面向低速清扫车的信息融合车辆跟踪方法
|
摘要
交通参与者运动的准确跟踪与预测对智能车行为决策的有效性至关重要。传统运动目标的跟踪系统多采用单一传感器,难以保证数据的精度与可信度。为提高系统的鲁棒性与可靠性,设计一种融合毫米波雷达和相机的目标跟踪方案,该方案针对多目标的特征级信息进行融合。首先,考虑低速行驶的自动驾驶清扫车所处环境杂波较多,方案选择基于IMM/JPDA的多目标跟踪方法估计局部航迹。为降低JPDA数据关联的计算复杂度,结合基于马氏距离构造的椭圆关联门和基于车辆非完整性约束构造的扇形关联门,实现关联门的自适应调整,减少关联杂波的干扰。其次,结合传感器的配置与特性,对目标的航迹状态进行空间对准和时间对准,按照航迹点间的欧氏距离和互协方差选择融合模式,进行局部航迹融合。最后,为验证多目标跟踪和航迹融合方法的有效性与实用性,分别设计基于MATLAB/PreScan环境的仿真试验和基于智能清扫车平台的实车试验。研究结果表明:在横、纵方向上,融合后的系统状态都比单一传感器的估计状态更为准确,融合结果对单一传感器的估计误差有35%以上的提升;实车试验证明,该方案能有效融合ESR毫米波雷达和Mobileye单目前视相机的状态估计信息,能基本正确地跟踪目标和估计航迹;融合状态的横、纵向误差都在可接受范围以内,且融合状态比单一传感器的估计波动更小。
Accurate tracking and prediction of traffic participant objects is critical to the effectiveness of smart car behavior decisions. The traditional tracking system of maneuvering targets utilizes a single sensor, which has difficulties ensuring the accuracy and credibility of tracking data. To improve the robustness and reliability of a tracking system, a target tracking scheme that combines millimeter wave radar and camera has been designed. The scheme integrates the feature-level information of multiple targets. First, considering the cluttering environment of a low-speed self-driving sweeper vehicle, a multi-target tracking method based on interacting multiple modules/joint probability data association(IMM/JPDA) was used to estimate the local track. To reduce the computational complexity of the JPDA, we combined the ellipse correlation gate constructed according to Mahalanobis distance with the sector-related gate based on vehicle non-integrity constraints, which can adjust the association gate adaptively and reduce the interference of the associated clutter.Second, the target tracks were combined with the configuration and characteristics of the sensor and were then spatially and temporally aligned. A fusion mode was selected according to the Euclidean distance and the cross-covariance between the track points to perform local track fusion. Finally, to verify the validity and practicability of the multi-target tracking and track fusion methods, simulation experiments based on a MATLAB/PreScan environment and a real vehicle test based on an intelligent sweeper vehicle platform were designed. Simulation results show that in both horizontal and vertical directions, the system state after fusion is more accurate than the estimation state of a single sensor, and the fusion result shows an improved estimation error of more than 35% as compared with a single sensor. The vehicle test proved that this scheme can effectively integrate the state estimation information of an ESR millimeter wave radar and a Mobileye monocular camera and can basically track and estimate target tracks. Both horizontal and longitudinal errors of the fused state are all within an acceptable range, and the fused track is less fluctuant than the tracks of a single sensor.
Accurate tracking and prediction of traffic participant objects is critical to the effectiveness of smart car behavior decisions. The traditional tracking system of maneuvering targets utilizes a single sensor, which has difficulties ensuring the accuracy and credibility of tracking data. To improve the robustness and reliability of a tracking system, a target tracking scheme that combines millimeter wave radar and camera has been designed. The scheme integrates the feature-level information of multiple targets. First, considering the cluttering environment of a low-speed self-driving sweeper vehicle, a multi-target tracking method based on interacting multiple modules/joint probability data association(IMM/JPDA) was used to estimate the local track. To reduce the computational complexity of the JPDA, we combined the ellipse correlation gate constructed according to Mahalanobis distance with the sector-related gate based on vehicle non-integrity constraints, which can adjust the association gate adaptively and reduce the interference of the associated clutter.Second, the target tracks were combined with the configuration and characteristics of the sensor and were then spatially and temporally aligned. A fusion mode was selected according to the Euclidean distance and the cross-covariance between the track points to perform local track fusion. Finally, to verify the validity and practicability of the multi-target tracking and track fusion methods, simulation experiments based on a MATLAB/PreScan environment and a real vehicle test based on an intelligent sweeper vehicle platform were designed. Simulation results show that in both horizontal and vertical directions, the system state after fusion is more accurate than the estimation state of a single sensor, and the fusion result shows an improved estimation error of more than 35% as compared with a single sensor. The vehicle test proved that this scheme can effectively integrate the state estimation information of an ESR millimeter wave radar and a Mobileye monocular camera and can basically track and estimate target tracks. Both horizontal and longitudinal errors of the fused state are all within an acceptable range, and the fused track is less fluctuant than the tracks of a single sensor.
