Stereoscopic road network extraction based on footprint graph model

Abstract:

A new three-step method called stereoscopic intersection extraction based on footprint graph model is proposed in this paper. Firstly, the primitives are extracted by footprint, then the road model is built based on the graph of primitives, finally, the road lamination is realized by referring and reasoning. Though the footprint algorithm presented by Hu can be used to extract road primitives, there exist two problems, one is that two toes which are adjacent can be easily considered as one toe, and the other is that troughs are not likely to be detected, we solved these two problems by making some improvements. Then we use stereoscopic inferring algorithm to link the roads, set the heights of the roads and realize lamination. At last, we experiment on the dataset of Google earth, as a result, this algorithm can reflect the levels of the roads effectively, what’s more, the correctness being guaranteed, this algorithm still has strong robustness.