Colour image segmentation and labeling through multiedit-condensing

Abstract A new method is proposed for detecting and locating objects of interest within a colour scene under very strong variabilities in lighting conditions, object shape and pigmentation. The method is based on Nearest Neighbour classification and Multiedit-Condensing techniques and is applied to implement the vision subsystem of a robotic citric harvesting device. Experiments and results are reported showing the effectiveness of the method and illustrating its appropriateness to the proposed task.