An Evaluation of Multi-Band/Multipolarised SAR Data For Vegetation Discrimination, In Malaysia
Clear-cut
Appears after cutting off rubber or oil palm plantation (before replanting )
Suburban/settlement area
It consist of single-double story terrace house or bungalow house. Between the houses there are grasses, small vegetation or fruit trees.
Satellite data and Software
The remotely sensed data comprised of multifrequency (P, L bands ) polarimetric SAR data acquired by JPL over an area called 'Tapah' on 3 December 1996. For the purpose of our experiment we used a sub-image of 1141 x 948 pixels (figure 1) as it is apparently quite representative of all the land cover types comprised in the original SAR image. The software has been used is ENVI 3.0 ( the Environment for Visualizing Images).
Figure 1: Images of the study area, Tapah Perak- LHH, LIT, LHV (R, G, and B)
Methodology
Decompress-Synthesized Images
Polimetric radar data typically comes as compressed scattering matrix format that, can not easily be viewed by most image processing software. ENVI'S synthesize functions provide an easy-to-use and flexible means of generating any combination of transmit/receive polarization images directly from JPL AIRSAR compressed Stoles matrix data (and AIRSAR portion of the TOPSAR data set ) or from SIR-C scattering matrix compressed data. This includes the standard HH, VV, HV and Total power images for each frequency.
Data correction-Slant to ground Range and Cross-Track Illumination Correction were done by JPL.
Identified region of interest-oil palm, natural forest, plantation forest, clear cut and suburban . Identification was done based on visual integration-(color, texture, pattern, size), field verification and topography map.
selection of sample size was done starting with 50 pixels for natural and plantation forest, oil palm and rubber and 750 pixels for suburban. The intensity vale was collected for each of the cover types. The pixels size was chose for each cover types and backscattering coefficients were recorded according to frequencies and polarizations .
Result and Discussions
The effect of the variation of incidence angel across AIRSAR (from 51° to 57°) on the
s° is negligible. The study focuses on pixel number related to intensity stability and the
s° of the cover types .
Pixel Number
The intensity stability for each class is summarized in figure 2. land cover, verification was done during aircraft flown and July 98 for identification the region of interest. Figure 2 ( a and b) show the intensity are higher at the smaller pixel number, and decrease when the pixels for suburban area. All the vegetation areas - natural forest, plantation forest, rubber and oil palm intensity stability were stable at 300 pixels and become uncertainty when the pixel size increase. The intensity stability was highly correlated with the homogeneity of the region of interest. For clear-cut and suburban areas the pixel number for intensity stability is higher than the vegetation area due to mixture of the properties such s houses, park and vegetation for suburban areas.
Figure 2: Graph (a) shows the intensity value of clear-cut. (b) Intensity value of suburban and (c) intensity values for vegetation
