Cucumber detection for precision agriculture applications

Abstract

The objective of this research was to explore the feasibility of detecting cucumber fruits in field conditions for autonomous robotic harvesting applications. A high resolution colour camera and a time-of-flight camera are proposed as primary sensors for the design of the sensory system. The preliminary detection algorithm includes a pixel-based classifier that labels areas of interest that belong to cucumber fruits and a registration procedure that combines the results of the aforementioned classifier with the range data provided by the time-of-flight camera. The detection algorithm is extremely simple and efficient, and provides a satisfactory discrimination of the cucumbers fruits with respect to the rest of the elements of the scene. Several experimental tests have been carried out in outdoor conditions in order to evaluate and demonstrate the capabilities of the proposed approach.