Inversion Model Design of Chlorophyll a Based on BP Neural Network and Remote Sensing Image

Abstract

With the continuous deterioration of water quality and gradually worsening of eutrophication in Chinese lakes, the sustainable socioeconomic development has been severely restrained in recent years. Against this backdrop, it is of great necessity to improve pollution regulation. As an important indicator of eutrophication, the concentration of chlorophyll a enables us to make an effective evaluation of eutrophication of lakes. Since the traditional method is time-consuming and energy-inefficient, this paper puts forth the inversion method of chlorophyll a based on BP neural network and remote sensing image. Specifically, this research involves radiometric calibration, geometric correction, atmospheric correction and region clipping as well as other pretreatments of the data downloaded from environment mini-satellites. The BP neural network model is built to extract the features of inverted chlorophyll a of remote sensing image. Based on the field investigation of chlorophyll a concentration in Taihu Lake, the parameters of the BP neural network model are optimized and trained. Finally, the author verifies the accuracy of the proposed model. Through the analysis of verified data, it is proved that the neural network has made an accurate and feasible prediction of the chlorophyll a concentration in Taihu Lake.