Pest Detection in Smart Farms: Integrating Sound, Deep Learning, and IoT

The importance of agriculture in the survival of human beings in the modern day makes it imperative to maximize crop yield. There is a risk that pest infestations can cause damage or retard the growth of crops and thus negatively impact agricultural productivity. It is essential to identify pests early and do so accurately to preserve crops and promote sustainable agriculture. The proposed work introduces a new paradigm for pest detection by integrating pest sound analysis, deep learning models, and IoT-enabled sensor networks to support real-time surveillance over large-scale crops. The system we developed uses distributed acoustic sensors to record insect and pest sound frequencies, send the sounds over an IoT network, and process them in a centralized processing unit. A deep model called Alex-Net is used to classify audio features extracted from test samples; thus, the pest species can be recognized accurately. The pest detection model of Alex-Net was trained by the pest audio sounds pre-processed by the Weiner filter, DPSS, non-overlap-add methods, and features extracted by LFCC and PNCC methods. This research reveals the dream of IoT and AI concerning pest control on agricultural land. The Alex-Net proposed achieved high performance in automating pest recognition from agricultural field experience. The proposed model outperformed benchmark models APNet, Pest-PVT, VOLOv5m6, G-DINO, and YOLO-YSTs B3 in pest detection with an accuracy of 99.76%, recall 99.95% and F1 score 99.95%.