based on the results of each algorithm is as follows:

[0031] When a pixel in a satellite remote sensing image is identified as a red tide by one of the following methods: Red Tide Index, Red Tide Detection Index, Improved Fluorescence Height Method, and Normalized Difference Chlorophyll Index Method, the Red Tide Deterministic Index (RTCI) is set to 1. If a pixel in the image is identified as a red tide by two of the four red tide remote sensing algorithms, the RTCI is set to 2. If a pixel in the image is identified as a red tide by three of the four red tide remote sensing algorithms, the RTCI is set to 2 or 3. If a pixel in the image is identified as a red tide by four of the four red tide remote sensing algorithms, the RTCI is set to 2, 3, or 4.

[0032] Preferably, in S5, the comprehensive evaluation index FM is expressed as follows:

[0033]

[0034] On the other hand, a red tide high-risk area identification device based on ocean color satellites includes the following:

[0035] The satellite remote sensing image acquisition and preprocessing module is used to acquire satellite remote sensing images of multiple time phases during the red tide, and to preprocess the acquired satellite remote sensing images to obtain the remote sensing reflectance of each band of the satellite remote sensing images.

[0036] The red tide discrimination algorithm module is used to perform red tide discrimination using several red tide discrimination algorithms based on the remote sensing reflectance, and to obtain the algorithm discrimination results of whether each pixel of the satellite remote sensing image is a red tide point;

[0037] The red tide deterministic index module is used to obtain the red tide deterministic index based on the discrimination results of each algorithm. It constructs the algorithm discrimination result matrix corresponding to the red tide deterministic index based on the algorithm discrimination results of all pixels in the satellite remote sensing image.

[0038] The measured data comparison module is used to obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix.

[0039] The comparison results include: both the algorithm's determination result and the measured red tide data indicate a red tide point; the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point; the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point; and both the algorithm's determination result and the measured red tide data indicate a non-red tide point. The total number of cases where both the algorithm's determination result and the measured red tide data indicate a red tide point is TP; the total number of cases where the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point is FP; the total number of cases where the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point is FN; and the total number of cases where both the algorithm's determination result and the measured red tide data indicate a non-red tide point is TN. The TP, FP, FN, and TN corresponding to a red tide determinism index value constitute the confusion matrix corresponding to that value.

[0040] The algorithm discrimination result evaluation module is used to evaluate the confusion matrix based on the F-measure algorithm, obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index, and take the red tide certainty index value with the highest FM value as the judgment threshold for high-risk areas of red tide occurrence.

[0041] The red tide high-risk area determination module is used to determine the area consisting of all pixels whose red tide certainty index value is greater than or equal to the judgment threshold as a red tide high-risk area.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] (1) This invention utilizes a variety of red tide discrimination algorithms to generate a red tide deterministic index. This index can provide a relatively accurate location and range of red tide occurrence in operational applications. After evaluation by on-site measured data, the threshold of the deterministic index is selected, and the location of high-risk red tide occurrence areas is further clarified, thereby providing a reference for red tide emergency decision-making and management.

[0044] (2) By constructing multiple red tide discrimination algorithms and constructing a red tide deterministic index based on multiple red tide discrimination algorithms, this invention can reduce the uncertainty of a single algorithm in red tide monitoring and provide a more stable and reliable red tide monitoring service.

[0045] (3) This invention constructs an operational and automated identification process for high-risk areas of red tides, which can improve the accuracy and efficiency of identifying high-risk areas of red tides, thereby reducing the harm of red tides. Attached Figure Description

[0046] The present invention will now be described in further detail with reference to the accompanying drawings;

[0047] Figure 1 This is a flowchart of a method for identifying high-risk areas of red tide occurrence based on ocean color satellites, according to an embodiment of the present invention.

[0048] Figure 2This is a technical process for identifying high-risk areas of red tide based on a water color satellite, according to an embodiment of the present invention.

