双动态条件下多传感器融合的车辆检测方法研究

Research on Vehicle Detection Method of Multi-Sensor Under Dual Dynamic Condition

运用双动态条件研究追尾事故中车辆防撞预警方法,提出一种基于毫米波雷达和PTZ摄像机相结合的前方车辆检测方法。解决摄像机、雷达、世界坐标系的变换关系,实现空间上的融合;采用线程同步方式实现时间上的融合。目标识别过程中,在图像中生成感兴趣区域,通过三阶Kalman滤波算法对当前周期内获取的初始目标信息进行预测,基于已训练好的级联分类器进行多尺度识别,得到感兴趣区域内的车辆。将单目测距获得的距离与毫米波雷达获得的实际距离进行比对,以验证上述算法的准确性,试验结果表明,测量距离与实际距离误差率为1.96%,可控制在用户可接受的区间内。

Presents a method of vehicle detection based on millimeter wave radar and PTZ camera,which is using double dynamic condition to study vehicle crash warning method.Solve the camera,radar,world coordinate system transformation relationship,to achieve spatial fusion;Using the thread synchronization way to achieve the time fusion.In the target recognition process,the region of interest is generated in the image,the third-order Kalman filter algorithm is used to predict the initial target information obtained in the current cycle.Comparing the predictive value of effective vehicle information of(n+1)th cycle with primary vehicle status to verify the effectiveness of the target.Based on the trained cascade classifier for multi-scale recognition to get the vehicle in the area of interest.The distance obtained from monocular ranging is compared with the actual distance obtained by millimeter-wave radar to verify the accuracy of the above algorithm.The experimental results show that the error ratio between measurement distance and actual distance is 1.96%,which can be controlled within the acceptable range of the user.