3D Object Detection for Autonomous Driving: A Survey
A comprehensive survey of 3D object detection for autonomous driving, encompassing all the main concerns including sensors, datasets, performance metrics and the recent state-of-the-art detection methods, together with their pros and cons.
Abstract
Autonomous driving is regarded as one of the most promising remedies to\nshield human beings from severe crashes. To this end, 3D object detection\nserves as the core basis of perception stack especially for the sake of path\nplanning, motion prediction, and collision avoidance etc. Taking a quick glance\nat the progress we have made, we attribute challenges to visual appearance\nrecovery in the absence of depth information from images, representation\nlearning from partially occluded unstructured point clouds, and semantic\nalignments over heterogeneous features from cross modalities. Despite existing\nefforts, 3D object detection for autonomous driving is still in its infancy.\nRecently, a large body of literature have been investigated to address this 3D\nvision task. Nevertheless, few investigations have looked into collecting and\nstructuring this growing knowledge. We therefore aim to fill this gap in a\ncomprehensive survey, encompassing all the main concerns including sensors,\ndatasets, performance metrics and the recent state-of-the-art detection\nmethods, together with their pros and cons. Furthermore, we provide\nquantitative comparisons with the state of the art. A case study on fifteen\nselected representative methods is presented, involved with runtime analysis,\nerror analysis, and robustness analysis. Finally, we provide concluding remarks\nafter an in-depth analysis of the surveyed works and identify promising\ndirections for future work.\n