基于车载视频的压线检测与车道偏移预警
|
摘要
车辆压线检测系统可对车辆运行过程中发生的压线行为进行检测并做出警告,避免由于司机注意不集中、疲劳驾驶、驾驶陋习等原因导致车辆偏移而造成交通事故。对此提出一种基于车载视频的压线检测与车道偏移预警方法。首先,利用合成数据方法构造丰富多样的压线检测数据集;然后,结合图像语义分割方法完成车道线检测;最后,利用当前车道双边线多个几何参数对车辆压线行为作出检测并做出预警。实验表明,单帧平均压线检测准确率为93. 7%,耗时78 ms,车道偏移预警召回率为93. 5%,该方法具备一定的实际应用价值。
Lane-crossinging detection system can detect line-crossing behavior of vehicle and raise warning,which can be used to avoid traffic accident due to lack of concentration,fatigue driving,bad habits of driving and so on. To tackle this issue,a lane-crossing detection and lane-departure warning method based on vehicle video is proposed. Firstly,the method uses synthetic data method to build a rich and varied lane-crossing detection dataset. Then,it accomplishes lane line detection with semantic segmentation. At last,it detects lane-crossing behavior and raises warning using several geometric parameters of 2 lines of current lane. Results show the method achieves an average precision of 93. 7% for lane-crossing detection,takes an average time of 78 ms,and achieves an average recall rate of 93. 5% for lane-departure warning,which means the method is robust and practical.
Lane-crossinging detection system can detect line-crossing behavior of vehicle and raise warning,which can be used to avoid traffic accident due to lack of concentration,fatigue driving,bad habits of driving and so on. To tackle this issue,a lane-crossing detection and lane-departure warning method based on vehicle video is proposed. Firstly,the method uses synthetic data method to build a rich and varied lane-crossing detection dataset. Then,it accomplishes lane line detection with semantic segmentation. At last,it detects lane-crossing behavior and raises warning using several geometric parameters of 2 lines of current lane. Results show the method achieves an average precision of 93. 7% for lane-crossing detection,takes an average time of 78 ms,and achieves an average recall rate of 93. 5% for lane-departure warning,which means the method is robust and practical.
引文
[1]余厚云,张为公.直线模型下的车道线跟踪与车道偏离检测[J].自动化仪表,2009,30(11):1-7.
[2]郭子逸.一种基于单目视觉的车道偏离检测与预警方法[J].机械制造,2012,50(5):86-90.
[3]胡海辉.基于车载摄像机的车辆压线与变道检测技术研究与实现[D].南京:东南大学,2015.
[4]王晓锦.基于单目视觉的结构化道路车道线识别和SAE算法的车道偏离检测研究[D].济南:山东大学,2017.
[5]杨萍,杨磊.车道偏离预警算法的研究[J].绿色科技,2017(6):173-175,180.
[6]青岛慧拓智能机器有限公司.一种车道压线检测方法及系统:中国,CN201810128866.9[P].2018-02-08.
[7]GAIDON A,WANG Q,CABON Y,et al.Virtual worlds as proxy for multiobject tracking analysis[C].Proceedings of2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Las Vegas,NV,2016:4340-4349.
[8]ROS G,SELLART L,MATERZYNSKA J,et al.The SYN-THIA dataset:a large collection of synthetic images for semantic segmentation of urban scenes[C].Proceedings of2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Las Vegas,NV,2016:3234-3243.
[9]SRAH S,DEY D,LOVETT C,et al.Air Sim:high-fidelity visual and physical simulation for autonomous vehicles[C].Proceedings of 2017 Conference on Field and Service robotics.Zurich,Switzerland:Field and Servic-e Robotics,2017:621-635.
[10]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2016,40(4):834-848.
[2]郭子逸.一种基于单目视觉的车道偏离检测与预警方法[J].机械制造,2012,50(5):86-90.
[3]胡海辉.基于车载摄像机的车辆压线与变道检测技术研究与实现[D].南京:东南大学,2015.
[4]王晓锦.基于单目视觉的结构化道路车道线识别和SAE算法的车道偏离检测研究[D].济南:山东大学,2017.
[5]杨萍,杨磊.车道偏离预警算法的研究[J].绿色科技,2017(6):173-175,180.
[6]青岛慧拓智能机器有限公司.一种车道压线检测方法及系统:中国,CN201810128866.9[P].2018-02-08.
[7]GAIDON A,WANG Q,CABON Y,et al.Virtual worlds as proxy for multiobject tracking analysis[C].Proceedings of2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Las Vegas,NV,2016:4340-4349.
[8]ROS G,SELLART L,MATERZYNSKA J,et al.The SYN-THIA dataset:a large collection of synthetic images for semantic segmentation of urban scenes[C].Proceedings of2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR),Las Vegas,NV,2016:3234-3243.
[9]SRAH S,DEY D,LOVETT C,et al.Air Sim:high-fidelity visual and physical simulation for autonomous vehicles[C].Proceedings of 2017 Conference on Field and Service robotics.Zurich,Switzerland:Field and Servic-e Robotics,2017:621-635.
[10]CHEN L C,PAPANDREOU G,KOKKINOS I,et al.DeepLab:semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected CRFs[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2016,40(4):834-848.