摘要
交通标志的有效分割是交通标志识别系统中的关键问题.针对禁令标志颜色特征,结合人类视觉系统的颜色感知特点,首先以颜色分量R(红)、G(绿)、B(蓝)作为输入特征量构造颜色特征粗分类器,再通过计算粗分类结果的特征量与红色特征类标准样本中心的矢量余弦,得到颜色相似度特征灰度图,用改进的Otsu方法实现标志的最终分割,最后给出了通过训练粗分类器优化分割结果的方法.实验结果表明,本文的分割方法可在不同气候条件下,有效地提高交通标志分割效率,具有较好的鲁棒性.
Segmenting traffic sign effectively is a key technique in traffic sign recognition system.A method of traffic forbidden sign segmentation based on color-apperceived of human visual system(HVS) is presented.First,the coarse classifier of color feature is constructed and the RGB color-components are taken as the most characteristic parameter.And then,the value of vector cosine is calculated between the characteristic parameter of classification result and standard sample of red feature,and the grayscale image is made by the similarity value of pixel.The accurate segmentation of traffic signs is achieved by the Otsu adaptive threshold.Finally,the segmentation results are improved by training the coarse classifier.Experimental result shows that the method of segmenting traffic sign can improve the efficiency of segmentation in different climatic conditions with more robustness.
出处
《武汉大学学报(理学版)》
CAS
CSCD
北大核心
2011年第3期236-240,共5页
Journal of Wuhan University:Natural Science Edition
基金
广西自然科学基金项目(2010GXNSFA013126)
广西科学基金项目(0832066
0991012)
广西教育厅科研项目(201010LX220)资助