摘要
针对智能交通系统中小尺度交通标志识别率低的问题,文中提出一种改进卷积神经网络的交通标志识别方法。该方法通过在Faster R-CNN算法的低层特征图上增加优化的RPN网络,提升了小尺度交通标志的检测率。该方法还利用Max Pooling方法实了现图像的局部细节特征与全局语义特征充分融合。在TT-100K数据集上稍微实验结果表明新方法可以明显提高小尺度交通标志的识别率。
Aiming at the low recognition rate of small scale traffic signs in intelligent transportation system, an improved convolution neural network method for traffic sign recognition was proposed in this paper.This method could improve the detection rate of small-scale traffic signs by adding an optimized RPN network to the low-level feature map of Faster R-CNN algorithm.In addition, Max Pooling method was used to fully fuse the local details and global semantic features of the image. The experimental results on TT-100 K data set showed that the proposed method could significantly improve the recognition rate of small-scale traffic signs.
作者
袁小平
王岗
王晔枫
汪喆远
孙辉
YUAN Xiaoping;WANG Gang;WANG Yefeng;WANG Zheyuan;SUN Hui(School of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221000,China)
出处
《电子科技》
2019年第11期28-32,共5页
Electronic Science and Technology
基金
江苏省自然科学基金(BK20170278)~~