The projective correspondence between planar point/line sets isdetermined using conic invariants. In each set six (4+2) randomly chosenpoints/lines define two conics to which two absolute projectiveinvariants can be associated. Similar invariant values yield six-tuplesmatched between the sets. The order of features on a conic is notimportant and thus the probability of finding a match is maximized.Feature correspondence is recovered with a dynamic programming typeanalysis of the contingency table derived from the ensemble of matchedsix-tuples. The algorithm is very sensitive to feature accuracy, raisingquestions about the straightforward use of conic invariants inrecognition of arbitrary objects. The procedures employed,however-affine invariant conic fitting to noisy data,-probabilistictwo-stage matching and are of general interest
展开▼
