Search

CN-122024044-A - Chlorophyll a concentration inversion method, chlorophyll a concentration inversion system, storage medium and terminal

CN122024044ACN 122024044 ACN122024044 ACN 122024044ACN-122024044-A

Abstract

The invention provides a chlorophyll a concentration inversion method, a chlorophyll a concentration inversion system, a storage medium and a terminal, wherein the method comprises the following steps of synchronously acquiring multispectral remote sensing images of a target area and chlorophyll a concentration data; the method comprises the steps of preprocessing a multispectral remote sensing image to obtain an effective multispectral remote sensing image, preprocessing chlorophyll a concentration data to obtain effective chlorophyll a concentration data, extracting characteristic band combination indexes based on the effective multispectral remote sensing image, training a chlorophyll a concentration inversion model based on the characteristic band combination indexes and the effective chlorophyll a concentration data, and carrying out chlorophyll a concentration inversion based on the trained chlorophyll a concentration inversion model. The chlorophyll a concentration inversion method, the chlorophyll a concentration inversion system, the storage medium and the terminal realize high-precision and high-robustness inversion of the chlorophyll a concentration.

Inventors

  • WU MENGQIAN
  • YANG XIAONAN
  • JING YI
  • YAO KAIYI

Assignees

  • 上海市信息技术研究中心

Dates

Publication Date
20260512
Application Date
20260116

Claims (10)

  1. 1. A chlorophyll a concentration inversion method, characterized in that the method comprises the steps of: Synchronously acquiring multispectral remote sensing images and chlorophyll a concentration data of a target area; Preprocessing the multispectral remote sensing image to obtain an effective multispectral remote sensing image; Preprocessing the chlorophyll a concentration data to obtain effective chlorophyll a concentration data; Extracting a characteristic band combination index based on the effective multispectral remote sensing image; And training a chlorophyll a concentration inversion model based on the characteristic band combination index and the effective chlorophyll a concentration data, so as to invert the chlorophyll a concentration based on the trained chlorophyll a concentration inversion model.
  2. 2. The chlorophyll a concentration inversion method according to claim 1, wherein preprocessing the multispectral remote sensing image to obtain an effective multispectral remote sensing image comprises the following steps: Performing quality optimization and restoration on the multispectral remote sensing image; performing geometric correction and orthographic projection on the multispectral remote sensing image after quality optimization and restoration; Performing radiation calibration and atmosphere correction on the multispectral remote sensing image subjected to geometric correction and orthographic projection; Performing image registration and seamless stitching on the multispectral remote sensing image subjected to radiometric calibration and atmospheric correction to obtain a multispectral orthographic stitching image; and performing noise suppression and boundary clipping on the multispectral orthomosaic image to obtain the effective multispectral remote sensing image.
  3. 3. The chlorophyll a concentration inversion method according to claim 1, wherein extracting a characteristic band combination index based on the effective multispectral remote sensing image includes the steps of: chlorophyll a multiband synergy index cmsi=extracted based on the effective multispectral remote sensing image Wherein Indicating the reflectivity of the blue band in the band range between 420nm and 480nm, Indicating the reflectivity of the red band in the band range 682nm-638nm, Representing the reflectivity of the red-side band in the band range between 710nm and 730nm, Representing wavelength; Extracting a wide-area response synergy index WDRSI = based on the effective multispectral remote sensing image - + Wherein Representing the reflectivity of the green band in the band range between 528nm and 582nm, Indicating the reflectivity of the red band in the band range 682nm-638nm, Representing the reflectivity of the red-side band in the band range of 710nm-730 nm; Extracting red-green band enhancement index GREI = based on the effective multispectral remote sensing image - + Wherein Representing the reflectivity of the green band in the band range between 528nm and 582nm, Indicating the reflectance of the red band in the band range 682nm-638 nm.
  4. 4. The chlorophyll a concentration inversion method of claim 1, wherein said chlorophyll a concentration inversion model employs CatBoost gradient lifting decision tree algorithm.
  5. 5. The chlorophyll a concentration inversion method according to claim 4, wherein said CatBoost gradient boost decision tree algorithm uses a random grid search strategy or a bayesian optimization strategy to obtain an optimal super-parameter combination.
  6. 6. The chlorophyll a concentration inversion method according to claim 5, wherein said optimal super-parameter combinations include one or more of maximum depth of decision tree, number of iterations, learning rate, L2 regularization coefficient, feature sampling scale.
  7. 7. The chlorophyll a concentration inversion system is characterized by comprising an acquisition module, a first preprocessing module, a second preprocessing module, an extraction module and a training module; the acquisition module is used for synchronously acquiring multispectral remote sensing images of the target area and chlorophyll a concentration data; the first preprocessing module is used for preprocessing the multispectral remote sensing image to obtain an effective multispectral remote sensing image; The second pretreatment module is used for carrying out pretreatment on the chlorophyll a concentration data to obtain effective chlorophyll a concentration data; The extraction module is used for extracting characteristic wave band combination indexes based on the effective multispectral remote sensing images; The training module is used for training a chlorophyll a concentration inversion model based on the characteristic wave band combination index and the effective chlorophyll a concentration data so as to invert the chlorophyll a concentration based on the trained chlorophyll a concentration inversion model.
  8. 8. A storage medium having stored thereon a computer program, which when executed by a processor implements the chlorophyll-a concentration inversion method according to any one of claims 1 to 6.
  9. 9. The terminal is characterized by comprising a processor and a memory; the memory is used for storing a computer program; the processor is configured to execute the computer program stored in the memory, to cause the terminal to perform the chlorophyll-a concentration inversion method according to any one of claims 1 to 6.
  10. 10. A chlorophyll a concentration inversion system, comprising an unmanned aerial vehicle, an image acquisition device, a chlorophyll a concentration acquisition device, and the terminal of claim 9; The image acquisition equipment is arranged on the unmanned aerial vehicle, is used for acquiring multispectral remote sensing images of the target area and provides the multispectral remote sensing images to the terminal; the chlorophyll a concentration acquisition equipment is used for synchronously acquiring chlorophyll a concentration data corresponding to the multispectral remote sensing image and providing the chlorophyll a concentration data to the terminal.

