In order to satisfy accuracy and real-time needs of special mechanical arm fetching dangerous goods in the complex environment, research method of using visual information to identify and orientate is proposed. A visual recognition and positioning strategy based on a single and binocular combination is presented. The color segmentation, opening and closing operations, edge contour detection and feature matching are used to complete the monocular vision algorithm. The stereo calibration, stereo correction and depth calculation are used to complete the binocular vision algorithm. According to the position information of the target object obtained by the homogeneous transformation, special mechanical arm is used to solve the joint rotation angle, and the optimal solution is obtained by using the angle threshold range, and through the target object pose measurement experiment and grasping operation to verify. The verification results show that the method is feasible and reasonable, and can achieve the actual fetching of the target.