In order to improve the accuracy and the stability of stereo matching, two kinds of stereo matching method using triplet image are proposed. Their performance in investigated using simulated linear array sensor data, and they are proved to improve the accuracy and the stability of stereo matching.
Triplet image has the redundancy that three bundles intersect on a target. One of the methods proposed is to eliminate mismatchings using this redundancy. The other is to search matching points under the restriction that three bundles should have the same intersection point.
As the result of the simulation, it is shown that the R.M.S.E. and the maximum error with these methods are about 5/6-1/5 and 4/5-1/13 of those with an ordinary stereo matching method.
View full abstract