SOM神经网络在树叶形状分类中的应用

使用目前较常用的自组织映射神经网络 (SOM神经网络 ) ,以几类比较多树叶图片的样本为基础 ,提取树叶形状特征值作为神经网络的输入 ,选适当的输出结点 ,应用自组织特征映射网络对各样本进行竞争学习 ,自动形成不同的输出值对应于该类别 ,从而得到分类结果 ,分类正确率达 86 .6 7%。

Application of SOM Neural Network on Leaves' Shape Classification

By using the presently general selforganizing feature map network (SOM neural network),and on the basis of a little more leaves' shape sample,this paper extracts the leaves' picture characteristic values to use as the inputting values of the neural network,selects the adaptable output node and applies the selforganizing feature map network to do competition learning for each sample,and then automatically to generate the different output that are relative with the classification so as to obtain the classification result,and the classification accuracy reaches to 86.67%.