The Research On Detection And Tracking Algorithms Of Moving Target In Video Sequences

Among the information people perceive from the outside, above80%of it is obtained from the sense of vision. Image is one of the most important ways people use to receive visual information. In recent decades, with the continuous progress of science and technology, remarkable advances have been made in the visual image processing field which takes computer as its core. As a core subject in visual image processing field, detection and tracking of moving target has been widely used in many fields such as medical science, military and scientific research. So, the research on detection and tracking algorithms of moving target has extremely important theoretical significance and application value.This paper regards intelligent video surveillance as the application background, and it focuses on the research on detection and tracking of moving target in video sequences which is shot by stationary camera. It makes some.improvements based on existing algorithms, and the improved algorithms are implemented by programming, finally, the advantages of the improved algorithms are verified by comparisons with the existing algorithms.In terms of detection of moving target, this paper especially study the background subtraction of Gaussian Mixture Model, to overcome its defect of the update strategy of the model parameters, this paper puts forward a new learning mechanism of the model parameters, so that the Gaussian model gets a convergent variance, and this mechanism avoids falling into the vicious cycle of too small variance, so finally the occurrences of misjudgment of target can be reduced to a certain extent.In terms of tracking of moving target, this paper especially study the tracking algorithm of Mean Shift, which is based on feature matching, to overcome its defects of unfavorable tracking effect of fast moving target and lack of update mechanism of target model, this paper puts forward a corresponding improved method. An update mechanism of target model is added to the improved algorithm, which can describe the target more precisely, thus, the accuracy of the algorithm is improved, and Kalman filter is also added to predict the location of the target in the current frame, so that the initiative location of the tracking iteration point is more close to the real location of the target, and so the searching area is narrowed and the iteration times are reduced, and finally, the speed and efficiency of tracking is improved.Finally, based on the platform of Visual C++6.0, this paper carries out the programming of detection and tracking algorithms of moving target in video sequences, using the OpenCV library functions. The experimental results show that, in terms of detection of moving target, compared with the traditional GMM, the improved background subtraction of Gaussian Mixture Model has a much higher interference immunity, which can reduce the possibility of misjudgment of target, and in terms of tracking of moving target, the accuracy and efficiency of tracking has been greatly improved by using the improved Mean Shift algorithm.