视觉同步定位与建图中特征点匹配算法优化

Visual Simultaneous Localization and Mapping Feature Point Matching Algorithm Optimization

为提高视觉SLAM图像间特征点匹配的准确性,提高自动驾驶的同步定位与建图精度,在ORB-SLAM2的基础上优化特征点匹配算法。采用由德国卡尔斯鲁厄理工学院和丰田美国技术研究院联合创办的自动驾驶场景KITTI数据集分别对ORB-SLAM2与基于ORB-SLAM2对特征点匹配算法优化进行实验验证。实验表明,优化后的算法生成的轨迹精度相较于ORB-SLAM2有所提高,绝对位姿误差平均值优化了14.29%,相对位姿误差平均值优化了9.67%。优化后的SLAM算法提升了轨迹精度,从而提升了自动驾驶的同步定位与建图的精度。

In order to improve the accuracy of feature point matching between visual SLAM images and improve the accuracy of synchronous positioning and mapping of automatic driving, the feature point matching algorithm is optimized on the basis of ORB-SLAM2. The autonomous driving scenario KITTI dataset jointly established by the Karlsruhe Institute of Technology in Germany and the Toyota American Institute of Technology was used to experimentally verify the optimization of ORB-SLAM2 and the feature point matching algorithm based on ORB-SLAM2. Experiments show that the accuracy of the trajectory generated by the optimized algorithm is improved compared with ORB-SLAM2. The average absolute pose error is optimized by14.29%, and the average relative pose error is optimized by 9.67%. The optimized SLAM algorithm improves the trajectory accuracy, thereby improving the accuracy of synchronized positioning and mapping of autonomous driving.