基于LSP与GLCM方法的碎米识别特性研究  

Research on Identification Characteristics of Broken Rice Based on LSP and GLCM Methods

作　　者：徐子龙 杨柳 肖轩 范雨超 罗洋 李瑜 裴后昌 张永林 Xu Zilong;Yang Liu;Xiao Xuan;Fan Yuchao;Luo Yan;Li Yu;Pei Houchang;Zhang Yonglin(School of Mechanical Engineering,Wuhan Polytechnic University,Wuhan 430023)

机构地区：[1]武汉轻工大学机械工程学院,武汉430023

出　　处：《中国粮油学报》2023年第10期196-203,共8页Journal of the Chinese Cereals and Oils Association

基　　金：湖北省自然科学基金项目(2022CFB944);湖北省教育厅科研项目(Q20211609);江苏省食品先进制造重点实验室开放基金项目(FM-202103);湖北省重点研发计划项目(2022BBA0047);武汉轻工大学校杰出青年科研项目(2020J06)。

摘　　要：随着农业机械化的不断推进,稻米加工过程中的碎米率逐步上升,碎米的快速准确识别对于大米加工与分级等具有重要意义,高效碎米识别方法的建立与应用是提高碎米的利用价值的有效途径。以米粒为研究对象,采用计算机视觉技术提取米粒特征,实现碎米自动分类识别。首先对米粒图像预处理,提取4维形状特征,利用局部相似模式(LSP)和灰度共生矩阵(GLCM)提取米粒4维纹理特征,建立支持向量机(SVM)、线性判别分类器(LDA)、K近邻(KNN),随机森林(RF),4种机器学习方法训练碎米识别分类器进行分类识别,利用验证集数据分析各类分类器的性能特征。结果表明,基于纹理特征及形状特征的支持向量机分类器的碎米识别精度最高,正确率达到97.56%。With the constant promotion of agricultural mechanization,the rate of broken fracture rice during processing gradually increases.The rapid and accurate identification of broken fracture rice is significant for rice milling companies.Selecting an appropriate identification and detection method of broken fracture rice is one of the effective ways to improve the utilization value of broken fracture rice.rice grains were taken as the raw materials,and computer vision technique was used to extract the features of rice grains to achieve automatic classification and recognition of broken fracture rice.Firstly,the images of rice grains were pre-processed,then,4-dimensional shape features were extracted,and 4-dimensional texture features of brown rice were by use of Local Similarity Patterns(LSP)and Gray-Level Co-occurrence Matrix(GLCM).Finally,three machine learning methods,such as support vector machine(SVM),Linear Discriminant Analysis(LDA),K-Nearest Neighbors(KNN)and Random Forest(RF),were established to train the classifier for broken fracture rice recognition.The performance of each type of classifier was analyzed and compared by use of the validation set data.The results indicated that the support vector machine classifier based on texture features and shape features has the highest accuracy in recognizing broken fracture rice with a 97.56%correct rate.

关 键 词：碎米 图像识别 LSP GLCM SVM 

分 类 号：S126[农业科学—农业基础科学]