摘要
为了利用智能算法来识别大米的种类属性,提出一种基于有导师学习神经网络的大米识别方法。该方法将训练样本的参数数据构造成测试集的标准数据,建立每种稻米的GRNN和PNN模型,并利用MATLAB软件对包含有不同几何特征参数的GRNN和PNN模型进行仿真。在已有训练样本数据的指导下,分析稻米米粒几何参数与稻米种类之间的关系,识别出其他样本的类别。结果显示,提出的方法能够精准地识别稻米的种属,而且GRNN模型的准确率要低于PNN模型,该方法容易实现,无庞杂计算,避免繁琐迭代过程。
The paper main focus on intelligent algorithms for rice classification, and establish a new method based on supervised learning neural network. The parameter of training sample is considered as standard data for testing sample, the model of GRNN and PNN are carried out respectively to each rice and MATLAB is employed to simulate the previous models with different geometric characteristic, and moreover, the relationship between rice geometric parameters and rice varieties is analyzed. The results evidently pointed out to identify accurately and a good effect with the method, and GRNN’s accuracy rate is a little lower than PNN’s. In addition, the vital advantage to this approach is not only the simplicity to implement without iteration, but also a small amount of calculation to employ.
作者
王正
WANG Zheng(College of Xinhua,Ningxia University,Yinchuan 750021,China;School of Civil Engineering and Hydraulic Engineering,Ningxia University,Yinchuan 750021,China)
出处
《机械与电子》
2018年第5期67-70,共4页
Machinery & Electronics
基金
宁夏高等学校科学研究项目(NGY2016215)
关键词
神经网络
神经元阈值
权值矩阵
高斯函数
neural network
neuron threshold
weight matrix
gauss function
