Automated High Precision Path Annotations on Digital Map from Crowdsourced GPS Traces

Developing reliable and well-informed mobile phone-based transportation is followed by growing popularity and accessibility of digital maps, which has definitely simplified navigation. But that comes with severe drawbacks in the case of narrow unknown paths which have no place in these maps. The inevitability of poor path geometry precision and fine-grained scarcity knowledge is hampering a route guidance system. In this paper, we are offering an approach to construct high-precision maps from traces of crowd equipped with differential GPS receivers with updated existing path along with addition of new path segment scaled according to the exact width and shaped on electronic map. As the cost of GPS-enabled devices is decreasing rapidly, we assume that large amounts of GPS traces can be obtained with sufficient confidence to build up a new road section. Our approach is to make a server side application where the crowd traces will be uploaded through an user-friendly graphical interface and a fine-grained map standardized by human labeling will be generated. This paper describes an offline algorithm which generates an enhanced map by a certain time gap in a contiguous manner. Our new contribution is a spatial community detection of GPS coordinates from all uploaded traces subjected to a threshold value which is crucial to select as it varies with the width of path and volume of collected information from crowd.