Research On Auto Capture And Focusing Of Interest Video Region Under Dynamic Scene

The target capture and focusing technology of interest video region underdynamic scene is very important and fundamental to the field of computer vision.Dueto the dynamic changes of the background, such as the weather, illumination, shadow,leaves, the improved target capture and focus evaluation algorithm are proposed.Based on the dual-camera system,the wide view capture and zoom-focusing of theimage are realized. The paper mainly concludes the following aspects:Firstly, the significance and the status of this topic are introduced. Thetraditional methods of target capture and background modeling are discussed. Animproved moving target capture method is proposed with an effective and capture ofinterest target and robust noise immunity.Secondly, Kalman filter and CamShift tracking algorithm are introduced whichare combined to present a target tracking method based on the color gradientbackground difference.Thirdly, the dual-camera system is introduced, the rotating angles of PTZ cameraare determined by the affine transformation matrix and the relationship of thecoordinates is discussed. This paper introduces and analyzes the traditional focusingtechnologies consisting of focusing window, evaluation function and searchingstrategy. An improving focus evaluation function is proposed with a low calculationconsuming and robustness.Finally, the system of interest target capturing and focusing is comprehensivelyanalyzed，the software and hardware environment of the system is introduced inwhich different targets are tested.The experiment demonstrates that the dual-camerasystem has better robustness and anti-interference ability.