Abstract:  In order to solve the problem of extracting navigation region in vision positioning and navigation of
unmanned aerial vehicle, a navigation region extraction algorithm based on superpixels’ saliency is proposed.
Firstly, the land image is divided into superpixel regions with high internal similarity by using SLIC (Simple
Linear Iterative Clustering), the saliency value of each superpixel is calculated to obtain the superpixel saliency
map. And the quasi guidance areas are extracted based on the navigation region filtering rules. Finally,
navigation areas are obtained through estimating the matching probability of the quasi guidance areas. The results
of our experiment show that the precision and recall of the proposed method are respectively 89% and 87% on
the test set, which means the method can basically meet the requirements of UAV (Unmanned Aerial Vehicle)
vision positioning and navigation.

Key words: vision-based navigation, superpixel segmentation, navigation region, saliency map, unmanned aerial vehicle (UAV)