Vision-based uncut crop edge detection for automated guidance of head-feeding combine

Abstract

This study proposes a vision-based uncut crop edge detection method to be utilized as a part of an automated guidance system for a head-feeding combine harvester, which is widely used in Japan for the harvesting of rice and wheat. The proposed method removes the perspective effects of the acquired images by inverse perspective mapping and recovers the crop rows to their actual parallel states. Then, the uncut crop edges are detected by applying color transformation and the edge detection method. The proposed method has shown outstanding detection performance on the images acquired under various conditions of the paddy field with an average accuracy of 97% and a processing speed of 33 ms per frame.