However, the order constraint is relatively weak compared to the overlap constraint.
Such a non-overlap constraint is accurate and differentiable, facilitating an analytical design sensitivity analysis.
Coherently with the view overlap constraint we set that threshold to 20%.
On the other hand, the overlap constraint makes it difficult to determine correspondence, resulting in a lack of flexibility.
For instance, imagine that an operator is using a scanner to acquire a human statue, trying to abide by the policy imposed by the overlap constraint.
Therein, the non-overlap constraint for all the embedded holes is imposed as a single explicit integral constraint over the design domain, thus avoiding the difficulties in the overlap detection of multiple arbitrary-shaped geometries.
Moreover, an effective model for the non-overlap constraint is proposed in a unified and systematic manner to avoid the overlaps between the holes and between each hole and the design domain boundary.
In fact, pairwise coarse alignment requires each couple of subsequent views in a dataset to possess a certain amount of overlap (from now on we address such requirement as the overlap constraint).
However, in order to comply with the overlap constraint he would need to capture more range data in regions he has already scanned to create a path between the back and the belly of the statue.
Nevertheless, in order to reconstruct an entire object through such a pipeline, we would need to either define a chained acquisition path for which every consecutive image pair complies with the overlap constraint (thus limiting the scanner usability), or manually reorder the scans prior to applying the alignment (a tedious, time consuming and necessarily off-line operation).
The adjacent PCR product overlap constraint in the proposed algorithm was set to 90∼200 nts as mentioned in the materials and methods section.
