Drunkard's walk

4.3k · February 25, 2009

This may be a silly question, but here goes...

Pick any section of road you know that runs in mostly a straight line for a half mile or so. Then look at the road in DeLorme's Topo 7 -- or Garmin's City Navigator -- or Lowrance's MapCreate -- or any other GPS map software you have.

Is the line representing the road straight? In most places I look, in different vendors' packages, the lines usual have slight zigzags between intersections. As if someone decided to measure the road's bearing by walking its length, but kept crossing from one side of the street to the other every block or so.

It's especially noticeably in DeLorme's Topo 7 when you can overlay aerials in hybrid map view. But I've noticed this for years in many different vendor's packages, so there must be a reason for it.

Why?

18.9k · February 25, 2009

Computers and coordinate systems have a tough time with curves. Don't really understand them very well at all. So they simulate them by joining together a series of line segments. If you had enough very small points and line segments it would be much better but then the "storage" size of the road would be much higher. So to save space the lines that represent the road are broken into as few segments as possible to still give the general layout of the road. Put very simply, they gathered as few datapoints as possible to still keep you on the road or had equipment that only gathered a limited number of datapoints.

4.3k · February 25, 2009

Put very simply, they gathered as few datapoints as possible to still keep you on the road or had equipment that only gathered a limited number of datapoints.

Okay, then why do the road layers in Google Earth and Google maps line up so beautifully with their aerial/satellite imagery?

6.4k · February 25, 2009

I really have no clue what you are on about. Roads are straight in CN when they are straight.

18.9k · February 25, 2009

Put very simply, they gathered as few datapoints as possible to still keep you on the road or had equipment that only gathered a limited number of datapoints.
Okay, then why do the road layers in Google Earth and Google maps line up so beautifully with their aerial/satellite imagery?

Good programming and error review.

Just to prove it - I know a guy here locally that contacted Google about a rather misplaced rural road that the images didn't stitch together well and the overlaid lines also had sharp zigzags. A few weeks later the line and road image looked perfect end to end.

4.3k · February 25, 2009

Good programming and error review.

Good answer! But what does say about brands D, G, L, M and others? :lol:

8.9k · February 25, 2009

Good programming and error review.
Good answer! But what does say about brands D, G, L, M and others?

To me, it says we're not sitting at home in front of our computers, we're in the field with our GPS units!

:lol:

9.5k · February 26, 2009

Good programming and error review.
Good answer! But what does say about brands D, G, L, M and others?

Taking the cheap and easy route?

February 26, 2009

I really have no clue what you are on about. Roads are straight in CN when they are straight.

Yeah, I have to second this (if the poster is referring to City Navigator). I think it's actually spot on, from what I can see after a month's daily use over a metro area of maybe 70 square miles.

February 26, 2009

I really have no clue what you are on about. Roads are straight in CN when they are straight.

I remember back when I was using Metroguide/city Select V6 and noticing all the roads would have angles between blocks where they should just run straight. If it was a perfectly East/West road & you were traveling West, one block it would point slightly SW, the next block it would point slightly NW, but never straight. Some roads were also off in position by 10-30ft, which you would notice when zoomed in.

I upgraded to City Select 7 and all these strange angles were gone & every Mapset since then have displayed roads straight like they should(also VERY close if not spot on to actual position).

Edited February 26, 2009 by hogrod

4.3k · February 26, 2009

To me, it says we're not sitting at home in front of our computers, we're in the field with our GPS units!

Try to stay on topic, eh? Look at a few areas in any map programs you likee and see if the streets line up with aerial imagery? Post some coords for locations you check so others can compare in theirs. Thanks. Edited February 26, 2009 by lee_rimar

February 26, 2009

I think this maybe how the data for your area was collected. I noticed the change for V6 to V7 of city select, I believe this is when Garmin first started using Navteq data for maps. I was really sad seeing google switch to Tele atlas data, in my area Navteq is much more accurate. At least for now garmin is sticking with Navteq.

