单目视觉的前车位置识别及安全车距监测

针对自动驾驶中的前车位置识别及安全车距监测问题,提出了基于单目视觉的解决方案:首先,通过图像中前车轮廓与本车道及同向车行道分界线是否相交判断前车是否为监测车辆;其次,根据图像中像素坐标系与空间坐标系的对应关系构建转换模型,计算前方监测车辆与本车的真实车距;最后,根据可跨越的同向车行道分界线在连续图像帧中的位置变化计算本车的行驶速度,并依据交通法规判断本车与前方监测车辆之间是否为安全距离.实验结果表明,该解决方案可以有效地识别前车位置、监测安全车距并降低系统成本.

Position recognition of the vehicle ahead and safety distance monitoring based on monocular vision

A solution based on monocular vision is proposed to recognize the position of the vehicle ahead and to monitor the safe distance in automatic driving. Firstly, whether the vehicle ahead is a monitoring vehicle can be judged by whether the contour of the vehicle ahead intersects with the roadway and the dividing line of the roadway. Secondly, according to the corresponding relationship between pixel coordinate system and spatial coordinate system in the image, the transformation model is built to calculate the real distance from the monitored vehicle. Finally, the speed of this vehicle is calculated according to the position change of the crossable delimitation line of the same direction roadway in the continuous image frames, and the safety distance between this vehicle and the monitored vehicle is determined according to the traffic laws and regulations. Experimental results show that this solution can effectively recognize the position of the vehicle ahead, monitor the safety distance and reduce the system cost.