In this thesis, the 3-D positioning problem for a quadruped walking robot is investigated. In order to determine the robot's exterior position and orientation in a world coordinate system, a 3-D positioning algorithm is proposed by using a stereo vision system with a proposed Guide-Mark Pattern (GMP). For implementation, Guide-Mark Pattern(GMP) is specially designed so that necessary information can be extracted in fast manner from other scences and that the effects of noise is reduced in real environment. Some experimental results along with error analysis and several means of minimizing the effects of vision processing error in the proposed algorithm are disscussed.