3D lidar data segmentation using a sequential hybrid method

This chapter proposes a sequential hybrid method for 3D Lidar data segmentation. The presented approach provides more reliable results against the under-segmentation issue, i.e., assigning several objects to one segment, by combining spatial and temporal information to discriminate nearby objects in the data. For instance, it is common for pedestrians to get under-segmented with their neighboring objects. Combining temporal and spatial cues allow us to resolve such ambiguities. After getting the temporal features, we propose a sequential hybrid approach using the mean-shift method and a sequential variant of distance dependent Chinese Restaurant Process (ddCRP). The segmentation blobs are spatially extracted from the scene with a connected components algorithm. Then, as a post-processing, the mean-shift seeks the number of possible objects in the state space of each blob. If the mean-shift algorithm determines an under-segmentation, the sequential ddCRP performs the final partition in this blob. Otherwise, the queried blob remains the same and it is assigned as a segment. Compared to the other recent methods in the literature, our framework significantly reduces the under-segmentation errors while running in real time.