4.3k · February 26, 2009

Red90 and Hogrod, you raise good points. While I've seen the zigzags in products from various vendors, it isn't universal. It may vary depending on who the GPS vendor license their map data from. Later versions of software may be better than older ones. For Garmin i(the only one with native viewers on Mac and PC), there may also be rendering differences between the PC's MapSource and the Mac's RoadTrip viewers.

Indulge me and look at a specific spot...

Here's one near where I grew up: 42.457149, -83.187432

Granzon Street in Oak Park Michigan runs east-west between Church and Coolidge.

Looking at it in RoadTrip on the Mac with CN 2009 data, it does look a bit ziggy.

Let me know how it looks to you, and what map software you're looking at.

Edited February 26, 2009 by lee_rimar

6.4k · February 26, 2009

Yes, all the maps show that street as not straight. It means the base data (i.e. Government data) is that way. Everyone gets the data from the same source.

4.3k · February 26, 2009

...all the maps show that street as not straight...

Okay, thanks. Now you, BlueDamsel, and others can see what I'm talking about. To me it seems mostly visible on roads that don't run PRECISELY north-south or east-west, and have lots of intersections over a relatively short distance.

But I'm not sure EVERYONE gets all of their data from ONE source? There is certainly some overlap, but the GPS vendors license map data from a lot of sources. I don't think Navteq, Telenav, DeLorme, and others are all pulling street maps from the Tiger Census database.

February 26, 2009

I really think that StarBrand is on the right track here. The road data is converted to vector form. On anything but a ruler straight segment, the vector shapes are only going to be an approximation. Even on a straight segment, if the vectors are broken at intersections, you have the possibility of some variation. The more you try to minimize the size of the data set, the worse the approximations get. Different amounts of space optimization in different formats might account for some of the observed variations, even if all the data is based on the same data set. For example, Google -- with terrabytes of storage -- might not compress their vector data as much as Garmin or Delorme -- who are dealing with mere GB of storage on the handheld.

In addition, you have plain vanilla display issues. For vectors which do not run in the same direction as the pixel matrix on the display, you will also get artifacts that are the result of the rendering process. These will always manifest as "jaggies" (another technical term ) -- diagonal stair steps in the rendered line. If your unit is set to true north and north up, you will see these artifacts on any vector line (even if the vector itself is perfectly straight) that does not run north/south or east/west. Depending on the pixel aspect ratio, there is probably one intermediate angle in each quadrant which will also appear straight. For square pixels, this would be 45 degrees -- NE, SE, SW, NW. But that's as good as it gets. This is not a flaw in the underlying vector data -- it's a characteristic of digital displays. You have to have pretty good eyes to notice this on the GPSr display -- eliminates old geezers like me

Edited in the hope of getting last post date updated.

Edited February 26, 2009 by twolpert

2.1k · February 26, 2009

It's especially noticeably in DeLorme's Topo 7 when you can overlay aerials in hybrid map view. But I've noticed this for years in many different vendor's packages, so there must be a reason for it.

Why?

The residential streets in my neighborhood are all N-S and E-W. Formerly, it was Topo 6 that had them rotated slightly and some with shallow "V"s in them at intersections.

However, now with Topo 7, they are in much better N-S, E-W orthogonal alignment. Possibly, the anomalies that you now see in Topo 7 will be corrected in the forthcoming Topo 8.

DeLorme does have a process by which users may submit corrections.

4.3k · February 26, 2009

The residential streets in my neighborhood are all N-S and E-W ... Possibly, the anomalies that you now see in Topo 7 will be corrected in the forthcoming Topo 8.

They might be. Did you look at the specific point I indicated in the earlier note? Granzon doesn't run perfectly E-W.

DeLorme does have a process by which users may submit corrections.

Hm, yes, but...

If it's a systemic problem with the data and/or rendering, that's gonna take a LOT of user submitted corrections How many? Any road anywhere that doesn't run perfectly N-S or E-W, and has lots of intersections?

