Improving The Efficiency of Oyster Mushroom Cultivation Using IoT and Deep Learning / Dayangku Azreen binti Pengiran Hj Tajudin

Dissertation Submitted in Fulfilment of the Requirements for the Degree of Master of Science

ABSTRACT
This research aims to enhance the quality and yield of oyster mushroom cultivation in Brunei by leveraging innovative technologies. Brunei's equatorial climate poses challenges for oyster mushroom cultivation, with constant heat and humidity hindering optimal growth conditions. As a result, Brunei relies on mushroom imports to meet domestic demand.
However, recognizing the potential of agriculture, particularly mushroom cultivation, the government has incentivized local youths to engage in agribusiness through grants and support. This study focuses on two primary objectives: Utilizing IoT for Environmental Monitoring and Deep Learning for Disease Detection. For the first objective, we deploy loT sensors to monitor temperature, humidity, and other relevant parameters within cultivation environments, collect comprehensive datasets, analyse data to identify correlations between environmental conditions and mushroom yield, implement adjustments to cultivation practices, and quantify improvements in yield and quality, stating the experiments have successfully shown a higher yield of 2.17 grams per bag for rack A with loT than rack B without IoT for a small number of bags. For the second objective, we train CNN models using annotated datasets of diseased and healthy mushroom samples, integrate the models into an automated disease detection system, evaluate the system's performance, implement preventive measures based on early disease detection, and quantify the reduction in crop damage and increase in overall cultivation success rates, stating the experiments have successfully shown a higher success rate for using Convolutional Neural Network in disease detection in mushroom cultivation. The experiments reveal a consistent decrease of 0.5% in disease detection percentage each day towards healthier bags, alongside fluctuations of 5.9% in detection percentage between consecutive days for specific bags. Despite these variations, our CNN-based system consistently outperforms manual labelling methods, validating its efficacy in improving disease management strategies in mushroom cultivation. By achieving these objectives, this research endeavours to address the challenges mushroom cultivators face in Brunei and pave the way for sustainable and technology-driven agricultural practices in the region.

Dissertation (Master of Science) - Universiti Teknologi Brunei (2023)

Include bibliographical references