基于激光测距雷达和机器视觉车载相机,对智能车道路的障碍物检测进行了研究。通过分析,选用激光雷达及CMOS车载相机,标定了雷达和相机,为改进低照度环境下提取图像存在的缺陷,制定了Retinex算法来增强低照度图像。因两种传感器频率不同,结合D-S证据理论,将车载相机和激光雷达数据进行融合,对真实环境中的行人和车辆信息进行准确识别,融合相关数据后可以发现,与单一传感器相比,该系统检测的概率更高。在验证过程中,通过比较数据融合前后智能车避障轨迹开始点及障碍物间距离,相比于障碍物间距离,第六个目标避障中数据融合后避障轨迹开始点距离最小,第四个目标避障中提前距离最大,验证了传感器数据融合后对障碍物检测更有效、及时。

Based on lidar and machine vision vehicle camera, the obstacle detection of intelligent vehicle road is studied. Through the analysis of lidar and CMOS vehicle camera, the radar and camera are calibrated. In order to improve the defects of image extraction in low illumination environment, Retinex algorithm is developed to enhance the low illumination image. Because of the different frequency of two sensors, combined with the D-S evidence theory, the vehicle camera and lidar data are fused to accurately identify the pedestrian and vehicle information in the real environment. After fusing the relevant data, it can be found that the detection probability of the system is higher than that of a single sensor. In the verification process, by comparing the start point of obstacle avoidance trajectory and the distance between obstacles before and after data fusion, compared with the distance between obstacles before fusion, the distance between the start point of obstacle avoidance trajectory of the sixth target after data fusion is the smallest, the distance in advance of the fourth target is the largest, verifying that the obstacle detection is more effective and timely after sensor data fusion.

TN957.51

河南省科技攻关项目;河南省产学研合作项目;河南省高等学校青年骨干教师培养计划项目