摘要
小麦种子品种分类对提高农作物产量和粮食品质至关重要。光谱技术在小麦种子品种分类中应用广泛,但有限的样本量常导致模型分类性能受限。因此,本研究提出了一种特征融合策略,利用相似性约束提取目标域(待检测数据集)和相关域(与目标域相似的数据集)之间的跨域共享特征,同时构建目标域特征提取器捕获特有特征,用于小麦种子品种分类。利用2019年和2020年4个小麦品种的数据,比较了3种相关模型,实验结果显示所提模型在2个年份的小麦种子品种分类任务中效果最优,分类精度达到了98.03%和92.35%。此外,研究表明适度增加相关域样本量可有效提高分类性能。本研究利用相关域数据提高了小麦种子品种分类性能,为小麦种子品种分类提供了新的思路。
The classification of wheat seed varieties is considered crucial for improving crop yields and food quality.Spectral technology has been widely used in wheat seed variety classification,but a limited sample size often limits the classification performance of models.Therefore,a feature fusion strategy was proposed in this study,in which cross-domain shared features between the target domain(the dataset to be analyzed)and the related domain(similar datasets)were extracted by employing similarity constraints,while a feature extractor for the target domain was constructed to capture unique features for wheat seed variety classification.Using data from four wheat varieties in 2019 and 2020,three related models were compared.The experimental results showed that the proposed model achieved the best performance in wheat seed variety classification tasks for both years,with classification accuracies of 98.03%and 92.35%,respectively.In addition,it was demonstrated that the classification performance could be effectively improved by moderately increasing the sample size of the related domain.In this study,the performance of wheat seed variety classification was improved by utilizing related domain data,which provides a new approach for wheat seed variety classification.
作者
刘硕
赵鑫
朱启兵
黄敏
郭新年
Liu Shuo;Zhao Xin;Zhu Qibing;Huang Min;Guo Xinnian(Key Laboratory of Advanced Process Control for Light Industry(Ministry of Education),Jiangnan University,Wuxi 214122;Jiangsu Province Poultry Intelligent Breeding and Intelligent Equipment Engineering Research Center,Suqian 223800)
出处
《中国粮油学报》
北大核心
2025年第9期209-216,共8页
Journal of the Chinese Cereals and Oils Association
基金
国家自然科学基金项目(62205128)
江苏省家禽智慧养殖与智能装备工程研究中心开放课题项目(2023KF01)。
关键词
小麦种子
分类模型
高光谱
特征融合
光谱重构
wheat seed
classification model
hyperspectral
feature fusion
spectral reconstruction
