Automatic insect identification, counting, and monitoring techniques: a review

Abstract

BACKGROUND Insect pests threaten agricultural productivity, while conventional surveillance is labour-intensive and excessive pesticide use creates environmental and health risks. This review evaluated automated insect detection, counting, classification and monitoring technologies, focusing on machine-learning integration, field transferability, and operational readiness.

RESULTS A PRISMA 2020-guided formal search covering publications from 2010 to 2024 identified 313 records relevant to the technological and methodological scope of the review. After title/abstract screening and full-text eligibility assessment, 116 publications were retained. The evidence encompassed camera- and vision-based systems; acoustic, infrared, and optoelectronic sensors; electrostatic and physical approaches; non-lethal photonic monitoring; radar and biosensing technologies; Internet of Things-enabled platforms; and machine- and deep-learning methods. Camera- and vision-based systems were among the most extensively documented and field-oriented approaches, particularly when integrated with remote image acquisition, communication networks, and automated classification. Acoustic systems provided complementary options for continuous or low-power monitoring, whereas radar, biosensing, and other emerging approaches broadened applications toward biodiversity assessment, large-area surveillance, and early detection. Reported performance was not directly comparable across publications because metrics, acquisition conditions, and validation levels differed substantially, ranging from laboratory or dataset benchmarks to semi-field trials, field validations, and operational deployments.

CONCLUSION No single technology is optimal for all monitoring objectives. Operational adoption requires standardized task-specific reporting, independent and sustained field validation, representative datasets, improved transferability across crops and environments, and system-level assessment of robustness, energy autonomy, connectivity, maintenance, and cost. Sensor fusion, edge artificial intelligence, and domain-adaptive or uncertainty-aware models are promising when they provide measurable decision-support benefits. © 2026 The Author(s). Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.