Research On Content-based Aerial Image Retrieval

With the rapid development of computer and information science,especially with the implementation of the plan of the digital earth,the processing,transmission and application of the information of the earth has reached a higher level.However,how to process,query and match the massive aerial images is still a challenging task while the remote sensing technology is developing rapidly with the demand of users and the quality and quantity of aerial images continues growth of user demands.Aerial image retrieval is one of the key technologies in various services using GIS,in which a match is to be identified in a large image database with a given query image with unknown location,orientation,and scale.To achieve aerial image discovery,especially in city area,there are three main requirements need to meet:accuracy,efficiency and robustness.Although there have been various research attempts made in this area,the performance of automated discovery is not sufficiently high.The common content-based image retrieval system cannot meet the requirement of aerial image discovery because the aerial images have much richer contents and more complex structure compared to common images.In this paper,we propose a new method for achieving city aerial image discovery based on texture feature calculated using GLCM combined with SIFT algorithm.This algorithm uses grey-level co-occurrence matrix for pre-selecting search candidates,and LSH forest to process the pre-selected result.Then the scale invariant feature transform algorithm and the random sample consensus algorithm are used to obtain the final matching result.Experimental results demonstrate that these sets of algorithms can greatly reduce the discovery time while meeting the requirements of precision and robustness.