Hyperspectral Image Analysis for Measuring Ripeness of Tomatoes

The latest developments in optics and sensors allow imaging spectrometry: the creation of hyperspectral images, i.e. images with wavelength-specific measurements over a large part of the spectrum. In this study, hyperspectral images of several stages of ripeness of tomatoes were recorded and analyzed. The electromagnetic spectrum between 450 and 850 nm was recorded in 80 bands (every 5 nm). Results show that hyperspectral images offer more discriminating power than standard RGB-images for discriminating ripeness stages of tomatoes, reducing the classification error of individual pixels from 40% to 7%. Using a reference tile, the images can be made invariant to the light source and even object geometry without losing discriminating power. This makes it possible to have comparable classification results under a large range of conditions.