摘要
针对目前交通标志检测识别和跟踪技术仍存在的问题,提出一种新的解决方案。首先基于RGB和HSV颜色空间的融合特征筛选出感兴趣区域,然后采用多尺度两阶段的卷积神经网络对感兴趣区域进行交通标志检测和识别,最后利用基于核相关滤波的交通跟踪算法进行目标跟踪。通过实验可知,该方法具有较好的检测和识别精度,能够有效地进行跟踪。
In view of the present traffic sign detection identification and tracking technology is still existing problems,and puts forward a new solution.Firstly,the regions of interest are screened out based on the fusion features of RGB and HSV color space,then the multi-scale two-stage convolutional neural network is used to detect and identify traffic signs in the regions of interest,and finally,the traffic tracking algorithm based on nuclear correlation filtering is used for target tracking.Experiments show that this method has good detection and identification ac⁃curacy and can be used to track effectively.
作者
陈阳
CHEN Yang(College of Computer Science,Sichuan University,Chengdu 610065)
出处
《现代计算机》
2021年第7期111-116,共6页
Modern Computer
关键词
交通标志
卷积神经网络
核相关滤波
目标跟踪
Traffic Signs
Convolutional Neural Network
Nuclear Correlation Filtering
Target Tracking