引文
[1] MAYYAS A,SHEN Q,MAYYAS A,et al.Using Quality Function Deployment and Analytical Hierarchy Process for Material Selection of Body-in-white [J].Materials & Design,2011,32 (5):2771-2782.
[2] 单云霄,郭晓旻,龙江云,等.渐优随机采样算法在结构化道路无人驾驶中的应用[J].中国公路学报,2018,31(4):196-205.SHAN Yun-xiao,GUO Xiao-wen,LONG Jiang-yun,et al.Asymptotically Sampling-based Algorithm with Applications to Autonomous Urban Driving on Structured Road [J].China Journal of Highway and Transport,2018,31 (4):196-205.
[3] XIE G,GAO H,QIAN L,et al.Vehicle Trajectory Prediction by Integrating Physics-and Maneuver-Based Approaches Using Interactive Multiple Models [J].IEEE Transactions on Industrial Electronics,2018,65 (7):5999-6008.
[4] FORTIN B,LHERBIER R,NOYER J C.A Model-Based Joint Detection and Tracking Approach for Multi-vehicle Tracking with Lidar Sensor [J].IEEE Transactions on Intelligent Transportation Systems,2015,16 (4):1883-1895.
[5] GUO Xiao-xiao,LIU Yuan-sheng,Zhong Qi-xue et al.Research on Moving Target Tracking Algorithm Based on Lidar and Visual Fusion [J].Journal of Advanced Computational Intelligence and Intelligent Informatics,2018,22:593-601.
[6] 郭磊,李克强,王建强,等.一种基于特征的车辆检测方法[J].汽车工程,2006,28(11):1031-1035.GUO Lei,LI Ke-qiang,WANG Jian-qiang,et al.A Feature-based Vehicle Detection Method [J].Automotive Engineering,2006,28 (11):1031-1035.
[7] 林明秀,刘伟佳,徐心和.基于模板匹配的多模式车辆跟踪算法[J].系统仿真学报,2007,19(7):1519-1522.LIN Ming-xiu,LIU Wei-jia,XU Xin-he.Algorithm for Multimode Vehicle Tracking Based on Template Match [J].Journal of System Simulation,2007,19 (7):1519-1522.
[8] SUN N,HAN G,DU K,et al.Person/Vehicle Classification Based on Deep Belief Networks [C] // IEEE.International Conference on Natural Computation.New York:IEEE,2014:113-117.
[9] WANG H,CAI Y,CHEN L.A Vehicle Detection Algorithm Based on Deep Belief Network [J].The Scientific World Journal,2014,2014:1-7.
[10] LEE M S,KIM Y H.Automotive Radar Tracking of Multi-target for Vehicle CW/CA Systems [J].Mechatronics,2004,14 (1):143-151.
[11] SCHUSTER M,REUTER J,WANIELIK G.Tracking of Vehicles on Nearside Lanes Using Multiple Radar Sensors [C] // IEEE.Proceedings of the 2014 International Radar Conference.New York:IEEE,2014:1-6.
[12] 曾杰.基于视觉与雷达信息融合的车辆检测算法研究[D].重庆:重庆大学,2017.ZENG Jie.Research on Vehicle Detection Algorithm Based on the Data Fusion by Vision and Radar [D].Chongqing:Chongqing University,2017.
[13] ALESSANDRETTI G,BROGGI A,CERRI P.Vehicle and Guard Rail Detection Using Radar and Vision Data Fusion [J].IEEE Transactions on Intelligent Transportation Systems,2007,8 (1):95-105.
[14] CHAVEZGARCIA O,BURLET J,VU T D,et al.Frontal Object Perception Using Radar and Mono-vision [C] // IEEE.Intelligent Vehicles Symposium.New York:IEEE,2012:159-164.
