Object Detection is an essential task in autonomous driving. The existing object detection methods are susceptible to the variable lighting, which leads to the decrease of detection accuracy. To address the above issue, we introduce LRPN, a robust detection network with multi-modal feature fusion based on prior knowledge of lidar points. LRPN aims to improve the accuracy of object detection in light changing scenes by using fusion method of sparse point cloud and image. We propose two fusion modules: 1) Sparse lidar feature are fused with image in the feature extractor network. 2) Guided anchors are created based on visual projection using lidar points. Our method is applied to Faster RCNN and achieved higher precision under different illumination conditions. Compared with the original Faster RCNN network, the mAP of the method proposed in this paper is raised by 2.2 % under good illumination and 3.7% under low illumination. The results show that the introduction of sparse point cloud can significantly improve the detection effect, especially in low illumination scenarios.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    An Object Detection Method Enhanced by Sparse Point Cloud for Low Illumination in Autonomous Driving


    Contributors:
    Li, Shuguang (author) / Liu, Bei (author) / Zhao, Yang (author) / Zheng, Ke (author) / Cheng, Hong (author)


    Publication date :

    2022-10-08


    Size :

    1662281 byte




    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    BEVDetNet: Bird's Eye View LiDAR Point Cloud based Real-time 3D Object Detection for Autonomous Driving

    Mohapatra, Sambit / Yogamani, Senthil / Gotzig, Heinrich et al. | IEEE | 2021


    3D Object Detection for Autonomous Driving

    Tan, Yihua / Chen, Siwei / Yan, Pei | Springer Verlag | 2022


    3D LIDAR point cloud based intersection recognition for autonomous driving

    Zhu, Quanwen / Chen, Long / Li, Qingquan et al. | IEEE | 2012


    Lidar Point Cloud Compression, Processing and Learning for Autonomous Driving

    Abbasi, Rashid / Bashir, Ali Kashif / Alyamani, Hasan J. et al. | IEEE | 2023