HIdentifier: A Method in Agriculture CPS Framework to Automatically Identify Disease Hotspots Using Message Passing in Graph

Plant diseases affect the health and yield of the crop which can in turn impact farmers economically and pose a threat to food security across the globe. To protect crops from disease and obtain high yield, many Agriculture Cyber-Physical Systems (A-CPS) have been developed with the ability to detect the presence of a disease. In addition, to effectively control the spread of the disease and reduce its impact on the crops, insights on the spread of the disease and locations of disease hotspots that act as sources for the disease must be identified. HIdentifier proposes a novel graph-based spatial analysis method that takes segmented images of the diseased farmland as input. All the identified disease segments in the input image are represented as a graph to capture the spatial relation between diseased locations. Message passing is performed over the graph to know about the neighbors of each node and effectively identify hotspots. The solution proposed has been validated on images of leaves with disease spots on them annotated and the results were similar to that of visual estimation by humans. These results can be instrumental in automating disease control with the help of Internet-of-Agro- Things (IoAT) for effective disease management.