The Study Of Image Registration And Moving Target Detection Based On Infrared Image In Complex Scene

In recent years, surveillance system is widely used in the field of traffic control, intelligent weapons, and so on. As the chief part of digital image processing, the technology of moving target detection and tracking become the important part of surveillance.In this paper, the research background and the present condition in infrared surveillance system as the special one is introduced. In the meanwhile, the system structure and research algorithm is analyzed. Some problems in the practical use are discussed and studied too.The difference of the study algorithms between infrared images and visual images depends on their characters. The features of visual image, infrared image and their noise are discussed first. The distinction with them is analyzed. By using several existing denoising algorithms to the two type of image to filter and enhance them and comparing the simulation experiment results of all the algorithms mentioned, the chief characters and the application fields of them are induced.A lot of research on surveillance system has been done in past years. Most of the discussions focus on the situation whether the video camera in the indoors scene or in the static status. Some common approaches in the field of image registration and movement compensation including the algorithms based on phase correlation, block matching, optic flow and a algorithm based on feature extraction are discussed following the analysis of infrared detector motion state and scene features. Several simulation experiment has been finished, and a improve algorithms based on the feature points extraction is proposed. Through preliminarily analyzed simulation experiment to the improve algorithm, the algorithm feasibility is proved.The technology of moving target detection and tracking is discussed and analyzed delicately as the chief technology of surveillance system especially moving target detection. The algorithm structure, implement procedure and application field etc of the temporal difference algorithms including the adjacent frame difference mean algorithm and three frame difference algorithm and the background statistics modeling algorithms including background subtraction and Gaussian Mixture Model are particularly discussed in this paper. The advantages and disadvantages are pointed according to the simulation experiment results of several algorithms. Some classical tracking algorithms are introduced tersely.