Research on Air Pollution Smoke Detection Method Based on Improved YOLOv8
The detection of atmospheric pollution smoke plays a crucial role in the field of environmental monitoring,enabling the accurate identification of smoke emission sources.This paper proposes a novel atmospheric pollution smoke detection mod-el,YOLOv8n-SC.Firstly,the Slim-Neck network is employed to enhance the neck section of YOLOv8n,significantly re-ducing redundancy,simplifying the model complexity,and accelerating the detection speed.Secondly,the original model's up-sampling operator is improved by replacing the nearest neighbor operator with CARAFE,resulting in more precise sampling results and finer localization information.Finally,a smoke dataset is established,and the Copy-Pasting method is used to augment the dataset,generating new samples with subtle variations to expand the training dataset and enhance model perform-ance.The research findings indicate that compared to the original YOLOv8n model,the YOLOv8n-SC model achieves a 6.38％reduction in the number of parameters and a 2.7％improvement in mean average precision.This model is not only com-pact and easy to deploy but also meets the requirements for detection accuracy.
air pollutionsmoke detectionYOLOv8 modelYOLOv8n-SC model
NETL
NSTL
万方数据