[0049] Figure 3 This is a spatial distribution map of the red tide determinism index at different time phases, representing an embodiment of the present invention for a method of identifying high-risk red tide areas based on a water color satellite; (a) shows the spatial distribution map of the red tide determinism index on May 21, 2019; (b) shows the spatial distribution map of the red tide determinism index on May 23, 2019; (c) shows the spatial distribution map of the red tide determinism index on April 29, 2020; (d) shows the spatial distribution map of the red tide determinism index on August 14, 2020; (f) shows the spatial distribution map of the red tide determinism index on August 15, 2020; each pixel only displays the highest red tide determinism index value;

[0050] Figure 4 This is a block diagram of a red tide high-risk area identification device based on a water color satellite according to an embodiment of the present invention. Detailed Implementation

[0051] The present invention will be further described below through specific embodiments.

[0052] like Figure 1 and Figure 2 As shown, a method for identifying high-risk areas for red tide occurrence based on ocean color satellites is described, with the following specific steps.

[0053] S1: Acquire satellite remote sensing images of multiple time phases during the red tide. Preprocess the acquired satellite remote sensing images to obtain the remote sensing reflectance of each band of the satellite remote sensing images.

[0054] During the peak red tide season (March to June each year), download the information products retrieved from the OLCI (Ocean and Land Colour Instrument) water color sensor carried by the Sentinel-3 ocean color satellite. The dataset format is OLCILevel 2Ocean Colour Full Resolution in NTC-Sentinel-3. The spatial resolution of the OLCI product can reach 300×300m. Preprocess the downloaded L2 files using the Sentinel Application Platform (SNAP) software, specifically by batch performing projection transformation and Mosaic stitching using the GPT command line. The projection method used is Equidistant Cylindrical, thereby obtaining the remote sensing reflectance of each band required for the red tide discrimination algorithm.

[0055] S2, Based on the remote sensing reflectance, several red tide discrimination algorithms are used to distinguish red tides, and the algorithm discrimination results of whether each pixel in the satellite remote sensing image is a red tide point are obtained respectively.

[0056] Specifically, the red tide discrimination algorithms used include the red tide index method, the red tide detection index method, the improved fluorescence height method, and the normalized differential chlorophyll index method. The expressions for each red tide discrimination algorithm are shown below:

[0057] The red tide index method utilizes water color remote sensing data, such as the remote sensing reflectance (R0) in the 443nm, 490nm, and 560nm bands of the Sentinel-3 OLCI sensor. rs It is constructed by subtracting R from both the blue and green bands in the RI calculation formula. rs (443) is to eliminate the influence of turbid water. The red tide index method is expressed as follows:

[0058]

[0059] Where RI represents the red tide index calculated for the current pixel; R rs (560) represents the remote sensing reflectance of the current pixel in the 560nm band; R rs (443) represents the remote sensing reflectance of the current pixel in the 443nm band; R rs (490) represents the remote sensing reflectance of the current pixel in the 490nm band.

[0060] RI>4.0 was used as the criterion for judging the occurrence of red tide in each pixel.

[0061] The expression for the Red Tide Detection Index is as follows:

[0062]

[0063] Where RDI represents the calculated value of the red tide detection index for the current pixel; R rs (λ1) represents the remote sensing reflectance of the current pixel in the λ1 band, where λ1 = 665 nm; R rs (λ2) represents the remote sensing reflectance of the current pixel in the λ2 band, where λ2 = 560 nm; R rs (λ3) represents the remote sensing reflectance of the current pixel in the λ3 band, where λ3 = 753 nm.

