Toward Autonomous Mobile Robot Navigation in Early-Stage Crop Growth

摘要

This paper presents a general procedure for enabling autonomous row following in crops during early-stage
\ngrowth, without relying on absolute localization systems. A model based on deep learning techniques (object
\ndetection for wide-row crops and segmentation for narrow-row crops) was applied to accurately detect both
\ntypes of crops. Tests were performed using a manually operated mobile platform equipped with an RGB and
\na time-of-flight (ToF) cameras. Data were acquired during different time periods and weather conditions, in
\nmaize and wheat fields. The results showed the success on crop detection and enables the future development
\nof a fully autonomous navigation system in cultivated fields during early stage of crop growth.