ingenieur Vol.84 Oct-Dec 2020 Vol 84 2020 | Page 55

( ii ) Radiometric Correction The images were acquired at varying times , hence there would be diverse atmospheric circumstances Ideally the atmospheric environment must be uniform as it is the criteria for monitoring changes . To resolve this issue , radiometric correction was done using dark object subtraction to ameliorate atmospheric effects . Radiometric correction of remotely sensed imagery is typically performed to mitigate the effect of anomalies that may impair the capacity to evaluate and view images in quantitative terms . Due to the absence of coexisting reference details on these images , this problem was resolved by the relative radiometric correction process . The first being that digital numbers ( DNs ) should be translated to satellite reflectance utilising normal reference values to eliminate temporal variations in sensor sensitivity and environmental variables across image datasets .
PROCESSING
Image processing is a way of conducting such operations on an image to produce an improved image or to acquire valuable output . In this research , the output was generated for NDVI , LST , DEM and Terrain , Tasseled Cap Transformation ( TCT ) and sub-pixel analysis with the aid of GIS and remote sensing software such as System for Automated Geoscientific Analyses ( SAGA- GIS ) version 3.2 and Environment for Visualizing Images ( ENVI ) version 5.3 .
( i ) Normal Difference Vegetation Index ( NDVI ) The NDVI is a simple graphical indicator which can be used to analyse remotely sensed measured data often from a space platform to evaluate whether the target observed contains live green vegetation . This method can also be applied to detect changes in vegetation over a period of time . To generate a NDVI map , the red band and the NIR band were used with a geoprocessing feature in SAGA-GIS version 3.2.2 . Once the map was generated , the pixel value of 20 points from each lot were taken and averaged to get the average NDVI values of each lot . A trendline graph was also plotted to show the NDVI trendline of the oil palm land scheme over a time span of 32 years .
( ii ) Land Surface Temperature ( LST ) Landsat 5 and Landsat 8 Surface Reflectance products are only provisional and TM Band 6 and TIRS Band 10 and 11 do not have surface reflectance . Due to this , a manual conversion for these three bands was done specifically . The conversion was done from DN to radiance , radiance to brightness temperature ( BT ) and lastly from BT to LST . The radiance thermal bands ( band 6 for Landsat 5 , band 10 and 11 for Landsat 8 ) were used to generate the BT using Equation1 as shown below .
BT = Effective at-sensor brightness temperature
(° C ) K2 = Calibration constant 2 ( K ) K1 = Calibration constant 1 ( W / m2 srμm )
Lλ = Spectral radiance at the sensor ’ s aperture ( W / m2 srμm ) ln = Natural logarithm
Once the BT map was generated , the LST map was then generated using Equation 2 as shown below .
BT = Brightness Temperature (° C )
λ = Wavelength of emitted radiance
ρ = Plank ’ s Constant ( 1.438 x 10-2 m K )
ε = Land Surface Emissivity ρ = h × c / σ ( 1.438 x 10-2 m K )
σ = Boltzmann constant ( 1.38 x 10-23 J / K ) h = Planck ’ s constant ( 6.626 x 10-34 J s ) c = velocity of light ( 2.998 x 108 m / sec )
( iii ) Unsupervised Clustering Classification This method was based on image analysis without the user providing sample classes . The software used several techniques to classify pixels into groups . Spectral classes were grouped based on numerical data and then matched with the information classes obtained by the analysts . Once the unsupervised classification had been done , the classes were then recoded as multiple classes of the same feature into one specific value . This research involved three land cover
53