Abstract Localization of a mobile robot is crucial for autonomous navigation. Using laser scanners, this can be facilitated by the pairwise alignment of consecutive scans. In this paper, we are interested in improving this scan alignment in challenging natural environments. For this purpose, local descriptors are generally effective as they facilitate point matching. However, we show that in some natural environments, many of them are likely to be unreliable, which affects the accuracy and robustness of the results. Therefore, we propose to ?lter the unreliable descriptors as a prior step to alignment. Our approach uses a fast machine learning algorithm, trained on-the-?y under the positive and unlabeled learning paradigm without the need for human intervention. Our results show that the number of descriptors can be signi?cantly reduced, while increasing the proportion of reliable ones, thus speeding up and improving the robustness of the scan alignment process.