Qin Dai, Jianbo Liu, Shibin Liu, and Caihong Ma
Content-based image retrieval, remote sensing image, multi-features, relevance feedback
The Content-based remote sensing image retrieval (CBRSIR) is one effect method of quickly acquiring remote sensing information, its efficiency and intelligent performance will affect universality and timeliness of the expanding remote sensing data applications. Based on reviewing and analyzing the current related technology on CBRSIR, this paper proposed an intelligent multi-features integrated remote sensing image retrieval model and designed its framework. And an improved Support Vector Machine (SVM) approach based on Particle Swarm Optimization (PSO) and Genetic Algorithm (GA) for Relevance Feedback (RF) is also proposed for the remote sensing image retrieval. The proposed new RF approach can take consideration of avoiding local maxima in the SVM kernel parameters optimization and the subset feature selection simultaneously by combining PSO and GA. The retrieval experimental results compared with Grid, PSO and GA showed that the proposed new approach has speedy convergence and good stability, and can achieve a promising result.
Important Links:
Go Back
