Research On Fast Processing Technology Of SAR Image Target Detecting

Synthetic aperture radar (SAR) is of important use both for military reconnaissance and civil activities. The background of this dissertation is SAR image guidance. This paper studies a series of fast processing methods for SAR targets detecting and extracting.This dissertation studies and analyses the geometric features of SAR images. It studies the geometric model of missile borne SAR images, which should be geometrically corrected. It is proved to be available based on experiments.It studies and analyses the speckle noise model of SAR images, which gives a conclusion that the model is true of Rayleigh distribution or exponential distribution. Based on the conclusion, it studies a series of classical space local statistical filters. It presents a new fast filter based on edge detecting, which uses an edge detecting operator and mean filter in the edge area and local statistical filter in the non-edge area. It is proved to be efficiency by experiments and comparison with other classical filters.Systemically, it studies the principles and processes of series classical algorithms on SAR image segmentation. It analyses the features and results of all the algorithms. After that, the passage studies a kind of method on SAR image segmentation evaluation.This dissertation presents some new methods of some typical targets detecting and extracting based on the features of SAR images. It designs a method based on mathematical morphology for bright targets. This method transforms images by mathematical morphology way and extracts the targets by calculating self-adapt threshold with acreage restriction. It designs a method based on traditional Hough transform for straight lines. It presents a concept of gray grads directional entropy, which could detect the areas contain edges and the possible directions of the edges. This amelioration could reduce calculation efficiently. It designs a method based on C-Means clustering algorithm for shorelines. It ameliorates region growing algorithm by presenting a best algorithm on calculating threshold based on assumption that both the sea and the land of the image are true of normal distribution. In the end of this passage, it evaluates and compares the methods by experiments, which proves all the methods are useful and efficiency.