2017/10/17 by Damien Matti, Matti, Damien, Hazım Kemal Ekenel +3
Computer Science · Engineering · #Advanced Neural Network Applications #Autonomous Vehicle Technology and Safety #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Video Surveillance and Tracking Methods
paper · pdf · doi:10.48550/arxiv.1710.06160
openalex publication_date 2017/10/17 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Pedestrian detection is an important component for safety of autonomous\nvehicles, as well as for traffic and street surveillance. There are extensive\nbenchmarks on this topic and it has been shown to be a challenging problem when\napplied on real use-case scenarios. In purely image-based pedestrian detection\napproaches, the state-of-the-art results have been achieved with convolutional\nneural networks (CNN) and surprisingly few detection frameworks have been built\nupon multi-cue approaches. In this work, we develop a new pedestrian detector\nfor autonomous vehicles that exploits LiDAR data, in addition to visual\ninformation. In the proposed approach, LiDAR data is utilized to generate\nregion proposals by processing the three dimensional point cloud that it\nprovides. These candidate regions are then further processed by a\nstate-of-the-art CNN classifier that we have fine-tuned for pedestrian\ndetection. We have extensively evaluated the proposed detection process on the\nKITTI dataset. The experimental results show that the proposed LiDAR space\nclustering approach provides a very efficient way of generating region\nproposals leading to higher recall rates and fewer misses for pedestrian\ndetection. This indicates that LiDAR data can provide auxiliary information for\nCNN-based approaches.\n