Fast Segmentation Method Of Clutter Scene Of Land And Sea

Target detection in sea clutter background has been widely used in military and civil field. When the radar worked under the sea mode, the scanned scene is complicated and huge. The echoes of radar include various types of clutters, such as sea clutter, land clutter, reef clutter and so on. The intensity of land clutter and reef clutter is much stronger than that of sea clutter, which severely interfere target detection in the sea surface. Therefore, it is a necessary pre-processing before sea-surface target detection to segment the land and sea and the quality of the scene segmentation directly affects the performance of sea-surface target detection.The segmentation of clutter scene of land and sea is to segment the scene into land plus reef region and sea region based upon the radar returns data. Different from the traditional image segmentation, the segmentation of radar clutter scene includes transferring the radar data into a gray image and segmentation of the gray image. Image segmentation is a technology and process to divide an image into different regions and to extract objects of interest. The classical segmentation methods include the three categories: thresholding-based segmentation methods, region-based segmentation methods, and edge-based segmentation methods. The thresholding-based method is to choose a suitable grey value as a threshold to partition the image into the target region and background. Due to simple and fast implement-ation, it is widely used to obtain coarse segmentations of images. The region-based method exploits the regional similarity to partition an image. An image is first divided into small size subregions, where the pixels in each subregion are regarded to have the same characteristics. And then, the subregions with the same features are immerged into large-size regions to obtain a segmentation of the image. The edge-based method generally consists of edge extraction, edge connection, and region labeling. Besides the three basic categories, there are many methods that combine the characteristics of at least two categories, for example, the image segmentation methods based on the mathematical morphology, wavelet theory, genetic algorithm, and neural network. A clutter scene of land and sea is often composed of range-azimuth resolution cells over 106. Therefore, the fast and real-time segmentation method is crucial for application. At this dissertation, thelinearity of radar returns at a resolution cell is introduced to transfer the radar data into a gray image and then the thresholding and mathematical morphology filtering are used to implement a fast segmentation of the gray image.When the radar worked for sea-surface surveillance, the transmitted pulses at one beam position are generally within ten and thus the phases of the returns at a resolution cell is easy to compute. The linearity of the phase series at each resolution cell forms a gray image of the scene, where small values often indicate the pixels at the land or reef because their returns have quite small Doppler bandwidth and large values imply the pixels in sea region because their returns have large Doppler bandwidth. According to the analysis of the linearity of the cells, thresholding is imposed on the gray image to implement a coarse segmentation of the clutter scene. Considering the connection of land regions and sea regions, the mathematical morphological filtering is operated on the coarse segmentation to complete the final scene segmentation. Owing to the fast extraction of the linearity and fast thresholding and morphological filtering, the proposed method can satisfy the requirement of the practical radars on real-time implementation.