The increasing use of the Internet with vehicles has made travel more convenient.However,hackers can attack intelligent vehicles through various technical loopholes,resulting in a range of security issues.Due to these...The increasing use of the Internet with vehicles has made travel more convenient.However,hackers can attack intelligent vehicles through various technical loopholes,resulting in a range of security issues.Due to these security issues,the safety protection technology of the in-vehicle system has become a focus of research.Using the advanced autoencoder network and recurrent neural network in deep learning,we investigated the intrusion detection system based on the in-vehicle system.We combined two algorithms to realize the efficient learning of the vehicle’s boundary behavior and the detection of intrusive behavior.In order to verify the accuracy and efficiency of the proposed model,it was evaluated using real vehicle data.The experimental results show that the combination of the two technologies can effectively and accurately identify abnormal boundary behavior.The parameters of the model are self-iteratively updated using the time-based back propagation algorithm.We verified that the model proposed in this study can reach a nearly 96%accurate detection rate.展开更多
地铁受电弓的燃弧会加剧受电弓滑板和接触网磨损,严重危害轨道交通安全。文章针对地铁车辆受电弓燃弧检测存在强光干扰和背景多变的问题,提出了一种基于yolov4-tiny模型改进的燃弧检测方法。为提升小目标检测能力,该方法在yolov4-tiny...地铁受电弓的燃弧会加剧受电弓滑板和接触网磨损,严重危害轨道交通安全。文章针对地铁车辆受电弓燃弧检测存在强光干扰和背景多变的问题,提出了一种基于yolov4-tiny模型改进的燃弧检测方法。为提升小目标检测能力,该方法在yolov4-tiny原有两个尺度的预测分支的基础上,添加第三尺度的预测分支以实现小燃弧在浅层网络中的定位,在主干网络后增加RFB(Receptive Field Block)模块以扩大网络的感受野,增强模型的特征提取能力。结果表明,改进的模型在测试集上的平均精度值(PAP)比yolov4-tiny提升了7.8个百分点,达到了98.2%,燃弧的定位效果与yolov4相当,但速度得到了极大的提升,单张图片的推理速度仅为6.5 ms,能有效、准确地完成地铁车辆中的受电弓燃弧检测任务。展开更多
基金This work was supported by Research on the Influences of Network Security Threat Intelligence on Sichuan Government and Enterprises and the Development Countermeasure(Project ID 2018ZR0220)Research on Key Technologies of Network Security Protection in Intelligent Vehicle Based on(Project ID 2018JY0510)+3 种基金the Research on Abnormal Behavior Detection Technology of Automotive CAN Bus Based on Information Entropy(Project ID 2018Z105)the Research on the Training Mechanism of Driverless Network Safety Talents for Sichuan Auto Industry Based on Industry-University Synergy(Project ID 18RKX0667),Research and implementation of traffic cooperative perception and traffic signal optimization of main road(Project ID 2018YF0500707SN)Research and implementation of intelligent traffic control and monitoring system(Project ID 2019YGG0201)Remote upgrade system of intelligent vehicle software(Project ID 2018GZDZX0011).
文摘The increasing use of the Internet with vehicles has made travel more convenient.However,hackers can attack intelligent vehicles through various technical loopholes,resulting in a range of security issues.Due to these security issues,the safety protection technology of the in-vehicle system has become a focus of research.Using the advanced autoencoder network and recurrent neural network in deep learning,we investigated the intrusion detection system based on the in-vehicle system.We combined two algorithms to realize the efficient learning of the vehicle’s boundary behavior and the detection of intrusive behavior.In order to verify the accuracy and efficiency of the proposed model,it was evaluated using real vehicle data.The experimental results show that the combination of the two technologies can effectively and accurately identify abnormal boundary behavior.The parameters of the model are self-iteratively updated using the time-based back propagation algorithm.We verified that the model proposed in this study can reach a nearly 96%accurate detection rate.
文摘地铁受电弓的燃弧会加剧受电弓滑板和接触网磨损,严重危害轨道交通安全。文章针对地铁车辆受电弓燃弧检测存在强光干扰和背景多变的问题,提出了一种基于yolov4-tiny模型改进的燃弧检测方法。为提升小目标检测能力,该方法在yolov4-tiny原有两个尺度的预测分支的基础上,添加第三尺度的预测分支以实现小燃弧在浅层网络中的定位,在主干网络后增加RFB(Receptive Field Block)模块以扩大网络的感受野,增强模型的特征提取能力。结果表明,改进的模型在测试集上的平均精度值(PAP)比yolov4-tiny提升了7.8个百分点,达到了98.2%,燃弧的定位效果与yolov4相当,但速度得到了极大的提升,单张图片的推理速度仅为6.5 ms,能有效、准确地完成地铁车辆中的受电弓燃弧检测任务。