As an important technical means to observe the earth's features, remote sensing has been commonly applied in many fields such as ecological environment monitoring, vegetation monitoring, crop yield estimation, territorial spatial planning and global change research. The change detection of satellite remote sensing image data provides more accurate and timely analysis and assessment of environmental changes for ecological and environmental supervision departments, and provides effective support for comprehensive evaluation. In this paper, we selected Landsat satellite remote sensing images of two different phases of Fangcheng golden camellia national nature reserve and realized the change detection by combining PCA algorithm and K-means clustering algorithm, and the detection results show that some surface cover changes during the crop cultivation season and a small amount of minor machine road changes are effectively detected. Meanwhile, the experimental comparison results also show that high quality remote sensing images and accurate geographic alignment are more helpful to improve the change detection accuracy.