Research Projects

Traditional farming methods often rely on manual decision-making for crop selection, irrigation, disease diagnosis and harvesting, which can lead to inefficient land use, water management issues, food security and suboptimal crop choices. These methods are inadequate to meet the growing global demand for food. This challenge is largely driven by rapid population growth, climate change, and poor agricultural practices. The low technological levels, widespread pests, diseases, weeds, and rapid population growth are negatively affecting this sector. These challenges can be addressed by adopting advanced technologies like IoT-based systems, machine learning and deep learning. Despite its high potential, the adoption rate of IoT-based smart farming solutions is still restricted to small-scale farmers in developing economies. This is mainly due to high implementation cost, maintenance complexities, and limited technological expertise among farmers. Hence, in this project, a lightweight, energy-aware and cost-effective method for crop recommendation and early identification of diseases in crops would be designed and developed. Furthermore, to facilitate the deployment of the proposed model, an android/web application would also be developed, which offers farmers a user-friendly interface. The application would also provide farmers with both organic and chemical remedial solutions for the detected disease and poor fertility levels of soil. The importance of the proposed work is listed below: •A cost-effective and lightweight model would be designed and developed that can be trained faster and with fewer parameters, without sacrificing performance. •It would facilitate efficient identification and classification of crop leaf disease in presence of low contrast information, noise and blurriness in the input images. • Thermal imaging would be used in addition to drones as it is best for temperature-related data while drones can capture high resolution data from different angles and altitudes. • Limited research has been done to study the impact of climate change on crop diseases. To address this, different sensors like soil moisture sensor, light sensor, and humidity sensors would be used to adapt to climatic changes. •Fog and edge-based energy-aware system would be designed that would address the challenge of non-availability of network connectivity or high bandwidth network because of the location of agricultural lands in rural areas. Data analysis and storage would be performed in real-time at edge or fog nodes. This would also reduce the latency and computation overhead. •Personalized local model would be designed based on the regional conditions of weather, climate and environmental along with the type and fertility levels of soil. An efficient global model would be designed collaboratively based on the gradients received from various local models. This would not only help in accurate predictions but also help in efficient usage of resources.