With the development of drone and autonomous driving technologies, the demand for large-scale scene reconstruction and rendering on movable platforms has grown significantly in fields such as disaster relief, modern military operations, 3D mapping, and surveillance monitoring. This paper first proposes a scene reconstruction algorithm based on a frame-interpolation recovery architecture for movable platforms. The algorithm leverages the continuity and high overlap of data acquired by movable platforms to recover motion poses and reconstruct scenes for drone aerial platforms and single-camera vehicle-mounted platforms. It improves multi-view matching 3D reconstruction algorithms, addressing the issues of excessive computational time and poor scene fitting in movable platform reconstruction scenarios. Secondly, for vehicle-mounted platforms with multi-stereo camera architectures (i.e., movable platforms equipped with multiple stereo camera systems), this study investigates a stereo camera sequence fusion calibration algorithm to achieve precise calibration of multiple stereo camera systems. Finally, an incremental dense reconstruction algorithm for multi-camera systems is developed to reconstruct scenes on vehicle-mounted platforms (as a type of movable platform). The reconstructed scenes are then registered into a world coordinate system to align the vehicle's local coordinate system with global coordinates. This research integrates motion recovery, calibration optimization, and incremental dense reconstruction techniques to enhance the efficiency and accuracy of large-scale 3D scene reconstruction on movable platforms, with potential applications in autonomous navigation, emergency response, and intelligent surveillance systems.