Quantitative retrieval of suspended solid concentration in Lake Taihu based on BP neural net
Abstract
A two-layer BP neural net model is constructed with four input nodes of TM1, 2, 3, 4 band reflectances, and one output node of suspended solid concentration (SSC) to retrieve SSC of Lake Taihu. The results demonstrated that BP neural net is very fit to quantitatively retrieve water quality of case II water with complex optic characteristic, and has much higher accuracy than the common linear model. A test was made and the results suggest that 13 had relative error (RE)RE of less than 30%, accounting for 81.25% of the total samples.
This paper has been withdrawn.