Resolution. Learning Objectives. What are the four types of resolution?

Transcription

1 Ladsat ETM+ image Resolutio Learig Objectives Be able to ame ad defie the four types of data resolutio. Be able to calculate the umber of pixels i a give area. Uderstad the trade-offs betwee differet types of resolutio. Uderstad the relatioship betwee SNR ad resolutio. Learig Objectives (cot.) Uderstad biary data ad the relatioship betwee radiometric resolutio ad data storage space. Uderstad the differeces betwee types of orbits. What are the four types of resolutio? 1

2 IFOV Spatial Resolutio Usually reported as the legth of oe side of a sigle pixel I aalog imagery (film), the dimesio (e.g. width) of the smallest object o the groud that ca be distiguished i the imagery Determied by sesor characteristics (for digital imagery), film characteristics (for air photos), field of view, ad altitude. Group Problem 1 pixel Raster pixel size Higher resolutio If you have a study area that covers 1 km2, how may 30 m Ladsat pixels does it take to cover it (earest whole umber)? How may 15 m Ladsat pachromatic pixels would it take to cover the same area? Lower resolutio Available Spatial Resolutio for Lad RS Satellites: ~ 0.3 m to1 km Air photos ~ cetimeters to meters Satellite data resolutio MODIS: Ladsat MSS: Ladsat 5, 7, 8: IRS MS: SPOT: ASTER: WorldView 3: m 80 m 30 m (15 m pachromatic) 22.5 m (5 m pa) 20 m 15m 1.24 m (0.3 m pa!) 2

3 Quickbird (Digital Globe, Ic.) MODIS 500 m spatial resolutio ~ 2.4 m spatial resolutio i multispectral bads. Spatial Resolutio Trade-offs Spectral Resolutio Data volume Sigal to Noise Ratio (SNR) Salt ad Pepper Cost Ca be described two ways, but they usually go had i had. Spectral resolutio Case 1 Related to the measured rage of EMR Wide rage - coarser resolutio Narrow rage - fier resolutio How may spectral bads a istrumet records How wide each bad is (the rage of wavelegths covered by a sigle bad) Measure the EMR across a wide rage of wavelegths E.g., a sigle pachromatic bad coverig the etire visible portio of the spectrum Assigs a sigle DN represetig all visible light hittig the sesor Aalogous to black ad white (pachromatic) film 3

4 red gree UV blue Case Near-ifrared From USGS Spectral Characteristics Viewer Case 2 Case 2 gree red UV blue Measure EMR across arrower rages of wavelegths E.g., Separate bads for blue, gree, ad red parts of the spectrum Assig a DN for each of these wavelegth rages to create 3 bads Near-ifrared Coarser (lower) Spectral Resolutio RGB Fier (higher) Spectral Resolutio Red Gree Blue From USGS Spectral Characteristics Viewer 4

5 High Spectral Resolutio Reflectace 0.8 Spectral reflectace curve for gree leaf usig 224 bads (high spectral resolutio) Wavelegth (m) Low Spectral Resolutio Spectral reflectace curve for gree leaf usig 6 bads (lower spectral resolutio) 0.8 Reflectace Could you distiguish Dolomite from Calcite usig Ladsat 8 spectral data? Wavelegth (m) Problem (with parter) Spectral Resolutio Trade-Offs Data Volume ad processig 1 DN for each pixel i EACH BAND Sigal to Noise Ratio Cost For your 1 km2 study area, if you use 7 Ladsat 8 bads, how may DNs will your computer have to store? Group Exercise Is higher spatial resolutio better tha lower spatial resolutio? Yes/No with reaso Is higher spectral resolutio better tha lower spectral resolutio? Yes/No with reaso Radiometric Resolutio How fiely does the satellite divide up the radiace it receives i each bad? How much light does it take to chage the DN from oe umber to the ext? Radiometric resolutio is usually expressed as umber of bits used to store the maximum possible DN value 8 bits = 28 = 256 levels (DNs 0 to 255) 16 bits = 216 = 65,536 levels (0 to 65,535) 5

6 Radiometric resolutio 26 = 64 levels (6 bit) 1 bit ( 0-1) 8 bit ( ) (older Ladsats, may others) 16 bit ( 0-65,535 ) (Ladsat 8) 32 bit ( 0-4,294,967,295 ) (ucommo) 22 = 4 levels (2 bit) For a 8-bit satellite: DN = 0: No EMR or below some miimum amout of light (threshold) DN = 255: Max EMR or above some maximum amout of light Covertig Base 10 to Biary Base 10 Base 2 (Biary) (etc.) Radiometric Radiatio Trade Offs Data volume Every 8 bits takes 1 byte to store o a computer. Oe 8-bit DN takes 1 byte Oe 9-bit DN takes 2 bytes Oe 16-bit DN takes 2 bytes Oe 17-bit DN takes 3 bytes Etc. Radiometric resolutio 8 bit data (e.g., Ladsat 5) (256 values) Everythig will be scaled from Subtle details may ot be represeted 16 bit data (e.g., Ladsat 8) (65,536 values) Wide rage of choices Required storage space will be twice that of 8 bit Calculatig Image Size Computer hard drives store data i boxes called bytes (e.g., 1 Mb = 1 millio bytes) 1 byte ca hold 8 biary (base 2) digits (0s or 1s or some combiatio of 0s ad 1s) Each bit is a sigle biary digit A 8-bit umber is made of of 8 biary digits ad fits ito 1 byte. A 9-bit umber wo t fit i 1 byte ad requires 2 bytes. 6

7 Group Problem If your are usig 7-bad, 16-bit Ladsat 8 data for your 1 km2 area, how may bytes are eeded to store your DNs o your computer? Temporal resolutio Time betwee two subsequet data acquisitios for a area All of the Ladsat satellites have a 16-day retur time MODIS has a 1-2 day retur time. Retur Time (Temporal Resolutio) Orbital Characteristics Geosychroous Depeds o: Orbital characteristics Swath width Ability to poit the sesor Polar Su sychroous Geosychroous Orbits Polar/Su Sychroous Orbits Satellite orbits the earth at a rate that allows it to match the earth s rotatio so the satellite is always over the same place Narrow rage of altitudes about 35,786 km above the equator. Useful for commuicatios, weather etc. Example: GOES satellite (weather) Pass roughly over the orth ad south poles Fly over the same place o earth at the same time of day Examples: Ladsat, AVHRR Good for lad remote sesig Retur time related to spatial resolutio, latitude, swath width, ad orbital altitude Geosychroous orbitig earth satellite 7

8 Retur Time Trade Offs Spatial resolutio Viewig geometry effects (off adir) Clouds ad other atmospheric problems Lack of archival repeat coverage for poitable satellites I summary, choosig a satellite is a exercise i weighig the relative trade-offs of resolutio agaist data eeds (ad budgets!). 8