Cascaded CNN For Early Detection of Plant Diseases Machine Vision Techniques in Smart Agriculture

Plant disease detection is important because plants absorb the greenhouse gases emitted by industries and vehicles and liberate oxygen, ‎to extend human life in the Universe. The biodiversity chain is fully reliant on plant growth and mainly purifies the air and makes it suitable for ‎respiration. Various diseases like Bacterial Spot (BS), Early Blight (EB), Late Blight, Mold (LBM), Leaf Spot (LS), Spider Mites (SM), ‎Target Spot (TS), and Yellow Leaf Curl (YLC) attack the plants, causing retardation in plant growth, thereby offering stress to the plants. ‎Plant stress reduces crop productivity and creates a deadly impact on the economy. Several studies have exhibited that the quality of ‎agricultural products is seriously affected due to plant stress. Plants are stressed due to many reasons like extended use of synthetic fertilizers-‎ers, soil nutrients, plant diseases, and the various physiological parameters like atmospheric temperature, sunlight, soil pH, and soil moisture ‎content, which may vary seasonally. The benchmark images of the leaves under normal conditions are gathered, and pre-processing is per-‎formed. Further, the feature extraction is done by various types of Convolutional Neural Network (CNN) models like Faster Region-based CNN (FR-CNN), R-CNN, Cascaded CNN, and compared with the standardized CNN model. An optimal solution is offered using ‎Artificial Intelligence (AI) for early-stage detection of chlorophyll content and as well as plant disease prediction. The important advantage ‎of AI AI-based detection approach is to identify the leaf area turning yellow or brown at the start with optimal accuracy‎.