Research On The Mechanism Of Surveillance Video Inter-frame Tamper Detection And Implementation Of System

With the development of Internet technology,video has become an important information carrier,and the authenticity of video content has been unprotected because of increasingly convenient and powerful video editing software.In the field of justice,surveillance video is often used as an important evidence material,and it is an urgent practical need to identify the authenticity of its content.In this paper,the application of surveillance video inter-frame tampering detection in realistic scene is researched,the inter-frame tampering detection algorithms based on LK optical flow and NTF are studied,the performance of detection algorithm is optimized based on CUDA,and a video inter-tampering detection system is designed and implemented.The main work of this paper includes:First,a detailed analysis of the inter-frame tampering detection algorithm based on the secondary encoding effect and the tampering point content continuity is given,and the existing problems in the practical application are summarized,Conclusion that the detection algorithm based on the content feature is more applicable in future is given.Second,the inter-frame tamper detection algorithms based on LK optical flow and NTF are studied.Analyze the working mechanism of the algorithm and study the reason of its low efficiency.Combining the architecture of NVIDIA GPU and CUDA programming model,the parallelism of LK optical flow and third-order tensor Tucker decomposition is analyzed.The improved algorithm on GPU is proposed.After that,the optimization methods of GPU application are deeply studied,the initial algorithm is optimized,and the application prospect of this kind of algorithm is broadened.Thirdly,relying on OpenCV,QT and other application platforms,a friendly and easy-to-use video inter-frame tampering detection system is designed and implemented,which can effectively carry out Inter-frame tampering detection.Based on this,functions to create,remove,update,delete data and visualize the detectioin results are provided.