A stereo matching method based on objects' character

In 3D machine vision metrology, it is important to estimate the disparity between stereomappings. We discuss some related methods and introduce a new stereo matching method based on the objects' character. The prior knowledge of the scene is presented in the form of characteristic vectors in the procedure of matching, and therefore the stereo matching is converted to the graylevel-geometric space matching from the graylevel only based feature. As a result, we realized a novel effective algorithm, which could effectively achieve high accuracy and wide disparity range.