Data Collection (How to Digitize the World)

What do you do when you've found all the pre-recorded data that exists, but need more detail?  There are many places for which such data (aerial, satellite, road, vegetation, structures etc.) are either too expensive, inaccurate, or unavailable.

In the perfect world, there would be a device that you simply carry around, and it would sense and measure everything and build a database automatically.  Alternately, the device could be a UAV and you could collect your own data aerially.

GPS

Consumer

Professional

• Consumer GPS devices are more widespread, cheap and accurate than ever, but are still not sufficiently accurate or versatile enough to gather good enough data for ground-level visualization.
• As of 2012, "professional" GPS is still unaffordable (thousands of US$).  The cost difference is due to two major differences over consumer GPS:
  1. They record the signal from all the satellites in range, so that each signal can be logged and post-processed independently later.
  2. They capture the "L2 band", a higher precision signal that requires a better chip, which isn't mass-produced so it's inherently more expensive.
• A typical consumer unit is the Garmin eTrex GPS: A handheld unit about as big as a cell phone, around $100, battery life of ~30 hours on two AA's. With good sky visibility, horizontal accuracy is 5-7 meters.
  • More expensive consumer units also have a built-in compass and barometric altimeter (GPS altitude readings are pretty inaccurate), but it's still nowhere useful for most surveying.
• A big problem with GPS is that it needs line of sight to at least 3 satellites to get an accurate fix, which means its of limited usefulness in places such as dense cities or dense forests.  Buildings and trees problems, both with obstruction, and the"multi-path" problem, where the signal bounces.
• DGPS - Differential GPS - see all about Differential GPS
  • Decreases error (to potentially less than 1m) by using another antenna and a transmitter at a known ground location.
  • The system is therefore more much expensive (typically >$5000)
• WAAS - the Wide Area Augmentation System
  • Improves accuracy to 1–2 meters horizontal and 2– meters vertically, similarly to DGPS.
  • Unlike DPGS, works with exiting GPS receivers, provided their software is upgraded.
  • Only covers the USA
• Even with DGPS or WAAS, there are still problems with obstruction and multi-path in some areas.
• Newer consumer units like the DeLorme Earthmate are WAAS-enabled, and the Professional model claims that post-processing software allows for sub-meter accuracy.
• There is a standard file format for GPS data like waypoints: GPX the GPS eXchange Format
• Garmin publishes their Communication Protocol Interface Specifications to enable custom software to communicate with their devices.
• As of 2011, nearly all smartphone have surprisingly good GPS capability, on par with standalone consumer GPS units.

Portable Data Collection Devices 

• MIT wearable sensor project, 2012
  • "sensor that digitally maps the area through which the wearer is moving, in real time"
  • received press as BBC: Wearable sensor maps disaster areas in real time
  • consists of accelerometers, gyroscopes, a Kinect, and LIDAR
  • main person is Maurice Fallon, perception lead on MIT’s Darpa Robotics Challenge team
• Surveylab - ike
  • Incorporates a GPS, an electronic compass, a laser distance meter, an inclinometer, a digital camera, and a Pocket PC handheld in a single unit.
  • The standard model's GPS has normal accuracy (~5m) but also allows some additional accuracy via DGPS Post Processing, RTCM 'Real Time' DGPS, and SBAS (i.e. WAAS and EGNOS)
  • Laser has a range of 100m, though 30m is more typical
  • costs $8500 (and up?)
    

Surveying

• Conventional surveying equipment (a Total Station, etc.) is expensive, and requires training, so generally it is only used only for small areas.  As a service, the cost to hire a surveyor or rent equipment varies greatly from place to place, but it's also generally too high for most people and projects.
• There are a very large number of companies in this field, Topcon is a fairly prominent one. Laser Technology Inc. makes the instruments in a wide range of products from high-end to low-end affordable laser rangefinders (as little as $150 as of 2012)

Video/Photo Processing

• for buildings and other structures
• MIT City Scanning Project
  • goal: given images of an urban environment, produce a CAD model of the structures in that environment
  • uses a special camera called Argus with global positioning (GPS), heading information (IMU), and dead-reckoning (optical encoders on the wheels)
  • at each position, takes up to 70 photos which are stitched into a high-resolution hemispherical image
  • a large number of these hemispheres are gathered, called a pose image dataset
  • feature extraction and texture extraction are used to produce the output models
• Automated reconstruction of 3D scenes from sequences of images (pdf), M. Pollefeys, R. Koch, M. Vergauwen, L. Van Gool, 2000

  • claims fully automatic extraction of textured 3D models from multiple photographs, using projective reconstruction, self-calibration and dense depth estimation.

LIDAR

• See Point Clouds, LIDAR

Voice Recognition

• it's possible that a trained user could use natural language to rapidly input 3D scene data
• the voice recognition should be fairly doable, since we can ensure a limited, fixed vocabulary, and train the system to a specific user's voice
• CMU Sphinx Open Source Speech Recognition
  • Sphinx-2 is a real-time, large vocabulary, speaker-independent system
  • Sphinx-3 is slower but more accurate

Amateur aerial photography

• a possibility, discussed on Ben's personal site: how to elevate yourself

Older

• Historical: Personal Pilot from GeoPerception
  • a (proposed) hardware/software solution that allows users to interact with GIS data in the field using a lightweight, wearable, hands-free system
  • included a Positioning Engine which combines a GPS receiver, a differential signal receiver (for <2 meter), a heading decoder (<2 degrees) and a computer with custom data and software
  • their DGPS used the free 300 KHz differential signal
  • hardware base was a Xybernaut wearable PC
  • the software enables you to query features by looking at them, to log data to your GIS by walking about, to have a continuous 'bird's eye' perspective of where you are in relation to your data, etc.
  • contact: Dave Caswell, estimated cost $10k
• Historical: The ART+COM UJP (1999)
  • Differential GPS (<1 meter), high precision electronic compass, stamps every single piece of information that is recorded by the UJP with the exact location and viewing direction from wich it was recorded
  • Records video from a head-mounted camera on MiniDV, transmits back to a Base Station in scalable resolutions. Mobile phone is integrated, wired to an audio switcher with connectors for the headset, the external handheld microphone, the MiniDV recorder and the computer's soundboard
  • Software: Linux, Java, Informix database
  • Full PC motherboard with a BT848 based video frame grabber and a 8 Gig IBM Travelstar hard disk
  • Worked with an Sony Glasstron or 10.4" TFT display with touchscreen
  • Battery  power for up to 6 hours
  • One prototype was made, estimated cost $30,000
  • The 'Trimedia Consortium' was a website (as of 1999) which seems to tie the UJP together with several other companies
• Geometrix [geometrixinc.com, no longer online?]
  • back in 1998-2000, was working on "software-based systems which automatically generate 3D simulation models of real-world objects and scenes"
  • they call this "VideoModeling" and the product "SoftScene"
  • they implied that it would be good for digitizing structures such as buildings
  • ***Geometrix Awarded $750,000 Development Contract (February 24, 1999)
    Geometrix announced it has been awarded a research and development contract valued at $749,470 for systems that automatically generate high-resolution 3D simulation databases from video imagery tagged with GPS metadata acquired by airborne platforms. The systems will generate 3D urban terrain models at levels of detail not currently possible with satellite-based imagers. [...]
  • the product was never finished and as of early 2001, the project was inactive