Detection Of Ship In Optical Remote Sensing Image Of Median-low Resolution

Ship detection and surveillance is a traditional task of defense, maritime management and fishery of the seacoast countries. As an effective means of ship detetion, remote sensing have been payed more and more attention and widely adopted in the world. This thesis studies ship detection in optical image of median-low resolution and also the detection of fleet formation. The thesis includes primarily following parts:1. Detection of ship under sea-land background. It is required to segment land and sea before ship detection. Based on the analysis of gray distribution of sea-region, a statistical classification method is proposed to segment sea and land.Then a gray and texture combined method is adopted to detect ship.2. Ship detection under a complicated sea background image.A complicated sea background image is a remote sensing image partly covered by clouds and shadows.A visual attention mechanism based method is proposed to detect ship in an image with a compicated background.For a moving ship, there will be a wake behind the ship. According to the analysis of the wake pattern, a region texture based method is proposed to detect wake in the region partitioned. There are always false alarms in a detect result, so target discrimination is a necessary step.A multi-feature combined sequential discrimination method is proposed to mitigate false alarm.3. Detection of fleet formation. According to the knowledge of fleet formation, organization rules of perceptual organization are applied to study ships spatial relationship, then a fuzzy function is used to meaure the regularity of spatial target primitive.At last, fleet formation detection is achieved by iterative spectral graph partitioning algorithm.All algorithms mentioned above are tested with data of satellite based remote sengsing systems and the results show the promising performances of the algorithms.