Millimeter-wave Passive Imaging System Modeling And Preprocessing Algorithm

Passive millimeter wave imaging system works by making use of the differences of radiant electromagnetic energy of objects without transmitting electromagnetic signals itself. It has the ability to image under poor conditions like sandstorms, fog and night scene. It can find camouflaged objects and fortifications, or concealed weapons under thick clothing. For its unique technical advantages, passive millimeter wave imaging in counter-terrorism, security and other needs become more and more prominent.However, the mathematic model for focal plane passive millimeter wave imaging system is not perfect by now, that restricts the design of passive millimeter wave imaging system; on the other hand, the research on the image preprocessing for passive millimeter wave image is rare, that directly impacts on the image process later. Therefore, by the study project of passive millimeter wave imaging, this thesis studied the system model of millimeter wave passive imaging and millimeter wave image preprocessing methods, it includes:1. Analyzed system realization structure for the focal plane passive millimeter wave imaging system, and discussed the system key technical indicators.2. Be aimed at the problem that the imaging model based on radiation measuring cannot accurately describe the imaging process, researched a passive millimeter wave imaging model based on quasi optical imaging. This model effectively reflected the antenna effect, non-uniform channel, scanning geometric distortion, system noise respectively. This model laid the theoretical foundation both for system design and image restoration such as super resolution.3. Be aimed at the problem that the image strip noise caused by non-uniform channels, this thesis proposed two improved algorithms for destriping. The improved neural network algorithms (INNA) modified the estimate expression on the hidden layer, make the algorithm converge faster, and destriped more effectively. The improved filtering method in frequency domain using the prior information of antenna beam to determine the cutoff frequency of the filter. Compared with the traditional algorithms, this method is better in destriping and more easily for system implementation.4. Be aimed at the problem that the geometric distortion caused by panoramic scanning, established a system imaging equation which the exterior orientation elements were considered. This equation laid the theoretical foundation of the geometric correction for onboard or vehicular imaging system. Under the big view condition, making use of the imaging equation and anchor correction method, geometric correction was effective.Through the simulation experiments and field measurement experiment, proved the accuracy of the model and the effectiveness of these methods. The experimental results showed that the passive millimeter wave imaging mathematics model can accurately simulates the system imaging process, the image correction based on imaging equation can effectively correct geometric distortion, improved neural network algorithm and frequency domain filtering method can remove the strip noise effectively.