Adaptive Object Learning for Robot Carinet

Shengtao Xiao, Shuzhi Sam Ge, Shuicheng YAN

Abstract: In this paper, an adaptive object learning method based on deep neural network is developed for a robot to learn features of moving objects, e.g., humans and vehicles, via observation. The proposed method provides a solution for the robot to learn unknown moving objects in a real-time scenario. A hybrid scheme of learning and identification is proposed to recognize the moving object by fusion of foreground segmentation and identification.