Research On Key Technologies Of Passive Millimeter Wave Image Processing

Via detecting and measuring the millimeter wave radiation energy of target scene,the mil-limeter wave interferometric synthetic aperture radiometer(MMW InSAR)which converts the received radiation signal into electrical signal is able to achieve the millimeter wave brightness temperature images of target scene.Millimeter wave radiation imaging system has the advantage over microwave in image resolution and anti-interference ability.And compared with infa-red(IR)and optical imaging,millimeter wave imaging system is superior in penetrability of dust,smoke and invisibility and could be operated under all time and all-weather conditions.Specially for the radiation imaging system which only receive the signal,has the favorable invisibility.Thus,it plays an important role in target detection and has been widely applied in military and civil area,such as battlefield navigation,remote sensing and security check.The purpose of millimeter wave radiation imaging system is to achieve a high quality millimeter wave brightness temperature image and applied in different areas via effective target detection and analysis.However,due to the imaging noise and the diffraction limited effect,the obtained passive millimeter wave(PMMW)image is often noisy and in low resolution.So how to improve the image quality of PMMW image,and further improve the accuracy of image classification and target detection is quite important.The image denoising,image super-resolution and classification technology is researched in the paper which mainly includes the following aspects:(1)The research on imaging theory,modeling and simulation of InSAR imaging system.The radiation characteristics of millimeter-wave target are firstly introduced,and the relationship between plural correlation output of millimeter wave interferometry and millimeter wave bright temperature distribution in far/near field condition is represented based on the binary interferometer.The imaging model of MMW InSAR is established in cartesian coordinate system and the variable imaging parameters is researched and analyzed.(2)A PMMW image joint filtering based on adaptive two-stage non-local means.To address the noise problem of millimeter wave radiation imaging system which is easily affect by imaging environment and circuits,via combining the non-local means filtering and the weak texture patches based noise estimation,an adaptive non-local means filtering is performed to improve its denoising performance.A two-stage strategy is also introduced to perform the two-stage non-local means filtering to increase the reliability and accuracy of weight distribution function,which further improves the denoising and edge-preserving ability.(3)A PMMW image denoising based on adaptive manifolds high-dimensional filtering.To solve the impulse and stripe noise problem induced by the inconsistency between channels,the window traversing method and the principal component analysis is exploited to extend image dimension and extract to the primary dimensional information respectively.The high-dimensional mean median filtering is then applied to improve the adaptive manifolds high-dimensional filtering and remove the noise in the PMMW image.Moreover,the Laplacian of Gaussian based edge enhancement method is used to enhance the object edge which further improve the quality of image.(4)The regularized nonlinear diffusion based multi-frame super-resolution method for single PMMW image.To extend the multi-frame super-resolution method to process the single PMMW image,the rolling guidance filtering which generates the structurally multi-scale PMMW sequence is performed.And the combination of regularized nonlinear diffusion based multi-frame super-resolution and the total variation convolution is exploited to achieve single image super-resolution under multi-image framework which restores the high frequency information.(5)The design and realization of PMMW image classification and recognition system.To perform the classification and recognition of PMMW image and to validate the effectiveness of proposed denoising and super-resolution method,the PMMW image dataset is built and a PMMW image classification and recognition system is designed in MATLAB GUI platform.Based on the multi-class 3D models,the main viewpoint selection and small-angle rotation is applied to capture the objects's main postures.The InSAR imaging theory is used to simulate the PMMW image and establish the database.The PMMW image denoising based on adaptive manifolds high-dimentional filtering and the PMMW image super-resolution based on gularized nonlinear diffusion are exploited to process the image.Combining the local binary pattern image feature and histogram intersection kernel support vector machine classifier,the classification and recognition of PMMW image is achieved.Finally,the effectiveness of denoising method and super-resolution method is validated by calcification and retrieval experiments.