Ocean Information Extraction Based On Space Scale Theory

With the "Digital Ocean" strategy being put forward, many oceanographers have been focusing on auto-extraction of ocean information. Among these ocean information, the research on ocean fronts (temperature fronts, salinity fronts and density fronts) not only makes important senses on ocean fishery, but also takes great impact on national defence and produce. And then the auto-extraction of ocean fronts has been paid attention to widely in recent years.The extraction of ocean fronts is to extract the edge information between ocean water mass, which belongs to the high frequency extraction from ocean fields. Since the strong edge and weak edge co-exist in the same image, it is difficult to attain satisfied results by classical edge extracting operators (Gauss operator, Robert operator, Sobel operator, Prewitt operator and Kirsch operator). If the threshold value is too large, weak edge information and noise are all filtered out, and the weak edge loses greatly; if the threshold value is too small, weak edge may be preserved, while noise interferes with the raw information severely.To begin with, from the recognition theory and space scale theory the characteristics about geography entity depending on space scale is narrated: it is realistic meaning and scientific meaning for the geography entity to analyze and extract only in certain scale. Based on the above, ocean fronts's multi-denoted and ocean fronts's features distribution on serial scale space are analyzed and the extracting ocean fronts by the optimal scale is discussed.Originating from 1960's, Mathematical morphology has unique properties on multi-scale analysis. Its core concept: structure element may form structure element sequence by self-basic concept and self-basic characteristics. There exists one-to-one relationship between the sequence structure element's size and the space scale of geography entity. Based on the relationship between structure element's size and the characteristics of morphological gradient, the auto-adaptive algorithm of the optimal structure element's size to extract ocean front is designed through the minus value image between the neighbor structure element's size and the information entropy of edge image. In the end, the remote sensing images: MODIS-SST and NOSS-SST are taken as experience data, and the experience results are compared with the results from the classic gradient operators. It shows that the auto-adaptive algorithm of the optimal structure element's size can ideally extract the ocean fronts on robust, continuity, position precise and contrast, and priors to classical gradient operators.6riginating from 1980's, wavelet theory is regard as "microscope of mathematic analysis". Based on the wavelet multi-scale analysis, the wavelets base-function selecting demands from the edge extraction and some basic characteristics about common wavelets are discussed, and the possibility to extract edge information through Harr-Wavelet is reached. The distribution feature of ocean and noise information on sequence scale and its distribution difference are analyzed, and the extracting algorithm on ocean fronts based on wavelet optimal scale is designed. Ocean fronts images can be decomposed into horizontal detail image sequence, vertical detail image sequence and diagonal detail image sequence by Han wavelet. By selecting edge points, filtering out noise points and linking edges, these sequence images can respectively form horizontal detail image, vertical detail image and diagonal image, which can be reconstructed the ultimate ocean fronts image by Anti-Harr wavelet. In the end, the remote sensing images: MODIS-SST and NOSS-SST are taken as experience data, and the experience results are compared with the results from theclassic gradient operators, which shows that the algorithm can extract ocean fronts on sequence scales.In a short, the paper chiefly solves the following three problems: (1) What scale are the ocean fronts on remote sensing fields (What); (2) What scale are the ocean fronts in real world (What); (3) How to select the optimal scale (operator) to match the scale on remote sensing fields and the scale in real world and attain the best extraction results (How).