Online Program

Adaptive 3D Segmentation of LiDAR Data for Object Detection and Localization for Autonomous Vehicles and Robots (304050)

Keywords: LiDAR, Segmentation, Big Data Processing and Modeling, Autonomous Driving, Object Detection, Adaptive Parameter

Light Detection And Ranging (LiDAR) is increasingly being used for object detection by autonomous vehicles. An essential step in LiDAR data processing is segmentation. Segmentation methods identify clusters corresponding to different objects (cars, cyclists, pedestrians, poles, trees etc.) in LiDAR data. However the greatest challenge in segmenting LiDAR data, is the uneven distribution of the LiDAR returns in the 3D space. As distance increases, the number of LiDAR returns decrease significantly posing a great challenge to any segmentation algorithm. Segmentation algorithms require user inputs defining a cluster: minimum number of signal return points (k) that should be detected within a neighborhood radius of the neighborhood (Eps). However, since the LiDAR data distribution changes with range, using a fixed k or Eps values do not provide satisfactory results at all ranges. In this paper, we present a method to compute both k and Eps parameters adaptively for each range, thus increasing detection accuracy by ~30%. Furthermore, the proposed method also takes into consideration the LiDAR data distribution, thus providing optimal solution for all scenarios and ranges from the car.