Structured Compressed Sensing Method And Implementation Of Infrared Small Target Image

Infrared small target image capture and transmission is the basic premise for various infrared technology applications.The Nyquist theorem constraints the traditional methods of image sampling and compression,some phenomenon like high cost and waste,will be caused in this process.Compressive sensing(CS)could combines the image sampling and compression,and it can overcome the above defects of the conventional methods.The compressive sensing technology can find a place in compression of many different kinds of images.CS can be used in infrared small target image collecting process.Some scientific studies have showed that the performance of signal CS reconstruction will be improve if the structured prior of signal is added to CS process.This dissertation launched a study on structured CS method for infrared small target images,and make an initial exploration on realization of this CS method.1)An improved Bayesian CS method to incorporate structured prior of infrared small target image is proposed.Inspired by the existed T-MSBL algorithm,our CS method mine the structured information of row-correlation in one infrared small target image.By using multiple measurement vectors(MMV)to model the row-correlation in reshaped small target image,CS measurements will contain the structural information of the image signal.When the compression rate remaining unchanged and image fidelity being secured,the calculation of CS reconstruction will reduce and the image rebuilding speed will accelerate.Inference from the mathematics angle and comprehensive simulation results have proved the validity of the new CS method.2)A design for infrared small target image acquisition system of hardware and software combination is proposed.This image acquisition system is an implementation of the above CS method.According to the characteristic of this new CS method,the whole system will composed of three modules.Row CS module is a hardware module,one kind of Analog-to-information Convertor(AIC)is used to realize its function.Row measurements reshaping module is completed with the use of DSP peripheral interface.Column CS module use software to accomplish its task in DSP.The feasibility of the system scheme is verified with the derived equivalent CS measurement matrix about the AIC in the proposed system.Frame frequency of the system is calculated reasonably,this also shows that the system has potential research values.In short,the CS method proposed in this dissertation took into account compression efficiency,reconstruction time,reconstruction accuracy and practical application,and it has big potential in the future.