面向未知区域深度测量的序列图像稠密点特征生成算法

对未知着降区平坦度测量是无人机在复杂地形下安全着陆的关键问题。首先,根据小孔成像原理推导出基于单目序列图像的未知区域深度计算方程;其次,针对稀疏匹配存在深度信息重构误差大而稠密匹配在平滑区域误匹配率高的问题,提出一种基于Delaunay三角剖分的稠密点特征生成算法;然后,分别对序列图像中的2帧图像提取亚像素级Harris角点和尺度不变特征变换(SIFT)特征点,并分别进行特征点匹配;再以2种特征点间的欧氏距离作为约束条件将2种特征点进行融合,生成准稠密特征点;最后,将准稠密特征点进行Delaunay三角剖分,并根据每个剖分三角形上3个顶点像素偏差的方差值制定稠密特征点的生成策略,并结合所提出的深度计算方程计算整个未知区域各点的深度信息。通过Vega Prime(VP)搭建仿真演示验证系统,实验结果表明在机载相机距地面400m处计算高度分别为90m和55m的物体深度信息时,其深度测量相对误差不超过0.89%,具有较高的精度。

Dense point feature generation algorithm based on monocular sequence images for depth measurement of unknown zone

It is essential to measure the flatness of an unknown zone for UAV landing in a complex terrain. Firstly, a depth calculation equation based on monocular sequence images is derived according to the pinhole imaging principle. Secondly, a dense point feature generation algorithm based on Delaunay triangulation is proposed to solve the problem that the large error of depth information reconstruction exists in sparse matching and the problem that high false match rate based on dense matching is high in the smooth region. Then, sub pixel Harris corner and scale invariant feature transform (SIFT) feature points are extracted and matched respectively in two frames which are selected from sequence images. After that, the two type feature points are fused under the conditions of Euclidean distance between them. So quasi dense feature points can be obtained. Finally, quasi dense feature points are Delaunay triangulated and dense feature points generating strategy is developed according to the variance of the three vertex pixel deviation in each triangulation triangle. Depth information of the whole unknown zone is calculated according to the proposed depth calculation equation. A simulation demonstration system is built by Vega Prime (VP) simulation and experimental results show that the relative depth measurement error of two objects whose height are 90 m and 55 m are less than 0.89% when the airborne camera is 400 m above the ground. The experimental results verify that the proposed algorithm has high accuracy.