Application Of SVM Model To Chlorophyll-a Nonlinear Quantitative Remote Sensing Retrieval

The water quality of the Great Lake Donghu area in Wuhan city is one of the major water environmental problems faced by Wuhan.It is undoubtedly a valuable value to analyze the Chl-a concentration of the remote sensing image by monitoring the characteristics of Chl-a.,And it is of great significance to promote the inland water body Chl-a and other quantitative remote sensing inversion of water quality,which is of great significance to the lake’s eutrophication.It is of great significance to the environmental protection of Wuhan East Lake and the construction of good ecological system.The early warning of the provision of technical support.Chlorophyll a concentration is an important factor in determining the reflectance spectrum characteristics of water bodies and is an important basis for assessing the health status of water quality and the impact of increasing human activities and external environmental changes on aquatic ecosystems.Because the conventional water quality monitoring method can not carry out real-time and large-scale monitoring of water quality,this paper makes use of the remote sensing to cover the above-mentioned shortcomings by covering the wide range of data source with stable and regular repetition period.Although the established technical route has been used for inversion of chlorophyll a concentration in inland lake water bodies,it is limited to regional development differences.These models are not applicable to the Great Lake Donghu area in Wuhan city water bodies.Therefore,it is necessary to establish a high-precision remote sensing inversion model for Wuhan East Lake.Based on the spectral characteristics of Inland Lake Class II water bodies,three kinds of remote sensing inversion models for chlorophyll a concentration in Donghu region of Wuhan were constructed by using the measured chlorophyll a concentration data and synchronous / quasi-synchronized multi-spectral Landsat8 images.A single band model with inversion accuracy of 77.11% and RMSE of 8.93 and a four-band model with an inversion accuracy of 88.01% and RMSE of 4.55 were established by analyzing the correlation between each band and chlorophyll a concentration.In order to further improve the accuracy of inversion,the inversion model with high accuracy is established by the input parameters,kernel function and parameter adjustment in SVM technology.The accuracy of the model is 94.46% and RMSE is 2.16.The results show that the SVM model has a better prediction effect and provides a reliable basis and method for rapid and accurate inversion of chlorophyll a concentration in waters using remote sensing technology.The distribution of chlorophyll a concentration was obtained by SVM model inversion,and the temporal and spatial variation of chlorophyll a was analyzed.