摘要
针对复杂天气下,目标受雨、雪、雾等影响,造成语义识别困难,误检率高等问题,提出了一种基于AlexNet融合改进的YOLOv3网络模型[1]。该网络基于卷积神经网络可以提取不同天气图像中判别性特征,然后基于不同天气状况使用不同方法去除干扰,最后使用改进的YOLOv3进行多目标判识。结果表明,复杂天气下该方法可以提高检测精度,并能满足实时性的需求。
Aiming at the problems of difficult semantic recognition and high false detection rate caused by rain,snow,fog and other effects on targets in complex weather,an improved YOLOv3 network model based on AlexNet fusion is proposed.The network can extract discriminant features in different weather images based on convolutional neural network,then use different methods to remove interference based on different weather conditions,and finally use improved YOLOv3 for multi-target identification.The results show that the present method can also maintain the existing detection accuracy under complex weather,and can meet the real-time requirements.
出处
《信息通信》
2020年第11期72-74,共3页
Information & Communications
基金
江苏高校品牌专业建设工程项目(PPZY2015B190)
江苏省大学生创新训练计划项目(202013108002Y)研究成果。
关键词
目标检测
深度学习
复杂天气
交通
target detection
deep learning
complex weather
traffic
