The Qinghai-Tibet Plateau holds a crucial position in the global ecosystem. As its core city, Lhasa City stands as a representative focal point for studying the delicate balance between urban development levels and ecosystem service capacities. This study was conducted on the Google Earth Engine (GEE) cloud platform, utilizing Sentinel-2, VIIRS, and SRTM remote sensing imagery data. Based on spectral data combined with terrain and texture features, the study extracted artificial surfaces in Lhasa City through Pixel-Based (PB) and Object-Oriented (OO)classification methods. To compare the performance of different parts of the method, this study conducted a comparative analysis of three groups: OO or PB classification, inclusion or exclusion of texture features, and using Random Forest (RF) or Support Vector Machine (SVM) classifiers. The results showed that Based on the same spectral features, the best extraction result can be obtained by not using texture features in the RF classifier in the OO method (OO_RF), with an Overall Accuracy (OA) of 98.03%, a Kappa Coefficient of 0.952 0, a User Accuracy (UA) of 0.944 4, and a Producer Accuracy (PA) of 0.988 4. The effect of texture features in extracting artificial surfaces is relatively weak, with only slight improvements observed in PB methods. Specifically, the OA increased by 0.51% when using the RF classifier (PB_RF) and 0.68% when using the SVM classifier (PB_SVM). The RF classifier performed the best in this study, avoiding over estimation and identifying non-artificial surfaces within cities more accurately. In conclusion, this study provides a reference for extracting artificial surface information regarding methods and parameter settings at the urban scale. Using result data allows for further analysis and dynamic monitoring, which has practical application significance.