基于证据有序极限学习机Bagging集成的苹果分级方法  

Apple Grading Method Based on Bagging Ensemble of Evidential Ordinal Extreme Learning Machines

作　　者：马荔瑶 卫鹏 范肖辰 徐元[1] 毕淑慧 MA Liyao;WEI Peng;FAN Xiaochen;XU Yuan;BI Shuhui(School of Electrical Engineering,University of Jinan,Jinan 250022,Shandong,China)

机构地区：[1]济南大学自动化与电气工程学院,山东济南250022

出　　处：《济南大学学报(自然科学版)》2025年第2期263-271,共9页Journal of University of Jinan(Science and Technology)

基　　金：国家自然科学基金项目(61803175);山东省自然科学基金项目(ZR2021MF074,ZR2023MF094)。

摘　　要：为了提高苹果自动分级分拣系统的性能,提出一种基于证据有序极限学习机Bagging集成的无损苹果分级模型。使用苹果近红外光谱中提取的特征作为输入,根据可溶性固形物含量将苹果分为3个等级;考虑等级类标的认知不确定性和有序性,在Dempster-Shafer理论框架内提出高斯质量函数生成方法和证据编码方案;构造以证据编码为输出的证据有序极限学习机作为基学习器,通过Bagging算法实现集成学习;选取435个红富士苹果作为实验样本生成数据集,并进行交叉验证。结果表明,所提出的证据有序极限学习机Bagging集成模型的分级准确率达到91%,该模型训练时间比证据有序神经网络集成模型的缩减至少2个数量级。To improve the performance of the automatic apple grading and sorting system,a non-destructive apple grading model was proposed based on Bagging ensemble of evidential ordinal extreme learning machine.The features extracted from the near-infrared spectra of apples were taken as inputs and apples were classified into three grades based on soluble solids content.Considering the epistemic uncertainty and order of the grading labels,Gaussian mass function generation method and evidence encoding strategies were designed within the framework of Dempster-Shafer theory.Using evidence-encoded training set as outputs,evidential ordinal extreme learning machines were constructed as base learners and further integrated as an ensemble by the Bagging algorithm.Selecting 435 Red Fuji apples as the experimental instances,data were collected and processed into the apple data set.The experimental results of cross-validation show that the grading accuracy of the proposed Bagging ensemble of evidential ordinal extreme learning machines is 91%,and the model training time is reduced by at least 2 orders of magnitude compared with the integrated model of evidential ordinal neural network.

关 键 词：苹果分级 近红外光谱 DEMPSTER-SHAFER理论 有序分类 集成学习 

分 类 号：TP274[自动化与计算机技术—检测技术与自动化装置]