基于机器视觉的小麦不完善粒快速分析仪适用性评价  

Evaluation of Applicability of a Machine Vision-based Rapid Analyser for Unsound Wheat Kernels

作　　者：李冰杰 王超群 赵小萌 陈晶 叶金 李丽 高树青 姜平 王松雪 Li Bingjie;Wang Chaoqun;Zhao Xiaomeng;Chen Jing;Ye Jin;Li Li;Gao Shuqing;Jiang Ping;Wang Songxue(Academy of National Food and Strategic Reserves Administration,Beijing 100037;COFCO International Trading Company Limited,Beijing 2100005;Beijing Science and Technology Innovation Promotion Center,Beijing Municipal Science&Technology Commission,Administrative Commission of Zhongguancun Science Park,Beijing 100036)

机构地区：[1]国家粮食和物资储备局科学研究院,北京100037 [2]中粮贸易有限公司,北京100089 [3]北京市科学技术委员会,中关村科技园区管理委员会,北京科技创新促进中心,北京100036

出　　处：《中国粮油学报》2024年第10期174-179,共6页Journal of the Chinese Cereals and Oils Association

基　　金：国家重点研发计划项目(2023YFF1104605);全国粮食和物资储备领军人才项目(LJ2022201);全国粮食和物资储备青年拔尖人才项目(QN2022202)。

摘　　要：小麦不完善粒是评价小麦质量的关键指标,本研究对基于机器视觉的小麦不完善粒快速分析仪开展全面的适用性评价。包括对开发的全自动仪器检测小麦不完善粒含量的可靠性、重复性、稳定性、台间差、与标准方法进行对比等参数进行分析。结果显示:可靠性检t值为-0.14,P值为0.887,大于0.05,仪器的测定结果与国家标准方法(GB/T 5494—2019)的人工测定结果之间无显著性差异,仪器检测可靠性通过检验;重复性反差分析P值为1.000,大于0.05,重复性X 2值均小于X 2临界值11.07,极差值均小于临界极差值0.71%,重复性通过检验;稳定性ADF检验统计量的t值为-6.48,小于显著性水平为1%情况的t值-4.58,稳定性数据的X 2值为12.8,临界值为16.92,稳定性极差临界值为0.8%,本次数据的极差为0.75%,稳定性通过显著性检验;台间差数据的检验结果的t值为-0.92,P值为0.373,P值大于0.05,说明在5%的显著性水平下不能拒绝2种方法差值为零的原假设;该仪器测定与人工方法对比P值为0.887,大于0.05,且克服了人工方法主观性强、费时费力等不足,极大提高了准确性和时效性。Unsound wheat kernels are a key indicator for evaluating wheat quality.In this study,a comprehensive applicability evaluation of a machine vision-based rapid analyser was implemented for unsound wheat kernels.The parameters,including reliability,repeatability,stability,inter-table variation,and comparison with standard methods,were analysed for the development of a fully automated instrument for the detection of unsound wheat kernels.The results indicated that:the t-value of the reliability test was-0.14,the P-value was 0.887,greater than 0.05,and there was no significant difference between the measurement results of the instrument and the manual measurement results of the national standard method(GB/T 5494—2019),so the reliability of the instrument test passed the test;the P-value of the inverse contrast analysis of the repeatability was 1.000,greater than 0.05,the repeatability of the X 2 values were all less than the X 2 critical value of 11.07,the extreme deviation values were all less than the critical extreme deviation value of 0.71%,and the repeatability passed the test;the t-value of the ADF test statistic for stability was-6.48,smaller than the t-value of the case of a significance level of 1%-4.58,the X 2 value of the stability data was 12.8,the critical value was 16.92,the critical value of the extreme deviation of the stability was 0.8%,the extreme difference of this data was 0.75%,and the stability passes the significance test;the t-value of the test result of the inter-station difference data was-0.92,and the P-value was 0.373,and the P-value was greater than 0.05,indicating that the original hypothesis that the difference between the two methods was zero and cannot be rejected under the significance level of 5%;the P-value of this instrumental determination compared with the manual method was 0.887,greater than 0.05,and it overcame the manual method.The P-value of this instrument was 0.887,greater than 0.05,and it overcame the shortcomings of the manual method,such as strong subjectivity,time-con

关 键 词：小麦不完善粒检测仪 可靠性 重复性 稳定性 台间差 人工 

分 类 号：TS210.7[轻工技术与工程—粮食、油脂及植物蛋白工程]