Browsing by Author "Ripa, Leon"

Filter results by typing the first few letters
Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Orchard yield estimation using multi-angle image processing : submitted to the School of Food and Advanced Technology in partial fulfilment of the requirements for the degree of Master of Engineering in Mechatronics at Massey University, Auckland
    (Massey University, 2023) Ripa, Leon
    The rise of autonomous robots and deep learning techniques in recent times has sparked a surge in complex multi robot system (MRS), leveraging these technologies to handle intricate tasks and complement human labour. As the agricultural landscape has evolved from labour intensive, small-scale farming to vast macro-managed expanses, precision agriculture (PA) has emerged to address the challenge. PA offers farmers micro-scale insights into their farms, enabling precise knowledge of pest presence, crop growth variations, and expected yields. For kiwi fruit farmers in New Zealand—spanning over 15,500 hectares and yielding more than 11.65 thousand trays per season—issues persist due to the absence of a mechanical harvesting solution. The inability to accurately estimate yields results in potential profitability concerns, often leading to over, or underemployment during fruit picking. Furthermore, the spread of viruses and diseases poses significant challenges, compelling the need to minimize human intervention and activity under kiwi orchards. Integrating PA techniques not only facilitates fruit counting but also provides crucial insights into fruit density, aiding in identifying underperforming areas for better farm management and potential yield enhancement. This thesis introduces current methods used for orchard yield estimation and presents a novel approach tailored for estimating yields in kiwi fruit orchards. It discusses established solutions for similar agricultural challenges and explores their integration to devise the most effective method for estimating kiwi orchard yields. The proposed solution employs object detection through a convolutional neural network to identify, track, and count kiwi fruits. This is facilitated by images captured by a hexa-drone UAV flown beneath kiwi orchards, ensuring smoother camera capture for increased accuracy in object detection throughout the orchard. This data not only enables farmers to estimate current kiwi production but also aids in identifying sections of orchards that may be overperforming or underperforming.