Node density

Posted by leuty on 23 February 2013 in English (English)

The node density (number of nodes per way) is a measure for trust (uncertainty) for the location of a way. The more nodes we add per way, the bigger the confidence we signal to data consumers and fellow mappers. IMHO, it is not wise to choose a high node density, if I am not very sure that I actually add information to the database and not only noise.

Particularity this holds true for data imports. Even if the data source claims a certain accuracy, we can not be sure that this holds true. If you import any data, please be aware of this and choose your node density very conservatively (unless you have verified it yourself). We can still improve the exact location later on by hand. I know this takes longer, but Rome wasn’t built within a day either. We wan't to build the best map of the world and not the largest database.

In the recent days I manually removed a huge number of nodes from CanVec imports. I know that I also removed valid information during this process. But the benefits (a node density that reflects uncertainty) are larger than the costs (loss of a small amount of valid data). Instead of arguing with each importer individually (which I have done over and over), I thought that I should raise awareness here.

Thx for understanding.

Comment from robert on 24 February 2013 at 18:24

Interesting point.

Comment from jutezak on 24 February 2013 at 21:34

Really really straight roads should not have too many points, I'd think. Curves need them, even if unsure on how they run.

Thus, trustworthiness is better gained from other aspects, right?

But I agree that when building a road from a noisy track it is better to err on the 'straight' side. Otherwise precision would be suggested that doesn't exist.

Login to leave a comment