自主车导航中提高目标行为预测能力的新方法

Improved Objects Motion Predication for Autonomous Vehicle Navigation

鉴于激光雷达目标识别能力较弱的问题,首先利用视觉较强的分类能力,采用协同学习粒子滤波算法实现前方车体的识别和跟踪;再将图像中车体特征区域坐标信息转换到激光雷达图像坐标系中,在线快速提取目标车体距离信息序列,再融合里程计和GPS信息,实现自主车前方车体的行为预测;实验表明,所提出的方法能够在激光雷达信息中快速准确地定位车辆距离信息,并实时估计自主车前方车辆的行驶趋势,得到的目标速度计算误差约为6.2%,使导航更趋合理。

Laser Scanner may provide excellent range information to different objects. However, it is difficult to recognize these objects as vehicles from range information alone. Therefore, an improved system intergrating the vision classification to compensate the Laser is pro- posed in this paper. Firstly, an vehicle tracking algorithm based on co--learn particle filter is presented using the vision. Then the vehicles are detected in the vision image, and this vehicles coordinates information is transformed into the Laser coordinates. Finally, the distance in- formation series of the vehicles are saved with the time, and the motion predication is derived. Experimental results are presented to illustrate that the motion predication system is reliable （ the velocity error is about 6. 2%）. It can be used in applications such as traffic surveillance and roadway navigation tasks.