I'm reading and trying to implement the paper Monofusion(V. Pradeep, C. Rhemann, S. Izadi, C. Zach, M. Bleyer and S. Bathiche, "MonoFusion: Real-time 3D reconstruction of small scenes with a single web camera," 2013 IEEE International Symposium on Mixed and Augmented Reality (ISMAR), Adelaide, SA, 2013, pp. 83-88.)
In this paper, in the map initialization phase the system extracts MxM patches of FAST features in 3-level pyramid images and while the user slowly moves the camera, it also matches feature points by ZNCC score from patches in the initial and current frames, computes the essential matrix, recovers the initial camera pose by SVD and triangulates the matched features to create 3D landmarks.
They said the system automatically bootstraps only once a sufficient motion baseline is reached. However I don't understand how to get the "sufficient motion baseline" from this phase.
As I understand, it seems like the system automatically finishes the initialization when the actual baseline between the initial camera frame and current camera frame has reached some threshold.
Does this mean it computes the relative translation between the initial and current frame using the 3D triangulated points?
If this question is silly sorry in advance, I’m a newbie in the SLAM world.
Any advice is welcome.