Background Modeling And Foreground Detecting Based On Codebook Model

With the development of society and technology, security monitoring is more andmore used in people’s life. This not only brings convenience for people’s daily life,but also creates a new field of study---monitor video processing for the researchers.Most of the surveillance video has one common property of single lens and fixedscene. This thesis try to improve the single-layer codebook modeling method basedon the effort of the others and reach a good result proved by the simulation.The main work of this thesis includes:The history of the moving object detecting is introduced. Several mainstreammethods of foreground detection are introduced, and the effects of these methods arestudied. For comparison, the advantages and disadvantages of each method isanalyzed deeply.The improved codebook model method and the formula of color distance in YUVare proposed to deal with the problem of luminance. The result images are gained bythe use of Opencv ffmpeg and VS10. Based on the data of the experiment the correctrate of method was improved, the influence of luminance was reduced.The thesis proposes a method that called b-yuv to deal with the problem of thelow speed of training and detecting and huge space of storage. The result images aregained by the use of Opencv ffmpeg and VS10. Based on the data of the experimentthe training time and detecting time were reduced in the same level of correct rate andaccuracy rate.