Segmentation Methods Of Clutter Scenes Of Sea And Land Based Upon Multiple Measures

Whether in the military field or civilian areas,the target detection technology in the background of sea clutter is a very important and concerned research topic.Sea environment is complex and changeable,when the radar scans the sea,the echoes of radar are often mixed with sea clutter,island reef clutter,land clutter and other types of clutter,of which the strongest echoing island reef clutter and land clutter constitute the main obstacle to the target detection technique in the sea clutter background.Therefore,before the target detection,the rapid segmentation of the sea and land clutter scene is particularly important.This paper mainly studies the problem of sea and land segmentation under different platform conditions.The segmentation of sea and land clutter is usually divided into two parts: the first part is based on the analysis and comparison of the clutter data in the radar echo,selecting the appropriate clutter data measure according to the platform characteristics,and then transforming it to the grayscale image;the second part is the segmentation of gray scale image gained in the first part.For the second part of the image segmentation,the current technology has developed for more than 40 years and becomes relatively mature.Therefore,this paper first introduces three kinds of classic,widely used segmentation methods,namely,threshold-based image segmentation method,region-based image segmentation method and edge-based image segmentation method.Among them,the threshold-based segmentation method is widely used in the initial processing of various types of images because of its simplicity,fastness and efficiency.In this paper,four threshold segmentation algorithms are introduced in detail,and the simulation results show that the Otsu segmentation method is accurate and the segmentation results are stable,besides,it is also suitable for large-scale images segmentation and hence it is used in this paper for rough segmentation of the grayscale image of the sea-land segmentation measure.However,there are a lot of isolated points in the image after rough segmentation,and also many holes in the land area of the contiguous piece.In order to ensure the connectivity of the ocean and land in the sea-land segmentation results,the morphological filtering and the region filling operation are applied.For the first part to select the appropriate clutter data measure,this paper analyzes the radar echo data characteristics of different platforms,constructs the corresponding sea-land clutter scene segmentation measure.With the application of the simple and stable Otsu segmentation method,morphological filtering process,and region filling operation,this paper puts forward a fast segmentation method of sea and land clutter scene based on energy accumulation measure on moving platform and a sea-land segmentation method based on the direct-current component ratio measure for stationary platform.Through the measured data,these two methods can generally obtain satisfactory results.In addition,the sea-land segmentation method based on morphological derivative is proposed to solve the problem of sea-land segmentation of echo data in non-coherent mode.The conclusion is that the sea-land segmentation results based on echo amplitude or intensity is much worse than that based on the phase information.Therefore,it is not recommended to directly segment sea and land clutter scene by the radar echo amplitude in the non-coherent fast scan mode.