Polarization detection technology is a scene multi-dimensional optical information acquisition technology. It can be used in the detection of small targets, and can overcome the shortcomings of traditional optical detection technology that are difficult to detect in low-light scenes and harsh weather conditions. However, the target size in the polarized image is small and the response is weak, and the distribution characteristics of background clutter and image noise are different from the intensity image, which are the problems that the detection algorithm needs to solve. At first, the principle of polarization imaging is analyzed, and the characteristics of polarization image and traditional intensity image are compared. And then, the intensity image and polarization image are fused and enhanced. At last, according to the local contrast saliency theory, a double-neighborhood gradient based on polarization fusion image target detection algorithm is proposed. Experimental results show that compared with the traditional algorithm, the algorithm makes full use of the characteristics of polarized images, and can more effectively detect small targets.