外文摘要：Mango is one of the most popular fruits in the world, with a wide variety of types that exhibit significant differences in texture, taste, color, and other aspects. In order to meet the simultaneous demands of the consumers and protect their rights, the identification of mango varieties is particularly important. Near-infrared spectroscopy (NIRS) technology has been widely used in agricultural product identification due to its simplicity, rapidity, efficiency, and environmental friendliness. Moreover, artificial intelligence technology is more conducive to improve the accuracy and efficiency of mango variety identification. This study proposes a convolutional neural network (CNN) model based on channel attention mechanism (MCNN), which can adaptively learn channel weights and can more effectively extract near-infrared spectroscopy features of mangoes by incorporating parallel networks. Compared with traditional deep learning and machine learning models, the MCNN model achieved better performance in mango variety identification with an accuracy of 98.67%. The results indicate that the MCNN model can quickly and accurately recognize mango categories without preprocessing and feature extraction. This method not only enables the identification of mango varieties, but also lays a theoretical foundation for the identification of other agricultural products and related products. In this study, the combination of near-infrared technology and MCNN model provides a new approach for intelligent modeling of spectra. In addition, this method provides a theoretical basis for establishing a fast, non-destructive, and high-precision near-infrared qualitative analysis model, promoting the information processing of agricultural product quality appraisal.
外文关键词：Near-infrared spectroscopy;Convolutional Neural Network;attention mechanism;Variety identification
作者：Li, Yan;Dong, Zhilin;Wang, Jiajia;Sun, Penghui;Ran, Wensheng
作者单位：Xinjiang Univ;Post Doctoral Workstat Xinjiang Uygur Autonomous R
期刊名称：JOURNAL OF FOOD MEASUREMENT AND CHARACTERIZATION
期刊影响因子：0.0
出版年份：2024
出版刊次：18(3)
原文传递申请：江苏省科技资源（工程技术文献）统筹服务平台