[Talk-us] TIGER corrections

Alan Mintz Alan_Mintz+OSM at Earthlink.Net
Thu Jul 9 01:08:07 BST 2009


I've been doing some work in my local area (San Bernardino County, CA), but 
the amount of work necessary to correct the positioning of roads generated 
from TIGER that are badly aligned is starting to become a drag in the more 
dense areas.

So, I started looking at the TIGER 2008 dataset and found that the 
alignment is almost perfect in most cases. So, I did some work on importing 
both TIGER and OSM files into a SQL database (MSSQLServer) so I could 
manipulate them and hopefully be able to spit out a position-corrected OSM 
file, taking care not to disturb features that were not TIGER-generated, or 
had been reviewed.

However, after playing with it for quite a while, I'm unable to find a way 
to relate the "edges" from the TIGER database to the nodes and ways in OSM 
using the TLIDs, finding that the tiger:tlid generally has too many TLIDs 
in it, including some that are not directly related to the node or way.

Examples:

1. Short road "N Laurel Ave" (way 7497408) ending at a T-intersection into 
"W 18th St" (way 7516644). TLID is 144674357. The node at the "T" has TLIDs 
144674357:144674376:144986081, which is fine - the two segments of "W 18th 
St" that join the T are 376 and 081. However, the node at the south (dead) 
end of the road has the same TLIDs (144674357:144674376:144986081) instead 
of just 357.


2. "N Palm Ave" (way 7490579) and "Clark St" (way 7397888) end at an "L" 
intersection with each other. The intersection node has just two TLIDs: 
144674353:145219223. Through elimination, the segment of "Clark St" leaving 
the intersection is 223 and the segment of "N Palm Ave" us 353. So far, so 
good.

Moving west along "Clark St" to the "T" intersection with "N Redding Way" 
(way 7425487, TLID 144674354), the intersection (node 54809056) has TLIDs 
144674351:144674353:144674354, allowing the determination that the next 
segment of "Clark St" to the west has TLID 144674351. Still all is well.

Moving further west along "Clark St" to the "T" intersection with "N Tulare 
Way" (way 7454089, TLID 144674352), the intersection (node 54809062) has 
TLIDs 144674350:144674351:144674352:144669579:144669581, which is two more 
TLIDs than it should have.

The next "T" intersection west (node 54809066) has the same TLIDs, too.


Should I give up on trying to relate things, and simply wipe out 
un-reviewed TIGER imports and re-import the newer TIGER data?





More information about the Talk-us mailing list