基于近红外高光谱的松子品质分类  

Pine Nut Quality Classification Based on Near-Infrared Hyperspectroscopy

作　　者：陈雨博 CHEN Yubo(College of Mathematics and Computer Science,Zhejiang A&F University,Hangzhou 311300)

机构地区：[1]浙江农林大学数学与计算机科学学院,杭州311300

出　　处：《食品工业》2025年第2期324-327,共4页The Food Industry

摘　　要：带壳松子品质的快速无损检测有助于指导松子销售。因此,建立一种快速、可靠的带壳松子品质检测方法显得尤为重要。此次研究探讨了近红外高光谱成像检测松子品质的可行性。研究利用近红外高光谱成像仪获取了500颗松子的900~1 700 nm的高光谱图像。对提取的光谱数据进行多种方法的预处理和特征选择,分别使用传统机器学习方法和卷积神经网络模型进行分类,并加入注意力机制提高分类精度。传统机器学习模型总体分类精度最高仅有60.80%,而使用自注意力机制的3D-2DCNN模型的总体精度能达到94.40%,对三类松子的分类精度分别为93.33%, 93.75%和95.83%。研究表明,高光谱成像、卷积神经网络和注意力机制相结合,能实现松子的质量评估,确保向消费者提供更高质量的产品。Rapid nondestructive testing of shelled pine nut quality can help guide pine nut marketing.Therefore,it is particularly important to establish a fast and reliable method for detecting the quality of shelled pine nuts.In this study,the feasibility of near-infrared hyperspectral imaging for detecting the quality of pine nuts is investigated.In this study,a near-infrared hyperspectral imager is utilized to acquire hyperspectral images of 500 pine nuts at 900-1700 nm.The extracted spectral data are preprocessed and feature-selected by multiple methods,and are classified using traditional machine learning methods and convolutional neural network models,respectively,with the addition of an attention mechanism to improve the classification accuracy.The overall classification accuracy of the traditional machine learning model is only 60.80%at the highest,while the overall accuracy of the 3D-2DCNN model using the self-attention mechanism can reach 94.40%,and the classification accuracies of the three types of pinecones are 93.33%,93.75%,and 95.83%,respectively.The study demonstrates that the combination of hyperspectral imaging,convolutional neural networks,and the attention mechanism enables quality assessment of pine nuts and ensures that higher quality products are provided to consumers.

关 键 词：近红外高光谱 松子 品质检测 机器学习 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]