The truth about mapping Philippine roads using satellite imagery
Saturday, August 04, 2007
After I took my Masters degree in Geo-information Science in the Netherlands and an actual project in Spain, my love for maps and satellite images kicked off. Since then, I have been involved in numerous Geographic Information System (GIS) projects and did a lot of research studies that concern spatial data and information.In the university where I used to teach, I taught undergraduate and graduate engineering students the basics of GIS and the big-league concepts of Remote Sensing - tools that, I believe, students must know. However, the resources the university has, or the Philippines in general, were so limited (and still are) that I could only sigh for my ideas that could never materialize.
Presently, I work for a NASA project that involves analyzing vegetation stress through satellite images, while at the same time finishing a PhD degree in Geo-Information. Doing this in the United States gives me a relief that somehow the concepts that have been stuck in my head for long could be of use to the public. Accessibility of resources is not a problem and spatial data are enormous in this side of the world.
The existence of the free satellite images in Google and Yahoo or their ease of access is a great help for urban and regional planners and all those in the mapping sectors. I admit, I was amazed at how the resolutions of these images are becoming quite high and more detailed that you could even see your house or your wife going to your neighbor's house. But we do not use these images in mapping. Be careful, these images are better off as visualization aids rather than the sources of public maps.
A post by vaes states:
This (OpenStreetMaps using Yahoo's satellite images) greatly simplifies the mapping process because before, you needed to use GPS to get copyright-free street data, or else use low-resolution LANDSAT satellite pictures (or image), or public-domain but outdated maps. Now, you don’t need to have a Nokia N95 or other GPS units to get data; just load the Yahoo! satellite image, trace the roads and presto! Instant vector maps!This is easier said than done. If that's how easy it is, then we could have produced complete and comprehensive road maps much earlier. But our mappers didn't do that. Why? Because there are big-deal issues involved by utilizing these images. What vaes meant when he said "trace" is "digitize". Digitizing the satellite images is not a walk in the park. No, it is not a just, voila, here are the vector maps! There's more to just producing road maps. Geographers would always question the accuracy of the outputs. Unless of course Vaes initially intended his post to non-geospatial acquaintances, his ideas were ambiguous and inaccurate.
Our mapping agency in Pinas and other sectors in the same field have been digitizing analog maps to convert the old yellowed maps in paper into digital forms. This way we could easily share and use them as inputs for temporal investigations. One common problem that digitizers usually encounter is the offset created in the output. Roads or other line features don’t always map accurately. A fraction of inch of offset can mean tens of meters in the actual! I used to work in a project with a guy in the Philippines who did nothing but digitize all day. Offsets made his head ached.
Mapping using satellite images/digital aerial photography takes some steps like orthorectification and more. I took a course about it months ago and will write more of it soon.
Vaes did a good job of opening this topic up. It makes me wonder if I would have to deal more with GIS and Remote Sensing in future posts. This is more of an academic blog, maybe I should.
As for Yuga's question, "Google, where’s our street maps?", it seems to me that almost everyone I know assumes spatial data, derived from GPS for instance, are inexpensive, such that Google could offer them for free. To recall, India is progressing so well in the space science field, in fact, India is now the 5th Country in the world with capabilities to launch foreign satellites on a commercial basis. No wonder Google is interested with them.
It is better to end this post with this article The Future of Imagery and GIS. To those interested of links to free Remote Sensing images, leave me a note. To anyone interested in pursuing studies in GIS and Remote Sensing, list of universities are here.
Labels: Remote Sensing and GIS
10 Comments:
Thanks for commenting and reacting to my post.
Of course, I addressed the people who are not into GIS or cartography with my post. After all, I'm not a geospatial niche blogger.
OpenStreetMaps does not aim to create extremely accurate street maps such that roads are put down correctly to the last meter. Their objective is to create maps can be used to navigate streets, which is what the general population needs most of the time. (Unless they're fighting over property areas, of course) So mapping which streets intersect with which avenues, which roads are one-ways, etc. is the aim of OMS.
