分 焦 平 面 偏 振 相 机 串 扰 的 固 定 线 性 叠 加 模 型

Objective The division of focal plane polarization camera is a widely used, integrated polarization imaging system. Crosstalk between the pixels of the micro-polarizer array is a unique interference factor in such systems, which introduces errors in the measurement of the target’s polarization information. The value and superposition ratio of the crosstalk light intensity vary with the polarization state of the incident light in the scene. Previous polarization crosstalk models generally regard crosstalk as random temporal noise, which assumes that it is difficult to eliminate the crosstalk effect through pre calibration methods. A mathematical model describing the relationship between crosstalk and the system’s polarization parameters has not been established or verified. The main solution to restraining the crosstalk effect is typically through the optimization of the polarization imaging system’s structure. Methods We consider the factors influencing crosstalk in application and establish a whole-process crosstalk model that includes the parameters of the hardware, crosstalk, and polarization analyzer. The result of crosstalk is regarded as the linear superposition of constant weights of multiple analyzed intensities at different polarization analyzed directions. The intensity response containing crosstalk can still be accurately characterized using the cosine form, which is equivalent to a constant deviation in parameters such as the extinction ratio and analyzed directions, and can be distinguished from the random error caused by temporal noise. Through our experiment correlating the polarization state and crosstalk, we demonstrate the calibrability of the crosstalk-influenced analyzer parameters, the non-correlation of the incident light’s polarization state, and the correctability of the crosstalk deviation. The experiment uses a Sony polarization sensor to sample polarized light emitted from the integrating sphere, which rotates through a full cycle via a motor-driven mechanism, to fit the analyzer parameters and inversely infer the crosstalk coefficient with high precision. The light intensity and degree of polarization correlation experiments are respectively judged by the linear characteristics of the denoised unpolarized-light and pure-polarized-light response, while the angle of polarization correlation experiment is judged by the cosine response characteristics of the full angle of polarization periods. Results and Discussions The simulation shows that the transverse electric (TE) wave transmittance can be ignored, and the diffraction crosstalk distribution only needs to consider the transverse magnetic (TM) wave component in the incident light. Therefore, the crosstalk coefficient can be regarded as constant (Fig. 6). Increasing the exposure time leads to an increase in dark noise and a decrease in the extinction ratio (Figs. 10‒12), which highlights the necessity of pre-calibration and denoising to eliminate interference in crosstalk measurement. The simulation shows that crosstalk is independent of the target’s angle of polarization and degree of polarization (Fig. 6). This can be verified by varying the target polarization state and measuring the change in the system extinction ratio coefficient. The average change in the crosstalk coefficient is <0.1% within the measuring range (Figs. 13‒15), which proves that no correlation exists. Conclusions Crosstalk will lead to degradation of the extinction ratio, the deviation in the analyzer polarization angle and non-uniformity of the light intensity response coefficient. However, as long as the crosstalk coefficients do not change during use, the system analysis model remains a simple cosine form with constant coefficients, and this crosstalk effect can be calibrated. The crosstalk coefficient only needs to consider the TM wave diffraction distribution, and the polarization state of the target/scene is not correlated with the crosstalk. The crosstalk coefficient is a function of the effective aperture (working F-number) and pixel position.