Next Best View Planning in Table-top Scenarios

In this work, we propose an active scene perception system through next best view planning for robots operating in table-top scenarios. Viewpoint evaluation, candidate viewpoint generation and next best view planning steps are implemented using a voxel based method for the proposed system. The performance of the system is analyzed in simulation experiments, and the system is shown to be applicable in real time environments. The proposed best n-view local search strategy provides the highest performance.