CoCoSL: Agricultural Solutions Using Image Processing for Coconut Farms in Sri Lanka

Uthayakumaran Uthayarakavan, Kamalanathan Thakshayini, Sivakumar Krushi Yadushika, Ananthasiri Gowthaman, Rrubaa Panchendrarajan

Abstract: Coconut production is the most vital and one in all the most sources of income within the Sri Lankan economy. In addition to being a staple food, it generates the third-highest amount of foreign exchange for the government. However, various factors such as diseases that affect the coconut at different stages, delayed identification of growth stages, and poor estimation of quantities of primary products of coconut can result in a great fall in the expected yield. Finally, the farmers may either give up their effort on the plantations or show less interest in nurturing them. While existing solutions focus only on the identification of diseases, we propose a comprehensive solution, CoCoSL for coconut farming to enhance the end-to-end process. The solutions are grouped into four major components, disease management, growth management, quality management, and quantity management. The proposed CoCoSL system can automatically predict the diseases in coconut leaf and coir plants, nutrient deficiency in coir plants, growth stages of coconut and bloom stem, the freshness of coconut meat shell, and quantity of coconut milk and oil. Evaluations performed using the data gathered across various coconut farms in Sri Lanka shows, CoCoSL perform the prediction task with an accuracy range of 81.25% to 93.75%.