Abstract: This paper proposes a novel algorithm for target detection, which is based on motion and region information. A fast optical flow estimation method and a region growing and merging technique are utilized to estimate the motion vector and the region information of the target respectively. Unlike other algorithms the method makes good use of the motion and region information, as a result the computational complexity of optical flow estimation and region segmentation method decreases distinctly. Simulation results show the algorithm achieves a good performance in most complex outdoor environments.

Key words: optical flow estimation, region segmentation, mobile robot, target detection

摘要： 介绍了一种在运动相机条件下的基于目标运动与区域灰度信息的运动目标检测算法。采用快速光流估计算法与基于颜色粗糙度的区域分割方法进行目标运动与区域信息的运算，利用待跟踪目标时间与空间信息进行目标定位，降低了目标运动与区域信息估算的复杂度。仿真结果表明，该算法在大多数复杂场景中能够获得良好的目标识别效果。

关键词: 光流估计, 区域分割, 移动机器人, 目标检测

CLC Number: 

• TP18