To conduct research on infrared polarization imaging technology, this paper designs and builds a long-wave infrared polarization imaging system to achieve target time-sharing imaging. An improved differential image correction method based on single-pixel inhomogeneity was proposed to remove cold reflections; the image was registered using the Sobel edge detection method, and the polarization characteristics of the target scene were analyzed. The results demonstrate that the proposed system can obtain the infrared polarization information of the target scene, and the pre-processed image meets the requirements of the experiment. The edge contour and detailed information obtained by the Stokes vector method for the polarization image of the target scene are more abundant. To further measure the performance of the experimental system, the intensity and polarization degree images were fused. Compared with the infrared intensity image, the fused image has a significantly improved image evaluation quality, which verifies the feasibility of the experimental system. This technology can potentially improve target detection efficiency in complex environments in the future.