Assessing UAV-based wild blueberry plant height mapping - A consideration for wild blueberry harvester automation

Abstract

Mechanical harvesting of wild blueberries on uneven topography is challenging for the operators, emphasizing the need for harvester header automation. This research aims to develop accurate Digital Elevation Models (DEMs) and map plant height using a DJI Matrice-300 RTK Drone equipped with two distinct DJI sensors, Zenmuse P1 and Zenmuse L1. Accuracy assessment involved comparing drone-generated DEMs with elevation data from a Global Navigation Satellite System receiver, Emlid RS2. Results indicate that the P1 camera outperformed the L1 sensor in estimating DEMs within wild blueberry fields, with root mean square error values ranging from 0.36 to 1.04 cm and 1.29 to 1.70 cm for the P1 and L1 sensors, respectively. Furthermore, P1 demonstrated a significantly lower standard deviation than L1 in selected fields. The P1 imagery dataset determined wild blueberry plant heights with standard deviations of 3.66 cm and 4.77 cm for Field-A and Field-B, respectively. The precise DEMs and plant height information can be integrated into the control system of the wild blueberry harvester, facilitating automation that accounts for spatial variations in plant height, slope, and fruit zone within fields. Implementing an automated harvester holds promise for reducing fruit losses, enhancing grower profitability, and alleviating operator stress.