利用IACCN算法快速检测不同贮藏湿度的稻米营养成分研究  

Study on rapid detection of nutrient components of rice under different storage humidity conditions by using the IACCN algorithm

作　　者：闫衡 YAN Heng(Hanzhong Vocational and Technical College,Hanzhong 723003,China)

机构地区：[1]汉中职业技术学院,陕西汉中723003

出　　处：《粮食与饲料工业》2025年第1期117-121,共5页Cereal & Feed Industry

摘　　要：针对不同贮藏湿度的稻米营养成分检测,采用电子鼻和高光谱技术分别采集稻米的信息,提出了一种基于交错注意卷积块(IACB)和知识蒸馏(KD)技术的交错注意力卷积压缩网络(IACCN)算法,用于提取和处理稻米特征。实验结果表明,针对电子鼻采集数据,在最高湿度70%时,卷积神经网络(CNN)的准确率在85.6%~88.1%,而IACCN的准确率在97.6%~99.55%。在稻米高光谱数据处理和识别中,CNN的平均准确率为65%,IACNN的平均准确率为81%。此外,通过对电子鼻和高光谱数据的融合处理,IACCN算法的性能得到了显著提高,准确率高达99.5%。结果表明,IACCN算法在稻米营养成分检测中具有较高的准确率,尤其在湿度较高的情况下,优势更为明显。此次研究为稻米检测提供了有力支持,有望为稻米产业的发展贡献力量。To detect the nutrient components of rice under different storage humidity conditions,the electronic nose and hyperspectral techniques were used to collect rice information separately.A novel Interleaved Attention Convolutional Compression Network(IACCN)algorithm,based on Interleaved Attention Convolution Blocks(IACB)and Knowledge Distillation(KD)technology,was proposed for extracting and processing rice features.Experimental results showed that for data collected by the electronic nose,at a maximum humidity of 70%,the accuracy of the convolutional neural network(CNN)ranged from 85.6%to 88.1%,while the accuracy of IACCN ranged from 97.6%to 99.55%.In rice hyperspectral data processing and recognition,CNN achieved an average accuracy of 65%,whereas IACCN achieved an average accuracy of 81%.Furthermore,through the fusion of electronic nose and hyperspectral data,the performance of the IACCN algorithm was significantly improved,reaching an accuracy of up to 99.5%.The results indicate that the IACCN algorithm exhibits high accuracy in detecting the nutritional components of rice,with particularly significant advantages under higher humidity conditions.This study provides strong support for rice detection and is expected to contribute to the development of the rice industry.

关 键 词：稻米品质 IACCN 贮藏湿度 电子鼻 高光谱 数据融合 

分 类 号：TS210.7[轻工技术与工程—粮食、油脂及植物蛋白工程]