The location precision by binocular camera sensitive to environment is not high, thus the grasping success rate and robustness are not sound in the existing grasping system. And in order to overcome this problem, this paper proposed a method of fusing vision of infrared sensors, tactile sensors and encoders, which can realize a reliable, robust system capable of adjusting grip force automatically. In order to avoid the situation of grasping-nothing, binocular camera coupled with monocular camera and infrared sensors were used to locate precisely. By integrating tactile sensors and encoders, and adding monitoring force and position information in the process of grasping, the phenomena of broken and slippage of the object were reduced to a great extent. The experiments demonstrate that the grasping success rate and the system performance are greatly improved by fusing multi-sensor data compared with signal sensor.