[0064] The modified fluorescence line height method is a spectral difference method that simultaneously uses the water-leaving radiance (nL) in three wavelength bands: 665 nm, 754 nm, and 681 nm. wIts expression is shown in the following formula:

[0065]

[0066] Where MFLH represents the calculated value of the current pixel using the improved fluorescence height method; nL w (681) represents the water radiance of the current pixel in the 681nm band; nL w (665) represents the water-free radiance of the current pixel in the 665nm band; nL w (754) represents the water-free radiance of the current pixel in the 754nm band; the water-free radiance nL in the λ band. W (λ)=R rs (λ)×F0,R rs (λ) represents the remote sensing reflectance of the λ band, and F0 is the solar irradiance outside the atmosphere.

[0067] After optimization through experiments using field observation data, the optimal MFLH value was determined to be >0.4 mW cm⁻¹. -2 μm -1 sr -1 And b bp / MFLH<0.2mW cm -2 μm -1 sr -1 As a criterion for judging the occurrence of red tides in each pixel, b bp This is the backscattering coefficient.

[0068] The Normalized Difference Chlorophyll Index (NDCI) is shown in the following formula:

[0069]

[0070] Wherein, NDCI represents the calculated value of the current pixel using the normalized differential chlorophyll index method; R rs (708) represents the remote sensing reflectance of the current pixel in the 708nm band; R rs (665) represents the remote sensing reflectance of the current pixel in the 665nm band.

[0071] NDCI>0 is used as the criterion for judging the occurrence of red tide in each pixel.

[0072] S3. Obtain the red tide determinism index based on the discrimination results of each algorithm, and construct the algorithm discrimination result matrix corresponding to the red tide determinism index based on the algorithm discrimination results of all pixels in the satellite remote sensing image.

[0073] Based on the four red tide discrimination algorithms mentioned above, a comprehensive discrimination index, namely the Red Tide Certainty Index (RTCI), is proposed. The index is defined as follows: if a pixel in a satellite remote sensing image is identified as a red tide by one of the following algorithms: Red Tide Index Method, Red Tide Detection Index Method, Improved Fluorescence Height Method, and Normalized Difference Chlorophyll Index Method, then the RTCI is set to 1; if a pixel in the image is identified as a red tide by two of the four algorithms, then the RTCI is set to 2; if a pixel in the image is identified as a red tide by three of the four algorithms, then the RTCI is set to 2 or 3; if a pixel in the image is identified as a red tide by four of the four algorithms, then the RTCI is set to 2, 3, or 4.

[0074] S4. Obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix.

[0075] Specifically, the procedure for obtaining actual red tide observation data is as follows:

[0076] Red tide events are collected from the China Marine Disaster Bulletin or red tide information released by various localities, and information such as the occurrence time, location, and affected area of ​​each red tide event is extracted. However, the red tide locations recorded in the Marine Disaster Bulletin generally do not have accurate latitude and longitude coordinates, and the area provided is usually statistical information for the entire red tide event period. Using only this information to evaluate the true value of the algorithm's effectiveness will obviously have a certain bias. This embodiment uses the red tide information provided by the Marine Disaster Bulletin, combined with remote sensing FRGB images (pseudo-color images), to jointly determine and select red tide observation data for evaluating the algorithm's effectiveness.

[0077] The specific procedure is as follows: First, select corresponding remote sensing images based on the date and location recorded in the bulletin, prioritizing cloud-free or minimally cloud-affected images. Then, use the three bands of OLCI: R (865nm), G (560nm), and B (443nm) to generate a false color (False RGB) image. Through observation and comparison, it was found that points in the FRGB image showing red or reddish-brown hues with significant differences from surrounding pixels corresponded well to the red tide locations reported in the bulletin. Therefore, using SNPP software, by drawing polygonal ROI (Region of Interest) areas, select corresponding points within the red or reddish-brown abnormal areas of the FRGB image as the true red tide values, and select points around these areas where the water color does not show abnormalities as the non-red tide true values.

[0078] Specifically, the following method is used to compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, and then classify and count the comparison results to obtain the corresponding confusion matrix:

[0079] The true value of red tide, i.e., the actual observed red tide, is denoted as RT; the true value of non-red tide point, i.e., the actual observed non-red tide, is denoted as NRT; the results of each red tide discrimination algorithm are also divided into two types: algorithm-discriminated red tide (denoted as rt) or algorithm-discriminated non-red tide (denoted as nrt). The resulting binary classification confusion matrix is ​​shown in the table below.

