First results of combining RGB segments and multi-spectral pixel classes for strawberry ripeness detection

Abstract

Autonomous harvesting of strawberries is dependent on the detection of fruits and their ripeness. For harvesting in commercial fields, at least unripe, ripe, and overripe fruits must be distinguishable. In this paper, we employ visual image-wise segmentation with multi-spectral pixel-wise ripeness classification on real field recordings. Our results show the feasibility and potential of this combination. Furthermore, we identify problematic parts and necessary improvements.