A novel keypoint detection technique is proposed.
Highly repeatable keypoints are detected by exploiting 3D vector field’s divergence.
A local surface descriptor (3D-Vor) is also introduced forsurface representation.
The proposed 3D-Vor exploits the vector field׳s vorticity.
A novel 3D object recognition algorithm is also proposed.
Proposed technique is tested on 3 popular 3D object recognition datasets.
Proposed technique achieves superior recognition rates on these 3D object datasets.