The model invocation and hypothesis construction processes are largely based on
processing symbolic structural descriptions and do not closely examine the
input data.
Further, tolerances are allowed in the matching process to overcome
noise, segmentation variations and imperfect descriptive processes.
Consequently, it is possible for coincidental scene arrangements
to lead to spurious object hypotheses.
Many of these false hypotheses will have been eliminated by the geometric
constraints examined during hypothesis construction, but some may remain.
This chapter discusses additional constraints on solid physical
objects that help guarantee object existence and identity.