Development of an intelligent pest diagnosis and control system for precision agriculture using image recognition and unmanned aerial vehicles

The rapid advancement of intelligent agriculture has intensified the demand for accurate pest diagnosis and precision pesticide application. This study proposes an integrated intelligent pest diagnosis and control system that combines distributed sensing nodes, deep learning–based image recognition, fuzzy pest severity assessment, and unmanned aerial vehicle (UAV)–based variable-rate spraying. Environmental data and pest images captured by sensor nodes deployed in farmland are transmitted to a cloud platform for real-time analysis. An improved YOLOv4 model incorporating squeeze-and-excitation networks (SENet) is developed to enhance pest detection accuracy and robustness under complex field conditions. To address uncertainty in pest severity evaluation, a fuzzy logic–based grading mechanism is introduced, enabling flexible and realistic pest severity classification. Based on the spatial distribution of pest severity, a UAV variable-rate spraying strategy is implemented to achieve site-specific pesticide application. Field experiments conducted on cruciferous crops demonstrate that the system achieves a pest recognition accuracy of 88%. Compared with conventional uniform spraying, the UAV-based variable-rate spraying approach reduces pesticide usage by approximately 27% while decreasing the pest population by about 20%. The results verify the effectiveness, practicality, and applicability of the system for real-world agricultural environments, highlighting its potential for sustainable and precision pest management.