Authors:
Mehmet Ali Çagri Tuncer
and
Dirk Schulz
Affiliation:
Cognitive Mobile Systems, Fraunhofer FKIE, Fraunhoferstr. 20, 53343 Wachtberg and Germany
Keyword(s):
Segmentation, Classification, Support Vector Machines (SVM), Distance Dependent Chinese Restaurant Process, 3D Lidar Data.
Related
Ontology
Subjects/Areas/Topics:
Image Processing
;
Informatics in Control, Automation and Robotics
;
Perception and Awareness
;
Robotics and Automation
Abstract:
This paper proposes a novel framework for the segmentation and classification of 3D point cloud which jointly uses spatial, temporal and semantic information. It improves the classification performance by reducing under-segmentation errors. The presented framework, which can determine the number and label of objects in each spatially extracted blob, is decomposed into three steps to acquire spatial, temporal and semantic cues. For the spatial features, blobs are extracted spatially with a neighborhood system on an occupancy grid representation. A smoothed motion field is estimated for the acquisition of temporal cue, where the grid cells are tracked using individual Kalman filters and estimated velocities are transformed to one dimensional movement directions. A support vector machine (SVM) classifier is trained to discriminate the classes of interest for the semantic information of the blobs. A confidence metric is defined to probabilistically compare the volume of each classified b
lob with the volume of an average object for that class. If this metric is below a predefined threshold, a sequential variant of distance dependent Chinese restaurant process (s-ddCRP) performs the final partition in this blob by using spatial and temporal information. If the s-ddCRP approach splits the blob, the partitioned sub-blobs are afterwards reassigned to new objects by the classifier. Otherwise, the queried blob remains the same. This procedure iteratively continues while searching each blob in the scene at each time frame. Experiments on data obtained with a Velodyne HDL64 scanner in real traffic scenarios illustrate that the proposed framework improves the classification performance of an SVM classifier by reducing under-segmentation errors.
(More)