Local Image Feature Extraction in the Context of Automated Valet Parking based on Simultaneous Localization and Mapping

With the recent advances in automated valet parking solutions on the mobility market for controlled environments, the attention on generalizing this solution in arbitrary conditions got into the focus as well. This relies on robust localization that mainly uses visual information, which is still challenging under harsh illumination and weather conditions. In this work, we present the results of the robustness analysis for image-based localization techniques in the context of automated valet parking based on Simultaneous Localization and Mapping. We evaluated the most promising methods from the state of the art, with a focus on the keypoint-feature descriptor robustness in challenging outdoor conditions. The evaluation benchmark, dataset, and framework are available on the author’s webpage.