Selection of the most efficient wavelength bands for ‘Jonagold’ apple sorting

This paper presents a method based on quadratic discriminant analysis to select the best filters for detecting a wide range of defects in ‘Jonagold’ apple fruit using a multi-spectral vision system. Reflectance spectra of damaged and sound tissue were recorded using a visible/NIR spectrometer. Analysed defects consisted of scald, hail damage (with and without skin perforation), limb rubs, russets, scab tissue, frost damage, rot, visible flesh damage and recent bruises. Camera filter effects were approximated by summing the reflectances of all the wavelengths within the filter bandwidth. Combinations of three and four filters were tested and evaluated for discriminating damaged tissues from healthy ones. If a three-filter combination appeared sufficient to detect most of the damaged tissue, a four-filter combination should be considered for the complete sorting automation of this bicolour apple variety. A fourth filter was necessary to quantify the ratio between the blush and ground colours. Regarding recent bruise defects which represented the major difficulty, an image segmentation algorithm based on local contrast variations can enhance their detection.