Online Robust Video Condensation Based On Structural Constraint

As the worldwide demand grows for the public security and precaution,surveillance cameras installed in residential block,park,square,corridor and so on increase at a dramatic rate,which makes the video data increase fast.How to store,retrieval and browse the massive surveillance videos efficiently is a burning question.Video condensation is good solution to solve this problem,which segment the moving object sequences out,taking full advantages of the aperture time between these sequences,rearrange and recombine them along the time axis,to make the videos display a great deal of information at a short time.Existing video condensation algorithm use the online ways,because of the drawback of the moving object detection algorithm,they cannot condense the video with high visual equality in some complex scenes.Therefore,we proposed a robust video condensation algorithm based on structural constraint for most complex surveillance scenes.The main works and contributions of this thesis are:(1)Proposed an online and robust video condensation framework based on structural constraint.Such framework can detect the background change in video import,segment the video into stable parts and unstable parts,divide and conquer them to get the condensed video with high visual equality.(2)Proposed a new moving object detection algorithm with feedback.The feedback loop is built between pixel based background model and segmental result which can reduce the uncompleted object detection problem and adapt to most weather variations and intermittent moving object.(3)Proposed tubes conglutination in sticky tracking.This method can eliminate the swollen tubes,accelerate filling speed and reduce the operating quantity.(4)Proposed an acceleration strategy,which makes the whole algorithm can process the video with high solution in real-time.The comparative experiments of moving object detection algorithm and video condensation algorithm demonstrate the effectiveness of our approach in the thesis.In future years,the video condensation will play a vital role in video import and storage.