Design And Implementation Of Video Stabilization System Based On Trajectory Clustering And Kalman Filtering

With the rapid development of image processing technology,people have put higher requirements on video image quality.However,the decline of imaging quality due to camera shake has become a major pain point for users.At present,the research on image stabilization technology also faces many challenges,such as feature tracking is sensitive to subject brightness,local motion is difficult to reject,poor adaptability of motion filtering and so on.All these problems hinder the effect and speed of video stabilization.Therefore,this thesis proposed an improved video stabilization system which can be described as following:(1)An improved video stabilization method is proposed.First,aiming at the problem that feature tracking is sensitive to brightness,combining the characteristics of LK optical flow method and NCC template method based on image pyramid,this thesis proposed a resisting brightness changes feature tracking algorithm.Second,a motion estimation method based on trajectory clustering was proposed.PCA and OPTICS algorithms were used to make motion estimation without manually setting parameters.Third,in order to solve the problem of poor adaptability of traditional motion filtering,the scene recognition factor is introduced,and an adaptive Kalman filtering algorithm is proposed to adapt to motion filtering in more complex shooting scenarios.(2)The video stabilization system was designed and implemented.The system includes the user management module,the video stabilization module,and the video management module.(3)Test analysis is applied to video stabilization systems.In terms of functional testing,this thesis used test cases to test the system sub-modules to verify the feasibility of the image stabilization system.In terms of non-functional testing,scientific evaluation criteria are selected to evaluate the image stabilization effect and speed.In this thsis,an improved video stabilization method is proposed and implemented,which solves the problems of video stabilization,such as sensitivity to brightness changes,local motion interference,and poor adaptability of motion filtering.Designed and implemented the image stabilization system,providing users with friendly and practical video stabilization,user management,video management and other services.Through testing,it was verified that the system reached the expected goal,and an video stabilization system with high real-time performance,good image stabilization effect,and suitable for rich scenes was realized.