3-D Data Acquisition by Trinocular Vision

Abstract

For recognition of 3-D shape and measurement of 3-D position of objects, it is important that a vision system can measure the 3-D data of dense points in the scene. One approach is to measure the distance on the basis of triangulation principle from the disparity of two images. This binocular vision method has, however, a difficult problem that is to find correspondence of features between two images. This correspondence problem can be solved geometrically by adding one more camera, that is, by trinocular vision.
This paper presents the principles and implementation details of this trinocular vision. Based on the proposed method, we have made several experiments. By these experiments, it is found that many correct correspondence can be established even for images of complex scene by only geometrical constraint of the trinocular vision. However, when there are dense edge points in the image, multiple candidate points are found and unique correspondence cannot be established. To solve this problem, we will discuss two approaches in this paper.