In this paper, the control system of intelligent grasping for a special operation manipulator is designed to complete disaster relief, fire fighting, explosive disposal and the like for replacing human beings. The control system of the special operation manipulator is needed to have the characteristics of multiple terminal functions, accurate end accuracy and high system reliability and redundancy within outdoor environment. It puts forward higher requirements for the manipulator performance and operator. After analyzing the work requirements of the manipulator, the D-H parameters and motion space is calculated for further research. The forward and inverse kinematics models are built to make the control system into reality. The target grasping is achieved by using the force and position hybrid control and single joint/end control methods. The intelligent recognition system based on machine learning is constructed. The simulation training based on machine learning is carried out by using the back image of the end camera of the manipulator to help the operator locate the target quickly and accurately. Finally the special operation manipulator is used to carry out the target intelligent grasping experiment based on visual guidance. The test results show the effectiveness of the control system design. Thereby it is suitable to be promoted to other robot or manipulator control system design.