Development of a multispectral imaging system for online detection of bruises on apples

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.