In auxiliary operations quadruped robots cannot satisfy the needs of the flexible human-computer interaction through remote or terrestrial control. Based on the basic gait control of quadruped robots， the functions of obstacle avoidance and personnel follow for robots are developed by combining the algorithms of wireless positioning and laser radar perceived environment. Then， by using the Openpose gesture recognition method， the control of basic motion and follow-up task with different attitudes is realized through human-computer interaction on vision sensors. And the interaction identification and control effect under different environmental conditions are tested under the ROS frame. The results show that the gesture recognition speed is 20 frame/s， the recognition accuracy in the ideal environment is greater than 90%， and the robot’s reaction time is less than 1 s.