5.3k · February 27, 2009

I'm not sure what you're seeing on other programs, but if I look at the imagery in Google Earth and/or Maps and compare the imagery (both aerial/satellite and Street View) with the roads, I think the roads actually are a little squiggly. The intersections across W 9 mile and and Kenwood for roads like Gardiner, Church, and Susses actually appear to have a little "jog" in them as they cross the big roads.

Oh, and in the U.S. almost every starts from the Tiger database, but the commmercial companies like Navteq and TeleAtlas all routinely augment that.

4.3k · February 27, 2009

... if I look at the imagery in Google Earth ... intersections across W 9 mile and and Kenwood for roads like Gardner, Church, and Sussex actually appear to have a little "jog" in them as they cross the big roads.

The ones you mention do shift when crossing 9 Mile. But the stretch of Granzon I mentioned is ruler-straight (which is why I used it as an example): You can see traffic on Coolidge even if you're half a mile away at Church Street. At least I could when I was a sharp-eyed young lad, some 40 years ago.

Nothing special about that street or any specific map program either. I've seen similar zigzag in lots of places in different vendors' packages for years. Which makes sense if (as you said)...

... in the U.S. almost every(one) starts from the Tiger database, but ... Navteq and TeleAtlas all routinely augment that.

I really didn't know that. I always thought of Tiger as the poor choice and that the other vendors you mentioned built their own databases from other sources. Edited February 27, 2009 by lee_rimar

5.3k · February 27, 2009

Granzon looks as straight as can be on Earth and Maps and the imagery sets to me.

The U.S. is a big place. You can get map data with rights to redistribute it for free, but it's not that great. Map companies follow the money. If you had a choice of driving/resurveying the Vegas strip or or a cornfield in Kansas, which is most likely to rely on Tiger data?

It's all about the value-add.

2.1k · February 27, 2009

Granzon looks as straight as can be on Earth and Maps and the imagery sets to me.

The U.S. is a big place. You can get map data with rights to redistribute it for free, but it's not that great. Map companies follow the money. If you had a choice of driving/resurveying the Vegas strip or or a cornfield in Kansas, which is most likely to rely on Tiger data?

It's all about the value-add.

Forty years ago, the strip, but nowadays.......?

2.1k · February 27, 2009

The residential streets in my neighborhood are all N-S and E-W ... Possibly, the anomalies that you now see in Topo 7 will be corrected in the forthcoming Topo 8.
They might be. Did you look at the specific point I indicated in the earlier note? Granzon doesn't run perfectly E-W.
DeLorme does have a process by which users may submit corrections.
Hm, yes, but...

If it's a systemic problem with the data and/or rendering, that's gonna take a LOT of user submitted corrections How many? Any road anywhere that doesn't run perfectly N-S or E-W, and has lots of intersections?

Actually, I did see a thing on TV about a mapper in a car with all this antenna stuff (antennas are for RF, bugs have antennae) and it is a full time job just to drive and record for the map companies.

4.3k · February 27, 2009

Granzon looks as straight as can be on Earth and Maps and the imagery sets to me.

By "the imagery sets" do you mean just those in the Google products, or other vendors' mapping programs as well? The ones I have currently on my computer (Garmin CN, Ibcyus USA, DeLorme T7) show little zigzags. USGS 1:24K topos show it as straight. I don't have any Magellan products loaded at the moment, but I know I've seen the same kind of effect there also. Edited February 27, 2009 by lee_rimar

4.3k · February 27, 2009

I think I'd like to define this question a little more precisely. Originally it was:

Q: Why do map programs from so many vendors show this zigzag effect?

A: Because they're all using bad data from one original source.

Okay, that answers "why so many vendors show the same thing." But what I really want to know is...

WHY is the data (or rendering method) wrong, in a way that manifests itself as small alternating changes in bearing (typically at intersections) in roads that are actually straight? Not just in one area, but throughout the entire data set?

