论文部分内容阅读
In the present paper, a method is proposed to improve the performance of Artificial Neural Network (ANN) based algorithms for the retrieval of oceanic constituents in Case Ⅱ waters. The ANN-based algorithms have been developed based on a constraint condition, which represents, to a certain degree, the correlation between suspended particulate matter (SPM) and pigment (CHL), coloured dissolved organic matter (CDOM) and CHL, as well as CDOM and SPM, found in Case Ⅱ waters. Compared with the ANN-based algorithm developed without a constraint condition, the performance of ANN-based algorithms developed with a constraint conditions is much better for the retrieval of CHL and CDOM, especially in the case of high noise levels; however, there is not significant improvement for the retrieval of SPM.
The present paper, a method is proposed to improve the performance of Artificial Neural Network (ANN) based algorithms for the retrieval of oceanic constituents in Case II waters. The ANN-based algorithms have been developed based on a constraint condition, which represents, To a certain degree, the correlation between suspended particulate matter (SPM) and pigment (CHL), colored dissolved organic matter (CDOM) and CHL, as well as CDOM and SPM, found in Case II waters. Compared with the ANN-based algorithm developed without a constraint condition, the performance of ANN-based algorithms developed with a constraint condition is much better for the retrieval of CHL and CDOM, especially in the case of high noise levels; however, there is not significant improvement for the retrieval of SPM .