A Reconstruction Method Study Of Ocean Color And SST Based On Remote Sensing Data

The proportion of current marine development in the national economy is bigger and bigger. As an important means of observation, satellite ocean remote sensing is universally applied. Ocean color and SST field, as one of its important applications, has been of great concern to scholars. Due to perennial cloud cover over the ocean, the ratio of remote sensing data is very high; the study of reconstruction algorithm of ocean remote sensing color and SST missing data becomes a research focus.In this paper, for the missing data problem of ocean color and SST image, first proposes the self-organizing map algorithm so as to be applied to reconstruct the missing data of ocean color and SST data, and then combines with the emerging empirical orthogonal function EOF algorithm in this field to propose SOM-EOF algorithm. SOM-EOF algorithm uses the SOM algorithm’s advantage of well reflecting the nonlinear structure of the remote sensing data sets, applies the nonlinear estimate results of SOM algorithm to initialize the EOF algorithm which effectively overcomes the initialization sensitive issue of EOF algorithm to remote sensing data set, and also improves the EOF algorithm’s defects of powerless to reconstruct the high missing rate images and low program efficiency. In the data processing process, the use of lanczos operator in matrix decomposition of singular value decomposition also improves the efficiency of the program. In addition, this method also uses the Monte Carlo cross validation set to determine the optimal EOF mode number of reconstruction, and finally get the high precision reconstruction missing data, and uses the optimal mode to analyze the physical phenomena of study area.The original data is SST and chlorophyll concentration products collected from MODIS sensor equipped with the AQUA remote sensing satellite. The experiments show that the two new proposed algorithms’reconstruction errors are little, and the reconstruction accuracy is high. In the reconstruction experiment of SST, the minimum reconstruction error of SOM, EOF and SOM-EOF is 0.2736, 0.6000 and 0.2304(℃) respectively, and in CHL, is 0.4650, 0.5272 and 0.4267(log10mg/L).