Two-dimension (2D) code could be used to recognize and correct position in the system of Automated Guided Vehicle (AGV). But the common code of Quick Response (QR) 2D code has a problem that the process of decoding QR code spends long time. So a simpler 2D code of marker code is designed to make the speed of extracting information faster. Through dealing in advance with the marker code image, the marker code’s 2D information was gotten from the image base on the Contour tree algorithm. Let the image match the normal image, the position and pose of AGV were gained. After that, the error of AGV coordinates and pose were fixed. The experimental result shows that the method of 2D marker code is robust, high accuracy and good in real time.