Object Classification of Remote Sensing Images Based on Rotation-Invariant Discrete Hashing

Object classification is one of the most fundamental but challenging problems faced for large-scale remote sensing image analysis. Recently, learning based hashing techniques have attracted broad research interests because of their significant efficiency for high-dimensional data in both storage and speed. Despite the progress made in nature scene images, it is problematic to directly apply existing hashing methods to object classification in very high resolution (VHR) remote sensing images because they didn't consider the problem of object rotation variations. To address this problem, this paper proposes a novel method called Rotation-invariant Discrete Hashing (RIDISH), which jointly learns a discrete binary generation and rotation-invariant optimization model in the hashing learning framework. Experimental evaluations on a publicly available VHR remote sensing dataset demonstrate the effectiveness of proposed method.