基于BP神经网络的西洋参等级分类方法研究

以西洋参外观图像为基础数据源,从形状、颜色、纹理3方面入手提取表征等级差异显著的特征向量,构造输入数据集.基于BP神经网络采用经验法与枚举实验法相结合构建参数合理的等级分类模型.从收敛标准、自适应步长等方面对传统算法进行改进.进一步提高了分类精准度及实时性,实验结果显示,基于传统BP神经网络所构建的等级评判模型识别率达83.67%,改进后识别率达90.82%,且收敛速度较快.

A study of evaluation methods for American ginseng based on the BP neural network

This is a study of the feasibility of the automatic evaluation for the grades of American ginseng based on the BP neural network. This method extracts the characteristic vector of American ginseng and analyzes its shape,color and texture by using the BP neural network. The model parameters are determined by the experience method and the enumeration method,and the traditional algorithm is improved. The experimental results show that the recognition accuracy rate of the grade evaluation model based on the traditional BP neural network is 83. 67%,whle the improved one is 90. 82% with a quicker convergence speed.