“…However, with the development of feature extraction methods for RGB-D data, especially the development of deep learning-based feature extraction methods, storage cost for these features increases dramatically (Gupta et al, 2014;Zeng et al, 2018, Schwarz et al, 2015Wu et al, 2016 ). Hence, to meet the requirements of high-speed computation for robotic applications, some researchers use hashing learning methods to the research field of robot vision (Yu et al, 2016;Chen et al, 2013). Hashing learning aims at representing original data with binary codes, which can improve the efficiency of data analysis (Weiss et al, 2008;Gong et al, 2013;Liu et al, 2018).…”