When it comes to Asian countries, rice is the most common type of food that is consumed daily. Due to that rice farmers face a huge amount of stress to supply according to the massive demand. This is happening while they are farming in poor conditions such as, amongst diseases and pests that harm rice crops with the inclusion of weeds that plague the field. They also have difficulties finding the correct fertilizers and the amount that is needed for the crops to grow properly. Another issue discovered, was that some rice plants are underdeveloped, and farmers lack the understanding when it comes to proper treatment. These topics were chosen according to a multitude of statistics including losses due to all insects, losses due to all diseases, losses due to all weeds, potential production harvested, and total potential production lost before harvest, being found respectively at 34.4%, 9.9%, 10.8%, 44.9%, and 55.1%. The main objective of this research is to provide solutions to the above-mentioned issues faced by the farmers. Four CNN models were compared in previously mentioned four areas to provide solutions. The used models were 2 customized CNNs that were best fit for pest and fertilizer management, a customized AlexNet for growth management, and a customized ResNet model for weed management. To map the weed, u-net, FCN architecture was used and calibrated, with it providing 94.88% training accuracy and 64.39% validation accuracy. Hight measuring and area calculation was implemented and then finetuned using a custom-made python operation. The aim is to develop a mobile application that will help farmers solve these problems using the chosen algorithms. The application will use image processing to analyze crops to find solutions stored in a cloud database. Then machine learning and deep learning will be used to recommend appropriate solutions.

Farming rice crops agriculture harvest