Font Size: a A A

Research On The Ship Target Detection In High Spatial Resolution Optical Remote Sensing Image

Posted on:2007-08-18Degree:MasterType:Thesis
Country:ChinaCandidate:L B JiangFull Text:PDF
GTID:2178360215470244Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Automatic ship target detection in high spatial resolution optical remote sensing image is of great importance in military application. The key problem of automatic ship target detection is how to distinguish target from massive remote sensing image data quickly and accurately. In this dissertation, study on offshore and in-shore ship target automatic detection was investigated against the background of sea target surveillance by remote sensing satellite. The paper mainly included three facets. Firstly, an off-shore ship target detection frame was proposed based on the attention mechanism, in which the procedure of detection was divided into two levels: rough detection and fine detection. In the phase of rough detection, the regions of interest (ROI) were quickly obtained by the guidance of gray and area characteristics. In the following fine detection phase, more characteristics such as contour, length to width ratio and parallel were extracted and integrated in order to improve the detection reliability.Secondly, contour and parallel characteristic extraction methods in the fine detection phase were discussed, and then a multi-character based ROI ship target detection algorithm was advanced. The modified histogram-based chain code was introduced to describe the closed contour characteristic, which made the characteristic template relatively invariant to scaling and rotation. With the advantage of template invariance, the database size was greatly reduced, which could speed up the template matching procedure. At the same time parallel extraction method using Hough transform was discussed. By using multi-character, the number of suspicious ROI, which could not be conformed by using only contour characteristic, was reduced.Thirdly, aiming at the difficulty the unclosed contour characteristic description of in-shore ship, alterable included angle code (AIAC) was introduced to solve the problem, and then an AIAC-based in-shore ship detection method was proposed. By using line evaluate parameter, AIAC generation procedure was simplified. In the following the open curve matching algorithm was advanced by using AIAC description. In the end of the dissertation, in-shore ship detection experiments were carried out and the results demonstrated the efficiency of the algorithm.Finally, this dissertation was summarized and the further work was pointed out.
Keywords/Search Tags:High Spatial Resolution Optical Remote Sensing Image, Ship Target Detection, Attention Mechanism, Target Contour, Alterable Included Angle Code
PDF Full Text Request
Related items