Probe vehicle fleets equipped with intelligent visual sensing would require a range of capabilities to interpret complex road scenes and collate a global picture of traffic conditions. This paper focuses on one such capability, mapping and describes a metric mapping algorithm, deferred segmentation Mapping, which uses range data from a scanning laser sensor. Unlike other methods, this approach treats the measurement scan as defining a closed region of unoccupied space, and builds the map by manipulating the boundaries surrounding the space. Its advantages are that it avoids assumptions about surface connectivity, and it creates map objects with closed surfaces so that maximum extent, area etc. can be estimated immediately. Further work will include reclaiming occupied map space, dealing with moving objects, error handling and scaleability. The method has direct application in real-time fixed environment mapping. (Author/publisher) For the covering entry of this conference, please see ITRD abstract no. E214133.
Samenvatting