Description

Chlorophyll a concentration inversion method, chlorophyll a concentration inversion system, storage medium and terminal Technical Field The invention relates to the technical field of chlorophyll a concentration, in particular to a chlorophyll a concentration inversion method, a chlorophyll a concentration inversion system, a storage medium and a terminal. Background With the acceleration of the urban process and the increase of the activity intensity of human beings, the production and life intensity of human beings are continuously increased, a large amount of nutrient substances such as nitrogen, phosphorus and the like are discharged into the water body, so that the problem of eutrophication of inland water bodies is increasingly prominent, and the problem of serious environment for restricting the sustainable development of regional economy and society is formed. Chlorophyll a is used as a key pigment for photosynthesis of phytoplankton, the concentration of the chlorophyll a directly represents the nutrition degree and the phytoplankton biomass of a water body, and the chlorophyll a is an internationally recognized water body eutrophication core evaluation index. Therefore, the chlorophyll a concentration is accurately and efficiently monitored, and the method has important guiding significance for water quality assessment, algal bloom early warning, water ecological health management and drinking water source safety guarantee. The traditional manual sampling method has the problems of sparse point positions, long period, high cost, difficulty in reflecting the space-time continuous change of large-area water quality and the like. Satellite remote sensing can realize large-scale monitoring, but is limited by space-time resolution, so that the high-precision monitoring requirements of long and narrow water bodies such as urban rivers and small micro water bodies are difficult to meet. The unmanned aerial vehicle remote sensing technology provides an effective technical means for fine water quality monitoring due to the characteristics of high flexibility, high spatial resolution, low operation cost, small influence of cloud layers on data acquisition and the like. The high-resolution multispectral camera carried by the unmanned aerial vehicle can effectively extract high-quality small micro water body spectrum information under the condition of not being interfered by cloud layers, and the inversion accuracy is high, so that the practicability of unmanned aerial vehicle remote sensing water quality inversion is proved. The inland water has complex optical characteristics, is often simultaneously influenced by phytoplankton, suspended matters and colored dissolved organic matters, has remarkable optical signal superposition effect, and has poor adaptability and limited precision in the traditional single inversion model. Traditional inversion models of chlorophyll concentration in water bodies mainly depend on empirical or semi-empirical methods based on biological optical characteristics. Such models are typically implemented by establishing a linear or nonlinear regression relationship between the reflectance combinations (e.g., waves Duan Bizhi, normalized difference indices, etc.) and measured chlorophyll concentrations for specific bands (e.g., red and near infrared bands). While these models are physically well defined and computationally simple, their modeling is highly empirical in nature. Their model parameters are severely dependent on the particular data set employed in modeling (e.g., sample data for a particular time, a particular body of water), resulting in a serious inadequacy in their universality (i.e., generalization capability). When the method is applied to urban fine river channels with complex and changeable optical characteristics, the inversion accuracy can be drastically reduced and the generalization capability is quite unsatisfactory because the space-time heterogeneity of water components (such as suspended matters and colored soluble organic matters) is quite strong and the interference of surrounding environments (such as building shadows and river bank vegetation) is remarkable, and the fixed parameters of the traditional model are difficult to adapt. In order to improve the fitting capability to complex nonlinear relationships, some research has begun to explore the introduction of advanced machine learning algorithms. The attempts mainly comprise (1) introducing an optical classification idea, namely classifying water bodies according to optical characteristics, and then establishing independent inversion models for different types of water bodies so as to improve pertinence, and (2) adopting a nonlinear machine learning model, such as a support vector machine (Support Vector Machine, SVM) and a back propagation neural network (Back Propagation Neural Network, BP neural network), so as to break through the limitation of a linear model. However, these improvement strateg