Polarized light navigation orientation method in cloudy weather

Abstract

The invention relates to the field of polarized light navigation, and discloses a polarized light navigation orientation method in cloudy weather, which comprises the following steps: introducing dynamic snakelike convolution and a multi-dimensional cooperative attention mechanism into a traditional Unet neural network to improve, and designing a DSCA-Unet model for sky cloud layer segmentation; jointly predicting a mask segmentation cloud layer by combining a DSCA-Unet model and threshold segmentation; removing a cloud layer region in the clouded sky image by using a binary mask based on a sample block repair algorithm to obtain a repaired sky image; and resolving a polarization azimuth image of the restored image by using a Stokes vector, and fitting a sun meridian to solve a course angle. According to the method, the problem that the navigation precision of polarized light navigation is reduced under the cloudy condition is solved, the interference-free sky image is generated by removing the cloud layer and is used for calculating the course angle, and support and reference are provided for the actual sky polarized light navigation orientation under the cloudy weather.

Images