Proposal of calibration error-tolerant 3D measurement method using stereo vision for robot vision

The accuracy of Stereo 3D Measurement Method strongly depends on the accuracy of calibration. But, it's very difficult to maintain optical systems in exactly the same. So, we are always pressed by painful calibration. For this, some self-calibration methods are developed, but they are not a panacea for all situation. To solve this problem, we proposed “The Relative Stereo 3D Measurement Method” so far. In this method, almost calibration error is canceled by overlap stereo images on reference point. So, this method has a good tolerance to calibration error and shows an impressively high accuracy in relative distance. Though, it has a disadvantage for close range measuring. The reason is that it uses scaling relationship which isn't strictly accurate. Therefore, this research proposes a new 3D measurement method using scaling relationship of triangles which is always strictly accurate. This method inherits a good tolerance to calibration error from “The Relative Stereo 3D Measurement Method” and solves conventional problems of it.