基于小波描述子和神经网络的形状识别  

The Shape Recognition System Based on Wavelet Descriptors and Neural Networks

作　　者：夏敏[1] 刘宏申[1] 

机构地区：[1]安徽工业大学计算机学院,安徽马鞍山243002

出　　处：《计算机技术与发展》2007年第3期106-108,共3页Computer Technology and Development

基　　金：安徽省教育厅自然科学基金项目(2006KJ018A)

摘　　要：提出一种把小波描述子和神经网络相结合的形状识别方法。通过小波描述子提取待识别物体的形状特征,所提取的特征值不受物体位移、缩放和旋转的影响,接着用多层分类器网络对物体的形状进行识别分类,并采用BP算法对神经网络进行学习和训练。最后得出令人满意的实验结果。The application of combining wavelet and neural networks to the 2 - D pattern recognition on the basis of its shape is presented, First, use wavelet descriptors to extract the shape features. The wavelet descriptors provide the appropriate invariances to the scale, rotation and translation. Then use neural networks to recognize and classify the shape of the 2 - D patterns, The type of neural networks is the multilayer perceptron. The shape recognitions systems constructed neural networks carry out studying and training by using the BP algorithm, Finally, a very good experiment result is obtained.

关 键 词：形状识别 小波描述子 多层感知器 BP算法 

分 类 号：TN911.73[电子电信—通信与信息系统]