摘要
本文针对三维车辆的定位和识别问题,提出了一种融合雷达和图像信息的新方法.结合雷达的滤波信息和图像的灰度信息建立视觉窗口,确定车辆的平移参数,并缩小了图像处理区域,降低了环境噪声.以改进的Hausdorff距离为依据建立目标的姿态评价函数,通过全局寻优确定车辆的旋转参数,降低了计算损耗,避免了噪声点对模型匹配的影响.车辆识别以定位技术为基础,各模型在最优姿态下的评价函数值决定了车辆的类型.三维仿真场景实验证明,该方法能有效地实现车辆的定位与识别.
A new three-dimensional (3D) vehicle location and recognition method based on radar and image fusion is proposed. A vision window, which is obtained by fusing radar filtered information and image intensity information, is employed to determine the translation parameters of vehicle location and to reduce the image processing area and the environmental noises. The pose evaluation function, whose optimum solution determines the rotation parameter, is established based on improved Hausdorff distance, and by this means the computational efficiency and the matching precision are improved. The vehicle recognition is based on location technology, where the vehicle type is determined by the evaluation value of each model at its optimum pose. Experiments on 3D simulation scenario confirm that the proposed fusion method can effectively realize vehicle location and recognition.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2005年第6期1105-1108,共4页
Acta Electronica Sinica
基金
国家重点基础研究发展规划(973计划)项目(No.2001CB309403)
关键词
多传感融合
目标定位
目标识别
三维模型
Evaluation
Functions
Image processing
Mathematical models
Optimization
Radar target recognition
Sensor data fusion
Three dimensional