In the study of the autonomous navigation of outdoor intelligent mobile robots, bionic polarization navigation technology has gradually attracted attentions due to its characteristics that it is not easily affected by accumulated errors and electromagnetic interference. In order to improve the accuracy and enhance the robustness, a heading angle extraction method suitable for urban and haze weather is proposed for bionic polarization navigation. The image intensity information that is not easily affected by occlusion and haze factors is introduced, and the optimization cost is constructed for the image intensity information and the polarized skylight information respectively, and joint optimization is carried out to obtain the optimal heading angle. Outdoor experimental tests show that the information fusion method performs well in scenarios such as haze or sky occlusion. The standard deviation is 0.10~0.24° in the hazing scene, which is 83.1~99.7% higher than that of the single polarized skylight heading angle extraction method; the standard deviation in the sky occlusion scene is 0.04 ~0.09°, which is 90.9~96.7% higher than the single polarized skylight heading angle extraction method.