基于全局和局部特征相融合的交通标志识别研究

项目名称： 基于全局和局部特征相融合的交通标志识别研究

项目编号： No.61301186

项目类型： 青年科学基金项目

立项/批准年度： 2014

项目学科： 无线电电子学、电信技术

项目作者： 袁雪

作者单位： 北京交通大学

项目金额： 24万元

中文摘要： 如何提高交通标志识别系统在室外复杂环境下的检测与识别能力是交通标志识别技术实用化的关键。本项目拟通过全局与局部特征相融合的算法提取交通标志的特征，进而完成交通标志的检测与识别。在全局特征提取方面，拟提出具有鲁棒性的颜色和空间结构分布特征作为图像的全局特征；在局部特征提取方面，将改进传统的局部二值模式(LBP)，并融入梯度方向分布信息作为图像的局部特征；其次，研究特征融合方法，对上述提取的全局和局部特征进行有效的融合，使融合后的图像特征描述子具有充分性、鉴别性和鲁棒性；最后，将图像特征描述子送入分类器，实现交通标志的识别，并采用仿真实验和理论分析相结合的方法评价系统的性能。最终目标是，在不明显增加运算量的前提下，利用全局及局部特征融合的方法实现比传统方法更好的识别效果，达到将识别率提高5%-10%以上的预期目标。通过本项目的研究，丰富图像特征提取技术，为推进交通标志识别系统的实用化奠定基础。

中文关键词： 交通标志检测；交通标志识别；图论；全局特征；局部特征

英文摘要： As an important subsystem in intelligent transportation system technology, traffic sign recogniton (TSR) systems based on computer vision have gradually become an important research topic. How to enhance the performance of TSR systems under outdoor complicated conditions is the key to its practical application. In this project, a TSR system based on fusion of global and local features is proposed. First, color and global spatical structure information are extracted as global features; Direction structure and local texture information are extracted as local features; Then, the global and local features are combined to build a feature descriptor by proposed feature fusion approach, which is satisfying the following three criteria: 1)sufficiency; 2)distinctiveness; and 3)robustness. Finally, an adaptive classifier is designed, and the efficiency of the proposed system will be validated by theory analysis and experiments. Comparing with traditional approaches, the research objective of this project is raising the recognition rate from 5% to 10%. By this project, it could be expected to speed up the develoment and practical application of TSR system.

英文关键词： Traffic sign detection；Traffic sign recognition；Graph；Global feature；Local feature