Abstract: The realtime information of the working environment is the important basis for intelligent decision of robot． For the instantaneity and adaptivity of 3D environment information detection，a detection method of the shape， size and location of the object or obstacle based on binocular vision is proposed． Firstly，the Otsu into Canny is integrated． It improves the efficiency of the target edge by down sampling and compressing gradient magnitude level． Secondly，an edge-point classification matching algorithm based on gray correlation is applied to classify and match the edge point． In addition， it improves the efficiency and accuracy of the algorithm at the same time． Then，based on the structure of point clouds， the automatic extract method for 3D geometry and the location information of contours based on edge curvature angle are proposed． The experiments of the robot autonomous operation in dynamic environment show that the methods proposed are able to obtain the 3D information of the object in the working environment． The planar size error is 0. 65%，height error is 1. 69%， and distinguishes the object or obstacle accurately． The robot completes the expected task according to the realtime location information of the object and the obstacle． 

Key words: robot, binocular vision, 3D information detection, adaptive edge-detection, stereo matching