Amid the global challenge of carbon emissions (CE), developing a method for accurately measuring city-level carbon emissions is crucial for crafting effective carbon reduction strategies. Therefore, the integration of urban nighttime light data (NTL) and urban industrial structure (IS) characteristics contributes to the characterization of urban carbon emissions. It uses the ISC to reflect urban industrial structure variances across 21 cities and counties in Sichuan Province, establishing a correlation between NTL-ISC and urban carbon emissions through a PSO-SVM model. This approach is evaluated against the conventional NTL method, aiming for precise urban CE quantification. Additionally, the impact of the tertiary sector, population density, and urbanization on CEI and CEC is analyzed. Findings indicate: (1) A comparison between the traditional linear regression method, solely based on NTL (adjusted R² = 0.86), and the machine learning method incorporating NTL-ISC (adjusted R² = 0.89), demonstrates the efficacy of the proposed carbon emission measurement methodology. (2) Population density exhibits divergent contribution rates to cities with medium and low IS, with the tertiary industry’s impact inversely related to population density. The urbanization rate significantly affects Panzhihua. This study enhances and broadens the methodologies for measuring urban carbon emissions using NTL, offering support for cities of varied industrial structures in devising tailored carbon reduction policies.