Research On The Method Of Tracking Drift Correction For Dim And Small Target In Infrared Imaging Detection System

As one of the core technology of computer vision, infrared dim and small target detection and tracking method is widely used in military field and commercial field. Traditional methods always bring the accumulation of errors, which is caused by the defect of tracking methods and background interference. Tracking drift will occur in these circumstances. Along with the development of computer vision technology, target tracking methods concentrates more on the approaches to improve tracking accuracy and reduce tracking error. The technique to achieve high precision target tracking has become a research hot point in computer vision field.The mechanism of tracking drift phenomena under complex background is analyzed, and the corresponding solutions to these problems are proposed. Experimental results on real data has shown the tracking drift suppressing and correction capability of proposed method.The main research content of this thesis can be summarized as follows:(1) Basic concepts and traditional methods of infrared imaging, preprocess, dim and small target detection and tracking is studied. For further research, we analyzed the causes of tracking drift under complex background, as well as studying the research status and trend of development.(2) The study of small target tracking drift correction method based on sparse ring representation(SRR). The main researches include basic theories of machine learning, small target detection algorithm based on SRR, small target tracking algorithm based on k-NN classifier combined with SRR features, and the small target tracking algorithm based on AdaBoost classifier combined with SRR features. Experimental results show that these methods are efficient in suppressing and correcting tracking drift.(3) The study on small target tracking drift correction method based on local contrast method(LCM). Main research contents include the small target detection method based on LCM, and the tracking drift correction method based on LCM constraints. Experiments are made to evaluate the efficiency of proposed method.(4) Image processing platform based on C# is build. The ‘.NET Framework’ and GDI+ is studied, as well as the building process of image processing platform. Methods mentioned above is implemented and tested in this software platform.