Research and verification of key technologies for infrared polarization of non-cooperative space targets

Based on the target detection requirements of spacecraft infrared detectors and aerospace engineering principles, this paper proposes a non-cooperative target polarization feature processing data fusion algorithm and completes the entire process. All the work is based on a self-developed wide-spectrum polarization infrared sensing system. This algorithm innovatively constructs an infrared polarization dataset for space equipment by acquiring infrared polarization images from multiple angles. Combined with the single-level target monitoring network YOLOv7, a method for inverting the infrared polarization features of non-cooperative targets in space is developed and applied for the first time, to the best of our knowledge, to space target identification. Experimental results show that this method achieves a 95% accuracy rate in identification and detects 10 types of targets, including solar panels, satellites, antennas, docking rings, etc., with an inference time of 20 ms.