This guide will walk you through the best practice to accurately collect data for your 3D reconstruction.
Collecting First Artifacts
Collecting a first artifact requires the user to record many perspectives of the space being scanned. The most valuable images are images that contain many features. Features come from texture and color, a monocolored flat wall is featureless and a fuzzy patterned carpet is feature rich.
The collection method will depend on the FoV of the camera being used. Here the guide will split into two sections describing methodologies for using Smartapp’s app and Retina. However, it should be emphasized that the best practices are universal between devices, it is only specific methods that change as the devices FoV increases.
Phone App
If the user is using a phone with a single wide angle camera or less then below describes a robust methodology that will lead to the best results.
Walk around the perimeter of the room close to the wall with the camera pointing inwards towards the center of the room.
Then walk around the center of the room three more times changing the direction of the phone each time.
If there are any large structures in the space such as pillars that block large positions of the scene or any objects of interest that you want to be sure to have come out well, then orbit the object/structure from far away trying to include the whole structure in the frame if possible.
To achieve the best results also record some video at multiple heights ideally a loop getting as close to the floor and the ceiling as possible.
The order in which you follow these steps does not matter, these are the highest value perspectives to capture.
The total path could look something like this:
Do not point the camera at close walls, when the camera is too close to a structure it acts as an obstruction. This will result in fewer feature overlap and can cause the reconstruction to take longer to resolve but wont cause the reconstruction to fail unless most of the images are like this.
Collecting Updates, Improvements, and Dollhouses
To collect an update or dollhouse point your camera at the new area or object and orbit it at multiple heights capturing many perspectives of the changed or new area. The same best practices as creating first artifacts apply here as well.
Ensuring Sufficient Overlap
A critical aspect of an update is that it can be related to the original scan meaning that there must be sufficient overlap between the latest 3DR and the current update. The best way to capture this overlap is to orbit the change from far away (as well as up close) to capture a large amount of the surrounding unchanged scene or in the same video rescan a section of the area that hasn't changed.
In the case where the user wants to dollhouse multiple rooms there must simply be enough video recorded at the junction of the rooms as if they were initially captured as one room. The highest value perspectives are ones that can see many features from both rooms.