Humanoid robots have become the best form of human services for human beings by the widespread concern. The mobile manipulation of the humanoid robot means that the humanoid robot moves by the footsteps. At the same time, it also uses the arm and hand to complete the crawling and other operations. In order to complete the mobile manipulation, the robot is required to have target recognition, online path planning and motion control. This paper describes the hardware, software design and implementation of the minitype humanoid robot mobile operating system, and has carried on the experimental verification. Firstly, it has carried out kinematics modeling and inverse kinematics for the NAO experimental platform of minitype humanoid robot. Secondly, the research on path planning and motion control of humanoid robot is studied. Thirdly, the target recognition and positioning based on monocular vision are studied. Finally, it carries out experiments on NAO robots. In the experiment, the robot completes the mobile manipulation task accurately and efficiently. Experiments show that the target recognition and positioning method are efficient and reliable, and can realize the mobile manipulation of the robot completely. In addition, it completes the research topic of minitype humanoid robot target recognition and mobile manipulation.