Automated detecting and placing road objects from street-level images

Abstract

Navigation services utilized by autonomous vehicles or ordinary users require the availability of detailed information about road-related objects and their geolocations, especially at road intersections. However, these road intersections are generally represented as point elements without detailed information, or are even not available in current versions of crowdsourced mapping databases including OpenStreetMap (OSM). This study proposes an approach to automatically detect road objects from street-level images and place them to correct locations according to urban rules. Our processing pipeline relies on two convolutional neural networks: the first one segments the images, while the second one detects and classifies the specific objects. Moreover, to locate the detected objects, we propose an attributed topological binary tree (ATBT) based on urban rules for each image in an image sequence to depict the coherent relations of topologies, attributes and semantics of the road objects. Then the ATBT is further matched with map features on OSM to determine the right placed location. The proposed method has been applied to a case study in Berlin, Germany. We validate the effectiveness of the proposed method on two object classes: traffic signs and traffic lights. Experimental results demonstrate that the proposed app roach provides promising results in terms of completeness and positional accuracy.

Funding

Statens vegvesen,