I'm doing an quality assessment of openstreetmap's street data. As a comparism dataset I have to use an administrative dataset, which is highly accurate and comes in polygons. Here's a sneak peek:
So far I've examined positional accuracy via goodchild and hunter's buffer method. But that's not enough content and I'm struggling to find more evaluation methods I could build some models with. I came across hausdorff distance, but a (required) conversion from polygons to polylines didn't work out well.