RESULTS AND DISCUSSION

Seven classifications were carried out with Landsat ETM+ bands 1 to 5 and band 7 of 2000 x 2000 size (acquired on March 14, 2000) using training data collected from field and validated using separate test data into agriculture, builtup, forest, plantation, wasteland and water bodies that are the six major categories in the forested and mountainous terrain of Uttara Kannada district in Central Western Ghats (Fig. 1). Figure 2 shows the classified images and the LU statistics are listed in Table I. Accuracy assessment was done by generating error matrix; producer’s, user’s, overall accuracies and kappa were computed (Table II). The highest two overall accuracies are highlighted in bold.

Table 1. LU estimates from ETM+ using advanced classifiers

Table 2. Accuracy assessment for ETM+ classified data

Landsat data having a spatial resolution of 30 m were classified most accurately using SMAP algorithm (89.03% overall accuracy as given in Table I). SMAP takes into account the intra class spectral variations and exploits spatial information among neighbouring pixels to improve classification results [17]. NN was difficult to train before it reached convergence as evident from the training RMS plot (not shown here). The area covered by this image has forested landscape, dominated by evergreen and semi-evergreen flora. Plantation was overestimated in DT and NN which had 4 hidden layers with 0.2 learning rate and 0.2 momentum with 20000 epochs, took 77 seconds to train. A second degree polynomial function with gamma as 0.167 was used in SVM (Poly), which gave lower accuracies in detecting water bodies in comparison to other seven techniques. DT, NN, RF and SVM (RBF) showed abnormal trends and have classified mountain ridges as narrow water channels (Fig. 2).

Figure 2. Classification of ETM+ Plus data through advanced classification algorithms.

Wasteland are often mixed with fallow land due to seasonal differences in crop practices, and also reflect similar to sand on the sea shores/sea beaches (Arabian sea on the west portion in the image), which were prominent in DT classification as depicted in Fig. 2. The above reported results are obtained with certain parameter settings and may not result in similar output when the parameters are altered or adjusted. At a regional scale, medium spatial resolutions such as Landsat TM/ETM+, Terra ASTER are most frequently used data. Eva et al., (2010) [18] demonstrates the usage of medium spatial resolution satellite imagery (Landsat-5 TM and SPOT-HRV) for monitoring forest areas from continental to territorial levels. Uncertainties involved in different stages of classification procedures influence classification accuracy, as well as the area estimation under different land use and land cover classes [19]. Understanding the relationships between the classification stages, identifying the probable factors that influence the accuracy and improving them are essential for successful image classification.