Research On Key Technologies Of Infrared Polarization Detection

The traditional infrared detection system can only obtain the intensity information of the target,and the infrared polarization detection system can also obtain the polarization information,which is helpful to study the polarization characteristics of the target.In this paper,based on the infrared division of focal plane polarimeters,the key technology of polarization detection is studied.The main research contents include:1.calculation method for degree of polarizationAfter obtaining the polarized information of different angles from the incident light,the degree of polarization and polarization angle of the incident light can be calculated by using the Stokes vector and the Mueller matrix.However,in practice,due to the limitations of the manufacturing process,the parameters of the division of focal plane polarimeters is not ideal.It is difficult to obtain accurate polarization information using traditional methods.A new polarization calculation method is proposed in this paper,which has a more concise expression form and can solve the encountered problems.In addition,this method has a wide range of applicability.2.High-speed infrared polarization imaging and processing systemAn infrared polarization imaging prototype was developed and an embedded graphics processing unit(GPU)was used as the processing core to process polarization information in real time.In addition,it provided a good hardware platform for target recognition.The system has faster processing speed and friendly development environment than traditional processing systems based on FPGA(Field-Programmable Gate Array)or DSP(Digital Signal Processing).3.Research on non-uniformity correction method for division of focal plane polarimetersAfter using division of focal plane polarimeters to obtain the target polarization image,the reasons that influence the polarization imaging quality are analyzed,including the crosstalk between pixels,the transmittance and angle uniformity of the micro polarizer,the uniformity of the detector response and the instantaneous field of view error.For the uniformity problem,two non-uniformity correction methods are proposed based on the new polarization calculation method,one is to separate the transmittance and the detector gain separately,and the other is to calibrate the transmittance with detector gain.Based on fusion calibration,two-point fusion correction and multi-point fusion correction methods are proposed,and good correction effects was achieved.The problem of inaccurate calculation of the polarization using previous correction method is solved by only considering the uniformity problem.4.Research on infrared polarization target recognition algorithmThe current mainstream target detection algorithms and detection algorithms based on deep learning are analyzed.The deep learning framework based on SSD(Single Shot Multibox Detector)is used to study infrared polarization target recognition.Due to the lack of a publicly available infrared polarization sample data set,and the detector in the article was used to completes the acquisition of the sample.In this data set,the recognition effect of infrared polarization fusion is better than infrared.