Abstract

The agriculture sector possesses significant potential to satisfy the increasing global demand for nutritious and wholesome food. Pests significantly damage harvests and diminish quality, making it challenging for farmers to identify them in their fields. Conventional pest detection methods necessitate scientists with substantial field expertise to recognize pests based on their morphological traits accurately. Pesticides adversely affect food and agriculture. The Internet of Things (IoT) technology employs several cost-effective sensors to gather information on pest-related agricultural growth attributes. The study article seeks to create an AI-enabled, real-time, IoT-based autonomous pest identification system utilizing acoustic pest analysis and IoT networks across extensive rural areas. The suggested method employed audio pre-processing techniques for denoising pest sounds, mitigating spectral leaking, transforming interspersed frames to non-overlapping structures, converting time-domain signals to a frequency field, ascertaining the band spectrum, identifying sinusoidal frequencies and internal components, and extracting features, accordingly. Attributes and other statistical data were gathered from 500 pest noises and subsequently trained, verified, and assessed utilizing Machine Learning (ML) networks. The extracted characteristics and several statistical metrics were contrasted during the testing phase. The suggested system incorporates ML training, validation, and assessment approaches.