- from open maps to open spaces-

Posted Tue Jan 29 19:28:33 -0800 2008

Here's my proposal for a session, or maybe a thread of sessions (?)

- from open maps to open spaces-

The world is getting really good at making and using 2D maps. Let's get up off these 2d planes of thinking!! I propose launching a serious effort to develop common open coding for 3D realspace!

What do I mean: 3D map views and spatial applications that are useful very near-field to use 3D geocoded data and media

Let's assume that very soon 3D data will be easily accessed through a handheld digital viewfinders, not glasses. (Take our your phone cam. look through the viewfinder and imagine you can see links, annotations, vectors, polygons and complex objects overlaid on the real world in 3space, )

Every infinite detail of every cubic volume of physical space can be digitally described detailed 3D and linked. Ideally this dense near-field infromation can be searched, blended and ideally viewed, and used casually

This is near-field geography for micro-locations - a great hacking frontier: Developing easiest ways to use and blend 3D data in near-field applications like augmented reality, sensorwebs, robotics, immersive media, realworld-virtual worlds . . .

Besides a stone simple 3D georss, and a spectrum of 'blended realities; dev tools, we need to understand semantics, mining, normalizing, renedering and interactiung with many legacy,, foriegn or discrete data imported from 3d gis, cad, rfids, and other local sensors, virtual worlds, gamespaces, . . .

If you are tired of mapping London, try mapping your living room in 3D, tired of mapping your park, try mapping your houseplant in 3D with annotations, tired of running around with your GPS, try mapping your desktop with unique object identifiers...


This sounds beautiful, "Let's assume that very soon 3D data will be easily accessed through a handheld digital viewfinders, not glasses. (Take our your phone cam. look through the viewfinder and imagine you can see links, annotations, vectors, polygons and complex objects overlaid on the real world in 3space)."

It seems that one will need more than a precise location of the viewfinder, but also the orientation of it. Not to say that we shouldn't have tiny gyroscopes in all our devices, but i wonder if the overlay process might be done with the contribution of a image to a image model, having the image matched, and then the overlay returned. Benefit is that the user is also a contributor, just as a GPS service user who broadcasts their location back to the system is providing data checks.

Large image modeling projects collected in this post: http://www.grey-cat.com/curious/?p=293

Not sure that all made sense....

 

Hey, maybe we could try something like this: http://www.david-laserscanner.com/ , but on a larger scale

 

something like the david laser scanner looks ideal. need to read more about ms windows dependencies. I'm in for a few $$ to build whatever we need if it doesn't get out of hand. if we're really going to do this, we need really sturdy tri-pod and maybe with a digital servo motor ( using midi?). & a way to precisely geoposition the sensor plane & lots of hacks to turn imagery into useful geodata data

 

agree, judelaine,

all kinds of image matching hacks are possible maybe in the cloud, with some - object recognition, returning overlays, etc.

need to precisely position the viewplane to. some nokia phones, and iphone have built in accelerometers that sense tilt. might be hackable to give some orientation. My guess is we'll need also a really rich array of wifi beacons. I bet Loki ( Ryan?) might be able to help with micorpositioning of our handhelds.

 

I've been taking high resolution panoramic images with a Gigapan unit. It has two stepper motors to allow pan and tilt movement to direct a camera.

See http://gigapan.org
and my panoramas at:
http://gigapan.org/viewProfile.php?userid=353

The firmware could be changed to do a continuous scan, so if we had a smallish rangefinder the gigapan could pan and tilt through the scene and collect the data.

 

attaching the laser to your motor and tripod sounds like a great idea

if we want to wrap images onto volumes, there will be distortion, so knowing precise postion of the focal plane will help properly 'rubbersheet' the pan tiles, on surfaces, and volumes

 

This is a good topic. In the logistics industry they have systems that do this kind of high-resolution 3D tagging and tracking on a warehouse scale.

The major problem with 3D sensor aggregation and spatial overlay out in the real-world is not the availability of data sources (they are myriad) but the inability to manage and index spatial data on the scale typical for such things. This has stymied a lot of efforts to build non-toy systems of this nature. There is also standards work that definitely needs to be done -- most of what has been done in this space so far is too narrow or does not scale up. Even among the people working on SDI standards, only a small minority seem to be thinking about architecture in addition to interoperability; we may get interoperability out of it, but it will not do us much good if the tacit architecture is hopeless for many real-world use cases.

I have considerable interest in 3D (and 4D) "mirror world" spatial overlay environments, but the technical and standards issues that remain to be solved are legion. Hence why it is a good topic. :-)

 

I did a little poking around and it looks like the data collection part could be much simpler if we could get our hands on a SICK LMS 200 LIDAR unit. It seems to be popular with the autonomous robot crowd so there's plenty of source and examples for interfacing with it., for example: http://www.pages.drexel.edu/~kws23/tutorials/sick/sick.html

The problem is that these units run around $6K, but they apparently turn up on ebay occasionally.

 

wow. great idea, but expensive ( beyond my range.) last year, one person brought a lidar unit with him. on that he used to collect data while driving -- but I don't remember his name. maybe he'll read this and volunteer to bring it again.

 

Focal length of a camera can determine the distance of objects, and are usually stored in the exif of the images)

Quick and cheap (but not totally accurate, I would think) - so could be good for measuring That Big Object from the Road.

 

The GPS traces collected on OpenStreetMap store 3D, GPS Elevation, but there's no use of them presently. Maybe we could look at creating crowd sourced DEMs...

 

Tried that before... you get a REALLY noisy surface. Might work if you smoothed the data and were happy with +- 5m vertical precision.

 

@wherecamp sorry can't be there to work on 3d geo after all. pre-empted by prior commitment for fun time with grandchildren 3 & 5, and now too , the passing of a close friend yesterday

Please have 3Dgeohacking fun!!