Graph Based Recognition of Grid Pattern in Street Networks

Abstract

ABSTRACT:Pattern recognition is an important step in map generalization. Pattern recognition in street network is significant for street network generalization. A grid is characterized by a set of mostly parallel lines, which are crossed by a second set of parallel lines with roughly right angle. Inspired by object recognition in image processing, this paper presents an approach to the grid recognition in street network based on graph theory. Firstly, the bridges and isolated points of the network are identified and deleted repeatedly. Secondly, the similar orientation graph is created, in which the vertices represent street segments and the edges represent the similar orientation relation between streets. Thirdly, the candidates are extracted through graph operators such as finding connected component, finding maximal complete sub-graph, join and intersection. Finally, the candidate are evaluated by deleting bridges and isolated lines repeatedly, reorganizing them into stroke models, changing these stroke models into street intersection graphs in which vertices represent strokes and edges represent strokes intersecting each other, and then calculating the clustering coefficient of these graphs. Experimental result shows the proposed approach is valid in detecting the grid pattern in lower degradation situation.

Linear street extraction using a Conditional Random Field model

“…A number of methods have been developed to detect geometrical patterns other than linear streets. The detection of grid patterns has been considered in Heinzle et al (2005); Yang et al (2010); Tian et al (2012). Heinzle et al (2005Heinzle et al ( , 2006 proposed methods for detecting ring and star patterns.…”