Geomorphometry research largely relies on 3-dimentional land surface models, especially for natural landforms with steep slopes and large complex artificial structures. The traditional Digital Elevation Model (DEM) encounters difficulties in representing detailed features in such scenes. To address this issue, this paper introduces a concept of a “true 3D surface” and proposes a multiple Unmanned Aerial Vehicle (UAV) collaborative method for high-quality reconstruction of a true 3D surface. We established an evaluation model of scene reconstruction based on the Multi-View Stereo (MVS) modeling rules, and utilized the Fibonacci approach to select the optimal viewpoints. We also adopted the method of finding the next-best-view for each UAV in turn to extend the multi-UAV trajectory. This method involved trade-offs between energy consumption and scene reconstruction to enhance the contribution of viewpoints for improved reconstruction quality of the true 3D surface. Real flight experiments were conducted over the Wuhan University Teaching Building and the Fangshiwan Bridge in Chibi, which revealed a significant improvement in the completeness and precision of the true 3D surface constructed by our proposed method when compared to vertical photography techniques.