Abstract
In epidemic prevention and control measures, unmanned devices based on autonomous driving technology have stepped into the front lines of epidemic prevention, playing a vital role in epidemic prevention measures such as protective measures detection. Autonomous positioning technology is one of the key technologies of autonomous driving. The realization of high-precision positioning can provide accurate location epidemic prevention services and a refined intelligent management system for the government and citizens. In this paper, we propose an unmanned vehicle (UV) positioning system REW_SLAM based on lidar and stereo camera, which realize real-time online pose estimation of UV by using high-precision lidar pose correction visual positioning data. A six-element extended Kalman filter (6-element EKF) is proposed to fusion lidar and stereo camera sensors information, which retains the second-order Taylor series of observation and state equation, and effectively improves the accuracy of data fusion. Meanwhile, considering improving lidar outputs quality, a modified wavelet denoising method is introduced to preprocess the original data of lidar. Our approach was tested on KITTI datasets and real UV platform, respectively. By comparing with the other two algorithms, the relative pose error and absolute trajectory error of this algorithm are increased by 0.26 m and 2.36 m on average, respectively, while the CPU occupancy rate is increased by 6.685% on average, thereby proving the robustness and effectiveness of the algorithm.