基于融合改进RANASC光流法的无人机视觉SLAM研究

Research on optimization algorithm based on UAV SLAM visual odometer

摘要:

针对同时定位与建图(SLAM)中存在定位精度不足、匹配特征点误差累积和特征匹配时间较长的问题，提出了一种融合改进RANSAC光流法的优化算法。该方法基于传统RANSAC算法，加入最小二乘法对模型迭代优化来估计最优模型，对光流法的误匹配点进行剔除，减少大量图像误匹配特征点；把融合改进后的RANSAC光流法与特征点通过卡尔曼滤波进行融合，最后使用改进后的算法在公开的EuRoC MAV数据集中进行SLAM定位精度实验。实验结果表明：本文的改进算法能够有效减小光流法特征匹配的误差，从而提高无人机视觉SLAM的定位精度，证明了改进算法的有效性与可行性。

Abstract:

In order to solve the problems in simultaneous localization and mapping (SLAM), such as insufficient localization accuracy, accumulation of error of matching feature points and long matching time of feature points, a fusion and improved RANSAC optical flow optimization algorithm was proposed. Based on the traditional RANSAC algorithm, the least square method was added to the model iterative optimization to estimate the optimal model, and the mismatching points of optical flow method were removed to reduce a large number of image mismatching feature points. The improved RANSAC optical flow method was fused with the feature points through Kalman filtering. Finally, the improved algorithm was used to perform SLAM localization accuracy experiments in the open EuRoC MAV data set. Experimental results show that the improved algorithm in this paper can effectively reduce the feature matching error of optical flow method, thus improving the positioning accuracy of UAV visual SLAM, which proves the effectiveness and feasibility of the improved algorithm.