| 92 | 2 | 135 |
| 下载次数 | 被引频次 | 阅读次数 |
为快速、精准、无损地识别苹果品种，促进苹果产业快速发展，凸显高品质品种苹果优势，本研究采用介电频谱技术，以阿克苏地区3个品种的‘红富士’苹果作为研究对象，基于LCR数字测试仪采集了120个苹果样本在0～100 kHz的介电特频谱数据作为原始输入参数，在全频和连续投影算法优选频率条件下，利用误差反向传播网络与极限学习机2种方法建立了品种识别模型，并对模型精度进行了分析比较。结果表明，所建立的模型平均准确率均在80%以上，频率优选条件下的2种模型分类准确率可达到90%。然而由于以上模型方法存在无用信息，导致建立的模型稳定性较差，需要数据预处理、降维等步骤，操作耗时，且过程繁琐，致使分类结果稳定性较差。基于此，设计了一维卷积神经网络品种分类模型，与其他模型相比，一维卷积神经网络分类模型以原始参数作为输入，校正集与预测集中的平均分类准确率分别为98.48%和99.26%,模型稳定性更好，且可简化模型复杂度，改善苹果分类的准确性和稳定性，更适宜于苹果品种的识别。
Abstract：In order to quickly, accurately and non-destructive identify apple varieties, promote the rapid development of the apple industry, and highlight the advantages of high-quality apple varieties, this paper adopted dielectric spectrum technology for apple variety identification. Three varieties of ‘Red Fuji' apples from Aksu region were selected as the research objects, and 120 apple samples were collected using LCR digital testing equipment with dielectric spectrum data at 0-100 kHz as the original input parameters. Under the optimal frequency conditions of full frequency and continuous projection algorithm, a variety recognition model was established by using two methods: error backpropagation network and extreme learning machine. Further, the accuracy of the methods was analyzed and compared. The results show that the average accuracy of the established models is above 80%, and the classification accuracy of the two models under frequency optimization conditions can reach 90%. However, the above model methods contain redundant information which undermines the stability of the model. Based on this, an one-dimensional convolutional neural network variety classification model was designed. Compared with other models, the new model uses raw parameters as input, and the average classification accuracy in the correction set and prediction set is 98.48% and 99.26%, respectively. The new model has better stability, simplifies model complexity, improves the accuracy and stability of apple classification, and is more suitable for identifying apple varieties.
[1] 李育农.苹果属植物种质资源研究[M].北京：中国农业出版社，2001:35-36.
[2] 王泽，谷海斌，梁燕，等.阿克苏两种红富士苹果土壤肥力、生长与品质调查研究[J].天津农业科学，2019,25(10):49-54.
[3] LI X N,HUANG J C,XIONG Y J,et al.Determination of soluble solid content in multi-origin ‘Fuji’ apples by using FT-NIR spectroscopy and an origin discriminant strategy[J].Computers and electronics in agriculture,2018,155:23-31.
[4] GUTHRIE J A,LIEBENBERG C J,WALSH K B.NIR model development and robustness in prediction of melon fruit total soluble solids[J].Australian journal of agricultural research,2006,57(4):411-418.
[5] BAI Y H,XIONG Y J,HUANG J C,et al.Accurate prediction of soluble solid content of apples from multiple geographical regions by combining deep learning with spectral fingerprint features[J].Postharvest biology and technology,2019,156:110943.
[6] RONG D,WANG H Y,YING Y B,et al.Peach variety detection using VIS-NIR spectroscopy and deep learning [J].Computers and electronics in agriculture,2020,175:105553.
[7] LUNA A S,DA SILVA A P,DA SILVA C S,et al.Chemometric methods for classification of clonal varieties of green coffee using Raman spectroscopy and direct sample analysis[J].Journal of food composition and analysis,2019,76:44-50.
[8] LLAVE Y,TERADA Y,FUKUOKA M,et al.Dielectric properties of frozen tuna and analysis of defrosting using a radio-frequency system at low frequencies[J].Journal of food engineering,2014,139:1-9.
[9] 李东博.基于介电特征的苹果品质分级方法研究[D].杨凌：西北农林科技大学，2021.
