During locomotion, one of various important tasks, a biped robot has to perform is the identification of 3D environment. The present work addresses the problem of offline 3D reconstruction of the environment using a stereo-vision system. The developed stereo-vision system works based on the concept of an epipolar geometry. The mathematical model that interpolates 3D position of a point with the help of an observed 2D point by two cameras, has been utilized to extract and virtually reproduce the environment, such as a staircase, a sloping surface and a ditch surface. The extracted information from the vision module may be used to plan the path of the biped robot in the said environment.