基于GPS和视觉同步定位与建图的无人驾驶车辆定位方法研究  

Research on Localization Method for Unmanned Vehicle Based on GPS and VSLAM

作　　者：高扬[1] 徐永贵 刘江[1] Gao Yang;Xu Yonggui;Liu Jiang(Chang’an University,Xi’an 710054)

机构地区：[1]长安大学,西安710054

出　　处：《汽车技术》2021年第6期19-24,共6页Automobile Technology

基　　金：长安大学研究生科研创新实践项目(300103703031)。

摘　　要：为提高无人驾驶车辆的定位精度,提出一种基于GPS与视觉同步定位与建图(VSLAM)相融合的定位算法,在VINS-Mono基础上进行改进,提出了VINS-FAST算法进行角点提取,并将VINS-FAST与GPS技术相结合,提出了GPS与VSLAM相融合的VINS-GPS算法。试验结果表明:VINS-GPS在KITTI数据集07和09子集上的均方根误差较VINS-Mono分别降低了18.16%和33.08%,改善了VSLAM在大范围环境中表现较差的问题,同时提高了在GPS信号薄弱或无GPS信号环境中的定位精度。In order to improve the localization accuracy of unmanned vehicles,a localization algorithm is proposed based on GPS and Visual Simultaneous Localization and Mapping(VSLAM)fusion.To improve the algorithm based on VINS-Mono,a corner extraction algorithm is proposed,namely the VINS-FAST.On this basis,through the VINS-FAST combined with GPS technology,a VINS-GPS algorithm that fuses GPS and VSLAM is proposed.The experimental results show that the root mean square error of VINS-GPS in KITTI 07 data set and 09 subsets is reduced by 18.16%and 33.08%respectively compared with VINS-Mono,which improves the performance of VSLAM in a wide range of environments,and also improves the localization accuracy in the weak or no GPS signal environment.

关 键 词：无人驾驶车辆 视觉同步定位与建图 角点均匀分布策略 GPS 

分 类 号：TP24.6[自动化与计算机技术—检测技术与自动化装置]