Image Enhancement Using Augmentation Technique for Pest Detection (Case Study: Ayesha Chili Peppers)

Abstract

Advances in image processing have progressed alongside developments in artificial intelligence (AI), enabling AI-based object detection across diverse domains, including agriculture. In crop cultivation, pests critically affect yields, particularly in bird’s eye chili (Capsicum frutescens). This study focuses on pest detection in the Ayesha Chili Peppers, targeting four pest classes: aphids, thrips, whiteflies, and caterpillars. The research develops an image augmentation technique integrated into the pre-processing stage of the YOLOv8 model to enhance detection precision for small-sized pests. YOLOv8 was chosen for its reliable performance, consistent small-object detection, and compatibility with cloud, edge computing, and IoT-based systems. A dataset of 14,419 images was used, divided into 70 percent training, 20 percent validation, and 10 percent testing. Experimental results demonstrate that the augmented YOLOv8 achieved a mean Average Precision (mAP) of 95.2 percent, compared to 88.7 percent without augmentation. These findings indicate that image augmentation significantly improves detection accuracy for small objects and provides a practical approach for integrating computer vision into precision agriculture.