[0080] Table 1. Confusion matrix of red tide discrimination results

[0081]

[0082] As shown in Table 1, comparing the algorithm's determination result with the actual observation value, four situations will occur: 1) Both the algorithm's determination and the actual observation value indicate a red tide point (rt-RT); 2) The algorithm's determination result indicates a red tide point, but the actual red tide did not occur (rt-NRT); 3) The algorithm's determination result indicates a non-red tide point, while the actual observation value indicates a red tide point (nrt-RT); 4) Both the algorithm's determination result and the actual observation value indicate a non-red tide point (nrt-NRT).

[0083] The total number of all comparison cases in the comparison results is summed to obtain the following: TP is the total number of cases where both the algorithm's judgment result and the red tide measured data are red tide points; FP is the total number of cases where the algorithm's judgment result is a red tide point but the red tide measured data is not a red tide point; FN is the total number of cases where the algorithm's judgment result is not a red tide point but the red tide measured data is a red tide point; and TN is the total number of cases where both the algorithm's judgment result and the red tide measured data are not red tide points.

[0084] S5. The confusion matrix is ​​evaluated based on the F-measure algorithm to obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index. The red tide certainty index with the highest FM value is used as the judgment threshold for high-risk areas of red tide occurrence.

[0085] Specifically, the method for evaluating the confusion matrix of each value of the red tide determinism index based on the F-measure algorithm is as follows:

[0086] Sensitivity represents the proportion of red tide detection algorithms that correctly identify red tide points. It measures the algorithm's ability to identify actual red tide seawater. The sensitivity formula is shown below:

[0087]

[0088] Sensitivity represents the calculated sensitivity value.

[0089] Specificity represents the proportion of times the red tide detection algorithm correctly identifies non-red tide points. It measures the algorithm's ability to identify seawater where red tides have not occurred. The specificity formula is shown below:

[0090]

[0091] Specificity represents the specific calculated value.

[0092] Precision represents the proportion of red tide points out of the total number of points correctly distinguished as red tide points and non-red tide points. The precision formula is as follows:

[0093]

[0094] Specificity represents the calculated precision value.

[0095] The comprehensive evaluation index is the harmonic average of sensitivity and accuracy, and its calculation formula is as follows:

[0096]

[0097] Among them, FM (i.e. F-measure) represents the comprehensive evaluation index, and the closer the FM value is to 1, the better the algorithm performance.

[0098] The false negative rate is the proportion of red tide points that the algorithm incorrectly identifies as non-red tide points; the false positive rate is the proportion of non-red tide points that the algorithm incorrectly identifies as red tide points. The formulas for the false negative rate and false positive rate are shown below:

[0099]

[0100] Wherein, False neg. (%) represents the percentage of missed detection rate; False pos. (%) represents the percentage of false alarm rate.

[0101] This embodiment takes the East China Sea as an example, selecting OLCI remote sensing images from five time phases with relatively low cloud cover to evaluate different RTCI values. These five remote sensing images record a red tide of Prorocentrum dinoflagellates / Noctiluca scintillans in the waters off Xiangshan-Yushan, Ningbo on May 21 and May 23, 2019; a red tide of Prorocentrum dinoflagellates in the waters off Nanji-Shipu-Yushan, Wenzhou on April 29, 2020; and a red tide of diatoms of Rhizophora filamentosa in the waters east of Zhujiajian, Zhoushan on August 14 and August 15, 2020.

