摘要
在分析了BP网络学习算法的缺陷基础上,引入动量项和遗传算法对BP网络学习算法进行改进。利用小波多尺度边缘检测对汽车车型图像进行分割和特征提取,利用矩不变量特征和改进BP神经网络对汽车车型进行自动识别和分类。提高了汽车车型自动识别和分类的速度和正确率。
BP network is improved by introducing momentum and genetic algorithm based on analyzing its demerits. We divide up image of automatic vehicle and pick up its characteristics with wavelet multi-scale edge detecting, and use the invariant quadratures and improved BP network to extract the features of image of Automatic vehicle. The study can improve automatic vehicle recognition, classifying speed and preciseness.
出处
《广西工学院学报》
CAS
2008年第3期23-26,共4页
Journal of Guangxi University of Technology
基金
广西教育厅科研项目(编号200707LX196)
广西工学院自然科学基金项目(院科自0704102)
关键词
BP网络
动量项
遗传算法
矩不变量
汽车车型自动识别
BP network
momentum
genetic algorithm
invariant quadrature
automatic vehicle recognition
