【作者】 李忠育； 段中兴；

【Author】 Li Zhongyu;Duan Zhongxing;College of Information and Control Engineering Xi’an University of Architecture and Technology;

【机构】 西安建筑科技大学信息与控制工程学院；

【摘要】 毫米波雷达与立体双目相机的融合对于自动驾驶系统中的目标跟踪、定位和路径规划至关重要。毫米波雷达具有较高的距离分辨率,但方位分辨率较低。而立体双目相机能够提供精确的方位测量和较高的目标检测率,但在目标距离估计方面存在困难。为了解决这一问题,本文提出了一种融合系统,该系统结合了毫米波雷达和立体双目相机。该融合系统利用时空校准技术处理雷达和立体双目相机的数据,实现有效的目标匹配。提出的融合系统利用毫米波雷达获取的目标纵向位置来推断目标在立体双目相机下的位置数据。随后,本文提出了一种改进的基于证据决策距离自适应滤波的方法,并结合均值滤波和最优加权滤波方法对毫米波雷达和立体双目相机的位置信息进行融合,从而有效提升目标的横向定位精度。统计实验结果表明,与毫米波雷达的横向定位均方根误差(RMSE)相比,均值滤波和最优加权滤波分别降低了横向定位测量RMSE 31.9%和32.9%,而所提出的滤波方法使横向定位RMSE下降了35.2%。该融合方法在精确目标定位方面具有良好的应用前景。更多还原

【Abstract】 The fusion of millimeter-wave radar and camera sensors is critical for target tracking,localization and path planning in autonomous driving systems.Millimeter-wave radar has high distance resolution but low azimuthal resolution.While stereo cameras can provide accurate orientation measurements and high target detection rate,they have difficulties in target distance estimation.To solve this problem,this paper proposes a fusion system that combines a millimeter-wave radar and a stereo camera.The fusion system utilizes a spatio-temporal calibration technique to process the data from the radar and stereo camera to achieve effective target matching.The proposed fusion system utilizes the longitudinal position of the target acquired by the millimeter-wave radar to infer the position data of the target under the stereo camera.Subsequently,this paper proposes an improved adaptive filtering method based on evidence-based decision distance and combines mean filtering and optimal weighted filtering methods to fuse the position information of millimeter-wave radar and stereo camera,so as to effectively improve the target’s lateral positioning accuracy.Statistical experimental results show that compared with the root mean square error(RMSE) of lateral localization of millimeter-wave radar,mean filtering and optimal weighted filtering reduce the lateral localization measurement RMSE by 31.9% and 32.9%,respectively,while the proposed filtering method reduces the lateral localization RMSE by 35.2%.The fusion method has good application prospects in precise target localization.更多还原