[15] 王宝锋,齐志权,马国成,等.一种基于雷达和机器视觉信息融合的车辆识别方法 [J].汽车工程,2015,37(6):674-678.WANG Bao-feng,QI Zhi-quan,MA Guo-cheng,et al.Vehicle Detection Based on Information Fusion of Radar and Machine Vision [J].Automotive Engineering,2015,37 (6):674-678.
[16] WANG X,XU L,SUN H,et al.On-Road Vehicle Detection and Tracking Using MMW Radar and Monovision Fusion [J].IEEE Transactions on Intelligent Transportation Systems,2016,17 (7):2075-2084.
[17] KHALIL M,ELTRASS A S,ELZAAFARANY O,et al.An Improved Approach for Multi-target Detection and Tracking in Automotive Radar Systems [C] // IEEE.Proceedings of the International Conference on Electromagnetics in Advanced Applications.New York:IEEE,2016,480-483.
[18] KIM B,YI K,YOO H J,et al.An IMM/EKF Approach for Enhanced Multitarget State Estimation for Application to Integrated Risk Management System [J].IEEE Transactions on Vehicular Technology,2015,64 (3):876-889.
[19] LEE K H,KANZAWA Y,DERRY M,et al.Multi-Target Track-to-Track Fusion Based on Permutation Matrix Track Association [C] // IEEE.Intelligent Vehicles Symposium (IV).New York:IEEE,2018:465-470.
[20] TUGNAIT J K.Tracking of Multiple Maneuvering Targets in Clutter Using Multiple Sensors,IMM,and JPDA Coupled Filtering [J].IEEE Transactions on Aerospace and Electronic Systems,2004,40 (1):320-330.
[21] 孙宁,秦洪懋,张利,等.基于多传感器信息融合的车辆目标识别方法[J].汽车工程,2017,39(11):91-96.SUN Ning,QIN Hong-mao,ZHANG Li,et al.Vehicle Target Recognition Based on Multi-sensor Information Fusion [J].Automotive Engineering,2017,39 (11):91-96.
[22] 郭睿利,郭云飞,张云龙,等.基于一种改进IMMJPDA算法的地面目标跟踪[J].信息与电子工程,2012,10(4):406-411.GUO Rui-li,GUO Yun-fei,ZHANG Yun-long,et al.Ground Target Tracking Based on an Improved IMMJPDA Algorithm [J].Information and Electronic Engineering,2012,10 (4):406-411.
[23] 梁凯,潘泉,宋国明,等.基于曲线拟合的多传感器时间对准方法研究[J].火力与指挥控制,2006,31(12):51-53.LIANG Kai,PAN Quan,SONG Guo-ming,et al.The Study of Multi-sensor Time Registration Method based on Curve Fitting [J].Fire Control and Command Control,2006,31 (12):51-53.
[2] 单云霄,郭晓旻,龙江云,等.渐优随机采样算法在结构化道路无人驾驶中的应用[J].中国公路学报,2018,31(4):196-205.SHAN Yun-xiao,GUO Xiao-wen,LONG Jiang-yun,et al.Asymptotically Sampling-based Algorithm with Applications to Autonomous Urban Driving on Structured Road [J].China Journal of Highway and Transport,2018,31 (4):196-205.
[3] XIE G,GAO H,QIAN L,et al.Vehicle Trajectory Prediction by Integrating Physics-and Maneuver-Based Approaches Using Interactive Multiple Models [J].IEEE Transactions on Industrial Electronics,2018,65 (7):5999-6008.
[4] FORTIN B,LHERBIER R,NOYER J C.A Model-Based Joint Detection and Tracking Approach for Multi-vehicle Tracking with Lidar Sensor [J].IEEE Transactions on Intelligent Transportation Systems,2015,16 (4):1883-1895.
[5] GUO Xiao-xiao,LIU Yuan-sheng,Zhong Qi-xue et al.Research on Moving Target Tracking Algorithm Based on Lidar and Visual Fusion [J].Journal of Advanced Computational Intelligence and Intelligent Informatics,2018,22:593-601.
[6] 郭磊,李克强,王建强,等.一种基于特征的车辆检测方法[J].汽车工程,2006,28(11):1031-1035.GUO Lei,LI Ke-qiang,WANG Jian-qiang,et al.A Feature-based Vehicle Detection Method [J].Automotive Engineering,2006,28 (11):1031-1035.
[7] 林明秀,刘伟佳,徐心和.基于模板匹配的多模式车辆跟踪算法[J].系统仿真学报,2007,19(7):1519-1522.LIN Ming-xiu,LIU Wei-jia,XU Xin-he.Algorithm for Multimode Vehicle Tracking Based on Template Match [J].Journal of System Simulation,2007,19 (7):1519-1522.
