Font Size: a A A

Compressive Imaging Theory And System Design In Optical Remote Sensing

Posted on:2014-07-08Degree:DoctorType:Dissertation
Country:ChinaCandidate:Z L WangFull Text:PDF
GTID:1108330479979542Subject:Systems analysis and integration
Abstract/Summary:PDF Full Text Request
As the most important method for mankind to obtain the information, optical remote sensing(ORS) imaging always suffers many neck-bottles that restrict the improvement of imaging quality, such as the difficulty of system optimization, complex imaging environment and high data rate. Compressive imaging(CI) provides a new idea to the above problems by changing the traditional imaging method. This thesis mainly researches on key issues in sampling, reconstruction and processing of CI in ORS; and investigates the program of system design in ORS CI. The main achievements are as follows:Firstly, compressive sampling theory on sparse signals in continuous domain is researched on. Compressive sensing theory is mainly used for sparse signals in discrete domain; and it is difficult to implement compressive sampling and reconstruction for that in continuous domain. This thesis studies the sampling theory of signals with finite rate of innovation(FRI), which suits for sparse signals in continuous domain. Especially for multiple signals with FRI and sparse common support, we obtain the theoretical low bound of the compressive sampling number for exact reconstruction. Meanwhile, sampling theory of signals with FRI in parametric domain is proposed to enlarge the sampling space; and the corresponding implementation is also presented.Secondly, we introduce three kinds of dynamic CI methods, i.e., progressive CI, motion compensation CI and motion super-resolution imaging. With respect to the unknown sparsity of the imaging scene, we propose the progressive CI with innovative reconstruction, which offers the online control between the sampling rate and reconstruction quality.For the relative motion between the imaging scene and imaging system, we establish the motion compensation CI by motion compressive sampling model and joint reconstruction,which overcomes the impact of relative motion to imaging quality. At the same time, we investigate the mechanism of capturing the sub-pixel information from motion sampling and propose the motion super-resolution imaging method by subdivision scheme and joint reconstruction, which makes a breakthrough in imaging system for sub-pixel imaging.Thirdly, manifold regularization model for image processing is proposed and the corresponding algorithm is designed. With respect to the inverse problems in image processing, this model generalizes the regularization model in Euclidean space. It not only provides a useful tool for investigating the mechanism of image degradation in the viewpoint of geometry and utilizing the coupling information among image channels; but also suits for the images defined on general surface. Meanwhile, with respect to the nonconvexity and non-smoothness of the model, the alternate optimization algorithm is also designed. This algorithm makes use of the splitting scheme to transfer the original model to its splitting format with good local properties; then the new format is solved numerically by alternate optimization method.Finally, we propose the pushbroom CI system design scheme in ORS as well as the method for analyzing its performance by mathematical simulations and semi-physical simulations. Based on the character and requirements of ORS, the proposed scheme not only produces advantages of CI, such as high signal-to-noise rate(SNR), high quantum efficiency and low data rate, but also suits for the motion environment of imaging platform.It also has high stability and good extendibility. Meanwhile, we make use of mathematical simulations and semi-physical simulations to quantitatively analyze the impact of the imaging factors, such as the sparsity of imaging scene and sampling rate, to imaging quality, which causes the optimization and improvement of the imaging system.
Keywords/Search Tags:Compressive sensing, Sampling of signal with FRI, Dynamic compressive imaging, Manifold regularization, Pushbroom system
PDF Full Text Request
Related items