摘要
神经网络具有较强的容错性和自适应学习能力,在交通标志图像识别中获得广泛的应用,但目前采用的多层感知器、BP神经网络以及径向基神经网络等几种神经网络存在一些不足,故本文提出采用改进概率神经网络进行交通标志图像识别的新方法,整个算法分两步实现:首先对交通标志图像提取它的T cheb ichef不变距并作为图像的特征;然后采用改进的概率神经网络进行识别。仿真实验表明,提出的方法比以前的方法有更好的识别效果。
Neural networks with higher fault-tolerant and good adaptive learning ability find great applications in classification of traffic signs. But at the recent times, the existed methods such as multilayer perceptron, BP networks and RBF neural networks have also some essential drawbacks and deficiencies. So this paper gives a new method for classification of traffic signs based on modified probabilistic neural networks （PNN）. Our classification algorithm is accomplished by two steps： the characters of traffic sign are extracted according to its Tchebichef moment invariants first, then based on modified PNN, traffic sign is identified. The simulation results validate the effectiveness of the proposed method.
出处
《系统工程》
CSCD
北大核心
2006年第4期97-101,共5页
Systems Engineering
基金
湖南省自然科学基金资助项目(05JJ30121)
中南大学博士研究生学位论文创新选题(040122)
关键词
交通标志
识别算法
概率神经网络
不变距
差异演化
Traffic Signs
Classification
Probabilistic Neural Networks （PNN）
Moment Invariants
Differential Evolution （DE）
