随着智慧城市、智慧交通和天网工程的不断发展，复杂城市环境下的车窗透反混叠图像解耦技术成为当前研究的热点，在卡口登记、安全驾驶、逃犯追踪和**反恐等领域需求迫切。为克服传统成像技术依赖光强信息、易受环境干扰等劣势，基于偏振成像技术建立玻璃透反光传输模型，提出车窗透射信息光和反射干扰光的解耦分离方法，面向室内模拟场景和户外真实场景开展实验研究。结果表明：室内场景下轿车模型和客车模型消反图像的信息熵较原始强度图像分别提升14.3%和9.8%；室外场景真实车辆消反图像因消除包含大量环境信息的反射光成分而使信息熵较原始强度图像下降2.7%；这几类场景消反图像的区域对比度相较原始强度图像分别提升40.1%、117.5%、237.8%。此外，研究图像消反质量与模型几何因素的相关关系，所得结论能对相机摆放高度和车辆停止线划线位置起到指导作用。

Owing to the development of smart cities, intelligent transportation, and the Skynet project, increased attention is being paid to the decoupling technology used in the transmission and reflection of the overlapping images of car windows, which is essential for checkpoint registration, driver monitoring, fugitive trailing, and military counter-terrorism. In this study, a light transmission and reflection model based on polarization imaging technology is established and a decoupling method for transmitting and reflecting light from car windows is proposed to overcome the drawbacks of traditional imaging technology such as the reliance on intensity information and susceptibility to disturbance. The results conducted under the simulated indoor and realistic outdoor scenes show that the image information entropies of the de-reflected images of a saloon car and passenger car model are 14.3% and 9.8% higher than those of the original intensity images, respectively. Moreover, the image information entropy of the de-reflected image of a real vehicle is 2.7% lower than that of the original intensity image because the reflection contains a large quantity of environmental information. Additionally, the region contrast of these de-reflected images is improved by 40.1%, 117.5%, and 237.8%, respectively, compared with those of the original intensity images. Therefore, the relationship between the decoupling quality of the overlapping image and geometric factors of the model is studied, and a conclusion aimed at providing a reference for the installment height of the camera and position of the vehicle stop line is provided.