Study On Key Technologies Of Spatial-temporal Super-resolution Imaging For Space Observation

The existing optical imaging measurement devices have limited spatial and temporal resolution,which are limited by the spatial density,frame-rate and exposure-time of the detectors and diffraction limit of optical system.Moreover,some factors such as environmental illumination changes,optical or motion blur,subsampling and noise disturbance can significantly degrade the image quality and resolution,and affect the performance of space moving targets recognition and tracking.It is difficult to achieve spatial and temporal super-resolution reconstruction of a dynamic scene using conventional sampling strategies.The spatio-temporal super-resolution technology based on compressive sensing and computational photography can overcome the bottleneck of conventional sampling strategies and performances of device.Therefore,it is important to study the spatio-temporal super-resolution theory and its key technology.Spatio-temporal super-resolution imaging technology is an interdisciplinary subject relevant to optical compressive sensing,computational photography,and conventional image processing.Based on elaborations in some current related technology studies,this dissertation focuses on spatio-temporal super-resolution imaging,and mainly studies the Coded Aperture Compressive Temporal Imaging(CACTI)and spatio-temporal super-resolution for multi videos.The main contributions and innovations of this dissertation are as follows:1.Aiming at joint spatio-temporal alignment of videos for real scenes,a algorithm using astronomical calibration and scale-invariant feature transform(SIFT)Flow is proposed,which can improve the accuracy of estimation and simplify the complexity of computation.In order to achieve the spatial registration of multi-sensor for space observation,a camera calibration and an attitude measurement algorithm are proposed based on the star image simulation.According to star observation model of multi-sensor,the simulated star image is produced in order to transform constellation features into image features.Based on the invariant collinearity of quadrilateral diagonal and the singular value decomposition method,image features between the observed star map and the simulated star image are matched,and the space registration parameters can also be estimated.Meanwhile,the similarity of SIFT flow field is used for solving the time-transformation parameters of spatio-temporal joint alignment.Experiment results show that the algorithm can achieve robustness,simpleness of calculation and high accuracy of estimation for complex motion scenes,satisfying the applications of spatio-temporal super-resolution imaging.2.Aiming at overcoming problems of current spatio-temporal super-resolution reconstruction algorithms such as oversimplified motion model,unknown blurring and noise level,a Maximum Posterior Likelihood-Markov Random Field(MAP-MRF)based super-resolution reconstruction method is proposed to achieve the real-world super-resolution imaging.In order to joint spatio-temporal alignment of video with large displacement,rotational movement and other complexities,the improved SIFT Flow algorithm based on sparsity in wavelet domain is proposed,which has high accuracy of estimation and robustness.A weighted 3D neighborhood system(WNS)MAP-MRF model is proposed,which uses the weighted total variation(TV)to illustrate the spatio-temporal sparsity and smoothness of multiple videos.The Belief Propagation(BP)algorithm is applied to joint estimation of the parameters of MAP-MRF model,such as motion vector and the super-resolution images,using an iterative coarse-to-fine scheme.With the proposed algorithms mentioned above,MAP-MRF based super-resolution reconstruction method has better capabilities to keep the sharpness of edge,detail of texture preserving,and the robustness of noise suppresing.The experimental result has confirmed the effectiveness of the proposed method under practical conditions.3.Aiming at overcoming the problem that CACTI can only afford low-cost temporal super-resolution(SR),which leads to reduction of reconstruction quality due to the noise and compression ratio,a robust reconstruction approach based on multi-reconstruction and MAP-MRF model is proposed.By utilizing sparsity in different domains and inter-frame non-redundant information from multiple observations,the proposed multi-reconstruction algorithm can simultaneously provide high visual quality in the foreground and background of a scene and enhance fidelity of the reconstruction results.A multi-reconstruction algorithm based on sparsity of TV and motion vectors is proposed,which has obvious advantages in robustness and quality of reconstructed images.The experimental results have verified the efficacy of our new optimization framework and the proposed reconstruction approach.4.The construction of camera array imaging system and CACTI syetem are used to verify multi-sensor spatial registration methods based on the astronomical calibration,the MAP-MRF based spatio-temporal super-resolution reconstruction algorithm,and the effectiveness of spatio-temporal super-resolution technologies under practical conditions.