In real life, the targets behind the window need to be detected on many occasions. For example, when a natural disaster occurs, the targets inside trapped vehicles and boats should be identified, and it is also necessary for museums to filter out glass display cases of stray light and highlight the true appearance of exhibits. Additionally, the driving status of drivers should be accurately identified in road video surveillance, and in the fight against crimes, it is a necessity to accurately distinguish the location relationship between criminals and hostages behind the window. These occasions require a more reasonable target through-window detection method, and traditional target through-window detection methods have certain limitations in avoiding image degradation. The current research mostly adopts polarization image processing and image fusion methods to solve the image degradation, but these methods are not applicable in the light spot obscuring the target. When reflection interferes with target identification, single polarization suppression reflection is the most commonly employed polarization suppression reflection method. However, in the face of strong reflected light interference, the acquired image is overexposed and it is difficult to distinguish the target from the background. Utilizing double polarization suppression reflection can overcome the strong reflected light interference, and the contrast between the target and the background is improved with lower overall image brightness, and the recognition ability for dark and weak targets is poor in practical applications. Therefore, a more reasonable polarization suppression reflection method is necessary for active imaging, and it can be applied to target recognition under strong reflection interference and to the recognition of dark and weak targets. Finally, better image information for subsequent image processing can be provided.

The specular reflection of glass is used as the interference object. Firstly, we analyze the ability of polarization suppression reflection and conduct polarization suppression reflection experiments by single polarization suppression method and double polarization orthogonal method respectively. Secondly, we analyze the change law of Mueller matrix elements under different incident angles of light source. M₂₂ and M₃₃ images in the Mueller matrix are different from other matrix elements in the distribution of gray values, and the depolarized images generated by M₂₂ and M₃₃ can effectively distinguish the target from the background. Finally, by comparing the target through-window detection image index, it is proven that the utilization of depolarized images can ensure higher overall brightness of the images and improve contrast between the target and the background simultaneously.

The double polarization orthogonal method suppresses the specular reflection well, with improved contrast between the target and the background and lower overall image brightness (Fig. 8). The M₂₂ and M₃₃ images are the most sensitive to the distinction between the target and the background, and the gray values of the M₂₂ and M₃₃ images are more dispersive than those of other matrix elements (Fig. 10). The mean gray value and standard deviation of M₂₂ and M₃₃ images are significantly higher than those of other matrix elements, which indicates that the brightness dispersion and gray values of M₂₂ and M₃₃ images are much higher than those of other matrix elements, and the incident angles of different light sources show the same pattern (Fig. 11). The depolarized images improve the contrast between the target and the background, and ensure the overall image brightness. Meanwhile, they have a sound image effect in dealing with the target through-window detection at most angles, and do not cause image information loss due to the excessive intensity of the light source (Table 4).

This study designs a Mueller matrix test setup. Firstly, the polarization ability to suppress reflection is analyzed. Secondly, the gray value distribution of Mueller matrix elements under different incident angles of the light source is analyzed to obtain the change law of different Mueller matrix elements. Finally, depolarized images are employed to improve the target saliency. The results show that the double polarization orthogonal method exploits the difference in polarization characteristics between the specularly reflected light and the target reflected light. The orthogonal polarization information is suppressed by the polarizer. The polarization information through the same polarization direction as the polarizer is applied to achieve the purpose of suppressing reflection. However, the overall image brightness obtained by the double polarization orthogonal method is too low, and it is difficult to identify the dark and weak targets. The double polarization orthogonal method is effective but defective in practical application capability.

In analyzing the variation pattern of Mueller matrix elements under different incident angles of the light source, the average gray values and standard deviation of M₂₂ and M₃₃ images in the Mueller matrix are significantly higher than those of other matrix elements. This indicates that the dispersion of brightness and gray values of M₂₂ and M₃₃ images are much higher than those of other matrix elements, and the incident angles of different light sources show the same pattern. The depolarized images generated by M₂₂ and M₃₃ can effectively distinguish the target from the background. Therefore, we further adopt Mueller matrix images to generate depolarized images and obtain images with better target through-window detection. The depolarized images generated by M₂₂ and M₃₃ improve the contrast between the target and the background and ensure the overall image brightness. They have a better image effect in dealing with target through-window detection at most angles and do not cause image information loss due to excessive light source intensity. There is a significant improvement over the single polarization suppression method and the double polarization orthogonal method. Therefore, the reasonable use of depolarized images for target through-window detection can effectively suppress reflection and improve the recognition of target detail information. Additionally, a new solution is provided for the relevant applications of product design and the unavoidable requirement for target through-window detection. However, we do not further optimize the depolarized images through image fusion, resulting in bright spots in some parts of the images. In the future, image optimization will be conducted on depolarized images, and image fusion will be employed to restore the images to bright spot-free ones.