2018/09/28 by B Kiran, Luis Roldão, Kiran, B Ravi +16 · 1 citation
Computer Science · Engineering · Environmental Science · Physics and Astronomy · #Advanced Neural Network Applications #Advanced Optical Sensing Technologies #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Remote Sensing and LiDAR Applications #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1809.11036
openalex publication_date 2018/09/28 · openalex created_date 2022/08/02 · openalex updated_date 2026/07/28
Lidar has become an essential sensor for autonomous driving as it provides\nreliable depth estimation. Lidar is also the primary sensor used in building 3D\nmaps which can be used even in the case of low-cost systems which do not use\nLidar. Computation on Lidar point clouds is intensive as it requires processing\nof millions of points per second. Additionally there are many subsequent tasks\nsuch as clustering, detection, tracking and classification which makes\nreal-time execution challenging. In this paper, we discuss real-time dynamic\nobject detection algorithms which leverages previously mapped Lidar point\nclouds to reduce processing. The prior 3D maps provide a static background\nmodel and we formulate dynamic object detection as a background subtraction\nproblem. Computation and modeling challenges in the mapping and online\nexecution pipeline are described. We propose a rejection cascade architecture\nto subtract road regions and other 3D regions separately. We implemented an\ninitial version of our proposed algorithm and evaluated the accuracy on CARLA\nsimulator.\n