Increase of the Navigation Landmark Detection Accuracy Based on the Computer Vision Methods in a Lack of Visibility

This paper presents the computer vision system for navigation of an unmanned vehicle based on the use of navigation landmarks in a lack of Visibility. The CNN-based image preprocessing algorithm has been developed to improve the accuracy of navigation landmarks recognition in conditions inherent to the far north (snow, fog). The modified CNN YOLOv4 was used for the recognition. The recognition accuracy of the pre-processed images resulted in 84%.