CN-121999387-A - Method and system for identifying cyanobacteria bloom risk based on double-star cooperation

CN121999387ACN 121999387 ACN121999387 ACN 121999387ACN-121999387-A

Abstract

The invention discloses a method and a system for identifying cyanobacteria bloom risks based on double stars, wherein the method comprises the following steps of S1, constructing an A star lightweight chromaticity boundary trigger model and a B star refined spectrum risk classification model based on historical data, S2, utilizing an A star received image to conduct pretreatment to identify a target water body area, converting a multispectral reflectance value of each water body pixel into CIE chromaticity space to obtain chromaticity coordinates, judging whether the chromaticity coordinates exceed a preset boundary, scheduling the B star through an inter-satellite communication link if the chromaticity coordinates exceed a trigger condition, S3, B star response scheduling instructions, analyzing task trigger reports, acquiring an image to extract pixel multidimensional feature system, calculating risk scores and classifying by utilizing a spectrum risk classification model, and outputting identification results.

Inventors

A YINGLAN
LI JIEJI
WANG GUOQIANG
ZHANG MENG
XUE BAOLIN
WU JIN
WANG YUNTAO

Assignees

北京师范大学

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (10)

1. The method for identifying the cyanobacteria bloom risk based on double-star cooperation is characterized by comprising the following steps of: S1, constructing a lightweight chromaticity boundary trigger model for an A star and a refined spectrum risk classification model for a B star based on historical remote sensing data and a water bloom event sample; S2, receiving an original remote sensing image by using an A star, identifying a target water body area after pretreatment, and converting the multispectral reflectance value of each water body pixel into CIE chromaticity space to obtain chromaticity coordinates; Judging whether the chromaticity coordinates of the pixels of each water body exceed a preset boundary or not by utilizing the chromaticity boundary triggering model, if the quantity or proportion of the pixels exceeding the preset boundary meets the triggering condition, constructing a task triggering report, and scheduling a B star through an inter-satellite communication link; S3, responding to the scheduling instruction by the star B, analyzing the task triggering report, acquiring an image of the target water body area, extracting a multidimensional feature system of the pixels after preprocessing, calculating risk scores of the pixels by using the spectrum risk classification model, classifying risk grades according to the score results, and outputting risk identification results.
2. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein the S2, the acquiring process of chromaticity coordinates specifically comprises: receiving an original remote sensing image by using the A star, and obtaining a preprocessed remote sensing image through radiation correction, geometric correction and atmospheric correction; generating a water mask map by classifying and identifying the target water region in the preprocessed remote sensing image through the on-orbit ground object; and extracting the multispectral reflectance values of the water pixels in the water mask graph, and converting the multispectral reflectance values into CIE chromaticity space to obtain chromaticity coordinates.
3. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein the step S2 of determining whether the chromaticity coordinates of each water body pixel exceed a preset boundary by using the chromaticity boundary triggering model specifically comprises the following steps: calculating the nearest neighbor distance between the burst type sample set and the non-burst type sample set in the chromaticity space, and selecting a sample pair with the distance smaller than a threshold value as a boundary candidate pair; Clustering burst samples in the boundary candidate pairs, dividing the burst samples into a plurality of boundary clusters by adopting a K-means algorithm, and extracting density center points or boundary center points in each cluster as representative samples to form a boundary sample set; And performing polynomial regression fitting or piecewise fitting on the boundary sample set to obtain a boundary function for judging the water bloom risk.
4. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein the S3, multidimensional characteristic system comprises a combination of basic band reflectivity characteristic, spectrum index characteristic and derivative characteristic; the spectral index features include normalized vegetation index, planktonic algae index, red-edge normalized index and spectral slope index.
5. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein in S3, the construction process of the decision boundary of the risk score is specifically as follows: firstly, screening dominant features in a multidimensional feature system by utilizing feature correlation analysis and random forest evaluation, and mapping high-dimensional features to a low-dimensional space by utilizing principal component analysis to construct a decision boundary; The derivative feature combination comprises a band ratio, a normalized difference and a difference feature.
6. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein in S3, the discrimination function of the risk score is specifically: f(x)=w 1 *x 1 +w 2 *x 2 +...+w n *x n +b; Where f (x) is a risk score, w is a weight vector, b is a bias, x i is an input feature vector, i=1, 2.
7. The method for identifying cyanobacteria bloom risk based on double star cooperation according to claim 1, wherein in the step S3, the specific strategy for carrying out risk classification according to the scoring result comprises the steps of setting a risk threshold alpha; When the risk score f (x) is more than or equal to +alpha, judging that the risk is high; when |f (x) | < α, a stroke risk is determined; And when the risk score f (x) is less than or equal to-alpha, judging that the risk is low.
8. A system for identifying cyanobacteria bloom risk based on double-star cooperation, which is used for realizing the method for identifying cyanobacteria bloom risk based on double-star cooperation according to any one of claims 1-7, and is characterized by comprising a ground system, an A star and a B star; The ground system is used for constructing a lightweight chromaticity boundary trigger model for the A star and a refined spectrum risk classification model for the B star based on historical remote sensing data and water bloom event samples; the A star is used for receiving an original remote sensing image, identifying a target water body area after pretreatment, and converting the multispectral reflectance value of each water body pixel into CIE chromaticity space to obtain chromaticity coordinates; Judging whether the chromaticity coordinates of the pixels of each water body exceed a preset boundary or not by utilizing the chromaticity boundary triggering model, if the quantity or proportion of the pixels exceeding the preset boundary meets the triggering condition, constructing a task triggering report, and scheduling a B star through an inter-satellite communication link; And the star B is used for analyzing the task trigger report, acquiring an image of the target water body area, extracting a multidimensional feature system of the pixels after preprocessing, calculating risk scores of the pixels by using the spectrum risk classification model, classifying risk grades according to the score results, and outputting risk identification results.
9. The system for identifying cyanobacterial bloom risk based on double star coordination of claim 8, wherein the a star comprises: the radiation correction and geometric correction module is used for eliminating systematic errors and mapping the systematic errors to a standard geographic coordinate system; The atmosphere correction module is used for inverting the atmospheric aerosol parameters and acquiring the surface reflectivity; The water body identification module is used for eliminating non-water body areas; the chromaticity space conversion module is used for converting R, G, B reflectivity of the water body pixels into CIE chromaticity coordinates; and the boundary judging and triggering module is used for applying the boundary function to judge and count the risk pixel proportion and generating a triggering instruction.
10. The system for identifying cyanobacterial bloom risk based on double star coordination of claim 8, wherein the B star comprises: The spectrum characteristic extraction module is used for calculating the reflectivity, the spectrum index and the derivative characteristics of the basic wave band of the pixel; The risk calculation module is used for loading the support vector machine model and calculating the risk score of the pixels; and the grading and visualizing module is used for grading the risk grades according to the threshold value and generating a risk thermodynamic diagram and a risk grade diagram.

