Font Size: a A A

The Study Of Ship Target Detection In Optical Satellite Remote Sensing Image

Posted on:2012-08-14Degree:MasterType:Thesis
Country:ChinaCandidate:J Y XuFull Text:PDF
GTID:2218330362960142Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
The technology in optical satellite remote sensing develops quickly, it's important to seek how to detect the ship target from large amounts of optical satellite remote sensing images (OSRSI). This thesis makes a study of this course. It mainly consists of pre-processing of OSRSI, ship target detection, ship target discrimination and ship wake detection.In the pre-processing of OSRSI aspect, the main researches focus on the removing of the image noise and the separating of the sea and land. Aiming at the lost detail problem in the noise removing using median filter, the thesis import a median filter method which could decide whether filter or not by a threshold, and develop a fast median filter based on local gray histogram. Finally it realizes the fast filtering and retains image detail consistently. And then, an Otsu method combined with morphology processing is studied to separate the sea and land in the OSRSI, feasibility of two algorithms is validated by experiments.In the target detection aspect, aiming at the problem that CFAR detection cannot detect the black polar target, a ship target detection method based on GLRT (Generalized Likelihood Ratio Test) is proposed. This method uses a sliding window to detect the target, but it differs from CFAR which considers only the distribution in the background window, this method combines the distribution in the target window and the GLRT theory to detect targets. The experimental results show that this method can detect the black polar target which cannot be detected by CFAR, and the false alarm rate is lower, the speed is faster.In the ship target discrimination aspect, Graph Partitioning Active Contours (GPAC) is adopted for the refining object segmentation. Then the shape features, gray and texture features of the object are extracted. Using these features, the SVM classifier is employed to classify and determine whether the object is a ship. By designing different schemes of the features choosing and combining, different SVM classifiers are obtained. Comprehensive consideration of the classifying accuracy, discrimination detection rate and discrimination false alarm rate, the best classifier is chosen.Finally, analyzing the theory of Radon transform, this paper points out that Radon transform has the disadvantage of sampling uniformity. Aiming at the problem, a ship wake detection algorithm based on Scan Method is proposed. For speeding up, the algorithm reduces the detection area by pre-segmenting and―grid method‖. Then, points belonged to one wake are merged, and false alarms are removed. The experimental results show that this algorithm is available.
Keywords/Search Tags:Optical Satellite Remote Sensing Image, Ship Target Detection, Fast Median Filtering, Discrimination, SVM, Feature, Wake, Radon Transform Local Scan
PDF Full Text Request
Related items