[0102] Figure 3Table 2 shows the spatial distribution of the red tide determinism index (RTCI) for the five time phases mentioned above, and presents the F-measure algorithm evaluation results for different RTCI values. Overall, when the RTCI is 2, the FM value is the highest (0.94), with relatively low false alarm and false alarm rates. However, as the RTCI value increases, the FM value does not increase but rather decreases. This is because increasing the RTCI value places higher demands on algorithm consistency; only when the consistency of the four algorithms is relatively good will the FM value increase with the RTCI value. When the RTCI is 1, using a single algorithm may overestimate the impact range of red tides in complex nearshore waters, resulting in a lower FM value. Therefore, a reasonable RTCI value (as shown in Table 2, the judgment threshold in this embodiment is RTCI = 2) can effectively reduce false alarms, thereby improving the accuracy of red tide detection (FM value).

[0103] Table 2. Statistical results of F-measure for different RTCI values.

[0104]

[0105] Here, Sen. is an abbreviation for Sensibility, and Spe. is an abbreviation for Specificity.

[0106] S6 defines the area consisting of all pixels whose red tide certainty index values ​​are greater than or equal to the judgment threshold as a high-risk area for red tide.

[0107] In this embodiment, after evaluation, it was found that when RTCI≥2, the uncertainty of a single algorithm in red tide monitoring can be reduced to a certain extent. The certainty of red tide occurrence is highest in these pixels, and they can be identified as high-risk areas for red tide occurrence.

[0108] This embodiment comprehensively utilizes multiple red tide discrimination algorithms to generate a red tide certainty index. This index can provide relatively accurate red tide occurrence location and range in operational applications. After evaluation based on field measurement data, the threshold of the red tide certainty index is selected, and the location of high-risk red tide occurrence areas is further clarified, thus providing a reference for red tide emergency decision-making and management. This embodiment also constructs an operational and automated discrimination process for high-risk red tide occurrence areas, which can improve the accuracy and efficiency of red tide high-risk area discrimination, thereby reducing the harm of red tides.

[0109] like Figure 4 As shown, this invention also discloses a red tide high-risk area identification device based on ocean color satellites, comprising:

[0110] The satellite remote sensing image acquisition and preprocessing module 401 is used to acquire satellite remote sensing images of multiple time phases during the red tide, preprocess the acquired satellite remote sensing images, and obtain the remote sensing reflectance of each band of the satellite remote sensing images.

[0111] The red tide discrimination algorithm module 402 is used to perform red tide discrimination using several red tide discrimination algorithms based on the remote sensing reflectance, and obtain the algorithm discrimination results of whether each pixel of the satellite remote sensing image is a red tide point.

[0112] The red tide determinism index module 403 is used to obtain the red tide determinism index based on the discrimination results of each algorithm. It constructs the algorithm discrimination result matrix corresponding to the red tide determinism index based on the algorithm discrimination results of all pixels in the satellite remote sensing image.

[0113] The measured data comparison module 404 is used to obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix.

[0114] The comparison results include: both the algorithm's determination result and the measured red tide data indicate a red tide point; the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point; the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point; and both the algorithm's determination result and the measured red tide data indicate a non-red tide point. The total number of cases where both the algorithm's determination result and the measured red tide data indicate a red tide point is TP; the total number of cases where the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point is FP; the total number of cases where the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point is FN; and the total number of cases where both the algorithm's determination result and the measured red tide data indicate a non-red tide point is TN. The TP, FP, FN, and TN corresponding to a red tide determinism index value constitute the confusion matrix corresponding to that value.

[0115] The algorithm discrimination result evaluation module 405 is used to evaluate the confusion matrix based on the F-measure algorithm, obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index, and take the red tide certainty index value with the highest FM value as the judgment threshold for high-risk areas of red tide occurrence.

[0116] The red tide high-risk area determination module 406 is used to determine the range of all pixels whose red tide certainty index value is greater than or equal to the judgment threshold as a red tide high-risk area.

[0117] A specific implementation of a red tide high-risk area identification device based on water color satellite is described in the same way as the red tide high-risk area identification method based on water color satellite. This embodiment will not repeat the description.

