Data

The idea of implementing the above classifiers (MLC, DT, KNN, NN, RF, SMAP and SVM) was to investigate their performance on high, medium and low spatial resolution sensor data with unknown distribution. So, IKONOS MS (4 m), IRS LISS (Linear Imaging Self Scanner)-III MS (23.5 m), Landsat ETM+ (Enhanced Thematic Mapper Plus) of 30 m spatial resolution and MODIS (Moderate Resolution Imaging Spectroradiometer) with 7 bands (2 bands at 250 m and 5 bands at 500 m, resampled to 250 m) were chosen in lieu of high, medium and low spatial resolution data (as given in table 1) because these are the major sensors commonly used for numerous applications in LULC mapping and monitoring. Training and testing data were collected using pre-calibrated hand held GPS (Global Positioning System) for each study area during several field visits and Google Earth data were used to obtain pre-classification information and for post-classification validation.

Table 1: Remote sensing data sets used for algorithms implementation

.