Creating an accurate POI file can be tedious

 

First, let me say that the most accurate way to create a POI file is to be there with a GPS and record the L/L of the location, then transfer the data to the file.

When I made the Williamsburg POI file, I tried a couple of ways. I used Google Earth and I also tried the U.S Geophysical Survey digital maps. Both were disappointing. Field checking the results showed that the accuracy was just not good enough for doors that were sometimes just a few feet apart. Not only that, but the doors are usually not visible in a satelitte photo or on a map. My Nuvi will give me 18 foot accuracy or better outdoors, which is more reliable than Google Earth, in my opinion.

Of course, to do the BoA ATM file, I was not going to tour the US to visit them all and record each one's location (can you imagine all the time and money to do that?). I have to rely on Google Earth and accept the coordinates as good enough to be able to drive to the ATM.

First battle: The BoA web site.
1. There are mis-spellings and incorrect zipcodes (confuses Google maps and Google Earth). Can put you miles away from actual location when entered.
2. There are streets that are listed on the BoA site, but Google doesn't recognize them. Most of these are on college campuses.
3. Most college ATM listings just give the main business address, not the actual building name. There could be several ATM's on the campus, but BoA lists only one address (sometimes, a college website lists all the locations - but usually not and BoA does have some buildings listed when you click on "Visit website").
3. Town names are mis-spelled. In Massachusetts, there is a town named Marlborough. BoA spelled it Marlboro - like the cigarette - in 2 of 3 listings. This too, did confuse Google.
4. BoA's map has the same errors that Google does (you can zoom in and out for more or less detail). Some of these errors are miles off of the actual location. Some have been corrected on the BoA site and not on Google.
5. Interactive map for Google Maps and BoA site requires you to clear the browser cache periodically, else browser will crash or hang.
6. BoA Pop-ups can drive you nuts, if you don't have a "Blocker".

Second battle: Google Maps
1. Address algorithms used for Google maps can be terrible. Seems that the higher the street address, the greater the the chance for error.
2. Make sure you go where you type.
At the BoA website is this listing: "82 Worcester Rd, Grafton, MA 01604" and it contains 3 mistakes:
a. Its "St" not "Rd"
b. It's "North Grafton" not "Grafton".
b. The zipcode is wrong. Should be "01536".
This took me to 82 Grafton Rd in Worcester and I wasted 20 minutes in street view looking for an ATM that wasn't there. Fortunately, the BoA site said "located at Koopman Lumber" and by going to their web site I got the address sorted out.
3. Image quality in Google "Street" mode is usually poor, at least it has been so far in New England. Seems that the lens was never cleaned after a rain. My area is much better and was done recently.

So, this project will take awhile, but the POI file should take you right the the ATM location when done.

So far, ME, NH and MA are complete, although I'm sure I have missed some.

--
Metricman Nuvi 660, GTM-20 Traffic Receiver Nuvi 3597 GTM-60 Traffic Receiver Williamsburg, VA

Locations with Tellers

In building the BofA file, would it be possible to put in some comment that indicates that it is an ATM only?

What I often want to do when on vacation is to find the nearest branch - not the nearest ATM. If there were a way to pull out of the file all of the ATM's only, then I would be left with branches.

ATM only

I tried to put the word "Remote" in for ATM only, but didn't get them all.

If I get an actual list from BoA - I will be sure to correct any that are missing and I will be more diligent for the states to come.

Will start putting "Branch" or "Remote" in for the next state which will be RI.

--
Metricman Nuvi 660, GTM-20 Traffic Receiver Nuvi 3597 GTM-60 Traffic Receiver Williamsburg, VA

Agree It Is Tedious

I agree with you. Even with its limitations, especially with street addresses, Google Earth is the best way available to geocode. It beats having to go to every location.

My policy is to start small and make sure locations are as accurate as necessary. I feel this is better than offering many locations with doubtful accuracy.

dobs108 smile Nassau County NY

I also agree starting small

dobs108 wrote:

I agree with you. Even with its limitations, especially with street addresses, Google Earth is the best way available to geocode. It beats having to go to every location.

My policy is to start small and make sure locations are as accurate as necessary. I feel this is better than offering many locations with doubtful accuracy.

dobs108 smile Nassau County NY

After numerous restarts on a file I found starting small and checking how it looks in a GPS is the best way. You also get to see what info is there or still needed as a "nice to have".

--
Nuvi 2460LMT.

re POI file creation

Metricman Quote "First, let me say that the most accurate way to create a POI file is to be there with a GPS and record the L/L of the location, then transfer the data to the file."

It's even more fun when locals have their name for a road (ie Danforth Rd), Google map, sat, aerial shows another (Route 30), web site gives another (Main St) and Google street view is shows County Rd 30 on the signs. Some places in my files are so remote there is no aerial, sat is useless and the map view is all you have and you may or not have street view. Or the mailing and delivery address is a PO Box and hours of web searching won't give you a physical address and you can't phone the place because the number isn't listed or is wrong. But then I guess you just hope that somebody who downloaded the file lives near the place and can send you the correct info.

