An Intelligent IoT and Deep Learning-Based System for Early Detection and Root Cause Analysis of Black Pepper Diseases

Abstract

Black pepper (Piper nigrum L.) is an economically important spice crop whose cultivation is severely affected by Quick Wilt/Foot Rot, Slow Decline, Anthracnose (Pollu Disease), Yellowing, Root Rot, Leaf Blight, and viral symptoms. Visual diagnosis is difficult because symptom expression changes with lighting, disease stage, environmental stress, soil conditions, and drainage. This article combines an integrative review of artificial intelligence, computer vision, Internet of Things (IoT), wireless sensor networks, rootcause analysis, and decision-support methods with preliminary prototype validation of the CropGuard framework. The prototype integrates black pepper leaf-image validation, YOLOv8 disease localization, environmental sensing, XGBoost-based root-cause inference, and structured recommendation generation. The image collection contained more than 15,000 black pepper leaf images across disease and health categories. On a validation set of 163 images containing 1,603 annotated disease instances, the YOLOv8 detector achieved a precision of 0.747, recall of 0.720, and mAP@0.5 of 0.739. Class-wise AP@0.5 was 0.9342 for Yellowing, 0.8512 for Leaf Blight, 0.7002 for Quick Wilt, 0.6246 for Pollu Disease, and 0.5844 for Yellow Mottle Virus. A separate tabular dataset of 15,000 environmental records was prepared for the XGBoost root-cause component. The current revision reports functional integration of disease detection, root-cause output, and recommendation generation, while quantitative held-out evaluation of the root-cause model, sensor-network performance, and farmer usability remains future work. The findings support multimodal, explainable, and farmer-oriented disease diagnosis, while also showing that real-field validation, class balancing, sensor calibration, and agronomist verification are essential before operational deployment.