An Cbrsir Features Compression Approach Based on Dpso and Svm

The number of image features used by content-based remote sensing image retrieval (CBRSIR) system is not less than one hundred, and the image amount is very large, at the same time, the time cost is very important to the retrieval system. So the image feature compression is a crucial subject to CBRSIR. This paper proposed a high dimensional feature’s compression approach based on discrete particle swarm optimization (DPSO) and support vector machine (SVM). This approach trained the SVM classifier by DPSO, and gained the particle’s fitness by both the train data and the verification data. By iterative processing, the optimized high dimensional feature compression result achieved. This paper addressed the theory and the flow of the new approach in detail and the experiment verified the effectiveness of the new approach. * Corresponding author.