基于MHT模型的毫米波雷达车辆检测方法

为了在智能车辆系统中检测前方车辆,该文提出了一种基于多假设跟踪(Multiplehypothesis tracking,MHT)模型的车辆检测方法。首先在多假设跟踪模型下,定义毫米波雷达量测集合与目标集合的对应关系,采用广义概率数据关联算法提取量测集合中的有效目标,从而得到有效目标集合。利用概率树模型估算目标的出现概率来维护检测的目标集合,保留稳定的检测结果。实验结果表明:该方法对于远近距离下和较差环境下黑夜灯光的前方车辆达到了准确的检测效果,克服了基于视觉的车辆检测方法对目标距离和环境光线敏感的缺点,同时在目标的保持维护上也取得了良好的效果。

Vehicle Detection Method Based on MHT Model Using Millimeter-wave Radar

To detect vehicles in intelligent vehicle system,a vehicle detection method based on multiple hypothesis tracking(MHT)is proposed.The relationship between measurement sets collected by millimeter-wave radar and target set is defined under the framework of MHT.A generalized probability data association(GPDA)algorithm is used to find the acceptable target set from all the measurement data sets.After that,a probability tree is designed to maintain the target sets.Results prove that this method can accurately detect the vehicle ahead far or near or that in dark night with a poor light environment,which overcomes the shortcoming of the vision-based vehicle detection method that is sensitive to target ’ s distance and ambient light.Besides,this method also achieves good results in the maintenance of targets.