摘要
提出了一种基于神经网络中的单输出BP(Error Back-Propagation)网络的自然风景类图像的分类检索方法,并在网络学习过程中改进了BP算法,使速率在学习过程中根据误差平方和大小动态可变,极大地提高了收敛速度,分类结果表明图像特征的选取、隐藏层的节点数对网络性能影响很大。
A kind of classifying and retrieve method of natural scenic images which is based on single output BP networks in neural networks is proposed in this paper. Networks use a feature or the combination of several features of images as input features and improve the BP arithmetic to make the rate alter according to the sum of error square in the process of study. The results of classify indicate that the selection of image features and the number of hidden layer nodes have a big impact on the performance of networks.
出处
《计算机与数字工程》
2006年第11期77-81,共5页
Computer & Digital Engineering
关键词
BP(反向传播)
图像检索
图像特征
单输出
BP （Error Back -Propagation）, retrieve of image, image features, single output
