A Target Detection Method In High Resolution SAR Image Based On CBM

SAR imagery has gained increasingly attention on both military and business as it is capacity to work whole time and whole weather. With the improvement of the technique of SAR imagery, the resolution of SAR image is also ascending. In high-resolution SAR image, compared with low-resolution one, ground targets are larger and have far more complicated contour as well as low spatial homogeneity, which directly lead the conventional method for low-resolution SAR image interpretation based on weighted features to a low accuracy. A new method with high detection or classification accuracy is needed.With the development of artificial intelligence, Case based reasoning system has becoming more and more widely used in various fields. CBR system resembles human cognition, which reasons new unknowns with empirical knowledge. The method this paper proposes, which integrated CBR system into SAR image interpretation, overcome the interference of blurred features of SAR image targets. This method firstly builds up a Case Library which is also known as a set of classified targets as well as their features, and then use decision tree system to detect new targets through comparing the similarity between them and the cases from the library, as the reasoning result, the most similar case ’s category is decided as the type of the new target. This paper is arranged as follow:Chapter one introduces the home and abroad research status of SAR image interpretation, of which the corresponding procedures will also be introduced; At last a overall frame of the main work will be presented to elucidate the whole idea of this paper.Chapter two analyses on the features of high resolution SAR targets. Firstly, some most widely used feature extraction methods are given; Then these method will be tested on real Mini SAR images according to different ground targets; At last, there will be a analysis on the experimental result to show the divisibility of these method according to different ground targets.Chapter three introduces the whole procedure of the establishment of case library, which includes two parts: the getting of cases; the principle of case library. In the former part, we will discuss how to get appropriate case-chip from the whole SAR image. Which will also be divided into two parts: the preprocessing of SAR image as well as the segmentation of cases. Then in the second part, the case restoration principle will be given which should take both convenience and regulation into account.Chapter four explains on the reasoning procedure based on decision tree model. Firstly, the working system of CBR method and the decision tree model are introduced. Then it describes how to utilize the decision tree method in CBR system during SAR image interpretation. At last, a test using this method on real SAR image with resolution of 0.1m will show the advance of the proposed method after the analysis on the experimental results.