Pre-processing of a sample of multi-scene and multi-date Landsat imagery used to monitor forest cover changes over the tropics

The project incorporated in the TREES-3 i.e. projects that cultivate more than 12,000 Landsat TM satellite image data and ETM + systematically distributed to tropical areas.

Writer :

Joint Research Centre of the European Commission, European Union:

1. Catherine Bodart,
2. Hugh Eva,
3. René Beuchle,
4. Rastislav Raši,
5. Dario Simonetti,
6. Hans-Jürgen Stibig,
7. Andreas Brink,
8. Erik Lindquist,

United Nation Food and Agriculture Organization (FAO):

1. Frédéric Achard

The purpose:

On paper it shows about the stages of different pre-processing's aims to produce the multi-volume set temporal data with a consistent geometric and radiometric, thus obtained statistical data on changes in forests which do not berbias to generate a reliable classification.

 Characteristics Data

1. Satellite imagery Data is 20 x 20 Km from Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM +)
2. The reference year 1990 (1984-1997) and 2000 (1997-2003)
3. A source from the United States Geological Survey's (USGS) http://glovis.usgs.gov
4. Resolution of 30 meters
5. Additional image Data of the Committee on Earth Observation Satellites (CEOS) is used to replace in case of incomplete data from Landsat.
   

   Pre-Processing

   
   1. Visual Assessment
   2. Radiometric Calibration
   3. Remove The Cloud Cover
   4. Remove Fog
   5. Normalization of color density forests

   Visual Assessment

   In order for multiple-date image data with greater precision to the actual position on the Earth, then the required adjustment in visual (spatial registration) with the other image that has possess precise geo-location.

   
   

   1. Image ofthe year 2000serve as themaster,
   2. Image of1990as a slave,
   3. Thendoa linearshift inX and Ysopixelnyafit/fit.
   4. Thenthe newimagedatataken20 x 20 kmoftheseadjustmentsresults.

    Radiometric Calibration

   Radiometric calibration is the first step to reducing the differences that emerged as a result of changes in the lighting conditions and the mistakes of the equipment.

   
   

   1. Convert raw digital numbers (DN) to the Sensor Spectral Radiance 
   2. Change the sensor radiance to the top-of-atmosphere reflectance 
      Calculate the distance the reflection of light from the Sun.
      Then for sensor Thermal (temperature) is used to measure brightness temperature (T) in kelvins.

   
   

   Remove The Cloud Cover

   To perform the removal of the cloud, to do two stages of the process, namely:

   1. Detect all of the potential of cloud and its shadow pixel with Automatic Spectral approach to Rule-Based Preliminary proposed by Baraldi et al. (2006). The method is a system of classification with decision rule which is designed to map the results of the Landsat data have been calibrated to some categories of spectral kernel.
   2. The process consists of two sequential application of post-processing algorithm based on morphology and topology method that will clear a cloud based on size, width and shape.

   
   

   Remove Fog

   To get rid of fog can be done as follows: TC4 method Lavreau method (1991), using The Fourth Component of tasseled cap transformation (TC4). This method uses a 6 band reflection of the Landsat satellite

   
   

   Normalization of color density forests

   Find the median value of the density of the forest with the following stages:

   1. Createamask (caps) forestfor everysatellite image of1990-2000)based ona comparison ofthe determination of the thresholdonNDVIand bands4 and 5does not includethe existingpixelclouds.
   2. To reducethe variedforestsform, then the result ofmaskingin 1990-2000-iriskan(intersected).
   3. In addition,generallymaskingforestdiintersectionagain witha map ofthe forest thatwas derived from theproceeds oflandsustainablevegetation(VCF)discovered byHansenet al. (2003).

    Output 

   

   
   

   • a. Beforeremoval ofthe fog
   • b. The resultofmaskingTasseledCapTransformation(TC4)
   • c. Afterthe process ofde-hazing