摘要
为了识别退化的交通标志图像,该文采用一种新的特征提取算法。该算法在处理图像退化问题时,采用模糊-仿射联合不变矩直接提取图像的特征,从而避免了需要较大计算量的图像复原处理过程。针对各阶模糊-仿射联合不变矩数量级差异较大问题,提出一种数量级标准化算法。在深入分析PNN与K-means聚类算法的基础上,提出采用全局K-均值算法优化设计概率神经网络分类器,并将其用于交通标志图像的分类识别。仿真结果表明:模糊-仿射联合不变矩是一种有效的处理退化交通标志图像的方法,所设计的概率神经网络分类器不仅具有精简的结构而且具有较好的推广性能。
A novel feature extraction algorithm is presented for the recognition of traffic sign symbols undergoing degradations in this paper. In order to cope with the degradations, the Combined Blur-Affine Invariants （CBAIs） are adopted to extract the features of traffic sign symbols without any restorations which usually need a great amount of computations. A new magnitude normalization method is proposed for the great differences of magnitude of combined blur-affine invariants. Under the deep discussion of PNN and K-means algorithm, a probabilistic neural network classifier is designed using global K-means algorithm and applied to the classification of degraded traffic signs. The simulation results indicate that CBAIs are efficient for the feature extraction of degraded images, and the designed network is not only parsimonious but also has better generalization performance.
出处
《电子与信息学报》
EI
CSCD
北大核心
2008年第7期1703-1707,共5页
Journal of Electronics & Information Technology
基金
高等学校博士学科点专项科研基金(20050213010)
国家自然科学基金(60674101)资助课题
关键词
模式识别
概率神经网络
交通标志
模糊-仿射联合不变矩
全局K-均值算法
Pattern recognition
Probabilistic Neural Networks （PNN）
Traffic sign
Combined Blur-AffineInvariants （CBAIs）
Global K-means algorithm
