基于YOLOv5的城市交通目标检测研究

Research on Urban Traffic Object Detection Based on YOLOv5

随着机动化的成熟,更多的行人、车辆甚至天气等因素导致城市交通场景日趋复杂,且智能化的发展使得无人驾驶技术快速发展。更好地监测城市道路交通和完善无人驾驶场景归结为提高目标检测算法精准度问题。本文为了更准确地检测真实场景下的城市道路交通图片,首先利用非局部均值滤波(NLM)去除图片中的噪声,突出目标信息和位置,然后利用YOLOv5算法对交通图片进行目标检测,得到了更加精准的定位效果。

With the maturity of motorization,more pedestrian,vehicle and even weather factors lead to the increasingly complex city traff ic scene,and the development of intelligent technology makes the rapid development of driverless technology.Better monitoring of city road traff ic and improving the unmanned driving scene can be attributed to the problem of improving the accuracy of object detection algorithm.In this paper,in order to more accurately detect the city road traff ic images in the real scene,we f irst use the n on-local mean f ilter(NLM)t o r emove t he n oise in t he image,h ighlight t he object information and location,and then use the YOLOv5 algorithm to detect the traff ic image,and get a more accurate positioning effect.