Research and Application of Automated Information Placement in Augmented Reality Precision Agriculture System

Augmented Reality (AR) and computer vision (CV) techniques are now increasingly utilised in precision agriculture (PA). The seamless combination of these two diverse fields offers a potentially new research direction that empowers PA. AR-based PA environment has the natural attributes of high positioning sensitivity of the information and lack of immersive interactive experience for users. Therefore, the accurate placement of information and the conciseness and efficiency of human-computer interaction is critical. Systems that fail to meet these two criteria are considered difficult to use and inefficient. This thesis proposes using CV techniques to assist automatic information placement, which addresses the two critical attributes of the AR-based PA system. The precise location of the crop area in the user's view identified by the CV algorithm aids in automatically placing information in an AR environment. Specifically, this thesis will demonstrate how to use semantic segmentation to determine the position of crops in the field of vision. Several state-of-the-art semantic segmentation algorithms are proposed and bench-marked against each other. The dataset includes 242 crop field images obtained from CONSUS. An AR automatic information placement prototype has been developed to run on tablets. This work demonstrates how AR user interfaces could be placed correctly within an AR-based PA application, which traditionally has been an understudied area of research within AR.