Description

Method and system for identifying cyanobacteria bloom risk based on double-star cooperation Technical Field The invention relates to the technical field of computer vision, in particular to a method and a system for identifying cyanobacteria bloom risks based on double-star cooperation. Background With the continuous aggravation of global climate warming and water eutrophication problems, cyanobacteria bloom has become one of the major ecological environment problems affecting the stability of the fresh water ecological system and the safety of drinking water in China. In recent years, in typical lake areas such as Taihu lake, chaohu lake and Dian pond, blue algae bloom frequently bursts, and the characteristics of strong burst, high expansion speed, uneven spatial distribution and the like are presented, so that the ecological balance of water is seriously destroyed, and harmful substances such as microcystins and the like can be released, and the health of residents, fishery cultivation and urban water supply systems are endangered. Therefore, the rapid sensing, dynamic monitoring and risk early warning of the cyanobacteria bloom are realized, and the method has become a core requirement for the construction of an intelligent monitoring system of the water environment. At present, the traditional blue algae monitoring method mainly depends on ground water quality sampling and manual inspection, and has higher accuracy in a local range, but due to the dependence on manpower and space coverage, the traditional blue algae monitoring method is difficult to meet the requirements of large-scale, high-frequency and cross-lake-area dynamic monitoring. The remote sensing technology is widely applied to blue algae identification, water classification, water inversion and other tasks by virtue of the wide coverage, high time resolution and continuous observation capability. However, the existing remote sensing blue algae identification method has the following general challenges that most methods depend on concentration inversion models (such as indirect indexes of chlorophyll a, phytoplankton concentration and the like) and are greatly influenced by atmospheric correction precision and water optical complexity, the algorithm models generally face the problems of poor regional migration capability, weak timeliness, high response delay and the like, the algorithm models are difficult to directly apply to on-board real-time calculation scenes, and the algorithm models lack of unified algorithm construction specifications and light deployment strategies, so that popularization of remote sensing algorithms to on-board embedded platforms is limited. In summary, there is a need for a system scheme with algorithm lightweight, model mobility and processing instantaneity for cyanobacteria bloom risk identification to support automatic identification, risk triggering and emergency response to a water body, so as to realize efficient dynamic perception of bloom risk. Disclosure of Invention The invention aims to provide a method and a system for identifying cyanobacteria bloom risks based on double-star cooperation, wherein A star is used for screening and scheduling B star by using a chromaticity boundary trigger model, and B star is matched with the B star to carry out fine evaluation by using a spectrum risk classification model, so that the cyanobacteria bloom risks are identified by double-star cooperation. In order to achieve the above purpose, the present invention provides the following technical solutions: A method for identifying cyanobacteria bloom risk based on double-star cooperation comprises the following steps: S1, constructing a lightweight chromaticity boundary trigger model for an A star and a refined spectrum risk classification model for a B star based on historical remote sensing data and a water bloom event sample; S2, receiving an original remote sensing image by using an A star, identifying a target water body area after pretreatment, and converting the multispectral reflectance value of each water body pixel into CIE chromaticity space to obtain chromaticity coordinates; Judging whether the chromaticity coordinates of the pixels of each water body exceed a preset boundary or not by utilizing the chromaticity boundary triggering model, if the quantity or proportion of the pixels exceeding the preset boundary meets the triggering condition, constructing a task triggering report, and scheduling a B star through an inter-satellite communication link; S3, responding to the scheduling instruction by the star B, analyzing the task triggering report, acquiring an image of the target water body area, extracting a multidimensional feature system of the pixels after preprocessing, calculating risk scores of the pixels by using the spectrum risk classification model, classifying risk grades according to the score results, and outputting risk identification results. Further, the step S2 of obtaining the chromatici