Hi, I'm new to Context Capture and have just run a reconstruction and it has lots of missing information in the model. This was made out of 290 DSLR photos and the subject is only about 30' long. As you can see there was lots of overlap. I took photos all the way around the subject from eye level, then took two more laps with the camera held high and then low. Any ideas why there is so much missing information? Is CC like LiDAR in that it doesn't like dark shiny objects? My scanner did a good job of capturing it. Thanks,
Scott
I got very similar results on a shiny outdoor sculpture. If it is too shiny and too smooth - it makes it hard for the software to find matching points.
You didn't show the photo of the object but it seems it is shiny. CC has the same limitation as LIDAR that shiny objects or water can't be reconstructed properly. What you could do to get at least some results is limit reflections(cloudy day or move it indoors) and also do a very large overlap like 95% then there will be less difference between pictures. Then afterwards set block type to orbit thin structure and run aerotriangulation with keypoints density:high. If this doesn't help then you will have to use masks and mask background so it doesn't try to aerotriangulate using background points. See under Surveys:view automatic tie points and you will see where are the matching tie points, your target is to have as much tie points as possible on the object. Also it would help to do camera calibration on different object and re-use it in this project. If you have LIDAR points you could also generate model using both photos+LIDAR.
Thanks Oto. I waited for a cloudy day to shoot this, so that should help. I'll try the thin structure setting. Where do I find the masks? If I mask for AT will I loose the background in the final model or do they just take it out for the reconstruction? Can I add tie points to the object manually to help things along?
I think masking won't loose background but if it does then just add two sets of same photos in same block one with masks and one without. Read help about masks it is simple black and white pictures which should be in same directory and same name as photos plus "_mask" suffix so it can be added automatically. For existing block masks can be added by right clicking on photo and choosing add mask.
Adding user points would also help but it should be placed accurately. What I have noticed that sometimes user points cause CC to drop lot of automatic tie points around user tie point because it can't fit together. So add some and check afterwards in automatic tie points dialog if it helped or not.