Remote Sensing and Image Processing: 4

Transcription

1 Remote Sensing and Image Processing: 4 Dr. Mathias (Mat) Disney UCL Geography Office: 301, 3rd Floor, Chandler House Tel: mdisney@geog.ucl.ac.uk 1

2 Image display and enhancement Purpose visual enhancement to aid interpretation enhancement for improvement of information extraction techniques Today we ll look at image arithmetic and spectral indices 2

3 Basic image characteristics pixel -DN pixels - 2D grid (array) rows / columns (or lines / samples) dynamic range difference between lowest / highest DN 3

4 Aside: data volume? Size of digital image data easy (ish) to calculate size = (nrows * ncolumns * nbands * nbitsperpixel) bits in bytes = size / nbitsperbyte typical file has header information (giving rows, cols, bands, date etc.) (0,0) ncolumns nbands (0,0) ncolumns nbands nrows nrows (r,c) (r,c) Time 4

5 Aside Several ways to arrange data in binary image file Band sequential (BSQ) Band interleaved by line (BIL) Band interleaved by pixel (BIP) From 5

6 Data volume: examples Landsat ETM+ image? Bands 1-5, 7 (vis/nir) size of raw binary data (no header info) in bytes? 6000 rows (or lines) * 6600 cols (or samples) * 6 bands * 1 byte per pixel = bytes ~ 237MB actually MB as 1 MB 1x10 6 bytes, 1MB actually 2 20 bytes = bytes see Landsat 7 has 375GB on-board storage (~1500 images) Details from 6

7 Data volume: examples MODIS reflectance 500m tile (not raw swath...)? 2400 rows (or lines) * 2400 cols (or samples) * 7 bands * 2 bytes per pixel (i.e. 16-bit data) = bytes = 77MB Actual file also contains 1 32-bit QC (quality control) band & 2 8-bit bands containing other info. BUT 44 MODIS products, raw radiance in 36 bands at 250m Roughly 4800 * 4800 * 36 * 2 ~ 1.6GB per tile, so 100s GB data volume per day! Details from and 7

8 Combine multiple channels of information to enhance features Image Arithmetic e.g. NDVI (NIR-R)/(NIR+R) 8

9 Image Arithmetic Combine multiple channels of information to enhance features e.g. Normalised Difference Vegetation Index (NDVI) (NIR-R)/(NIR+R) ranges between -1 and 1 Vegetation MUCH brighter in NIR than R so NDVI for veg. close to 1 9

10 Image Arithmetic Common operators: Ratio topographic effects visible in all bands FCC 10

11 Image Arithmetic Common operators: Ratio (ch a /ch b ) apply band ratio = NIR/red what effect has it had? 11

12 Image Arithmetic Common operators: Ratio (ch a /ch b ) Reduces topographic effects Enhance/reduce spectral features e.g. ratio vegetation indices (SAVI, NDVI++) 12

13 Image Arithmetic Common operators: Subtraction An active burn near the Okavango Delta, Botswana NOAA-11 AVHRR LAC data (1.1km pixels) September Red indicates the positions of active fires NDVI provides poor burned/unburned discrimination Smoke plumes >500km long examine CHANGE e.g. in land cover 13

14 Top left AVHRR Ch3 day 235 Top Right AVHRR Ch3 day 236 Bottom difference pseudocolur scale: black - none blue - low red - high Botswana (approximately 300 * 300km) 14

15 Image Arithmetic Common operators: Addition + Reduce noise (increase SNR) averaging, smoothing... Normalisation (as in NDVI) = 15

16 Image Arithmetic Common operators: Multiplication rarely used per se: logical operations? land/sea mask 16

17 Basis: Monitoring usingvegetation Indices (VIs) 17

18 Why VIs? empirical relationships with range of vegetation / climatological parameters fapar fraction of absorbed photosynthetically active radiation (the bit of solar EM spectrum plants use) NPP net primary productivity (net gain of biomass by growing plants) simple (understand/implement) fast (ratio, difference etc.) 18

19 Why VIs? tracking of temporal characteristics / seasonality can reduce sensitivity to: topographic effects (soil background) (view/sun angle (?)) (atmosphere) whilst maintaining sensitivity to vegetation 19

20 Some VIs RVI (ratio) RVI nir = ρ ρ red DVI (difference) = ρ ρ DVI nir red NDVI NDVI = ( ( ρ ρ nir nir ρ + ρ red red ) ) NDVI = Normalised Difference Vegetation Index i.e. combine RVI and DVI 20

21 Properties of NDVI? Normalised, so ranges between -1 and +1 If ρ NIR >> ρ red NDVI 1 If ρ NIR << ρ red NDVI -1 In practice, NDVI > 0.7 almost certainly vegetation NDVI close to 0 or slightly ve definitelyy NOT vegetation! 21

22 why NDVI? continuity (17 years of AVHRR NDVI) 22

23 limitations of NDVI NDVI is empirical i.e. no physical meaning atmospheric effects: esp. aerosols (turbid - decrease) direct means - atmospheric correction indirect means: atmos.-resistant VI (ARVI/GEMI) sun-target-sensor effects (BRDF): MVC? - ok on cloud, not so effective on BRDF saturation problems: saturates at LAI of

24 24

25 saturated 25

26 Practical 2: image arithmetic Calculate band ratios What does this show us? NDVI Can we map vegetation? How/why? 26

27 MODIS NDVI Product: 1/1/04 and 5/3/04 27