Research On Image Compressed Sampling And Recovery Algorithm Based On Compressive Sensing In Deep Space

The acquisition and transmission of deep space image is faced with difficulties of high sampling rate, huge storage requirement and low transmission rate. Although traditional deep space image compression and transmission scheme can relieve the pressure on equipment storage and transmission of deep space communication system to a certain extent, Nyquist high speed sampling cannot be avoided. Meanwhile, the way of deep space image compression encoding which discards large amounts of data after high speed sampling is an extreme waste of time and resource. As a new theory of signal sampling and processing, Compressive Sensing can achieve the compressed sampling of signal at far less than the Nyquist rate. Therefor, it will be of great significance to study the correlative technique problems of deep space image compressed sampling based on the theory of Compressive Sensing.This thesis launches the research on measurement matrix and recovery algorithm according to the specific requirements of deep space image compressed sampling and achieves two innovations as following:1. In deep space image compressed sampling, measurement matrix must satisfy the recovery condition of Compressive Sensing as well as the specific requirements of deep space image acquisition including low storage and computation complexity, findredship to imaging equipment hardware implementation and the ability of fast encoding for large deep space image. Based on Weighing Matrix, this thesis utilizes the construction method of Structurally Random Matrix and the excellent pseudo-random property of chaotic sequence to construct a new measurement matrix with properties of block structure, simple elements, controllable sparsity and findredship to hardware implementation. Theory analysis and simulation results show that the proposed measurement matrix can sample deep space image with low storage and computational cost and simultaneously ensure high accuracy of image recovery to provide some reference for future application of Compressive Sensing in acquisition of deep space image.2. Deep space image compressed sampling puts forward high requirement for precision of recovery algorithms because of the great scientific value of deep space image. Meanwile, the data to be recovered in deep space image compressed sampling is always huge. Therefor, designing image recovery algorithm with high precision and the ability to process large-scale data is another problem to be solved in deep space image compressed sampling. Based on the frame of Gradient Pursuits algorithm, this thesis proposes a new recovery algorithm with high precision and low complexity. By using a nonmonotone supermemory gradient algorithm to calculate search direction and stepsize, the overall performance of proposed recovery algorithm in Compressive Sensing is significantly improved. Theory analysis and simulation results show that the proposed recovery algorithm has high speed and precision for deep space image recovery if appropriate parameter is choosed to provide a feasible choice for future efficient recovery of deep space image.