基于近红外光谱技术的大豆种子老化级别快速鉴别方法研究  

Research on Rapid Identification Method of Soybean Seed Aging Grade Based on Near Infrared Spectrum Technology

作　　者：时闯 杨冬风[1] 吕晨曦 SHI Chuang;YANG Dongfeng;LYU Chenxi(College of Information and Electrical Engineering,Heilongjiang Bayi Agricultural University,Daqing Heilongjiang 163319)

机构地区：[1]黑龙江八一农垦大学信息与电气工程学院,黑龙江大庆163319

出　　处：《现代农业科技》2023年第21期126-132,共7页Modern Agricultural Science and Technology

基　　金：黑龙江省自然科学基金(C2018050);大庆市科技局科技项目(zd-2019-38);黑龙江省省属高校基本科研业务费科研项目(ZRCPY201914)。

摘　　要：大豆种子容易发生老化并丧失活力,大豆种子活力检测对目前农业生产具有重要意义。以2020年收获的大豆种子为样本进行人工老化试验,老化时间设置为1、2、3、4、5、6 d,以未老化的种子作为对照组,每个老化等级30个样本。扫描获取全部210条近红外光谱数据,以4∶1的比例划分样本集。对原始光谱数据建立BP网络模型1,再分别采取多元散射校正和标准正态变量对原始光谱进行预处理,建立模型2、模型3。比较3种模型可以发现,预处理技术能缩短模型训练时间,同时可以消除部分噪声,提高模型预测能力,且经过标准正态变量处理后的模型结果较优,由于预处理后的数据维度并未发生变化,模型的训练时间较长,不利于实际应用。因此,采取主成分分析、连续投影法、竞争自适应重加权法对经过标准正态变量处理后的数据进行特征波长变量提取,将光谱数据由原来的1845维降到10维、23维和150维。对经过特征波长变量提取后的数据分别建立BP网络模型,得到模型4、模型5、模型6。综合鉴别上述6种模型,其中模型6的分类准确率达到93.43%,训练时间2.25 s,说明该模型可以较好地实现对7类不同老化级别的大豆种子快速、无损鉴别。Soybean seeds are prone to aging and losing vigor,and soybean seed vigor detection is of great significance to the current agricultural production.Artificial aging experiments were conducted on soybean seeds harvested in 2020,with different aging times(1,2,3,4,5,6 days),non-aged seeds were used as the control group,and 30 samples were set for each aging grade.All 210 near infrared spectral data were scanned and obtained,and the sample sets were divided by the ratio of 4∶1.BP network model 1 was established for the original spectral data,and then multiplicative scatter correction and standard normal variable were used to preprocess the original spectrum respectively to establish model 2 and model 3.Comparing the three models,it could be found that preprocessing techniques could shorten the training time of the model,eliminate some noise,and improve the prediction ability of the model at the same time.The model preprocessed by standard normal variable was superior to that preprocessed by the multiplicative scatter correction.Due to the unchanged data dimensions after preprocessing,the training time of the model was longer,which was not conducive to practical application.Therefore,principal component analysis,continuous projection method and competitive adaptive reweighting method were adopted to extract characteristic wavelength variables from the data after standard normal variable,and the spectral data were reduced from the original 1845 dimension to 10,23 and 150 dimensions.BP network models were established for the data extracted from the characteristic wavelength variables,and model 4,model 5,and model 6 were obtained.By comprehensive appraisement of the above six models,the classification accuracy of model 6 reached 93.43%,and the training time was 2.25 s,which could effectively realize the rapid and nondestructive identification of seven types of soybean seeds with different aging grades.

关 键 词：大豆种子 近红外光谱 活力检测 BP神经网络 

分 类 号：S126[农业科学—农业基础科学] S565.1