摘要
本文提出了一种针对交通标志的特征提取方法，改进了Hilditch骨架提取算法，并组合骨架的水平和竖直直方图作为标志特征。以这样的直方图数组作为特征，可以消除95％以上的不相关和冗余信息。实验表明，在相同构架的TSR（Traffic Sign Recognition）系统和BP神经网络下，相对于传统特征提取方法，该方法使识别正确率有显著提升，并增强了抗干扰能力。
This paper presents a feature extraction of traffic signs ways to improve the Hilditch skeleton extraction algorithm,and the combination of horizontal and vertical histogram of the skeleton as a symbol of features.With such an array as a histogram feature,you can eliminate more than 95% of irrelevant and redundant information.Experiments show that the same architecture TSR （Traffic Sign Recognition）system and the BP neural network,relative to the traditional feature extraction method to recognition accuracy was significantly improved and enhanced anti-jamming capability.
出处
《计算机光盘软件与应用》
2011年第2期47-48,共2页
Computer CD Software and Application
关键词
交通标志
骨架
直方图
Traffic sign
Skeleton
Histogram
