Detection of fruits slight bruises based on hyperspectral imaging technology

Abstract

An identification method to detect the slight bruises and bruises area of fruits based on hyperspectral imaging technology has been proposed. Fragrant pears were used as the experimental object. The fragrant pears' hyperspectral images between 500 nm and 800 nm wavelength were analyzed to acquire the previous six principal component images by the principal component analysis. The weight coefficients of the second, third and fourth principal component image (PC-2 to PC-4) can be used to select six optimal characteristic wavebands (541.7, 625.3, 641.6, 678.2, 725.2 and 777.1 nm). Then, principal component analysis was again performed based on feature wavelengths and the fourth principal component (PC4) image was found to be the most suitable for identifying the slight bruises, the feature of bruise was recognized through appropriate image processing and the bruises on fragrant pears were detected with an accuracy of 89.4%. The methods of direct pixel point and pixel transform were used to detect the bruises area without losing of the fragrant pears slightly bruise regions, the average relative error were 8.17% and 6.97% respectively. The study results show that the hyperspectral imaging technology can qualitatively and quantitatively be used to detect slight bruises on fruits.