How are point clouds computed?

Hei, I am trying to find out what the internal process in context capture is regarding point cloud generation. Say I am planning to produce only a point cloud from the 500 images I entered into context...
How to accept the position of proposed user tie points just as the GUI interface dose by python API?

I already give three measurements of one user tie point to the GUI interface and the interface is able to show the responding position in other images. But I have to click the "Accept Position" button...