摘要：Infrared-based detection methods have the advantages of non-contact, fast speed, and wide range. In this paper, a gas leakage source localization method combining the Gaussian diffusion model with YOLOv8 is proposed. This approach combines data from a thermoelectric infrared thermometer to enhance detection capabilities. First, the background and foreground images are segmented using the adaptive ViBe algorithm, and YOLOv8 is employed to extract effective gas information. Then, the image frames are overlaid to enhance the details of the gas leakage process, resulting in a normalized heatmap. Finally, the Gaussian diffusion model and gas centroid inversion algorithm are combined to achieve precise localization of the leakage source. Experimental results show that the proposed method improves the accuracy by 2.24 % compared to the original YOLOv8 network. Additionally, it reduces the false positive rate by 11.5 %, outperforming other YOLO-based methods and the DeeplabV3 + network. Under different distance conditions, the method is capable of detecting small gas leakage amounts, demonstrating strong applicability. The localization accuracy of the method has an error of no more than 3 % compared to manually labeled results, indicating a high level of localization accuracy. This method provides a new, high-precision gas leakage detection approach for gas safety monitoring systems in oil fields, gathering stations, and other applications.

关键字：Gas leak source localization; YOLOv8; Background Modeling; Flow Centroid Inversion

ISSN号：0263-2241

卷、期、页：卷258子辑A

发表日期：2026-01-30

影响因子：0.000000

期刊分区（SCI为中科院分区）：二区

收录情况：SCI（科学引文索引印刷版）,EI（工程索引）,SCIE（科学引文索引网络版）

发表期刊名称：MEASUREMENT

通讯作者：沈昊旸,徐鲁帅,于海洋,李凤

第一作者：董绍华,徐晴晴,王明月

论文类型：期刊论文

论文概要：沈昊旸,徐鲁帅,董绍华,徐晴晴,于海洋,王明月,李凤，Source localization of infrared gas leakage based on YOLOv8 and Gaussian diffusion models，MEASUREMENT，2026，卷258子辑A

论文题目：Source localization of infrared gas leakage based on YOLOv8 and Gaussian diffusion models