Infrared Polarization Imaging Key Technology Research

As a new detecting technique, polarization imaging is a combination of the polarization measurement and image processing.By measuring the polarization angle and the degree of the polarization of the target's radiation and reflection, it can effectively solves the problems that the conventional photometry cannot. It can compensate for the lack of traditional thermal imaging and obtain accurate results. This new detecting technique in both civil and military fields, especially in military, has broad application prospects.In this paper, our purpose is to complete a compact structure, small size, reliable, high spatial resolution infrared polarization associated systems and use the new system to make research of image fusion, clustering recognition. The main point is the polarization grating design, system building, wavelet fusion, and clustering recognition.This paper provides a polarization grating designed by rigorous coupled-wave analysis (RCWA) with 8-14um infrared band. And verify that sub-wavelength grating has a significant polarization feature. Combined with the Stokes polarization representation, the design of real-time infrared polarization single-channel system is provided. The rotation angle of polarization gratings is accurately controlled by the stepper motor and video capture card to obtain the polarization parameters of image acquisition. The polarization image has properties of sharpening object's fringe and distinguishing roughness and material variety. With the fusion of the polarization image and infrared image, it can not only significantly improve the image contrast, the capability of discovering hidden goals, but also improves the image entropy for observation. Adopting dual-tree complex wavelet(DTCWT) to achieve multi-resolution fusion can overcome the defect of 0°,60°,120°linear polarization images with no parameter registration. Fusion algorithm combined with simulated annealing algorithm and genetic algorithm (GA) to can fix important metric adaptively which based on image performance indicators. Finally, the adaption of kernel-FCM (KFCM) can partition the region of polarized image so as to establish database accurately with infrared polarized characteristics between diverse natural objects and artificiality. The BP neural network design enables the identification of the unknown target.