Towards Autonomous Visual Navigation in Arable Fields

Alireza Ahmadi,Michael Halstead,Chris McCool

DOI: https://doi.org/10.48550/arXiv.2109.11936

2022-07-01

Abstract:Autonomous navigation of a robot in agricultural fields is essential for every task from crop monitoring to weed management and fertilizer application. Many current approaches rely on accurate GPS, however, such technology is expensive and also prone to failure (e.g. through lack of coverage). As such, autonomous navigation through sensors that can interpret their environment (such as cameras) is important to achieve the goal of autonomy in agriculture. In this paper, we introduce a purely vision-based navigation scheme that is able to reliably guide the robot through row-crop fields without manual intervention. Independent of any global localization or mapping, this approach is able to accurately follow the crop-rows and switch between the rows, only using onboard cameras. With the help of a novel crop-row detection and a novel crop-row switching technique, our navigation scheme can be deployed in a wide range of fields with different canopy types in various growth stages with limited parameter tuning, creating a crop agnostic navigation approach. We have extensively evaluated our approach in three different fields under various illumination conditions using our agricultural robotic platform (BonnBot-I). For navigation, our approach is evaluated on five crop types and achieves an average navigation accuracy of 3.82cm relative to manual teleoperation.

Robotics,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve autonomous visual navigation of robots in farmland. Specifically, many current methods rely on accurate Global Positioning System (GPS), but this technology is costly and may be affected by insufficient coverage. Therefore, achieving autonomous navigation through sensors (such as cameras) that can interpret the environment is crucial for agricultural automation. This paper proposes a purely vision - based navigation scheme, which can use computer vision and signal processing techniques to guide the robot through crop rows without human intervention. This method is independent of any global positioning or map building, can accurately follow crop rows and switch between rows, using only on - board cameras. In addition, this method is applicable to a variety of crops at different growth stages and with different canopy shapes, and has been tested under different lighting conditions, demonstrating its reliability and accuracy in practical applications.