江苏大学学报（自然科学版）2024，Vol.45Issue(1)：38-45,8.DOI:10.3969/j.issn.1671-7775.2024.01.006
基于机器视觉与光谱融合的柑橘品质无损检测分级系统设计与试验
Design and experiment of non-destructive testing and grading system for citrus quality based on machine vision and spectral fusion
摘要
Abstract
For the three key quality indicators of citrus fruit diameter,coloring rate and internal sugar level,the citrus comprehensive quality non-destructive testing and grading system was designed based on the double-cone roller fruit cup transmission line with feeding part,machine vision detection module,near-infrared spectroscopy detection module and grading execution part.In the machine vision detection module,the single camera was used to capture videos of rolling citrus for obtaining large number of citrus images in different postures and performing contour extraction.The fruit diameter was calculated with the average value of the minimum external circle diameter of all single citrus frame images,and the average value of its two-dimensional yellow proportion obtained by each frame image was used as the full surface coloring rate.In the near-infrared spectrum detection module,the transmission light path was designed to collect the citrus transmittance spectrum,and the mixed attitude sugar detection model was established according to the two high-frequency attitudes of citrus during on-line detection.Comparing the modeling results under different pretreatment methods,the partial least squares(PLS)model established after applying the more effective multi-scattering correction(MSC)was selected.The on-line testing results show that the maximum absolute error of fruit diameter detection is-1.42 mm with the maximum absolute error of coloring rate detection of 0.048,and the correlation coefficient of the sugar test results is 0.817 with the root mean square error of 0.658%.The joint detection and grading method of internal and external quality determines the joint grading methods of three qualities according to the decision-making method of the discriminant tree.At the sorting speed of 5/s,the average accuracy of the comprehensive grading can reach 91.16%.The overall structure of the detection and grading system is simple with strong applicability for sphere-like fruits,which has great potential for industrial application.关键词
柑橘/无损检测/机器视觉/近红外光谱/分级Key words
citrus/non-destructive detection/machine vision/near infrared spectrum/classification分类
轻工纺织引用本文复制引用
文韬,代兴勇,李浪,刘豪..基于机器视觉与光谱融合的柑橘品质无损检测分级系统设计与试验[J].江苏大学学报（自然科学版）,2024,45(1):38-45,8.基金项目
湖南省自然科学基金资助项目(2020JJ4142) （2020JJ4142）
湖南省林业杰青培养科研项目(XLK202108-7) （XLK202108-7）
湖南省重点研发计划项目(2022NK2048) （2022NK2048）
湖南省教育厅科学研究重点项目(20A515) （20A515）
