Circular 3D Feature Descriptor for RGB-D Images
Jun Qiu, Lina Wu, Chang Liu · The Japan Society of Applied Physics · 2018
The complementarity between the color information and the geometry information promotes the application of RGB-D images in object detection, object recognition and classification feature description and 3D reconstruction. The feature vectors generated by the conventional RGB-D feature descriptor have high dimensionality and computational complexity. We proposed a circular descriptor of the 3D feature points of the RGB-D image by utilizing the rotation invariance.