[10] LI C H,LI L L,WU Y,et al.Apple variety identification using near-infrared spectroscopy[J].Journal of spectroscopy,2018,2018:6935197.
[11] 李小龙，于世辉，靳小真，等.基于电学特性的库尔勒香梨果实重量检测研究[J].塔里木大学学报，2021,33(3):85-89.
[12] 于世辉，唐玉荣，牛希跃，等.基于电学特性的库尔勒香梨振动损伤检测[J].塔里木大学学报，2022,34(1):77-87.
[13] 王转卫，赵春江，商亮，等.基于介电频谱技术的甜瓜品种无损检测[J].农业工程学报，2017,33(9):290-295.
[14] LIU D Y,WANG E F,WANG G L,et al.Non-destructive sugar content assessment of multiple cultivars of melons by dielectric properties[J].J Sci Food Agric,2021,101(10):4308-4314.
[15] CHA Y J,CHOI W,BüYüK?ZTüRK O.Deep learning-based crack damage detection using convolutional neural networks[J].Computer aided civil & infrastructure engineering,2017,32(5):361-378.
[16] CAO X G,WANG P,MENG C,et al.Region based cNN for foreign object debris detection on airfield pavement[J].Sensors,2018,18(3):737.
[17] FENG J,LI F M,LU S X,et al.Injurious or noninjurious defect identification from MFL images in pipeline inspection using convolutional neural network[J].IEEE transactions on instrumentation and measurement,2017,66(7):1883-1892.
[18] TONG Z,GAO J,ZHANG H T.Innovative method for recognizing subgrade defects based on a convolutional neural network[J].Construction and building materials,2018,169:69-82.
[19] 周世平，张海红，李海峰，等.基于果品介电特性的无损检测技术研究综述[J].食品研究与开发，2015,36(1):131-134.
[20] 王转卫.基于介电频谱与光谱技术的水果内部品质无损检测方法研究[D].杨凌：西北农林科技大学，2018.
[21] 洪涯，洪添胜，代芬，等.连续投影算法在砂糖橘总酸无损检测中的应用[J].农业工程学报，2010,26(增刊2):380-384.
[22] 郭文川，林碧莹.牛奶含水率介电谱结合化学计量学检测方法[J].农业机械学报，2016,47(9):249-255.
[23] 张楠楠，张晓，王城坤，等.基于高光谱和连续投影算法的棉花叶面积指数估测[J].农业机械学报，2022,53(增刊1):257-262.
[24] 郭文川，朱新华，郭康权，等.桃的电特性及新鲜度识别[J].农业机械学报，2007,38(1):112-115.
[25] 许娜，周炜明.水果分拣机械手控制系统研究：基于BP神经网络和图像识别[J].农机化研究，2023,45(12):35-39.
[26] 王杰，毕浩洋.一种基于粒子群优化的极限学习机[J].郑州大学学报(理学版),2013,45(1):100-104.
[27] HUANG G B,ZHU Q Y,SIEW C K.Extreme learning machine:theory and applications[J].Neurocomputing,2006,70(1/3):489-501.
[28] DENG C W,HUANG G B,XU J,et al.Extreme learning machines:new trends and applications[J].Science in China:information science (English),2015,58(2):1-16.
[29] 李倩倩.无信息变量消除法在三种谱学方法中的定量分析研究[D].北京：中国农业大学，2014.
[30] 商亮，谷静思，郭文川.基于介电特性及ANN的油桃糖度无损检测方法[J].农业工程学报，2013,29(17):257-264.
[31] YOON K.Convolutional neural networks for sentence classification[J].Eprint Arxiv,2014:1746-1751.
[32] ZHANG M I N,WILLISON J H M.Electrical impedance analysis in plant tissues8[J].Journal of experimental botany,1993,44(8):1369-1375.
基本信息:
中图分类号:TP18;S661.1
引用信息:
[1]郝玉梅,花元涛,李文凤,等.基于介电频谱技术的苹果品种识别[J].塔里木大学学报,2024,36(03):98-106.
基金信息:
塔里木大学校长基金硕士项目(TDZKSS202131);塔里木大学大学生创新创业训练项目(22000033635)
2023-09-08
2023
2023-11-22
2024-02-21
2024
1
2024-09-15
2024-09-15