Vehicle Pose Detection and Application Based on Grille Net
Zhenxin Yao, Xinping Song · 2019 3rd International Conference on Electronic Information Technology and Computer Engineering (EITCE) · 2019
With the rapid development of technology, autonomous driving technology has also received widespread attention. Among them, the vehicle's pose information provides important information for automatic driving decision and vehicle active safety. To solve this problem, this paper proposes to use the deep learning method to improve the traditional pose network and use the random area connection network to reduce the difficulty of vehicle orientation detection. In order to improve the performance of the random area connection network, this paper innovatively proposes raster convolution and designs a grid network based on raster convolution. Experiments were performed on the famous autopilot data set KITTI, achieving an average angular similarity of 89.12% and an average similarity of 87.32%. The experimental results show that compared with the traditional pose detection method, the proposed method has better robustness and detection effect.