Precision detection and management of aphid clusters in sorghum using AI and computer vision

Sorghum (Sorghum bicolor) production is seriously constrained by the sorghum aphid (Melanaphis sorghi), which can cause complete crop failure under severe infestations. Current control strategies depend largely on repeated field scouting and the use of broad-spectrum insecticides, both of which are costly, labor-intensive and environmentally unsustainable. This review summarizes recent progress in precision agriculture approaches for aphid management, with emphasis on automated pest detection using computer vision and data-driven models. Recent studies indicate a clear shift toward high-resolution imaging systems, including unmanned aerial vehicles and ground-based robotic platforms fitted with RGB, multispectral and hyperspectral sensors. Comparative analyses show that modern deep learning models consistently outperform conventional image analysis techniques. In particular, object detection frameworks such as YOLO variants and RT-DETR report high detection accuracy, with mean average precision values exceeding 85% for real-time identification of aphid colonies. In parallel, semantic segmentation models, including FESW-UNet, enable accurate pixel-level estimation of infestation intensity, achieving mean intersection-over-union values above 80% by effectively handling background complexity and variable lighting conditions. These developments support a transition from blanket insecticide applications to site-specific interventions, such as variable-rate spraying, which reduces chemical inputs and helps conserve natural enemies. Although challenges remain related to field-level variability, occlusion and model transferability across environments, the use of multi-scale datasets and attention-based learning strategies shows strong potential to address these limitations. Overall, the integration of automated sensing, on-device computation and autonomous field platforms represents a practical pathway toward improving the economic viability and ecological sustainability of sorghum production systems.