| 摘 要: 随着社会经济的发展，我国的道路交通系统发展迅速，人们生活变得便利的同时，也随之带来了相应的安 全隐患。道路交通识别系统为上述问题提供了一种解决方法，因此受到学者们的广泛关注。TSR通过安装在机动车上的 摄像机提取自然场景图像，系统会对图像进行交通标志检测与识别，最后将识别结果告知驾驶员，以提高交通运行速 率，降低交通事故的发生。本文对多年来各位学者的研究结果加以总结得出结论，如何有效利用交通标志的多种特征、 融合线性以及非线性子空间特征提取方法的优势，研究出具有高鲁棒性和高实时性的交通标志识别方法，将是今后的主 要发展方向。 |
| 关键词: 交通标志；ITS；模板匹配；神经网络；深度学习 |
|
中图分类号: TP391
文献标识码: A
|
| 基金项目: 国家自然科学基金资助(6154006)；2015年大学生创新项目资助(“基于子空间降维的交通标志识别研究”)；基于多源探测的室内公共场所火灾预警关键技术研究，省 自然科学基金资助(2014020116). |
|
| A Review of the Research on Traffic Sign Recognition |
|
NI Yuting，LIANG Yufeng，HAO Bowen，ZHONG Ling
|
( School of Software, Shenyang University of Technology, Shenyang 110023, China)
|
| Abstract: With the development of social economy,the road transportation system has developed rapidly in our country. While the people's life becomes more convenient,potential security risks also come along.The road traffic identification system provides a solution to the above problem and arouses wide concern of scholars.After TSR extracts natural scene images through the camera installed on the motor vehicle,the system detects and recognizes the images with traffic signs.The recognition result will be sent to the driver,in order to increase the traffic operation speed and decrease the traffic accident rate.In this paper to summarize the results of years of research of the scholars concluded,how to effectively use the various features of a traffic sign,the advantages of integration of linear and non-linear subspace feature extraction method,developed with high robustness and high real-time traffic sign recognition method will be the main future direction of development. |
| Keywords: traffic signs;ITS;template matching;neural network;deep learning |
