基于颜色信息与SVM网络的交通标志检测  

The Traffic Sign Detection Based on Color Information and SVM Network

作　　者：朱双东[1] 刘兰兰[1] 

机构地区：[1]宁波大学信息科学与工程学院,浙江宁波315211

出　　处：《自动化仪表》2009年第3期69-72,共4页Process Automation Instrumentation

基　　金：浙江省科技计划项目(编号:2007C21179)。

摘　　要：交通标志检测过程中,颜色复杂性造成的颜色失真是影响识别率的主要因素。遵循"简化复杂问题、基于颜色信息、采用智能方法"的基本思路,提出了一种新的基于颜色规格化的交通标志分类模型,先通过颜色规格化将交通标志中复杂的颜色信息简化为5种基本颜色,再利用两级智能分类器对交通标志进行分类。采用M-SVMs网络实现了颜色规格化,粗分类的识别率达到了100%,细分类的平均识别率也达到了70%。同时这种方法具有很好的鲁棒性。In traffic sign detecting process, the color distortion caused by color complexity is the key factor affecting recognition rate. On the basis of the thinking＂simplifying the complex issue; based on color information; adopting intelligent method＂, a novel traffic sign classifying model based on color standardization is proposed. The complex color information in traffic sign is simplified into five basic colors first, then the traffic signs are classified by adopting two-stage intelligent classifier. M-SVMs network is used for color standardization 100% recognition rate in approximated classification and 70% average recognition rate in sophisticated classification are achieved. The proposed approach also shows better robustness.

关 键 词：道路交通标志识别 颜色规格化 支持向量机 分类器 图像检测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]