[0118] The above are merely specific embodiments of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention using this concept shall be considered as infringing upon the protection scope of the present invention.

Claims

1. A method for identifying high-risk areas for red tide occurrence based on ocean color satellites, characterized in that, The specific steps are as follows: S1. Acquire satellite remote sensing images of multiple time phases during the red tide, preprocess the acquired satellite remote sensing images, and obtain the remote sensing reflectance of each band of the satellite remote sensing images. S2, Based on the remote sensing reflectance, several red tide discrimination algorithms are used to distinguish red tides, and the algorithm discrimination results of whether each pixel of the satellite remote sensing image is a red tide point are obtained respectively; S3. Obtain the red tide determinism index based on the discrimination results of each algorithm, and construct the algorithm discrimination result matrix corresponding to the red tide determinism index based on the algorithm discrimination results of all pixels in the satellite remote sensing image. S4. Obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix. The comparison results include: both the algorithm's determination result and the measured red tide data indicate a red tide point; the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point; the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point; and both the algorithm's determination result and the measured red tide data indicate a non-red tide point. The total number of cases where both the algorithm's determination result and the measured red tide data indicate a red tide point is TP; the total number of cases where the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point is FP; the total number of cases where the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point is FN; and the total number of cases where both the algorithm's determination result and the measured red tide data indicate a non-red tide point is TN. The TP, FP, FN, and TN corresponding to a red tide determinism index value constitute the confusion matrix corresponding to that value. S5. The confusion matrix is ​​evaluated based on the F-measure algorithm to obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index. The red tide certainty index with the highest FM value is used as the judgment threshold for high-risk areas of red tide occurrence. S6 defines the area consisting of all pixels whose red tide certainty index values ​​are greater than or equal to the judgment threshold as a high-risk area for red tide.

2. The method for identifying high-risk areas of red tide occurrence based on ocean color satellites according to claim 1, characterized in that, In S1, the preprocessing is projection transformation and Mosaic stitching.

3. The method for identifying high-risk areas of red tide occurrence based on ocean color satellites according to claim 1, characterized in that, In S2, the various red tide discrimination algorithms include the red tide index method, the red tide detection index method, the improved fluorescence height method, and the normalized differential chlorophyll index method. The results of the algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point using the red tide index method are as follows: Where RI represents the red tide index calculated for the current pixel; R rs (560) represents the remote sensing reflectance of the current pixel in the 560nm band; R rs (443) represents the remote sensing reflectance of the current pixel in the 443nm band; R rs (490) represents the remote sensing reflectance of the current pixel in the 490nm band; RI>4.0 was used as the criterion for judging the occurrence of red tide in each pixel; The results of the algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point using the red tide detection index method are as follows: Where RDI represents the calculated value of the red tide detection index for the current pixel; R rs (λ1) represents the remote sensing reflectance of the current pixel in the λ1 band, where λ1 = 665 nm; R rs (λ2) represents the remote sensing reflectance of the current pixel in the λ2 band, where λ2 = 560 nm; R rs (λ3) represents the remote sensing reflectance of the current pixel in the λ3 band, where λ3 = 753 nm; RDI>0.16 was used as the criterion for judging the occurrence of red tide in each pixel; The results of the improved fluorescence height method algorithm for determining whether each pixel in a satellite remote sensing image is a red tide point are as follows: Where MFLH represents the calculated value of the current pixel using the improved fluorescence height method; nL w (681) represents the water radiance of the current pixel in the 681nm band; nL w (665) represents the water-free radiance of the current pixel in the 665nm band; nL w (754) represents the water-free radiance of the current pixel in the 754nm band; the water-free radiance nL in the λ band. W (λ)=R rs (λ)×F0,R rs (λ) represents the remote sensing reflectance of the λ band, and F0 is the solar irradiance outside the atmosphere; Using MFLH>0.4mW cm -2 μm -1 sr -1 And b bp / MFLH<0.2mW cm -2 μm -1 sr -1 As the criteria for judging the occurrence of red tides in each pixel, b bp The backscattering coefficient; The algorithm results for determining whether each pixel in a satellite remote sensing image is a red tide point using the normalized differential chlorophyll index method are as follows: Wherein, NDCI represents the calculated value of the current pixel using the normalized differential chlorophyll index method; R rs (708) represents the remote sensing reflectance of the current pixel in the 708nm band; R rs (665) represents the remote sensing reflectance of the current pixel in the 665nm band; NDCI>0 is used as the criterion for judging the occurrence of red tide in each pixel.

