Development of a multispectral imaging system for online detection of bruises on apples
Development of a multispectral imaging system for online detection of bruises on apples
复制标题
开发用于在线检测苹果碰伤的多光谱成像系统
DOI：
10.1016/j.jfoodeng.2014.09.002
复制
发表时间：
2015-02-01
影响因子：
5.5
通讯作者：
Chen, Liping
中科院分区：
文献类型：
--
作者：
Huang, Wenqian;Li, Jiangbo;Chen, Liping
苹果损伤检测是苹果自动分拣系统的重要而又困难的一环。本研究的目的是开发一个原型的多光谱成像系统，用于在线检测苹果上的瘀伤。建立了一套波长范围为325-1100 nm的高光谱成像系统，以筛选富士苹果损伤检测的有效波长。分别对450-1000 nm、450-780 nm和780-1000 nm高光谱图像进行了分段主成分分析（PCA），并对所得主成分图像进行了比较。通过最佳PC图像的加权系数确定了780、850和960 nm三个有效波长。选择的有效波长的性能进行了评估，由183个苹果。对500 mm高度的钢球跌落损伤苹果在1、12和24 h内的检测准确率分别为90.4%、92.3%和92.3%。对200、300、400和500 mm高度的钢球在1 h内跌落造成的苹果损伤的检测准确率分别为92.4%、96.2%、96.2%和94.7%。在此基础上，研制了一套由两个二向色分束器、两个带通滤光片和三个基于棱镜的2CCD多光谱逐行面扫描相机组成的多光谱成像系统。在线图像采集速度约为每秒3个苹果。所开发的系统的性能进行了评估，使用59个独立的苹果静态和在线测试。在静态和在线测试中，压伤苹果的分类准确率分别为91.5%和74.6%。结果表明，基于高光谱成像系统和分段主成分分析方法，基于所选有效波长的多光谱成像系统可用于苹果损伤的在线检测。(C)2014爱思唯尔有限公司版权所有。
Detection of bruises on apples is important but difficult for an automatic apple sorting system. The objective of this study was to develop a prototype multispectral imaging system for online detection of bruises on apples. A hyperspectral imaging system with the wavelength range of 325-1100 nm was built to select the effective wavelengths for detecting bruises on 'Fuji' apples. Segmented principal component analysis (PCA) was conducted on the 450-1000 nm, 450-780 nm and 780-1000 nm hyperspectral images respectively and the resultant principal component (PC) images were compared. Three effective wavelengths 780, 850 and 960 nm were determined by the weighing coefficients of the best PC images. The performance of the selected effective wavelengths was evaluated by 183 apples. The detection accuracies of apples with bruises created by dropping a steel ball from a 500 mm height in 1, 12 and 24 h were 90.4%, 923% and 92.3%, respectively. The detection accuracies of apples with bruises created by dropping a steel ball from 200, 300, 400 and 500 mm heights in an hour were 92.4%, 96.2%, 96.2% and 94.7%, respectively. Then a multispectral imaging system with two dichroic beamsplitters, two band-pass filters and three prism-based 2CCD multispectral progressive area scan cameras was developed. The online image acquisition speed was about 3 apples per second. The performance of the developed system was evaluated by static and online tests using 59 independent apples. The classification accuracies of bruised apples in static and online tests were 91.5% and 74.6%, respectively. Results showed that the developed multispectral imaging system based on the selected effective wavelengths using a hyperspectral imaging system and segmented PCA could be used for online detection of bruises on apples. (C) 2014 Elsevier Ltd. All rights reserved.