Dealing with image data

From Latent Variable Methods in Engineering
Revision as of 19:01, 7 February 2017 by Kevin Dunn (talk | contribs)
Jump to navigation Jump to search
Class date(s): 04 November 2011
Video material (part 1)
Download video: Link (plays in Google Chrome) [183 Mb]


Video material(part 2)
Download video: Link (plays in Google Chrome) [255 Mb]


Video material (part 3)
Download video: Link (plays in Google Chrome) [287 Mb]

Class 9: Dealing with image data for product appearance monitoring (04 November 2011)

Download the class slides directly

MIA exercise

Background

If you have MATLAB, please download and install MACCMIA for to complete the questions below.

  • Data file: a 7-channel Landsat image of Washington D.C. and the corresponding Google map.
    • The image is a 1024 x 1026 x 7 cube, using 7.3 Mb of storage
    • Channel 1 = blue/green [450 to 520 nm]
    • Channel 2 = mainly green [520 to 600 nm]
    • Channel 3 = red [630 to 690 nm]
    • Channel 4 = Near infrared [760 to 900 nm]
    • Channel 5 = Moisture and vegetation sensitive mid infrared [1550 to 1750 nm]
    • Channel 6 = Thermal infrared [10400 to 12500 nm]
    • Channel 7 = Mineral-sensitive mid infrared [2085 to 2350 nm]

Questions

  1. Identify some water pixels from the Potomac river in the score plot. Where are they located?
  2. Mask these water pixels to highlight the entire river. Notice how you have to adjust the mask to get both the shallow and the deep water pixels.
  3. Notice how the pixels from half of the Georgetown Reservoir show up with the Potomac River's pixels, and the other half show up entirely different (sediment has settled out).
  4. Continue investigating features from the Google map and try to locate them in the satellite map and score space. For example:
    • The Reflecting Pool
    • Forests in The National Arboretum
    • The network of roads and highways