Interactive Segmentation Of Airport Target Via Super-pixel

The SAR Image Segmentation is the process of extracting needed ground object target by processing remote sensing data.Quality of segmentation directly determines the accuracy of later processing works including the area expression of image, extracted characteristic matching, target recognizing and so on.Therefore, study of method of Image Segmentation is an essential step of image processing.As one of the important military and civil ground object targets, it is also significant to study recognition and extraction method of airport runway.This thesis mainly focuses on segmentation method of airport targets in SAR images. There are two main steps:At first, we divided original images into several super-pixel with great homogeneity.By the comparison of super-pixel center points and artificial characteristic marks, we finished the classification of super-pixel in order to implement the extraction of airport targets.The experiments prove that method in this thesis is efficient during the process of extracting airport targets in the experimental images.The contents of this thesis mainly include:(1)An extracting segmentation method of airport based on super-pixel has proposed.First by Mean shift method, it divided SAR images into several similar super-pixels and replaced pixels by super-pixels to get results of segmentation. Result of experiments suggests that method of this thesis can reduce data complexity of images to be processed and extract airport targets accurately. In those SAR images with large size,complex backgrounds and low quality, runways and parking apron with irregular outlines,we still obtained great segmentation result.(2)A segmentation method of airport target by improving Mean shift ultra-pixel is proposed.We realized the purpose of accelerated optimization to Mean shift calculation process by reducing the number of initial points participating iteration, so that we can implement the optimization of time efficiency of whole segmentation process.Result of experiments suggested that this method can not only speed up the segmentation, but also make the right segmentation in airport target area. Most segmentation results of SAR images are ideal and have high detection rate and low false alarm rate. This can keep the accuracy and integrity of airport targets.