针对在无人机视觉导航过程中光流信息易受前景运动物体与噪声影响造成测速精度下降的问题, 基于Mean-Shift算法提出了一种光流测速优化方法。该方法使用Lucas-Kanade光流算法计算出稀疏光流后, 采用改进的Mean-Shift算法对光流数据进行聚类处理, 根据速度的一致性剔除光流数据中的干扰项。实验结果表明, 该优化方法可以有效地减少前景运动物体和噪声对无人机光流测速的影响, 提高无人机测速的精度。

Considering that the velocimetry accuracy will decline when the optical flow information is affected by foreground moving objects and noise in UAV visual navigation, we proposed an optical flow velocimetry optimization method based on the Mean-Shift algorithm.The method uses the pyramid Lucas-Kanade algorithm to calculate the sparse optical flow and the improved Mean-Shift algorithm for the clustering of the optical flow data, which eliminates the interference items in the data according to the consistency of speed.Experimental results show that the proposed method can reduce the effects of foreground moving objects and noise on optical flow velocimetry in UAV visual navigation, and improve the precision of UAV velocity measurement.