Bridges can be recorded, measured, diagnosed, and managed better by UAV vision sensing data such as still image or video associate with location and gesture data for both maintenance and disaster emergency missions. Recent technology makes it possible and easier to reconstruct bridge structural in 3D point cloud and textured model using general purpose UAV with optical camera or even LiDAR camera. Structure from Motion can be applied using frames from video combined with UAV sourced GPS Exchangeable image file format (Exif). However, unlike riverbank or dam, the bridge structure reconstruction is quite difficult using UAV and Structure from Motion methods as it is important to fetch information on its side and under surface. Thus, not only images token from above, but side view and under view photos are needed. And they must be aligned properly to prevent disturbances in the final model product. In this study the reason and some basic treatment in flight path design, such as flight high and frame intervals combination of far and close view, for miss alignment of image photos from UAV videos for bridges was discussed. With the implementation 2D image segmentation the projection of damage in 3D model and its application was also proposed for a next generation bridge maintenance platform. The recent improvements of multi damage segmentation including background reinforcement learning and human in the loop and their advance in practice of a few real bridge damage detections were also introduced.