基于颜色规格化的交通标志识别  

TRAFFIC SIGN RECOGNITION BASED ON COLOR STANDARDIZATION

作　　者：朱双东[1] 蒋甜甜 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 [2]浙江蓝天求是环保集团有限公司,浙江杭州310012

出　　处：《计算机应用与软件》2008年第1期220-222,共3页Computer Applications and Software

摘　　要：采用神经网络群构成的分类器解决实景交通标志识别问题时,识别率普遍较低。分析可知,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,提出了一种新的解决方案:先通过颜色规格化方法将交通标志中复杂的颜色信息简化为5种基本颜色,然后再利用两级智能分类器实现分类。采用BP网络实现了颜色规格化;实验表明,这种方法具有很好的鲁棒性。Nowadays ,for outdoor traffic sign recognition problems based on the neural networks, the recognition rate is not generally ideal. Through the analysis, it is shown that the key factor affecting recognition rate is the color distortion caused by the color complexity, A new solution to the simplification problem is presented based on color information and the intelligence approach. The complex color information is simplified to 5 kinds of standard color, and then a two-hierarchical intelligence classifier is employed for classification. BP network is used for col- or standardization. The experimental result shows that the proposed approach indicates better robustness.

关 键 词：道路交通标志识别(TSR) 颜色规格化 神经网络 图像检测 分类器 机器识别 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]