单目视觉行人身体多部件距离融合估计算法

针对现有自主汽车辅助驾驶系统中,单目摄像头行人距离估计精度不高的问题,提出一种单目视觉行人身体多部件距离融合计算法。联合脚部件与头部件共同对行人目标进行距离估计,分别采用半水波形等距线脚部部件距离估计与头部部件高度插值距离估计算法来获取行人的距离估计结果,使得算法能够同时具备精准性和稳定性。经仿真验证,该算法在非遮挡条件下,35 m内能够将误差控制在1 m以内,在半遮挡条件下,35 m内能够将误差控制在2 m以内,满足车载辅助驾驶对行人距离估计精度的要求。

Multiple Body Parts Distance Fusion Estimation Algorithm Based on Monocular Vision

This paper presents a distance fusion estimation algorithm based on multiple body parts, which can solve the problem of low accuracy of pedestrian distance estimations with vehicle-mounted monocular camera. The distances of head parts and foot parts are used to calculate the actual distance between the pedestrian and vehicle. In this algorithm, the distances of these two parts are fused by the distance estimation method of half water wave isometric line and the estimation method of height interpolation distance for head harts, which make our algorithm a good performance of accuracy and stability. In simulation, the error can be controlled within 1m in 35m under non-occlusion conditions. The error can be controlled within 2m in 35m under semi-occlusion conditions. The experiment shows that the error satisfies the accuracy of pedestrian distance estimation in vehicle assisted driving.