基于三维模型的多传感车辆定位与识别研究

车辆的定位与识别是智能车研究的重要环节。针对二维算法及单传感的不足,提出了一种基于三维模型的多传感车辆定位与识别算法。定位系统中,通过将雷达的位置估计和图像的灰度信息进行融合来获得位移参数,并将改进的Hausdorff距离和模拟退火算法相结合得到旋转参数。在此基础上,识别系统计算每个模型与目标车辆的匹配值,根据其在采样时段内的平均值确定车辆类型。实验证明,该方法能有效地实现目标车辆的定位和识别。

3-D model-based multi-sensor vehicle location and recognition

The location and recognition of vehicle is one of the principle problems in the intelligent vehicle research.The two dimensional-based and single sensor-based methods have their defects.To enhance system performance,a new vehicle location and recognition method based on three-dimensional(3-D) model and multi-sensor fusion is proposed.Translation parameter of vehicle location is obtained by fusing radar-based position evaluation data and image intensity data,and rotation parameter of location is obtained by integrating improved Hausdorff distance and simulated annealing algorithm.Recognition is then achieved by computing the average optimal matching between each model projection and target image during sampling time.The experiments confirm the validity of the proposed algorithm.