I’ve seen several discussions regarding coordinate acquisition and the inaccuracies of some of the coordinates. I personally have been directly responsible for geolocating millions of addresses in my career. Hopefully this discussion will be helpful to some…..sorry for the length.
My geolocating (obtaining the geographic location) experience is based on the use of “street centerline files.” The streets that you see in Google Map or MapQuest, for example, are based on a centerline file. Ie, the streets are represented by a series of lines segments. Associated with each line is a beginning and ending point, along with other points that define the shape. Associated with each point is a distinct lat/long value (depending on the coordinate system used). Also associated with each line is a range of addresses; either potential address ranges, or actual address ranges. As you know in nearly all cases in the US, the odd addresses are on one side of the street; evens on the other. If you go one direction the address numbers increase; they decrease going the other direction.
Looking at the sample street segment below, the even potential addresses are on the north side with odds on the side. This is called address parity.
Even though the highest numbered “actual” address in the 100 range might be 148, the full potential is assigned to the street. If you enter 148 Main St into Google Map, it approximates it’s location in the middle of the street (between 100 and 198) rather than at the end of the street where is might actually exist. The US Postal Service uses potential ranges to assign the Zip+4 numbers. Potential address ranges will incorporate the potential newly constructed houses without having to update address ranges although the downside is a less than accurate location coordinates.
If you search for an address that doesn’t have a corresponding street segment with address ranges, the system may return the coordinates for the middle of the zipcode that you entered. Hence, you may enter a new address and the site shows a point in the middle of a vacant field.
Satellite images are simply photographs and do not contain information about street names or address ranges. When you input an address into Google maps, the software obtains a coordinate from the street centerline file (based on the concept shown above) and zooms into the satellite image for that location.
Here are a couple other issues impacting the accuracy of the lat/long coordinates.
1) What source was used to create the lat/long coordinates of the street centerline? Ever see the streets on the map not following the streets in the satellite image? I’ve seen centerline files that were 100s of meters off and some that were submeter. The less accurate the file, the less accurate the lat/long values you receive. By the way, imagery has to be rectified and registered in order for it to be displayed in the correct location.
2) Accuracy of the GPS unit itself can also be a factor although at the scale used by a Navigation system, it’s kind of irrelevant. This is more noticeable if you are Geocaching.
3) Technically speaking, in the centerline file, the address is literally on the street and that is the coordinate most likely returned by the system. In reality, the business maybe be addressed to that street but be physically located 100s of yards away. You can see this in Google Earth where the icons are on the street instead of on the building. The only way for someone to get the coordinate on the structure is to know exactly where it is located.
4) Differences in lat/long values returned by Google Map vs. MapQuest or other sources can be explained by differences in the centerline files or the algorithm used to generate the lat/long.
Anyway, sorry again for the long messages but hopefully this clarifies some of the issues people are encountering.
I for one found it very interesting...
Thanks. My text version of the street/address range didn't display correctly so I had to link the image separately.
As someone who worked with nodes, block faces, et al back in the dark ages (pre graphical computer)
It brought back fond memories of assigning nodes, etc. obtained from thousands of pages of line printer output from the good old mainframe, filling out sheets for key punch operators.....
I thought you kept it quite short considering the complexity of the subject
I found it to be interesting and for the subject - was not that long...
Thanks for taking the time to explain.
Just another reason to keep coming back to POI Factory. The wealth of knowledge shared here is beyond compare. Thank you for the technical insight.
When I create a POI and get to the part of fine tuning its coordinates, where is the best place to put it?
I try to place it 20-100 feet from the road to try to force "Dorothy" or "Dort" for short to say "arriving at destination on left" or "arriving at destination on right". I've heard her say the wrong side of the street a couple of times. Granted, this is not an exact science, but are there any suggestions for the BEST place to select for your coordinates (i.e the building, parking lot, driveway)?
I found that most instructive. Thank you.
Very intuitive ind intriguing. Thanks so much, And BTW welcome to our community please feel free to input when you see it is needed.
Yes, welcome to the group, it is always nice to have new people like yourself join. I can tell that you are going to be a star in no time.
When I create a POI and get to the part of fine tuning its coordinates, where is the best place to put it?
