Polarization Orientation Method Based on Remote Sensing Image in Cloudy Weather

Autonomous navigation technology is a core for intelligent operation, allowing the vehicles to perform tasks without relying on external information, which effectively improves concealability and reliability. In this paper, based previous research bionic compound eye, multi-channel camera array with different polarization degrees was used construct atmospheric state measurement platform. A trough threshold segmentation algorithm applied study distribution characteristics characterization methods of states in remote sensing images. extracted feature map, tilting suggestion box obtained multi-direction window extraction network (similarity-based region proposal networks, SRPN) rotation (Rotation Region interests, RRoIs). Fast Convolutional Neural Networks (RCNN) screen boxes, Non-maximum suppression (NMS) method select angle, corresponding label highest score, as solar meridian azimuth vehicle coordinate system. The angle system can be calculated by astronomical formula. Finally, final heading according conversion relationship between systems. By fitting measured data least Square method, slope K value −1.062, RMSE (Root Mean Error) 6.984, determination coefficient R-Square 0.9968. Experimental results prove effectiveness proposed algorithm, an autonomous high concealment precision, providing new idea technology.