In this paper, based on the feature of high accuracy of GIS in the field of construction and land, we adopt GIS to obtain the initial data of land cover change, and find that there is a lot of interference information in the initial data. In this regard, the smooth polygon procedure in Arc Toolbox is used to smooth and thin it. The processed data were input into the support vector machine model, and the detection results were output after iterative training, thus completing the construction of the land cover change detection model based on SVM. Determine the research area and experimental equipment, synthesise the research data and the model in this paper, and carry out example analysis of land cover change. After analysis, it can be concluded that when the smoothing parameter and thinning coefficient are set to 40m and 1.0m, respectively, the processed data can be used for subsequent analysis of land cover change detection. In addition, the overall accuracy of this paper’s model in detecting the change points of land cover conversion and land cover gradual change is 0.7368 and 0.7430, respectively, which verifies the superiority of this paper’s model in land cover change testing. Compared with 1995, the area of forest land from 2000 to 2020 is increased by 9.81km²~115.6km², which reveals the development direction of land cover change in the future, and explores the application of land cover change detection in the field of construction. This study has practical and realistic significance for urban planning and building construction to make land cover change more rational.