Correction of converging verticals distortion due to camera pose and position

Abstract

This paper discusses the development of a vision-based intra-row navigation system for an agricultural robot. The development of this system was used to reduce the effects of converging verticals distortion due to the inclined, forward facing pose and position of the navigation vision system of the agricultural robot. The mitigation of this distortion was achieved through relating the homography between the image plane view and the ground plane projection of the vision system's field of view (FOV) to create an auxiliary view of the crop. This auxiliary view represented an aerial view of the visible crop area within the camera's FOV. This aerial view was used to discern crop row geometry when used in conjunction with a dual segmentation technique to isolate crop plants from foreign vegetation. Through the implementation of this system, mitigation of converging verticals distortion and successful crop geometry detection for the intra-row navigation of an agricultural robot was achieved.