Design Of Coding Matrix For Millimeter-wave Single-pixel Imaging And Research On Image Block Sampling Method

Traditional passive millimeter-wave imaging mostly uses mechanical scanning imaging or phased array scanning imaging working systems,which often has the disadvantages of longer sampling time or higher system cost,which limits the large-scale application in practical places.In order to effectively reduce the sampling complexity and the number of receiving channels of the system,this paper presents the framework of a passive millimeter-wave single-pixel compressed sensing imaging system.The research on the improvement of block effect and the sampling method of different image blocks are studied.The main work includes:(1)Based on the compressed sensing theory framework and passive millimeter wave imaging system,the passive millimeter-wave single pixel compressed sensing imaging system framework has been studied,and it has been verified that the millimeter wave image has good sparsity and is suitable for millimeter wave compressed sensing processing;(2)In order to further shorten the image reconstruction time and improve the image coding quality compares with traditional coding matrix,this paper further studied the deterministic coding matrix and random coding matrix based on Gabor filter,which are more suitable for the characteristics of millimeter wave images.The Matlab simulation comparison of one optical image and two millimeter-wave images,verifying that under the coding of10%?90% of the total data amount,the two coding matrices based on Gabor filters are better than the traditional Bernoulli coding matrices and Toplize coding matrices;(3)Aiming at solving the problem that the coding matrix size and image reconstruction time increase exponentially with the increase of the target scene image,the block compressed sensing framework is presented,which can effectively reduce the image reconstruction time and the coding matrix size.Because of the problem of image block effect when millimeterwave image are sampled less,an effective reconstruction image improvement method has been researched,which can reduce the image block effect and improve the quality of image reconstruction;(4)Due to the shortcomings in the traditional block compressed sensing theory of allocating the same amount of encoded data to different image blocks,four sampling methods based on the characteristics of millimeter wave image blocks have been studied: sampling method based on wavelet transform,local variance,saliency and local variance weighted adaptive allocation method,saliency and two-dimensional information entropy weighted adaptive allocation method.Simulation results verify that the above four image block sampling methods are more in line with the characteristics of millimeter wave images,and the quality of the reconstructed images are better.The wavelet transform-based sampling allocation method has the most stable reconstruction result.Based on the saliency and local variance weighted and based on saliency and two-dimensional information entropy weighted are more in line with human vision requirements under low sampling.