Design Of Passive Millimeter Wave Imaging System And Research On Image Super-resolution Restoration Algorithm

Millimeter-wave (MMW) is the electromagnetic wave that frequence is between microwave and infrared ray. MMW has the excellent transmission performance in the clouds and dust, and is sensitive to metal. MMW is widely used in military, civil areas because of its unique properties. Passive Millimeter-wave imaging system (PMMWIS) has the advantages of high resolution, good stealth, good counter stealth and compact structure, belonging to the forefront topic of imaging detection field, becoming a hotspot in recent years.PMMWIS is headed towards multi-channel, integrated. Because of the constraint of electronic technology and costs, image resolution is not high enough. Image Super-resolution (SR) technology is aimed to get higher resolution by restoring the information outside the cut-off frequency, which is determined by the limit of optical diffraction.The background of SR and PMMWIS is introduced firstly. Then a single-channel feed scanning PMMWIS is designed based on 3mm PMMW radiometer and TMS320F2812 DSP, while the principle of PMMW imaging is analyzed. At last, two algorithms, named SR Restoration for Image Based on Entropy Constraint and Projection onto Convex Set (POCS) , SR Restoration for Moving Image Based on Scale Invariant Feature Transform (SIFT) and Artificial neural network (ANN), are proposed.Super-Resolution Restoration for Image Based on Entropy Constraint and POCS Algorithm adds entropy constraint to original constraints. Entropy-difference threshold is set to achieve integrity constraint for a recovery image, and the rate-step, which is associated with entropy- difference, is set to control the speed and accuracy in iterative adaptively. Experiments with MMW images and optical images show that the improved algorithm significantly improves the resolution and speed for restored imagesSuper-Resolution Restoration for Moving Image Based on SIFT and ANN Algorithm is proposed for images with complex motion. SIFT, which is a popular local invariant feature extraction algorithm in the field of computer vision, is used for the motion compensation for non-rigid movement. Additionally, BP ANN is used for classing PSF before POCS. Experiments show that the new algorithm has better recovery effect for image sequences with complex motion.Two improved algorithms take advantage of the classical SR algorithms, can be applied in SR for MMW images and optical images.