To address the challenges of UAV navigation and positioning in GNSS-denied environments, this paper proposes an auxiliary solution based on the fusion of Ultra-Wideband (UWB) and vision for 3D scene modeling. A local UWB coordinate system is first established to obtain UAV position information, while an onboard vision module captures images. The geotagged images are imported into 3D modeling software for scene reconstruction. The resulting 3D model is then rotated and aligned so that its coordinate units possess real-world geographic semantics. Utilizing the complete map information from the reconstructed scene, combined with a UWB trilateration algorithm and Kalman filtering, the system enables controllable and command-based visualized flight in GNSS-denied conditions. Experimental results show that the maximum deviation in stretched distance within the 3D model is 0.29 m. When target coordinates generated by Unity3D are sent to the UAV, it navigates to the designated waypoint, achieving a final landing point deviation of 0.28 m from the modeled position, thereby demonstrating the feasibility of command-based flight using 3D map data in GNSS-denied environments.