Abstract: Residual plastic film negatively impacted soil quality and crop growth, making efficient assessment of its distribution a crucial issue. A deep learning-based detection method using YOLOv8 for the automatic identification of residual plastic film in UAV-captured RGB images was proposed. The research utilized farmland data collected from Huaxing Farm in Changji, Xinjiang, with annotations performed using LabelMe, followed by data preprocessing and augmentation. Experimental results demonstrated that YOLOv8 performed well in residual plastic film detection, achieving an mAP@0.5 of 86.5%, enabling efficient and accurate detection, providing technological support for agricultural environmental monitoring and pollution management.

Key words: residual plastic film, UAV, monitoring, YOLOv8