I try to put it basically at the point where you leave the last mapped road. Usually that's a driveway off the main road. That way, you actually cross the point when you're arriving, and the GPSr isn't just giving you a long straight line off the road directly to the point (where you may not be able to drive, anyway). And, as you suggest, this makes it clear which side of the road you want.
If the POI isn't clearly visible from that point, it may make sense to move the point closer, perhaps the actual entrance to a building.
In the case of specific objects instead of businesses, it may more sense to put the point exactly on whatever the interesting object is, or the nearest parking area, again depending on how obvious it will be once someone arrives.
Thanks now something new to research.
Very helpful information! I have been wondering how many streets coordinates do not overlay on the center of the satelite street. This happens to be the case for the street in front of my home. I'm beginning to think these occurences are widespread.
Thanks for the great description of address ranges.
Another problem related to these ranges is when the POI isn't physically at the provided address.
For example, a local shopping center has an IHOP who's address is on the main street. If you go to the geo-located point for the IHOP, you can't physically see it because it is on the back side of the shopping area. The physical location is more than 1000m away.
Another example is when a business owner uses her/his home address for the business.
The business in either case is listed in the phone book, or D&B using a mailing address.
I have found this the case with Parks too. The park administrators address is used instead of the "real" address.
Perhaps you could phrase what I am tying to say with similar succinctness.
Thanks for your post, and welcome.
Granted, this is not an exact science, but are there any suggestions for the BEST place to select for your coordinates (i.e the building, parking lot, driveway)?
If the building is very close to the road, then you can place the point on the building. However, for large malls, you should always place the point at the closest driveway. Otherwise you may get incorrect driving directions.
For example, lets say you have a huge square mall. The entire north half of the "square" is the parking lot, and the entire south half of the square is the stores. The only entrance to the mall is on the north side, since the south side is blocked by stores.
If you place your point on the south half of the square, the GPS may direct you to a back street which does not actually connect to the mall.
Thanks for the education. For the reasons you describe, when I give coordinates for areas that I know I will punch in the address and move the marker or push pin to the building location and give these.
thanks for the info. very informative!
not it make some sense
Thanks Duane and welcome!
Click, and the light bulb goes on. Your detailed and concise explanation really helps explain some of the mysterious routing we've all experienced. Thanks, Duane, for takine the time and effort. And welcome to POI-Factory!
this is very good information that everyone using google maps and such should know.
Thanks and welcome to POI-Factory. That does clear up some issues I was trying to research.
Welcome! The message was very informative, not long at all for what it contained, and very much appreciated. Then it leads me to thinking and one thought leads to another...
I work in a tech field, and one in a while I'll actually look through the piles of Design News and other free trade rags that clog my mailbox, and in one was an article about automotive grade hard drives and it said that they were being adopted earlier in Japan than the U.S. because of the very large databases needed for the vehicle navigation systems. Whoa! I thought, that's approaching 40GB (the rest is for MP3s ), when a Nuvi has only ~1GB worth of information that covers the whole of North America. Then, I found out that addresses in Tokyo can be non-sequential down the street, because the buildings are numbered in the order in which they were built, since ancient times I suppose. So, does that mean that each and every Tokyo address would have to be geolocated individually? No interpolating possible? That would explain the large map database sizes. Then, I looked at actual Tokyo addresses; they have a number, like 3-32-5 in it, which turns out to mean the address is in a large block numbered 3 and a city block numbered 32 within that other block. So, maybe it isn't that bad at all.
Your message also explains (for me) why Google often doesn't locate an address very accurately, but lets me login and move the marker for an address I've just looked up, if I know the actual point closest to the entrance of the house or building. They're letting me geolocate the address for them, for free.
Could you reiterate in monosyllabic iterations?
I few months ago sent in an adjustment to the WalMart POI coordinates for a newly constructed store. The new store happened to be just off the interstate. The access roads were adjacent to and infront of the Walmart. The road in front was a one way street. The location chosen for the point was NOT the main front entrance. Instead, it was the side entrance. If the main entrance in front of WalMart were chosen, when coming from the west "Crissy" would pass the first exit opportunity and drive to the second exit. At this exit one would need to backtrack (head west) to arrive at the main WalMart entrance. Using the side entrance to WalMart provided the most direct routing to the store when coming from BOTH directions.
I'm not familar with the Japaneese addressing system but they could list each individual address on each street segment side. That would give some insight to the larger files sizes.
