Minimum-constraint BBA: Bundle Block Adjustment (BBA) is a relatively complex optimization process with many small details. One of the most important aspects of BBA is choosing a coordinate system. Definition of a co.sys. could be as simple as choosing a center and 3 directions. A minimum definition for a co.sys. is to select a 3D frame aligned on the local co.sys. of the first image or any image. This is called minimum constrained BBA. This definition still lacks a scale, since during the optimization process, a gradient could happen that could potentially scale-up or down the whole network. Therefore, defining a base with constant length seems necessary in most cases.
Inner-constraint BBA: A better definition of a co.sys. is when we choose the center of gravity to be the center of co.sys. This definition led to a better convergence during the optimization process, since the gradients are defined from the center. The same scaling problem still happens for an inner constraint BBA, therefore, it is necessary to include a base to solve this issue.
Over-constraint BBA: There are cases where we can define co.sys. based on ground control points (GCPs). This is the state that we have access to at-least 3 GCPs in the model space, therefore, the degree of freedom (dof) is more than what we actually need. This state is called over-constraint BBA.
We should notice that shaped, sizes, and orientations of 3D error ellipses in the model space completely depends on the definition of co.sys. This is due to the change that happens in the value and direction of 3D gradients.