摘要
提出了一种应用单目视觉进行车辆检测的方法。该方法以车辆阴影以及边缘作为检测的主要特征。在图像预处理中采用自适应双阈值以满足不同光照条件下的使用要求;利用能量密度验证提高车辆垂向边界识别的准确性;用可变模型用来规划车辆检测范围,以满足不同距离车辆的检测需要。为提高车辆检测的准确性及效率,在算法中融合了雷达的探测数据。试验验证表明,该方法有较高的车辆检测准确率,并且能满足智能车应用中的实时性要求。
A vehicle detection method using monocular vision is proposed. The shadow under vehicle and the edges of vehicles are chosen as the main features for detection. The adaptive double thresholding is used for shadow detection. The energy density validation is adopted to improve the accuracy of vehicle vertical edge detection. The deformable model is introduced to plan the vehicle detection range for meeting the detection requirements of vehicles at different distances. Radar detection data is also fused into the algorithm to improve the accuracy and speed of vehicle detection. Experiments show that the method has high accuracy rate of vehicle detection and the algorithm can meet the real-time requirements for its application to intelligent vehicles.
出处
《汽车工程》
EI
CSCD
北大核心
2006年第11期1031-1035,共5页
Automotive Engineering
关键词
车辆检测
机器视觉
雷达
基于特征
Vehicle detection, Machine vision, Radar, Feature-based