The Japanese street numbers are assigned by when the structure is built, thus #1 could be next to 1904 which could be next to 2133 and 2 could be miles from 1 etc.. So if that is true, it would seem each address would have to be entered separately. This may be why they would have a block system.... Of course this all could just be urban myth ...
Edit think a gps would be of use??
There are also some anomalies in the US. As I recall, there are some areas in WI with hypenated addresses and a few areas in the US with alpha-numeric addresses.
Interesting. I saw a little blurb here about many roads in Japan not named:
I should ask my sister who's an ex-pat living in Japan if her car has a navigation system, and how useful it is. I never did talk to her about how she manages not to get lost.
I grew up in Queens, NY. Most addresses were of this style:
149-03 61st St or 61-10 149th St.
Most houses were easy to locate, just by knowing the address, especially since *most* streets were numbered, not named.
149-03 61st Rd would be the 2nd house from the corner of 149 St. and 61st Rd, on the odd side of 61st St.
61-10 149th St would be the 5th house from the corner of 61st Rd and 149th St, on the even side of 149th St.
Granted, 150-15 Reeves Ave is a little tougher because you have to find Reeves Ave, but at least you know it intersects 150th St.
I'm involved in providing geographical tools for census research to genealogists and academics. Street name changes and street naming conventions are one of my interests. Queens NY is interesting. For a look at the conventions see:
We also have up the street naming conventions for Salt Lake City.
Also at the Street Change utility we have, for some cities, tables that show when a street is renumbered the old and new numbers (for example...for Detroit).
I'm involved in providing geographical tools for census research to genealogists and academics.
Interesting......I work for the Census Bureau....
You just never know who you will meet here who has lost their census....
i helped startup a "summary tape processing center" for census data.
I worked with GBF/Dime Files (which were improved to become Tiger Files).
We used punched cards, then computer tape.
Then, as previous users suggest, I lost my census and went into engineering.
But I still have a fondness for mapping and now GPS, et al.
Thanks for the stroll down memory lane.
I started on the front end of TIGER creation. Worked with a few GBF/Dimes myself as they were the initial centerline data used for the metro areas in TIGER. Spent the last few years working to improve the spatial accuracy. No more of those stick looking roads now in the metro areas.
Is there a script which can change addresses (in text files) to lat/long ?
Thanks Duane for the very informative post.
I live in a block of 7 townhouses all with odd street numbers. The nuvi shows the distance to my address some ways off and on the opposite side of the street. I set my home location to right in front of my garage which is across the private street I live on.
Are there "rules" for street numbers, like being based on what side of the street the garage is on, rather than the house? I frequently go to a location who's address is West Spring Street. They do have an entrance on Spring Street, but the main entrance is at the parking lot between 3rd and 4th Streets. The only way to get to parking is to enter either from 3rd or 4th Streets. I'm wondering how they figure addresses and what some of the "rules" might be>
We don't need no stinking rules!!!!
It's my experience that the addresses are typically based on the location of the front door of the unit. However, the assignment of addresses is typically done at the city or county government level so that leaves lots of wiggle room. I'm surprised that an address would be assigned to a garage on the opposite of the street but who knows, the local addressing authority may have a reason for that?
Most mapping programs asign numbers by one of two means. One is the 911 guide, which has 528 addresses per mile. 1000 is one mile from the beginning. 1500 is 1.5 miles from the beginning. Provided the database provider, Tiger Streets in most cases, gets which end is the start point.
The other scheme is in a lot of cities the start and end numbers for a block, or street are given. If there are 100 numbers on a street, then number 1 will be at one end and 100 at the other and 50 will be in the middle, regardless of the actual location of #50.
As far as the side goes, for both 911 and block numbering it's usually North and East side of the street for even numbers, but sometimes it varies.
Or you could just Google for the correct info.
Lol...Thanks, for the info Duane. I guess there are no rules, except at the local level. A lot of the rural areas of California have 5 digit street numbers, even on a very small street, and I could never figure those out either. Maybe I should live in my garage, and park in the house....
A lot of the rural areas of California have 5 digit street numbers, even on a very small street, and I could never figure those out either.
My guess is that in this case, there is a county wide addressing system in place. I live in an area where one can see the influences of city based addressing, county based addressing, and now the new metro area addressing.
terms | privacy | contactCopyright © 2006-2022