8.9k · February 27, 2009

I think I'd like to define this question a little more precisely. Originally it was:

Q: Why do map programs from so many vendors show this zigzag effect?

A: Because they're all using bad data from one original source.

Okay, that answers "why so many vendors show the same thing." But what I really want to know is...

WHY is the data (or rendering method) wrong, in a way that manifests itself as small alternating changes in bearing (typically at intersections) in roads that are actually straight? Not just in one area, but throughout the entire data set?

I'd add my thought, but got my hand slapped for doing so last time.

4.3k · February 27, 2009

I'd add my thought, but got my hand slapped for doing so last time.

If you have an on-topic thought, I'd love to hear it.

On topic, have you noticed the zigzag in any areas you look at in DeLorme T7? It shows up especially whell when you turn on hybrid view to overlay T7's street layer with aerial or topo imagery.

8.9k · February 27, 2009

Maybe showing an example would be helpful since I don't see the problem. Like I had said before, maybe it's your GPS (well, that was basically what I was hinting at at least). You are talking about on the GPS, right?

When I look at my maps, I see fluid straight roads and what appears as matching roads in the hybrid mode as well.

I hope this is on-topic as I'd rather not be rude like some can be when asked to stay on topic!

4.3k · February 27, 2009

... an example would be helpful since I don't see the problem...You are talking about on the GPS, right?...

No, I'm talking about map software on the computer. See post #1 for the original problem description and post #13 for a specific example:

... one near where I grew up: 42.457149, -83.187432

Granzon Street in Oak Park Michigan runs east-west between Church and Coolidge.

Edited February 27, 2009 by lee_rimar

8.9k · February 27, 2009

... an example would be helpful since I don't see the problem...You are talking about on the GPS, right?...
No, I'm talking about map software on the computer. See post #1 for the original problem description and post #13 for a specific example:

... one near where I grew up: 42.457149, -83.187432

Granzon Street in Oak Park Michigan runs east-west between Church and Coolidge.

I'll check it out and get back to you on this, but my experience thus far is that the orads match up pretty good! Of course, I haven't bothered to check it too closely as I never thought the maps were off.

eta: I did read those posts, not a word said about on the computer or GPS. Since I have the TOPO7 product on my GPS, I just assumed you meant GPS, sorry.

Edited February 27, 2009 by Rockin Roddy

February 28, 2009

Google Maps

MapCreate7

The road does not run directly East / West but is at a slight angle, so it dips up a bit.

If you look closely the Google Maps line is not perfectly straight. But I think they use a higher resolution line so it looks cleaner than the low res line used in MapCreate7.

4.3k · March 1, 2009

Duh... I *knew* it was a silly question.

I figured it out myself. TWolpert came pretty close to explaining it correctly, I can show and tell in very simple terms, PRECISELY how it happens, and why it happens in so many different vendor's products (the data doesn't have to come from one source, everyone will have this problem).

Draw some vertical lines on a sheet of graph paper. Then draw a straight and mostly (but not perfectly) horizontal line across the sheet. Then draw an X through each square where your horizontal line intersects with the vertical ones you drew.

If you draw another set of lines through the centers of those X's, it won't follow the original path perfectly. You get the exact kind of zigzag line you see in map programs, for the same reason.

That's an over simplification, but it's how street data gets collected in the first place: Recording intersection locations with a finite amount of precision. Creating a map from those recorded points, each intersection location may have a position error of up to 1/2 the unit size of the measuring grid. In simpler words -- a straight line gets drawn in zig-zag segments.

The only way to avoid it would be what Google apparently does: Clean the data up by comparing against aerial imagery.

Edited March 1, 2009 by lee_rimar

March 3, 2009

Do the jagged lines have more to do with pixils than data input? R pixils square or round? do they set side by side and on top of each other? Early morning 3:16AM thoughts. Terry <_<

Drunkard's walk

Recommended Posts

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Link to comment

Join the conversation