In this paper a trinocular stereo vision method using pyramidal graph propagation matching is proposed. The choice of the starting triplet in the algorithm is based on perspective projective geometry and matching through propagation greatly limits the searching scope. Therefore, it can accelerate the computing and avoid the possible false targets. Matching through propagation makes use of the structural similarities of stereo triplet images in pyramidal graphs of BLI's, whereas the overall examination and verification of matching results is done by a comparison of the propagation success ratio. This algorithm can effectively overcome some essential limitations of binocular stereo vision methods and assure the matching results precise and reliable.