An Efficient Technique for Filtering of 3D Cluttered Surfaces

3D Object recognition is a rapidly growing research field in the area of computer vision. The presence of cluttered surfaces is the main obstacle to object recognition. In this paper, we propose to present a technique for automatic filtering of cluttered surfaces. First, the technique clusters a point cloud and then based on three features that are size, distance and spatial information, cluttered surfaces are separated. To the best of our knowledge, this is the first technique that can remove any cluttered surface (plane or irregular) from a point cloud. We have experimented on two complex RGBD datasets containing heavily cluttered surfaces and using the proposed metric, we measure the remaining cluttered surfaces after filtering of a point cloud. The proposed clutter filtering has removed 87.60% and 89.68% cluttered surfaces for Challenge and Willow datasets respectively.