Study Of Water Depth Retrieval Based On Artificial Neural Networks Under The Influence Of Multi-factor

Ocean remote sensing is a focus direction of current scientific research. Water-depth is the basal terrain of coastal zone. It is important to develop the fast and accurate methods of bathymetry for economic development, military affairs and national defense. Water depth retrieval has developed for a long time, but owing to the turbid characteristic of near shore areas water, which belongs to Class II water, the bathymetry there by means of visible-light remote sensing technology is comparatively difficult. We should take multi-factor influence into account. A method of HJ-1A CCD and HSI remote sensing of shallow water bathymetry based on artificial neural network(ANN) technique is studied in this thesis.In this paper, Dalian coastal water is used as the study area. First of all, analysised the mechanism of depth inversion based on remote sensing; Secondly, to study the effects of various factors to multispectral remote sensing inversion depth, summarized the main factors: suspended sediment and chlorophyll a, and suspended sediment and chlorophyll factor model were established by the measured data in accordance with remote sensing image; Again, used SPSS to establish a single-band linear regression model, using Matlab neural network toolbox to build two models to train the net: directly or add suspended sediment and chlorophyll parameters to the input range. Finally, by comparing the measured values and the depth of the inversion data on the performance of the three models of the algorithm is analyzed.The results of this study show that, BP artificial neural network, because of its good self-organizing, self-learning ability, nonlinear and other advantages, build depth inversion model significantly better than the linear regression modeling, and joined the Aqua factor, the retrieval depth accuracy can be further improved.