Push in helps if you capture a tunnel. Overlaps and Arc "Resolution" (frame number) depends on how detailed the scene, if you have a lot of thin objects (fence) the more is better. If you have something a pole it is better to arc around it and keep it in the middle of the frame. Also there is a movement like pano when you capture corners, but if you move while you capture, you get something like arcing but in the other direction. So if you go around in a "wave" path combining arc and "reverse arc" you can get smother motion while capturing, for example if you want to extract from video. If you capture video try to have high shutter speed to have less blur. And use 4K video so you have 8,3 mpixels. If you use 8K you get 33,2 mpixels per image. Set to 24 fps to get less frames or you can capture slow time lapse (2x real speed or 4x at most, depends on the desired spatial resolution). But then you really have to get the right speed to get optimal overlaps. If you have a camera or app that captures the frames of the timelapse you don't need a 8K camera to get high megapixel images. If you can think in 3D it helps if you try to imagine in your head what you can see from the images while capturing. (It helps alot if you have a 3D scanning app. (I use the free version of 3D camera on android from Luboš Vonásek it is great and easy to use if you use these thecniques from this video, for free it is amazing, but there is a more advanced paid version that van use the inbuilt TOF sensor if you have one. There is Scann3D it captures images from the camera not video like the other and shows the recognized similar points to help create great overlaps, but it is harder to use it)
This is very helpful. Although you missed full object scans (including positioning below) for like stand-alone statues, art, rock-and repetitive objects of similar proportions (like large tulip beds)
This is a GREAT video! Something I wonder is that if camera was set to rotate on Nodal Point (using something like a Nodal Ninja) how would this type of pano effect the scan.. because there is no parallax.
That is a great alternative term. We co-opted terms from professional filmmaking camera movement terms. Ultimately what matters is that you move in a manner that creates parallax in the imagery.
i always take a video and set the number of frames per seconds depending on the number of images i want. My problem is that my 3D reconstruction is not always in the unit box but outside which cause a lot of problem to the mesh.
I don't agree to panoramic, if you take multiple 360 deg panoramic pictures for one room for example 8 panorama, one for ech corner and one for ech wall, all combine have a large amount of data to calculate parallax.
Insanely useful! Thanks a lot! Is it possible to write off in a messenger convenient for you, ask a couple of questions? Сhat, maybe we will be useful to each other.
