Video Capture And Reconstruction System Based On Compressive Sensing

Traditional image and video capture devices rely on large-scale sensor arrays for image or video capture.The raw data collected from images or video is huge,which is not convenient to the storage and transmission.At the same time,obtaining video with high frame rates and high resolution is very expensive.Compressed sensing combined with image reconstruction provides a new idea for video capture.Compressive sensing uses the sparse signal to reconstruct the original signal at a much smaller rate than the Nyquist sampling rate.The single-pixel camera is a new type of imaging device,and the image is modulated by the observation matrix to obtain the observation value.The image is restored by the follow-up reconstruction algorithm.The imaging device based on compressive sensing effectively eliminates the redundant information of the original signal and provides a new way for the storage and transmission of massive image or video signals.The focus of this dissertation is to achieve multi-bands,high frame rate and high quality video capture and reconstruction with a small amount of data.The compressive sensing is taken as the theoretical basis,and the single-pixel camera is used as the hardware platform,and the recovery algorithm is used to achieve high frame rate video acquisition and reconstruction.Different from the original single-pixel camera,the single-pixel camera in this dissertation adopts dual-fiber collection and performs sample reconstruction on the actual platform.The results show that the dual-fiber acquisition greatly suppresses the influence of environmental noise.At the same time,based on the actual single-pixel camera platform,the dynamic scene is sampled in the visible and near-infrared bands respectively.The reconstructed video results show that the single-pixel camera system can effectively realize multi-bands video acquisition and reconstruction.In order to achieve high frame rate,high spatial resolution and time resolution video acquisition and reconstruction,combined with the actual single pixel platform,a dual-scale matrix is designed.The dual-scale matrix consists of low-dimensional matrix and high-dimensional matrix.The low-dimensional matrix of the dual-scale matrix is used to quickly reconstruct the low-resolution preview video,and the optical motion constraint between adjacent frames is obtained by the previewing video.Finally,the dual-scale matrix is used as the observation matrix,and the optical flow motion is used as a limit.Conditions,using a nonlinear reconstruction algorithm to achieve super-resolution video reconstruction.The reconstructed video results show that the proposed method can be applied to the fast-changing target scene for reconstruction,and at the same time realize the trade-off between spatial resolution and temporal resolution of reconstructed video.Therefore,the new camera with compressive sensing theory can realize multi-band,high frame rate,spatial resolution and time resolution trade-off video acquisition and reconstruction with a small amount of data,which is important for solving the shortcomings of current image or video acquisition equipment.