Research On Target Detection Of Large Format Ocean Satellite Remote Sensing Images

Ocean remote sensing technology is one of the most important research topics on global changing detection and military reconnaissance.How to extract the vital parts from a huge number of large format ocean remote sensing images effectively and accurately,is an important research direction in ocean remote sensing technology.The detection of ship target is key research contents and has military and civil significance.However,the spatial resolution of remote sensing images is getting higher and higher,and its size is getting larger and larger.It is difficult to detect the sea targets from large range remote sensing images.So in this paper,the author explored a series of key issues in target detection of large format ocean remote sensing images,including the analysis of the background characteristics of large format ocean remote sensing images;the presentation of the large range background modeling and the double refine estimation method of model parameters;and the presentation of the target detection methods of high-resolution remote sensing images that are based on large range background model.The research results of this paper are mainly in the following three aspects.1)Analyzed the local and large range statistical characteristics of seawater background for satellite high-resolution remote sensing images.Analyzed the local background statistical properties of satellite high-resolution remote sensing images based on local feature statistics and verified the local similarity of the statistical mean and variance for ocean background.Analyzed the large range background statistical properties of satellite high-resolution remote sensing images based on global feature statistics and verified the large range continuous change of the statistical mean and variance for ocean background2)Established the large range surface Gaussian distribution statistical model for highresolution remote sensing images background and presented estimation method of model parameters.Tested the gray level distribution characteristics of local ocean remote sensing background images by K-S hypothesis test and verified that the gray level distribution characteristics of the local ocean remote sensing background image is the most consistent with the Gauss distribution.Based on the local Gauss distribution of ocean remote sensing images,the local similarity and the large range continuous change of the statistical mean and variance for ocean background,established the statistical distribution model for each position in the seawater background surface by taking the large range surface fitting method.Introduced the coarse estimation and the refined interpolation of internal block parameters of sub-images parameters in the double refine estimation method of model parameters.Then obtained the gauss variation surface of the large range remote sensing images.3)Presented a method of high-resolution remote sensing image target detection based on the large range background model.Firstly,the mean and variance of the image sub blocks are taken as the classification vectors,and classified the image sub blocks of the pure and non-pure ocean by the fuzzy C means clustering method.Then used the large range background statistical model of non-pure ocean image blocks to detect the large format ocean remote sensing image,and obtained candidate target detection results of large format remote sensing image.Finally,used fuzzy inference with centroid distance and overlapping rate characteristics between any two candidate targets to fuse the candidate targets,and got the final detection results.