Google Maps, I believe, also aims for this level of accurateness. That way they can dominate the local search space by informing users where the local pizza parlor is, and where there is traffic, and what are the driving directions from L.A. to New York. Heck, Google should know that GPS derived data has an inaccuracy of about 10 meters, so they know the limits of the accurateness of the data they obtain in India.
In this regard, using Yahoo! satellite images to extract street data is very much a viable option for the aims of OpenStreetMaps. Very precise orthorectification or offsetting is not a goal. After all, the average Joe does not really care that his office is located 1,345m due 14.5°ENE of his house, right?
04 August, 2007
Thanks for the insight. But if they can do it in India using volunteers running kits, I assumed they can do that anywhere else.
04 August, 2007
Oh, by the way, despite what I said that I'm not a geospatial niche blogger, I do blog about maps and related topics. But take note that my perspective is more on the recreational side, so I really don't delve into high-end topics. (After all, I didn't take up geology, geography, or geodetic engineering in college.)
Probably the most technical post I did about mapping was "Topographic Maps Using SRTM3 Data".
04 August, 2007
Eugene, I am not sure how OSM data are actually produced. But I am afraid that if precise offsetting is not done, many new road intersections might appear in the map where they're not suppose to be. Imagine our streets separated only by few meters and then think of the offsets that digitizing could create. Could you see the difference?
05 August, 2007
Yuga~ yes, google can do it anywhere else. The point is, they do it to countries where their data matter most, where data could be fully utilized/maximized. With India's GIS and RS capabilities in full swing for years, who doesn't want to invest?
Thanks for the comment Yuga.
05 August, 2007
That's why the data is annotated and fixed after tracing it. (Yes, in some non-trivial cases, creating street maps from satellite/aerial images is not necessarily a "viola!" moment.) For one thing, you can't get one-way street info from satellite images, and you can't get street names, vehicular traffic restrictions (e.g., no PUVs allowed), etc.
For example, I'm quite familiar with the UP Diliman campus. Using satellite images, I can create a practically accurate map of the campus and annotate where the gates are, etc. In those cases precise location of streets is not needed.
Anyway, OSM relies on GPS trails primarily and encourages contributors to write notes while obtaining GPS data to provide annotation. The Yahoo! imagery is a big bonus because you can correlate your GPS data with the imagery to provide a more accurate feel of the streets. This also encourages people who don't have GPS units to add data for their neighborhood since they are intimately knowledgeable about their locations. That's why I think Yahoo! permitting OSM to use their satellite imagery is a really good thing.
This doesn't mean that high-end GIS mapping solutions is not needed anymore. But creating generally usable street maps aided by satellite imagery is something I don't think should be scoffed at, right?
05 August, 2007
Eugene~ First, I am not against the efforts of OSM. Second, the concept is not really unique but it is expedient.
Sure, creating maps with the aid of satellite images is a good thing. I don't question that. If the accuracy of the output maps in terms of accurate points, lines and polygon spatial locations is not part of OSM's goals, then there is no point to argue.
I am just after for accuracy because satellite images can be so deceiving - what you see is not what it may seem. Satellite images is not always the RGB (red, green, blue) colors that our human eyes could detect. It can go beyond that since satellite images are produced by EM radiations (quite different from aerial photos or any other lower altitude products).
Thus, in the end, plotting or mapping a road or a campus cannot be done "accurately". You may see greens on one side of a building and map it out as a grass class when in fact they are actually something else just colored green. Annotating the features wouldn't help in this case.
05 August, 2007
About satellite images not being accurate, there's also the fact that steep terrain throws the location of some places a bit off, I think this is what you mean by offsetting.
The thing is, any source of data, be it satellite, aerial, ground-level photos, or even an observer's own memories can be deceiving. (BTW, light is EM radiation, so aerial photos still use EM radiation. Maybe you meant to say that such remote sensing using satellite imagery is subject to more distortion because of the distance and obviously has lower resolution than lower-altitude photos.)
