A Design Approach for Identifying, Diagnosing and Controlling Soybean Diseases using CNN Based Computer Vision of the Leaves for Optimizing the Production

Abstract Leaves express the initial symptoms of many diseases in soybean crop. CNN based Computer-Vision (CV) of the leaf-images identifies, enables the diagnosing and controlling the soybean-crop diseases. All three actions need proactive maintenance and Operational Intelligence (OI) for the model used for the control measures. The OI optimizes production. The paper describes a new design approach. Proposed design deploys Computer-Vision (CV), Convolution Neural Network (CNN), Edge AI Computing and the IoT. The approach consists of four-stage framework. Four stages are (1) Cameras in a chain of two sets of four cameras, each at two plants. (2) Edge-AI computing unit for each pair of plants at the field. (3) Edge-Server, which serves multiple networked edge units, one each of the two, sets of cameras. The server also functions as gateway for Internet connectivity with Cloud services. (4) Cloud services for the OI, proactively control the updating CNN model feature metrics, CV functions and camera operations.