There is no much need to have GPS information as the frames are in sequence. GPS would just give approximate coordinates for the model. But you could add one additional photo from same device with same resolution as video so CC extracts EXIF metadata from it which could be then applied to video frames. Also provide sensor size and focal length parameters. You can increase tie points density by setting "Keypoints density:high". Also add Tangential distortion to "compute". If that doesn't help then relax initial parameters even more and set that camera model is fisheye for the Photogroup. Are the photos with at least 70-85% overlap? If not create even more frames. Also it would be best to extract only keyframes from the video as some inter-frames between keyframes could be blurry.
↧