I. Introduction
In recent years, with the development of computer technology, electronic technology, sensor technology and artificial intelligence technology, people are no longer satisfied that robots can only complete repetitive operations in a fixed environment and follow a set procedure, and start to work Research on intelligent robots with strong environmental adaptability and high sensitivity [1]. The mathematical model of the robot has the characteristics of severe nonlinearity, strong coupling, and contains uncertain factors such as friction and load changes [2]. The traditional control method based on object model cannot achieve satisfactory results. With the development of science and technology, machine vision is also developing rapidly. It has made great progress in three-dimensional scene analysis, image analysis and synthesis, target recognition, stereo vision, contour analysis, etc., and the tracking and recognition of moving targets has also been achieved. With a wide range of applications, the processing and analysis of dynamic images has received increasing attention [3]. Humans perceive the world mainly through eyes, and 70% of the information comes from vision. This illustrates the importance of vision for robots with self-control capabilities. Moving target detection and tracking is a process of real-time determination of whether there is motion information in the field of view and the state of target motion by analyzing the image error data from the environment without knowing the target motion information in advance [4].