If your point is that we cannot create accurate maps only from satellite images, I agree and I don't think anyone with any modicum of smarts will argue with that.
My point, and I think I didn't elaborate about it in my post because I assumed the readers can infer it themselves, is that you can use the satellite image to create basic maps and complement it with other sources of information (observer's notes, memories, ground-level photos, aerial photos, public domain maps, etc.) to make it more accurate. OSM, after all, lets anyone edit the existing data, in the style of Wikipedia (but subject to more editorial control).
Going back to my campus example, I've explored the campus to an extent that I can hand-draw a map from memory. Using the satellite imagery I can make my hand-drawn map more accurate since basic distances and more precise road curves can be extracted. That's how I think satellite images can be useful in creating accurate maps.
If I know that the green thing on one side of the building is actually the roof of a shed and not grass because I've personally seen the shed, then I won't be deceived by the satellite image. But the satellite image will help me place the orientation/location of that shed correctly (subject to acceptable error margins), right?
If one is creating maps from satellite images without correlating or corroborating what one sees with other sources of information, then I agree that one shouldn't be creating maps at all.
05 August, 2007
Eugene, your point is much clearer now. Accurate maps can be produced by integrating different sources of data, like those you mentioned above. If one solely relies on imagery to produce maps, he/she has to do field checking to confirm the output classes. However, there are still issues involved - inaccessibility of features due to terrain constraints and other risks/circumstances. In addition, images should be seen and analyzed in different "pseudo" colors to detect unique feature characteristics.
To elucidate further, yes, satellite images and aerial photos are the same because they're both "remotely sensed". But I used "quite different" because the way EM radiations are collected, stored and processed in RS imagery are intricate and they are the top-rate products. In terms of resolution [you might want to browse about this] satellite images can have higher resolutions as well, such as the panchromatic ones produced by QuickBird and IKONOS.
There are a lot to argue and discuss about this topic. But I think we're one in saying that "If one is creating maps from satellite images without correlating or corroborating what one sees with other sources of information, then I agree that one shouldn't be creating maps at all" - which are your own words that I agree too.
05 August, 2007
Some good discussion here and likewise some good lessons learned by me. Thanks.
However I would submit that we are once again locked into the gridlock argument mode by the conundrum of "what is accuracy?"
For the PhD's of the GIS world (sorry, talksmart but I've been fighting this battle at the consumer/user level for years now) it seems to the outsiders that no map and no data is ever accurate enough. Yet at the level of people who want to make a simple road trip, a few meters here and there are of no consequence. There has to be some common meeting ground, and in the past 20 years I have been frustrated by the professional attitude that no map at all is better than a map with minor inaccuracies.
Take your point regarding one way street data. certainly that is an important criteria, especially for _routing_ which some people understandably feel is an integral part of _mapping_. When I am approaching a road junction in Bulacan which doesn't even show on my map, the fact that NAMRIA is stymied in gathering information on one-way streets in Makati just doesn't really have much importance to me.
The situation as I see it today is that because the agencies .. such as NAMRIA do not have the data and the budget to collect these ancillary data they will, instead, retreat into that world-wide bastion of government employees (I was one for 38 years so I know the routine well) of releasing nothing, because they don't have the resources to release everything. In the meantime, Filipino businesses who pay taxes to support NAMRIA are forced to make their own (accuracy deficient, yet commercially usable) maps because NAMRIA can't/won't.
Since you are in the US, take a look at Google Maps and audit them for quality from a GIS professional standpoint. I would assume you can create a lengthy list of deficiencies. But take a look at them from a standpoint of their intended purpose as a publicly accessible, profit-making business tool and I think (correction I _know_) that they are indeed valuable.
Projects such as the open source initiative come into being out of frustration and personal and business necessity. Again, my point is not to knock the professionals in the GIS industry but to make a plea to promulgate standards that would allow the public access to usable maps at perhaps reduced standards of "perfection".
If Columbus had been sponsored by ESRI, he would still be in port waiting for an accurate map.
05 August, 2007
Post a Comment
Links to this post:
Create a Link