Neuro-Textural Classification of Indian Urban Environment

Experiments were conducted to see the effects of a set of factors on the Resilient backpropagation (Rprop) artificial neural network classification of an Indian urban environment using IRS-IC satellite data. Factors investigated were sample size, number of neurons in hidden layers and number of epochs. The effect of including texture information in the form of neighbourhood information and grey level co-occurance matrix (GLCM) features in the classification process has been explored. Statistically similar overall classification accuracy is achieved for Rprop and Gaussian maximum likelihood classification (GMLC). Investigations have revealed that a large sample size gave higher test accuracy; variation in number of neurons in hidden layer did not affect the overall classification accuracy significantly; lesser number of epochs resulted in higher overall test accuracy. Incorporation of texture information by both approaches improved classification accuracy in a statistically significant manner. number of training iterations etc. (Foody et al. 1995a, 1995b, Foody and Arora 1997). Most methods of classification use the grey scale values of a set of corresponding pixels taken from different spectral bands of the same scene to determine land use. However, a single ground cover usually occupies a region of neighbouring pixels and improved identification may be obtained by considering an entire region rather than a single pixel. The variability of grey values within the region can be taken into account together with the actual grey values. This variability constitutes the texture of the ground cover. Texture is a fundamental characteristic of image data and is often crucial to target discrimination (Woodcock and Strahler 1987). For spatially complex and spectrally mixed classes, the classification accuracy could improve if the spatial properties of classes were incorporated into the classification criterion (Lee and Philpot 1991). Texture methods are most appropriate under condition of high local variance like urban environments. Most of the studies reported for land use classification have used standard back propagation methods, which are based on gradient descent algorithms with inherent limitations (Kanellopoulos et al. 1992; Paola and Schowengerdt 1995a; Shaban 1999). A better algorithm Rprop (Riedmiller and Braun, 1993) that overcomes the disadvantages of gradient