基于车载双目立体视觉的运动障碍检测

Surrounding moving obstacles detection using onboard binocular stereo vision

提出一种基于多视几何约束的车载双目立体视觉运动障碍检测算法。研究结果表明:多视极线约束有效地解决了当相机和障碍同方向运动时对极几何无法检测运动障碍的问题。在粒子滤波框架下,以多视极线约束为状态观测,可同时检测和跟踪多个运动障碍。不同场景实际数据集上的实验结果证明:该方法具有鲁棒性和有效性。

A multiview epipolar constraint based algorithm was proposed for moving obstacles detection using onboard binocular stereo vision.The results show that a multiview epipolar constraint is derived from the relative camera positions in pairs of consecutive stereo views for independent motion detection.A dimensional variable particle filter for joint detection and tracking of multiple moving obstacles is presented.Experimental results on real-world driving sequences demonstrate that the method is effective and robust.