Font Size: a A A

SAR Image Processing Based On Genetic Optimization Neural Networks

Posted on:2007-10-21Degree:MasterType:Thesis
Country:ChinaCandidate:L L KouFull Text:PDF
GTID:2178360182478016Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Nowadays in the field of ocean pollution, it is a problem eagerly needing to abstain to hurt human health. Therefore it's the time to develop a key technology of the effective real-time software system for inspecting ocean environment. Synthetic Aperture Radar (SAR) is a modern high-resolution factor microwave remote sensing imaging radar. It can inspect oil-spill on the sea at anytime and any weather. In this paper, a new method of oil-spill condition detection on the sea is proposed based on the analysis of the SAR images and the present situation of oil-spill condition detection. This method is a algorithm using genetic optimizing neural networks. This method realizes real-time detection of sea oil-spill condition. It is improtant to prevent sea environment pollution and control. It is also necessary to the realization of intelligent sea environment system.SAR images are extensively used for the determination of oil spill in the marine environment, most of study found that oil spill appeared as dark slick or dark spot in SAR image due to the dampening effects on the sea. Utilizing the characteristics of abundant texture information in SAR image, this paper investigates a texture analysis and Neural Network method for the classification of SAR images with spilled oil, with particular emphasis on its applicability to get the accurate result of classification. In the progress of texture analysis, four pixel texture parameters that are sensitive to SAR image of oil spill are calculated by gray level difference statistics. The SAR image is classified by using the feature vector that is composed of the Gray Level Co-occurrence matrix features and gray of pixel. We use two types of algorithm;Back-Propagation Neural Network (BPNN) and Higher-order Neural Network (HONN). Using the data in form of 5-level feature vector as inputs. The ANN istrained and tested using sample data set to the network. The results of the above 2 types of network are compared in this paper. All of them have a good performance of classification;but HONN is more effective and accurate one as classifier for SAR images of oil-spill.
Keywords/Search Tags:Texture Analysis, Back-Propagation Neural Network (BPNN), Hopfield Neural Network, Higher-order Neural Network (HONN), Gegetic algorithm
PDF Full Text Request
Related items