An intelligent laser processing robot with active vision composed of binocular stereo camera and structure light is developed in this paper. It can measure a variety of complex surface and reconstruct its 3D shape. The system consists of a robot with six degrees of freedom, binocular cameras and structure light. Cameras and structure light emitter are fixed on the robot and follow the robot movement, which accomplishes multi-angle measurement to eliminate “dead space”, as well as transforms point cloud data to the robot coordinate. In this paper, measurement principle and mathematical model are described. Inside and outside parameters of single-camera calibration, the relative relationship between binocular camera calibration, binocular camera and robot hand-eye relationship calibration are carried out to obtain the parameters of camera matrix, hand-eye matrix, robot-world matrix. The target is extract by big step segmentation, and the image is smoothed to reduce noise. Sub-pixel center of structure light is extracted by gravity algorithm, and the left and right image is registrated according to the epipolar constraint. Point coordinate calculated by 3D algorithm, can be easily converted to the world coordinate to achieve the unity of the global coordinates.