Paper Title
Fast Retrieval on Remote Sensing Imagery Based on Orb Feature

Abstract
The Content Based Image Retrieval (CBIR) systems have been an active area of research in remote sensing image precessing. Local invariant features, such as SIFT, have been applied to remote sensing image retrieval. In this paper, we employ ORB feature as an alternative to SIFT, to improve the exibility and e_ciency of BoWs model in the codebook training, high-dimensional feature exaction, and quantization. Our method can achieves a reduction in the number of computations of about one order of magnitude, while obtaining the more accuracy. The e_ciency of ORB descriptor are evaluated with UC Merced Land Use-Land Cover data set. Index Terms—Image Retrieval, Local invariant features, BRIEF feature, ORB feature, Bag of visual words.