[8] SUN N,HAN G,DU K,et al.Person/Vehicle Classification Based on Deep Belief Networks [C] // IEEE.International Conference on Natural Computation.New York:IEEE,2014:113-117.
[9] WANG H,CAI Y,CHEN L.A Vehicle Detection Algorithm Based on Deep Belief Network [J].The Scientific World Journal,2014,2014:1-7.
[10] LEE M S,KIM Y H.Automotive Radar Tracking of Multi-target for Vehicle CW/CA Systems [J].Mechatronics,2004,14 (1):143-151.
[11] SCHUSTER M,REUTER J,WANIELIK G.Tracking of Vehicles on Nearside Lanes Using Multiple Radar Sensors [C] // IEEE.Proceedings of the 2014 International Radar Conference.New York:IEEE,2014:1-6.
[12] 曾杰.基于视觉与雷达信息融合的车辆检测算法研究[D].重庆:重庆大学,2017.ZENG Jie.Research on Vehicle Detection Algorithm Based on the Data Fusion by Vision and Radar [D].Chongqing:Chongqing University,2017.
[13] ALESSANDRETTI G,BROGGI A,CERRI P.Vehicle and Guard Rail Detection Using Radar and Vision Data Fusion [J].IEEE Transactions on Intelligent Transportation Systems,2007,8 (1):95-105.
[14] CHAVEZGARCIA O,BURLET J,VU T D,et al.Frontal Object Perception Using Radar and Mono-vision [C] // IEEE.Intelligent Vehicles Symposium.New York:IEEE,2012:159-164.
[15] 王宝锋,齐志权,马国成,等.一种基于雷达和机器视觉信息融合的车辆识别方法 [J].汽车工程,2015,37(6):674-678.WANG Bao-feng,QI Zhi-quan,MA Guo-cheng,et al.Vehicle Detection Based on Information Fusion of Radar and Machine Vision [J].Automotive Engineering,2015,37 (6):674-678.
[16] WANG X,XU L,SUN H,et al.On-Road Vehicle Detection and Tracking Using MMW Radar and Monovision Fusion [J].IEEE Transactions on Intelligent Transportation Systems,2016,17 (7):2075-2084.
[17] KHALIL M,ELTRASS A S,ELZAAFARANY O,et al.An Improved Approach for Multi-target Detection and Tracking in Automotive Radar Systems [C] // IEEE.Proceedings of the International Conference on Electromagnetics in Advanced Applications.New York:IEEE,2016,480-483.
[18] KIM B,YI K,YOO H J,et al.An IMM/EKF Approach for Enhanced Multitarget State Estimation for Application to Integrated Risk Management System [J].IEEE Transactions on Vehicular Technology,2015,64 (3):876-889.
[19] LEE K H,KANZAWA Y,DERRY M,et al.Multi-Target Track-to-Track Fusion Based on Permutation Matrix Track Association [C] // IEEE.Intelligent Vehicles Symposium (IV).New York:IEEE,2018:465-470.
[20] TUGNAIT J K.Tracking of Multiple Maneuvering Targets in Clutter Using Multiple Sensors,IMM,and JPDA Coupled Filtering [J].IEEE Transactions on Aerospace and Electronic Systems,2004,40 (1):320-330.
[21] 孙宁,秦洪懋,张利,等.基于多传感器信息融合的车辆目标识别方法[J].汽车工程,2017,39(11):91-96.SUN Ning,QIN Hong-mao,ZHANG Li,et al.Vehicle Target Recognition Based on Multi-sensor Information Fusion [J].Automotive Engineering,2017,39 (11):91-96.
[22] 郭睿利,郭云飞,张云龙,等.基于一种改进IMMJPDA算法的地面目标跟踪[J].信息与电子工程,2012,10(4):406-411.GUO Rui-li,GUO Yun-fei,ZHANG Yun-long,et al.Ground Target Tracking Based on an Improved IMMJPDA Algorithm [J].Information and Electronic Engineering,2012,10 (4):406-411.
[23] 梁凯,潘泉,宋国明,等.基于曲线拟合的多传感器时间对准方法研究[J].火力与指挥控制,2006,31(12):51-53.LIANG Kai,PAN Quan,SONG Guo-ming,et al.The Study of Multi-sensor Time Registration Method based on Curve Fitting [J].Fire Control and Command Control,2006,31 (12):51-53.