In the compound eyes of several insects, the dorsal rim area has regularly arranged polarization sensitive ommatidium. Insects rely on this special structure to extract the polarization information of skylight to facilitate navigation when the insect's eye view was partially occluded. An array polarization sensor was designed for navigation by imitating this special structure. In practical applications, the sensor was disturbed by many factors such as random partial occlusion of the view, multi-orientated wire-grid polarizer defect, and sharp noise. To address these problems, this paper proposes a novel and robust polarization intensity extraction algorithm, which was based on linear gray stretch, Otsu threshold segmentation, and the 3σ theorem. The polarization algorithm was discussed and the structure of the sensor was presented. Based on the MFC and OpenCV, a real-time control and monitor software interface of the sensor was developed, which can display polarization information with data update rates up to 10 Hz. The angle polarization measurement test results indicate that the sensor is stable and the angle error is ±0.25°. Robust tests under outdoor conditions with and without occlusion indicate that the sensor can adapt to changes in random shade environments, which is indicative of the robustness of the approach.