4. The method for identifying high-risk areas of red tide occurrence based on ocean color satellites according to claim 3, characterized in that, In S3, the red tide determinism index is obtained based on the judgment results of each algorithm, as follows: If a pixel in a satellite remote sensing image is identified as a red tide by one of the following methods: Red Tide Index, Red Tide Detection Index, Improved Fluorescence Height Method, and Normalized Difference Chlorophyll Index Method, then the Red Tide Deterministic Index (RTCI) is set to 1. If a pixel in the image is identified as a red tide by two of the four red tide remote sensing algorithms, then the RTCI is set to 2. If a pixel in the image is identified as a red tide by three of the four red tide remote sensing algorithms, then the RTCI is set to 2 or 3. If a pixel in the image is identified as a red tide by four of the four red tide remote sensing algorithms, then the RTCI is set to 2, 3, or 4.

5. The method for identifying high-risk areas of red tide occurrence based on ocean color satellites according to claim 4, characterized in that, In S5, the comprehensive evaluation index FM is represented as follows:

6. A device for identifying high-risk areas for red tide occurrence based on ocean color satellites, comprising the following: The satellite remote sensing image acquisition and preprocessing module is used to acquire satellite remote sensing images of multiple time phases during the red tide, and to preprocess the acquired satellite remote sensing images to obtain the remote sensing reflectance of each band of the satellite remote sensing images. The red tide discrimination algorithm module is used to perform red tide discrimination using several red tide discrimination algorithms based on the remote sensing reflectance, and obtain the algorithm discrimination results of whether each pixel of the satellite remote sensing image is a red tide point; The red tide determinism index module is used to obtain the red tide determinism index based on the judgment results of each algorithm. It constructs the algorithm judgment result matrix corresponding to the red tide determinism index based on the algorithm judgment results of all pixels in the satellite remote sensing image. The measured data comparison module is used to obtain measured red tide data in the same area, compare the measured red tide data with the corresponding pixel values ​​in the discrimination result matrix of each algorithm, classify and count the comparison results, and obtain the corresponding confusion matrix. The comparison results include: both the algorithm's determination result and the measured red tide data indicate a red tide point; the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point; the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point; and both the algorithm's determination result and the measured red tide data indicate a non-red tide point. The total number of cases where both the algorithm's determination result and the measured red tide data indicate a red tide point is TP; the total number of cases where the algorithm's determination result indicates a red tide point, but the measured red tide data indicates a non-red tide point is FP; the total number of cases where the algorithm's determination result indicates a non-red tide point, but the measured red tide data indicates a red tide point is FN; and the total number of cases where both the algorithm's determination result and the measured red tide data indicate a non-red tide point is TN. The TP, FP, FN, and TN corresponding to a red tide determinism index value constitute the confusion matrix corresponding to that value. The algorithm discrimination result evaluation module is used to evaluate the confusion matrix based on the F-measure algorithm, obtain the comprehensive evaluation index FM corresponding to each value of the red tide certainty index, and take the red tide certainty index value with the highest FM value as the judgment threshold of the high-risk area of ​​red tide occurrence. The red tide high-risk area determination module is used to determine the area consisting of all pixels whose red tide certainty index value is greater than or equal to the judgment threshold as a red tide high-risk area.

Images