Precision Agriculture Revolutionized: UAV-Based Detection of Leaffolders and Stem Borers in Ricefields using EfficientNetB1

The need for efficient pest management in rice cultivation is crucial due to the significant damage caused by pests like leaffolders and stem borers. Conventional manual inspection and sensory detection methods are labor-intensive and often inaccurate, leading to delayed and ineffective pest control measures. This study developed a system that uses unmanned areal vehicle (UAV) and machine learning algorithms to track the leaffolder and stem borer in rice fields. This system allows farmers to track the leaffolder and stem borer effectively and reliably without physically inspecting their entire field to determine the percentage affected by leaffolder and stem borer in the ricefield. Results indicate varying degrees of performance across different machine learning algorithms, with Convolutional Neural Network (CNN), consistently demonstrating superior accuracy, notably achieving 90.89% accuracy with EfficientNetB1. SVMs also exhibit commendable performance, excelling in complex decision frontier description. Decision trees, random forests, and transfer learning present nuanced performance characteristics, each offering unique strengths and limitations. The developed system tracked the disease between 73.22% and 90.89% using five groups with five tests per group. This study demonstrates the system’s effectiveness and opens avenues for further research and application in other areas of precision agriculture.