2019/04/18 by Christoph Schöller, Schöller, Christoph, Maximilian Schnettler +11 · 4 citations
Computer Science · Engineering · #Advanced Neural Network Applications #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Image and Object Detection Techniques #Robotics and Sensor-Based Localization
paper · pdf · doi:10.48550/arxiv.1904.08743
openalex publication_date 2019/04/18 · openalex created_date 2020/10/01 · openalex updated_date 2026/07/28
Most intelligent transportation systems use a combination of radar sensors\nand cameras for robust vehicle perception. The calibration of these\nheterogeneous sensor types in an automatic fashion during system operation is\nchallenging due to differing physical measurement principles and the high\nsparsity of traffic radars. We propose - to the best of our knowledge - the\nfirst data-driven method for automatic rotational radar-camera calibration\nwithout dedicated calibration targets. Our approach is based on a coarse and a\nfine convolutional neural network. We employ a boosting-inspired training\nalgorithm, where we train the fine network on the residual error of the coarse\nnetwork. Due to the unavailability of public datasets combining radar and\ncamera measurements, we recorded our own real-world data. We demonstrate that\nour method is able to reach precise and robust sensor registration and show its\ngeneralization capabilities to different sensor alignments and perspectives.\n