--
All the worlds indeed a stage and we are merely players. Rush

metricman wrote: Of course,

metricman wrote:

Of course, to do the BoA ATM file, I was not going to tour the US to visit them all and record each one's location (can you imagine all the time and money to do that?). I have to rely on Google Earth and accept the coordinates as good enough to be able to drive to the ATM.

I know how tedious it can be. Try finding a boat 20 feet long in Google Earth.

--
"Anyone who is capable of getting themselves made President should on no account be allowed to do the job." --Douglas Adams

Another way

Another way is to use the GPS-Equalizer (name?) for filling in L/L blanks, but that has the same errors as Google and the rest.

I'll just keep plugging away at it. I hope that no other state is as concentrated as MA was.

I did find 2 50 ft long planes in the Atlantic, off of Pea Island NC.

--
Metricman Nuvi 660, GTM-20 Traffic Receiver Nuvi 3597 GTM-60 Traffic Receiver Williamsburg, VA

A thought

This is just a thought, as I sit here putzing with the various Googles, EPE, and MapSource myself...

Out of all of the variables in getting to an accurate location with a Garmin, at least MapSource can show you the map that's on your unit, and the street numbers according to it.

Not that it won't have errors as well. I'm just throwing it out here.

--
It's about the Line- If a line can be drawn between the powers granted and the rights retained, it would seem to be the same thing, whether the latter be secured by declaring that they shall not be abridged, or that the former shall not be extended.

Google Earth accuracy, et al

Metricman wrote:

When I made the Williamsburg POI file, I tried a couple of ways. I used Google Earth and I also tried the U.S Geophysical Survey digital maps. Both were disappointing. Field checking the results showed that the accuracy was just not good enough for doors that were sometimes just a few feet apart. Not only that, but the doors are usually not visible in a satelitte photo or on a map. My Nuvi will give me 18 foot accuracy or better outdoors, which is more reliable than Google Earth, in my opinion.

I would beg to differ with a couple of Metricman's observations. First of all, Google Earth is extremely accurate with its latitude and longitude, all the way down to six decimal places. Jeez, how many more decimal places do you need before GE is "good enough" for you? Now, having said that, I also concede that the skill in getting good lat/long numbers often takes a fair amount of skill. Many of the sat pictures are extremely fuzzy, and therefore practically useless. While it would be nice for businesses to provide their customers the lat/long of all of their locations, that's not how it is yet. So it becomes the responsibility of the person wanting that information to track down the locations and put GE's crosshairs where they belong.

The second point in Metricman's post that I don't agree with is his comment about some doors being mere feet apart and therefore not distinguishable by GE. I would argue that that degree of accuracy is unnecessary. I've got news for you: if I'm in my car and my GPSr gets me to a door a few feet away from the door I really need, that's close enough. I'll find what I'm looking for.

I generally agree about his other points, especially about street and highway names being inconsistant. One of the problems that I have is that in rural areas, all maps show the street name as being the numeric route number, while the internals of the GPSr use the local street names, which are only known by the locals living in the immediate area.

--
Phil in Mentor, Ohio -- Garmin Nuvi 1450

Addresses not Accurate

Whether you are using Google Maps or Earth or the GPSr itself, addresses are not always accurate because the map does not have the coordinates of each address. I assume the coordinates of intersections are accurate and the position of an address is estimated by comparing it with the other addresses between two intersections.

This works when intersections are close together but fails when intersections are far apart. Many business websites have a mini Google Maps or Mapquest map to show where they are located, and these aren't always accurate.

I predict that Enhanced 911 systems will someday have the coordinates of every building in their area. When used with the real-time coordinates of emergency vehicles, computerized dispatching would be possible. The coordinates of addresses would be made available to businesses (for a fee) and Google Maps and Earth would then be able to obtain this info.

dobs108 smile

Accuracy

plunder wrote:

I would beg to differ with a couple of Metricman's observations. First of all, Google Earth is extremely accurate with its latitude and longitude, all the way down to six decimal places. Jeez, how many more decimal places do you need before GE is "good enough" for you? Now, having said that, I also concede that the skill in getting good lat/long numbers often takes a fair amount of skill. Many of the sat pictures are extremely fuzzy, and therefore practically useless. While it would be nice for businesses to provide their customers the lat/long of all of their locations, that's not how it is yet. So it becomes the responsibility of the person wanting that information to track down the locations and put GE's crosshairs where they belong.

The second point in Metricman's post that I don't agree with is his comment about some doors being mere feet apart and therefore not distinguishable by GE. I would argue that that degree of accuracy is unnecessary. I've got news for you: if I'm in my car and my GPSr gets me to a door a few feet away from the door I really need, that's close enough. I'll find what I'm looking for.

