Quantifying the spatial distribution of avocados for optimising yield estimation : a thesis presented in fulfilment of the requirements for the degree of Master of Engineering in Mechatronics at Massey University, Manawatū, New Zealand

Abstract

This thesis addresses the need for accurate yield estimation in commercial avocado orchards, leveraging insights from prior research on AI technology for avocado counting. The literature review identified a gap in understanding the spatial distribution of avocado fruit, leading to the development of a 3D positional data collection methodology. In this project a coordinate mapping device employing a laser distance measure and high precision rotary encoders was developed, tested and later implemented in a commercial Hass avocado orchard, resulting in a unique 3D positional dataset of 2,909 measured avocados from 27 trees. The results provide valuable insights into how avocados are distributed on the trees, leading to recommendations for an ideal image capturing procedure. It was found that 98% of all fruit would be included in a field of view extending out to half the row spacing along the row direction and 2.4 m in any other direction from the tree's base. Challenges such as occlusion prompt the need for further research, including simulations and trials based on the collected dataset. Recommendations are made to guide future research towards refining image-capturing procedures for implementing image-processing AI, leading to efficient and accurate individual tree counts. These counting methods can be employed in conjunction with traditional or satellite-based block sampling techniques, facilitating an accurate overall block yield estimation and providing growers with the necessary insights to make informed commercial harvest decisions.