Abstract:
Inefficient rubber tapping techniques among inexperienced farmers in Sri Lanka are
causing significant challenges in the natural rubber industry, leading to reduced
yields and tree trunk damage. This issue threatens the sustainability and economic
viability of rubber production in the country. While various automated tapping
technologies exist, there remains a critical gap in accessible, intelligent tools to
guide farmers toward precise, tree-friendly tapping methods. This research aims to
develop the Rubber Tapping Assistant Program (RTAP), an innovative application
leveraging image recognition and machine learning to provide real-time guidance
on optimal rubber-tapping techniques. The primary objectives include collecting a
comprehensive dataset of rubber tree images, developing accurate prediction
models for tapping techniques, and creating a user-friendly mobile application for
farmers. The study uses a multi-layered methodology, combining data collection
from rubber plantations, the development of machine learning models using the
YOLO algorithm, and the creation of a mobile application interface. RTAP includes
a tapping location prediction model using Random Forest, XGBoost, and AdaBoost
regressors. A tree distance prediction model utilizes the YOLOv8 classification
model. Tree Region of Interest (ROI) point prediction and tapped area damage
prediction models employing YOLOv8s, YOLOv9m, and YOLOv10s detection
models. These models work collectively to provide comprehensive guidance on
optimal tapping practices. Initial results show a mean square error of 0.0073 for
tapping location, 99% accuracy for tree distance, and average accuracies of 80%
and 60% for ROI and damage models, respectively. The damage prediction model
is designed to assess trunk health and suggest appropriate tapping strategies. This
research is expected to bridge the gap between traditional knowledge and modern
technology in rubber farming, potentially transforming tapping practices and
improving the economic stability of Sri Lanka's natural rubber sector.
