基于BP神经网络算法的红枣分级技术应用

BP neural network algorithm in the classification of jujube

分别从几何形状特征、颜色特征及纹理特征三方面入手,并对所有特征进行优化选择,最终得到3个表征红枣特性的特征量;建立红枣样本数据库,以BP神经网络为模型,设计合理的BP神经网络训练和测试方法,对红枣进行分级研究,通过不同样本条件下的实验测试,可以得到高达94%的红枣等级划分正确率,能较好地满足红枣分级的需求,对红枣产品的生产、销售均具有一定的理论和实际意义。

Feature extraction method was researched deeply from three aspects such as geometric shape feature,color feature and texture feature,and 21 features were extracted.After that,feature optimization method was studied,and 3 feature values were finally optimized to stand for the characters of red jujubes.A sample database of red jujube is set up for BP neural network,and a rational BP neural network training and test model was designed to study the classification of red jujube.Experiments under different condition of samples showed that 94% classification rate of red jujube could be achieved,and could meet the classification requirement of red jujube preferably,and it was significant in theory and practice for sale and manufacturing of red jujube.