Scale, Rotation and Pose Invariant 3D Object Detection using Log-polar Transform and Singular Value Decomposition

摘要：

This paper proposed a novel biologically motivated 3D object detection algorithm which works robustly in the presence of large scale changes, rotation, occlusion, and pose changes. The robustness to scale and rotation is achieved through log-polar transformation procedure. Singular value decomposition (SVD) in conjunction with the distance-tofeature space measure is used to handle the pose variations and reduce the computational complexity. The system was evaluated using 3D object with varying scale and pose on different complex backgrounds. The experimental results show that the object detector was quite robust to all these variations.

作者: Ren xianyi Dan tangren Ding tianhuai Liu yan Zhang wei Zhang guilin Zhang tianxu

作者单位: RITS Optomechatronics Key Lab Research Institute of Tsinghua University in Shenzhen Shenzhen, Guangd Institute of Image recognition and artificial intelligence Huazhong University of Sci & Tech Wuhan,

会议类型: 国际会议

会议名称: 2006 International Conference on Communications,Circuits and Systems(第四届国际通信、电路与系统学术会议)

会议地点: 广西桂林

会议语种:英文

页码: 480-484

在线出版日期: 2006-06-25（万方平台首次上网日期，不代表论文的发表时间）