Research On Imaging Method And Application Technology Based On Compressed Sensing

The photoelectric viewing system is one of the core system of army reconnaissance and perception equipment,which plays an essential role in target detection and battlefield situational awareness.In this paper,a imaging technology based on compressed sensing was put forward to deal with the requirements of high-resolution reconnaissance of land based photoelectric viewing system.A conjugate image plane coding compressed imaging method was proposed,and the feasible technical scheme was designed and verified.The performance of this technology was characterized using an experimental system and verified its advantages,and a parallel scheme of image reconstruction algorithm was designed with Jetson TX2 embedded platform.The main work can be described as follows.1)The theoretical framework of compressed sensing was investigated and analyzed with three components.Meanwhile,the feasible technical method for developing new compressed sensing algorithms was also discussed.2)After analyzing and discussing the imaging model of the traditional imaging system and the factors that influence the resolution of the traditional photoelectronic imaging system,a conjugate image plane coding compressed imaging method was put forward to deal with existing technology's limitations of combining the compressed sensing with optical imaging technology.Besides,a theoretical model and a hardware realizable coding mask matrix of the imaging method were built and designed,respectively.The steps and principles of the TVAL3 algorithm also were analyzed.According to the model,further simulation work was done to verify the performance of this imaging method.3)A technical test scheme of the conjugate image plane coding compressed imaging system was designed,and a matching relationship and a selection analysis of each module of the system were carried out in more detail.A Digital Micromirror Device(DMD)was adatped as a hardware of the coding mask,and a optical path structure of the compressed imaging system was designed based on the DMD.In order to overcome matching difficulties between DMD and detector,a new method for reducing the error between DMD micromirrors and detector pixels was proposed,the method recorded coding mask with high-resolution camera.Finally,performance of the system was evaluated by a test experiment.4)The application technology of embedded image processing was studied on Jetson TX2 embedded platform to deal with the miniaturization requirements of a compressed imaging system and the parallelization of image reconstruction algorithms.A feasibility of parallel solving '?i subproblem' and '? subproblem' in the image algorithm was analyzed,and the'? subproblem' which requires high computing requirements was decomposed at the parallel level.Besides,a parallel algorithm with blocks was also proposed for matrix vector multiplication operation with the largest amount of computation and the most time-consuming operation in the '? subproblem'.