PCA和改进BP神经网络的大米外观品质识别  

Rice Appearance Quality Recognition Based on PCA and Improved BP Neural Network

作　　者：仲伟峰[1] 马丽霞[1] 何小溪[1] 

机构地区：[1]哈尔滨理工大学自动化学院,黑龙江哈尔滨150080

出　　处：《哈尔滨理工大学学报》2015年第4期76-81,共6页Journal of Harbin University of Science and Technology

基　　金：2014年度黑龙江省教育厅科学技术研究项目(12541148)

摘　　要：为了研究大米品质的优劣,采用主成分分析算法和BP神经网络相结合的方法对大米米粒进行识别.将米粒的特征变量进行主成分分析以提取代表性的主成分分量,将获得的主成分作为输入神经元来建立自适应BP神经网络进行学习,并将训练完毕的神经网络用于实际过程中的大米品质判别,同时采用附加动量法和自适应调整速率策略优化网络结构.仿真和实验结果表明,此方法可以使得大米识别的准确度达到95%以上并且有效的减少识别所需时间.In terms of the pros and cons of rice quality, the recognition method that combines the BP neural network with PCA （Principal Component Analysis） algorithm is proposed to recognise rice. Principal component a- nalysis is adopt to analyze the rice characteristic variables so that representative principal component can be extrac- ted. Then the adaptive BP neural network is trained to learn that with principal component as its input. Finally, the trained neural network is used to recognize rice quality in actual manipulation process, meanwhile, the additional momentum method and adaptive adjustment rate strategy are used to optimize the whole network. The simulation and experimental results show that the proposed recognition method makes the recognition accuracy reach over 95 % and shortens the learning time efficiently.

关 键 词：大米品质识别 PCA算法 BP神经网络 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]