TL-AGRO: Transfer Learning-based Perilous Insect Detection Framework for Smart Agriculture Underlying Edge Computing

Remote sensing (RS) technology has revolutionized agriculture by harnessing advanced satellites with exceptional temporal and spatial resolution capabilities. These satellites capture Earth’s emitted energy and employ specialized sensors for data collection. Pre-processing techniques refine this data, which is seamlessly integrated into the Edge Device TL-AGRO AI system, transforming precision agriculture. Simultaneously, Unmanned Aerial Vehicles (UAVs) have become vital tools, capturing high-resolution data for plant ecology and Perilous farm insect detection. Our proposed framework TL-AGRO is an advanced agricultural analytics method that leverages meticulously preprocessed data to make precise predictions. It detects insects on agricultural properties and provides accurate classification using CNN Architectures, such as Visual Geometry Group 16 (VGG16), Xception Version 2 (XceptionV2), Residual Network 50 Version 2 (ResNet50V2), and Mobile Network Version 2 (MobileNetV2). Integration of coordinated alert system offers timely recommendations and warnings, empowering farmers to make informed decisions and mitigate insect-related risks. TL-AGRO can even identify individual insect infestations and recommend suitable pesticides, revolutionizing farming with an integrated framework that enhances productivity and sustainability. Performance evaluation of TL-AGRO includes various metrics such as accuracy and loss curves, confusion matrix, F1-score, recall and precision, demonstrating its effectiveness in advancing smart agriculture and ecological studies.