The Research And Application Of Moving Target Detection Aigorithm Under Complicated Background

Currently, computer vision research is a hot topic. It can help the user to achieve a good human-computer interaction and relief people from heavy equipment operation. The moving target detection is the basis of many fields of the computer vision. The result of the moving target detection is directly related to the follow-up work. The Gaussian mixture model is the mainstream of the moving target detection algorithm. It is one kind of background subtraction method.But the classic Gaussian mixture model is sensitive to the state changes of the moving object and the light mutation. It also has a problem of the convergence of variance. The paper proposes an improved algorithm to solve these problems. The algorithm uses the position information of the object, two parameters of the time accumulator and light mutation coefficient to supervise the object’s moving state and the situation of light. Then the improved algorithm adjusts the background model and the learning rate adaptively according to these parameters. It also controls the convergence of variance. These ways help the background to respond to the environment and the state changes of the moving object.Finally, it proves that the improved algorithm can suppress noise caused by the state changes of the moving object, the light mutation and the convergence of variance effectively through experiment. The algorithm’s robustness has improved.