Plunder,

There is some error involved when a "Flat" image of a curved surface is projected back onto a curved surface. Also, there is an error if the satellite is shooting at an angle other than absolute "0 Degrees" down. Go to any large city and note that the buildings are usually "Tilted". You can usually see one or more sides of the building in the image. I've seen this "tilt" on one story buildings. The further you go from directly overhead (0 degrees), then the greater the angle of tilt will be (plus - add in the curve of the earth to add to the tilt). and then you have to add in the altitude of the location, which adds to the error. These are "Commercial" images, not Military, and therfore not as accurate or as high a resolution.

Would you trust Google positioning to put a Cruise Missle into a 6 ft X 6 ft hole in the side of a 12,000 ft high mountain? I wouldn't! But we aren't doing that here. Just because Google Earth displays to 6 decimal places, doesn't mean that it is always that accurate.

Take your GPS to any landmark in your area - like the end of your driveway (please be outdoors and be sure the landmark is visible in Google E). Record the L/L from the GPS and then check it against Google E. My Garmin read (-)XX.72086, XX.36337 (claiming 18 ft accuracy) and Google E says -XX.72077, XX.720774. Entering the Garmin coordinates into Google E puts the indicator about 25 ft away. You can see some of the West walls of the houses in my neighborhood (but not much of the wall), so the satelitte was not directly overhead when the image was made. This observation throws the 6 decimal places accuracy right out the door.

Anybody else out there try this? How much error did you get?

If you go to:

http://www.poi-factory.com/node/21554

and if you read what I wrote, you will see that you have to "walk" in the restored area. You can't drive during the day and the POI is intended for those that have handheld GPS devices, and yes some doors to different buildings are mere feet apart. I made each location distinct so that users could see the proximity of each location when "zoomed in".

Not trying to start a big debate or argument, but check out the accuracy of Google E yourself before making any reply.

FYI, I did visit one University where they actually listed the L/L of some locations. I guess that the more GPS equipped IPOD's, etc. become popular - the more we will see the L/L being given.

--
Metricman Nuvi 660, GTM-20 Traffic Receiver Nuvi 3597 GTM-60 Traffic Receiver Williamsburg, VA

Addresses

dobs108 wrote:

I predict that Enhanced 911 systems will someday have the coordinates of every building in their area. When used with the real-time coordinates of emergency vehicles, computerized dispatching would be possible. The coordinates of addresses would be made available to businesses (for a fee) and Google Maps and Earth would then be able to obtain this info.

Where I live, the 2 counties and the city have maps already in Digital form. These maps can supply Street Address, owner(s) name, history of owners, how much assesed for by county/city (tax assessment) and how much property has been sold for, for each sale.

They also can display a map of the parcels and you click on the parcel to get the information. Visit your local government Commissioner of Revenue "online" to see if they have a map. I'm curious as to how many localities have done this.

All they need is the money to do it, if they already haven't done it.

--
Metricman Nuvi 660, GTM-20 Traffic Receiver Nuvi 3597 GTM-60 Traffic Receiver Williamsburg, VA

Accuracy vs Alignment

I have a feel for what Metricman is saying, and it's not as much accuracy as it is 'alignment' and completeness in some cases. Although in mapping all are the same I suppose, inaccurate.

For example, I have mapped my yard boundaries in GE and it's very accurate in linear measure and coordinate 'granularity'. However the coords don't exactly convert & match to my state's state plane GIS data, and we all know my GPS coords aren't going to match exactly either.. and neither my Magellan or Garmin do. I could use them pretty well for driving, but not so accurately walking.

And then the other data attached to the coords (like address number, road width & placement, etc) in the different data sources differs.

One of the reasons I suggested MapSource as an alternative is that I'm working on a poi 'test' file with a route attached. There is a difference in the 'roads' on Google Maps & Google Earth, and my Garmin map. Both in exact position as well as in existence. I found a great spot in GE to place my test pois and driving route only to find that the roads (not exactly public..Watkins Glen Motorway!) were not on the Garmin map set.

BTW, Maryland does the land ownership maps statewide & there is a LOT of various state GIS data & maps available.

--
It's about the Line- If a line can be drawn between the powers granted and the rights retained, it would seem to be the same thing, whether the latter be secured by declaring that they shall not be abridged, or that the former shall not be extended.

I wouldnt go so far as

I wouldnt go so far as staying such project is very difficult

I use geepeex and use the geocode function.
It gets me roughly to where I wanna be..then I just fine tune it using a hybrid map.

Granted, doing this on a block of buildings your not familiar with can be difficult...
But for something like drag strips that I added to a master file, it was pretty easy.
Generally its the long 1/4 strip. grin
Plop the location at the entrance and your done. grin

--
Nuvi 350 Born Oct 07 - Nuvi 660 Unit #2 (re)Born Sept 08 - Nuvi 360(Gift to 'the chick' yet maintained by myself) Born July 08

Easy to see

dood, dragstrips show up on Google Earth like they were airports!

dobs108 razz