Application of YOLO Object Segmentation in Agricultural Pest Monitoring

Agricultural pest monitoring plays a crucial role in the industry's upgrading and development. With the advancement of artificial intelligence (AI) technology, particularly image object detection techniques based on convolutional neural networks (CNNs), the use of AI for rapid and accurate pest identification has become a growing trend. This study constructed a dataset comprising 13 categories of agricultural pests and a total of 3,599 images, and explored two distinct image annotation strategies (pure individual annotation and individual-group hybrid annotation), using YOLOv8x, SSD, and Faster R-CNN models for training. The training results indicate that the pure individual annotation method outperforms the mixed annotation method across all metrics. The YOLOv8x model achieves 0.79 and 0.52 at mAP0.5 and mAP0.5:0.95, respectively, demonstrating the best performance. However, the training results also indicate that insect pests prone to clustering (such as aphids and red spiders) remain a challenge for object detection model. The research findings provide valuable references for enhancing the automation and accuracy of agricultural pest monitoring.