Abstract: Mechanical damage surface refers to the bare surface, which has texture characteristics and has not been restored by vegetation after large-scale mechanical engineering construction. Due to the influence of open-cut mining and engineering construction, the mechanical damage surfaces in Tang Langchuan basin are densely distributed. It is of great significance to extract the spatial distribution information of the mechanical damage surface quickly and accurately, which strengthen and protect the ecological environment construction in the watershed.However, there are currently few effective methods for extracting mechanical damage surfaces. In response to this problem, a typical area with densely distributed mechanical damage surfaces in the basin is selected as the research object. The remote sensing image of GF-2 is used as the data source and the object-oriented classification method is adopted. After dividing the research area with the optimal scale, fuzzy classification rules are constructed for the typical features in the research area. The information is extracted hierarchically according to the type of the features and got the object types in the study area distribution, achieving mechanical damage surface of remote sensing information extraction. The confusion matrix method based on pixel file (TTA Mask) was used to evaluate the classification accuracy. The overall classification accuracy reached 90% and the Kappa coefficient is 0.78. Compared with other traditional classification methods, the Maximum Likelihood Classification has the best classification method. The overall classification accuracy is 75% and the Kappa coefficient is 0.66, which is far lower than the accuracy of the proposed method in this paper. There are some errors such as dividing the mechanical damage surface into woodland and traffic land, incomplete road classification and so on, the classification effect is general. By comparing the results with the classification accuracy, the classification accuracy is improved obviously and the misclassification of ground objects is reduced, which indicates that the method proposed in this paper has better feasibility and superiority in extracting mechanical damage surface. The research results provide a technical support for the rapid extraction of mechanical damage surfaces in this basin and other similar areas, providing a reference for similar studies.