Machine vision for orchard navigation

Developing a machine vision based autonomous utility vehicle for agricultural application is a challenging task due to changing physical landmarks. While most research thus far has developed algorithms that take advantage of ground structures such as trunks and canopies in the orchard, this research uses the combination of the canopy with the background sky. By focusing on the tree canopy and sky of an orchard row, an unmanned ground vehicle can extract features that can be used for autonomously navigating through the center of the tree rows. This was attempted by using a small-unmanned ground vehicle platform driven by four motors and guided by a machine vision system. The machine vision system is composed of a multispectral camera to capture real-time images and a personal computer to process the images and obtain the features used for autonomous navigation. Laboratory field tests showed that the small vehicle platform system was able to navigate autonomously with an RMS error of 2.35 cm. Field tests using a peach orchard showed that the small vehicle platform system could navigate the rows autonomously with an RMS error of 2.13 cm. The machine vision algorithm developed in this study has the potential to guide small utility vehicles in the orchard in the future.