摘要
在无人驾驶和辅助驾驶领域,交通标志牌检测识别是重要的。针对目前基于YOLO的检测方法能够达到实时的检测效果,但在准确率方面有所降低的问题,提出了基于感兴趣区域(ROI)的交通标志牌检测方法。首先根据交通标志牌的颜色特性得到候选区域;再利用交通场景图像规则确定交通标志牌的ROI;最后在交通标志牌的ROI,基于YOLO v3实现对交通标志牌的检测识别。实验结果表明:由于本文提出的方法去除了图像中部分干扰因素,使得算法在检测精度上得到了提升,也能满足实时性的需求,并在无人驾驶车辆上进行了验证。
In the field of self-driving and driver assistance system,traffic sign detection and recognition is very important.Aiming at that current detection method based on YOLO can achieve real-time detection effect,but the accuracy is decreased,problem,a traffic sign detection method based on region of interest(ROI)is presented. color features of traffic sign are firstly used to get the candidate regions.Secondly,ascertain the ROI of traffic sign based on the rules of traffic scene image.Finally detect traffic sign in ROI based on YOLO v3.The experimental result show that self-driving or intelligent vehicle show that the proposed approach eliminate partial interference which makes the detection precision of algorithm is improved and can meet the needs of real-time performance.It is verified in remotely pilotod vehile.
作者
潘卫国
刘博
陈英昊
石洪丽
PAN Weiguo;LIU Bo;CHEN Yinghao;SHI Hongli(Beijing Key Laboratory of Information Service Engineering,Beijing Union University,Beijing 100101,China;College of Robotics,Beijing Union University,Beijing 100027,China;College of Applied Science and Technology,Beijing Union University,Beijing 100101,China)
出处
《传感器与微系统》
CSCD
2019年第11期147-150,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(61802019,61871039)
北京市教育委员会科技计划资助项目(KM201711417005,KM201911417001)
关键词
目标检测
感兴趣区域
深度学习
交通标志牌
object detection
region of interest(ROI)
Deep learning
traffic sign
