Synthetic Aperture Radar Image Line Feature Extraction Algorithm

As a tool of acquiring information, SAR is of significant strategic sense in national defense, environment and so on. The interpretation of SAR imagery is a very important task. Compared with the capability of acquiring the data, the processing technology of SAR imagery has lagged and can't meet the requirement of fast process to a great deal of data, so it needs to be improved imminently. We classify SAR targets into point targets, liner targets and region targets roughly via the analysis of SAR imagery, and classify the image features into the point features, liner features and region features. The feature extraction from SAR imagery is an important step in the interpretation based on SAR imagery, it can depress dimension of the feature space to interpret SAR imagery accurately and fast. Edge extraction is not only the basis of the interpretation of liner targets in SAR image, but also can improve the performance of image segmentation.The main work of this paper has three points:(1) Several feature extraction methods used frequently are discussed in the paper. Because the classic edge detectors can't be applied to SAR imagery directly, the algorithms adapting to SAR imagery are analyzed. Experimental results confirm that edge extraction must use other algorithms, considering having speckle noises in SAR imagery.(2) This paper carries out edge extraction of SAR image using two-steps detectors, and discusses detectors D1 and D2 as well as the fusion of their results. This approach accumulates the eight probably direction, and has a constant false-alarm rate.(3) Lastly, considering that Hough transformation has good capability of linking snatchy linear feature, so this paper uses it.Experimental results obtained from SAR images confirm that the proposed approach can bring an improvement, as compared to the Sobel, Prewitt and Canny, and the Hough transformation acquires the ideal results to link linear feature.