• An image-based framework is developed to analyse methane leakage dispersion under varying environmental conditions using infrared thermography. • Thirty-four quantitative morphological and texture features are constructed to characterise methane plume structure and diffusion behaviour. • Mutual information analysis reveals non-linear and complementary effects of leakage flow rate, wind speed, and temperature on image features. • Leakage flow rate is identified as the dominant factor governing texture and structural variation under moderate wind conditions. • The proposed approach provides an interpretable alternative to CFD-based analysis for environmental impact assessment of methane leakage in energy systems. In response to the complex evolution of infrared image texture features caused by variations in environmental conditions during methane pipeline leakage, this study conducts an environmental impact analysis based on image texture characteristics. First, a semantic segmentation approach is employed to accurately extract methane leakage regions. On this basis, a feature system comprising 34 grey-level co-occurrence matrix (GLCM) texture features is constructed to quantitatively characterise the structural properties of the leakage regions. Subsequently, in conjunction with experimentally recorded leakage flow rate, wind speed, and ambient temperature, correlation analysis, significance testing, mutual information analysis, and multivariate regression models are applied to systematically investigate the relationships between environmental factors and texture features from both linear and non-linear perspectives.The results indicate that, under experimental conditions with moderate wind speeds, all three environmental factors exert statistically significant effects on texture features, while exhibiting a stable hierarchy of relative importance: leakage flow rate has the most pronounced influence on texture variation, followed by wind speed, with temperature showing the weakest effect. The consistency of conclusions across multiple statistical analyses demonstrates that texture variations in methane leakage images are jointly driven by multiple factors; however, under controlled conditions, leakage flow rate remains the dominant determinant. The findings of this study provide empirical support and a theoretical basis for understanding the formation mechanisms of methane dispersion image features and for subsequent image-based assessment of leakage severity.
Xu et al. (Fri,) studied this question.