基于摄像机与激光雷达的车辆识别技术

Technology of Vehicle Identification Based on Camera and Laser Radar

提出了一种基于摄像机与激光雷达的车辆识别方法;该方法首先建立一个ROI(感兴趣的区域),对该区域内的雷达数据进行滤波及聚类处理;聚类后通过计算类的宽度排除一部分不可能是车辆的障碍物,再通过坐标转换得到可能是车辆的障碍物在图像中的位置,然后通过图像处理的方法进行识别;对于图像中可能是车辆的区域首先进行边缘检测,确定了车辆边缘后通过计算车辆区域的信息熵对障碍物进行车辆验证;实验表明,该方法能够有效地检测前方目标车辆,且每一帧数据的处理时间在8ms～17ms之间满足实时性的要求。

this paper proposes a Vehicle Identification approach by camera and two--dimensional laser radar. The approach we proposed uses the result of laser radar to instruct camera. Initially, the approach filters and clusters the laser radar data in a ROI （region of interest）. After filtering and clustering, exclude the obstacles which can not be vehicles by calculating width of the clusters. Then verify the obstacles by vehicle＇s image features. The approach of verify the obstacles is detecting edge and calculating the entropy of the obstacle region in image. Experimental results show that this approach can recognize vehicle effectively.