Research On Image Processing Technology Based On Wollaston Prisms Polarization Camera

In recent years,target recognition in low contrast environment is the key research direction of "key technology fields of detection and recognition" in the new generation of national development strategy.Polarization imaging technology uses the differences of radiation intensity and polarization characteristics of the target itself to distinguish the target and background,which is not easily affected by the changes of external environment.It can be used as a new detection method under the conditions of low illumination,complex background,target hiding and so on.In view of the problems of low precision and slow processing speed in the current polarization imaging detection system,this paper proposes an image processing research method based on the Wollaston prisms polarization camera.The work includes the following points:Firstly,the definition and classification of polarized light are introduced,and the state of polarization of light is described by Stokes vector method.The characterization of Muller matrix of different optical elements and the important parameters DOP(degree of polarization)and AOP(angle of polarization)in polarization measurement are studied.Finally,the common polarization imaging systems are compared and analyzed,including time-sharing and simultaneous polarization imaging systems.Secondly,the overall scheme of polarization imaging system is designed.It is mainly divided into the image acquisition part of Wollaston prisms polarization camera and FPGA image processing part,and then it is divided into modular design,focusing on the polarization imaging optical system module design based on Wollaston prisms,CCD camera acquisition module design,DDR3 image storage module design,HDMI image display module design.Finally,the polarization image processing algorithm is studied,including the realization and optimization of registration and fusion algorithm.In the aspect of polarization image registration,this paper combines the eight direction Sobel operator and SIFT(scale invariant feature transform)algorithm,using this improved SIFT algorithm can reduce the number of useless feature points,reduce the complexity of feature point matching algorithm,and improve the matching accuracy and operation speed.In the aspect of polarization image fusion,the selected source image is one of DOP image and Stokes parameter image.Three fusion methods are used: weighted average method,energy value maximization method and improved Curvelet transform method.The results show that the imaging effect of the three fusion images is better than that of the intensity image,and the objective evaluation index is used to verify that the fusion algorithm based on Curvelet transform can fully retain the information of the polarization image.A large number of experiments are carried out in different environments by using the imaging detection system.The experimental results show that the obtained polarization image contains rich feature information of the target,especially in the haze environment,it can greatly increase the amount of information contained in the detected target,which is expected to be applied to the detection and recognition of the target.