大型车辆右转弯盲区预警系统设计  

Design of right turn blind area warning system for large vehicles

作　　者：杨炜[1] 张志威 周凯霞 刘佳俊 YANG Wei;ZHANG Zhiwei;ZHOU Kaixia;LIU Jiajun(School of Automobile,Chang’an University,Xi’an 710064,China)

机构地区：[1]长安大学汽车学院

出　　处：《中国科技论文》2019年第7期737-742,共6页China Sciencepaper

基　　金：中央高校基金科研业务费专项资金资助项目(300102229112);陕西省自然科学基础研究计划项目(2017JQ6045)

摘　　要：为弥补传统大型车辆右侧盲区预警系统夜间工作效果差、危险预警准确率低等不足,设计了一种应用于全天候环境的大型车辆右转弯盲区预警系统。当采集到右转向灯开启信号时,系统启动林柏视C600高清红外视觉传感器拍摄盲区图像。应用已训练的Inception-v3迁移学习模型对图像进行行人和骑行者检测。若检测到危险情况,系统立即显示盲区状况并报警。试验表明,所设计系统对每帧右侧盲区图像检测的平均准确率为97%,平均耗时为74.8 ms,能在夜间无光照条件下进行准确识别,实现及时、可靠的右侧盲区预警。To compensate the poor night working effect and low risk warning accuracy of traditional right-side blind warning system,a right-turning collision warning system of large vehicles was designed for all-weather environment.When the signal of right turn signal was collected,the system started the RMONCAM-C600 high definition infrared vision sensor to shoot the blind area image.Blind images were detected for pedestrian and cyclist by using the trained Inception-v3 migration learning model.If dangerous situation was detected,the system would immediately display the blind area and give an alarm.The results show that the average accuracy of the system is 97%and the needed average time is 74.8 ms for the detection of the right blind area of each frame.The system can accurately identify the right blind area in the absence of illumination at night and provide timely and reliable collision warning of right blind area.

关 键 词：汽车工程 行人检测 Inception-v3模型 深度学习 高清红外视觉传感器 

分 类 号：U471.15[机械工程—车辆工程]