With the development of light-field acquisition technology from two-dimensional (2-D) to three-dimensional (3-D) sensors, point clouds are currently being used in the 3-D reconstruction of light-field imaging. The computational requirements of point cloud processing decrease the efficiency of 3-D object recognition in a light field reconstruction. An optimization strategy is proposed to improve the efficiency of object feature recognition in a 3-D light field reconstruction. The proposed method involves a normal estimation, uniform keypoint sampling, random Monte Carlo sampling, signature of histograms of orientations descriptor extraction, k-dimensional tree matching, and geometric consistency clustering estimation. During the experiments, all scenarios corresponding to each model are tested, 2711 times in three virtual and real international standard databases (i.e., Kinect, Mian, and Clutter). The experimental results indicate that the efficiency of the proposed method is improved by 9.26% on an average with the same accurate recognition rate of 84.67%. |
ACCESS THE FULL ARTICLE
No SPIE Account? Create one
CITATIONS
Cited by 3 scholarly publications.
Clouds
Object recognition
3D displays
3D image processing
3D modeling
3D acquisition
Data modeling