Studies On 1-Bit Coded Synthetic Aperture Radar Sparse Imaging

The range profile's resolution is one of the most important parameters in Synthetic Aperture Radar(SAR)imaging.And the range profile's resolution is related to the transmitting's bandwidth.Although increasing the transmitting's bandwidth can improve the range profile's resolution,it also increases the system's Analog to Digital Converter(ADC)burden.The One-Bit Compressed Sensing(1-bit CS)can not only ensure the SAR system's range profile's resolution,but reduce the required receiver ADC bandwidth.Since the one-bit sampling is much easier to realize than traditional ADC,it can also reduce the system's cost.Besides,because the one-bit sampling only keeps the signal's sign information during the quantization period,so it behaves better in anti-noise than traditional signal sampling theory.Although applying the one-bit theory to SAR imaging can reduce the amount of data,there are some new challenges to overcome.On the one hand,due to the limitation of CS theory,when the scale of scene grows the CS reconstruction algorithms'time consuming grows tremendous faster.Thus those algorithms based on CS will not suitable for SAR imaging with large scale scenes.On the other hand,there are two main kinds of disturbance,one comes from the SAR system and the other comes from background noise existed in the echo waves.Considering only the latter disturbance or normalizing both to the latter disturbance is the traditional way.By considering them as one disturbance,the algorithm can be very simple,but the accuracy of the SAR imaging model decreases.This paper mainly focus on these two problems.Our first goal in this work is a segmented SAR imaging algorithm based on one-bit compressed sensing.Our algorithm solves the problem that the reconstruction's time complexity is large.The algorithm consists of four steps,namely,reconstructing the range profile,splitting the scene into pieces,reconstructing every sub-scene,and assembling all sub-scenes together.Traditional segmented algorithm may cause error when there are targets on the boundary of two sub-scenes.However,segmented algorithm based on one-bit gains extra profit for these targets due to the anti-noise property,and can significantly reduce the error.Our second goal in this work aims to improve the SAR imaging 's quality.In radar's echo waves,the disturbance from the SAR system and the noise in the background are the main disturbance.Because the former disturbance's statistical property is not predictable,so algorithms based on dictionary learning are chosen to better decrease the influence to the image's quality.Generally speaking,the statistical property of the latter disturbance is easy to gain,so algorithms based on Bayesian Learning is used to lower the impact to the image's quality.