Detection of apple bruises caused by picking with the manipulator based on hyperspectral images

The objective of this research was to investigate the potential of using hyperspectral images to detect Red Fuji apple bruises caused by picking with the manipulator. In the test, apples were hold by a manipulator and applied different forces. The contact areas of the apples to the manipulator were marked, and the hyperspectral images were taken, then the apples were stored in the refrigerator. The hyperspectral camera with the spectral region between 372 nm and 1007nm was used. A series of hyperspectral images were taken. Average gray values of pixels selected from the bruise region (contacted region) in the marked contact area and the normal region (non-contacted region) outside marked contact area were calculated and then calibrated to obtain the reflectance respectively. To find the relationship between bruises and the forces, the spectral characteristics from hyperspectral images of the apples applied different forces at the same time were compared. The spectral characteristics from the hyperspectral images of apples from different storing time were also compared to observe changes over time after the forces applied by the manipulator. In general, the reflectance data from bruise regions of the apple were generally lower than those from normal region on the same color side at the same storing time. The standard deviation of the mean gray value was calculated and then extracted the maximum and minimum value. The images at 420 nm, 588 nm, 674 nm, and 900nm were then combined to draw outline of the apple and find the bruise area.