Abstract: Objectives:  In 3D scenes, because the bottom surface of 3D models is usually horizontal, and terrain models are mostly irregular surfaces with continuous undulations, which leads to problems such as gaps or terrain flooding when matching 3D models with terrain models, etc. In the current research on matching 3D models with terrain, there is a lack of scene complexity constraint, so some matching methods are not applicable to 3D scenes with multiple complexities. Therefore, how to complete model-terrain matching while taking into account the scene complexity constraint and retaining certain terrain integrity is the current problem to be considered and solved.   Methods:  Through research on 3D model data organization, scene complexity evaluation, and grid reconstruction, etc., a 3D model and terrain model matching method that takes into account the scene complexity is proposed to achieve a seamless matching between the model and the terrain. Firstly, the method improves the data organization of multi-source heterogeneous 3D model by combining the characteristics of 3D model; then, it constructs the 3D scene complexity evaluation model by combining the scene complexity evaluation factors, and calculates and classifies the 3D scene complexity level; finally, according to the scene complexity classification results, it realizes the accurate matching of multi-source heterogeneous 3D models and terrain considering the scene complexity.   Results:  Different 3D scenes contain different amounts of information, and their corresponding scene complexity is also different. For the 3D scenes with low scene complexity, the elevation value fitting method is used, and the 3D scenes with high scene complexity use the grid reconstruction method. The visual perception and real-time frame rate are used as reference indicators to compare and analyze this method with the traditional method. The frame rate is stable at 58 frames per second, and it can be seen that the scene complexity constraint of this method can optimize the system loading smoothness and improve the visualization effect.   Conclusion:  This paper constructs a scene complexity evaluation model based on scene information entropy, designs a model and terrain adaptation algorithm adapted to the scene complexity level, organizes and expresses the 3D scene model data with different scene complexity more efficiently while ensuring the real-time frame rate when rendering the 3D model, and effectively optimizes the problem of accurate matching between model and terrain.