Abstract—Satellite image classification plays an important
role in remote sensing, where information of an object or
phenomenon is acquired from real time sensing devices, such as
satellites and spacecrafts. In satellite image classification, the
goal is to correctly classify vegetation, agriculture, water bodies,
urban and open areas. The features like mean, Euclidean
distance, RGB and slope values are extracted for each pixel in
the input image. The image is classified using back propagation
algorithm which reduces the misclassification that occurs in pixel
based classification. This paper emphasizes on the classification
of IRS 1-D LISS-III images using neural network, k- nearest
neighbor and k- nearest neighbor with subwindows. An
experimental comparison of neural network approach with
back propagation algorithm was made with other
considerations of the k-nearest neighbor and with subwindows.
The results show that the k-nearest neighbor with subwindows
has better overall accuracy and kappa coefficient when
compared to neural networks.
Index Terms—Image classification, neural networks, remote sensing, spatial data, classification accuracy.
The authors are with the Dept. of CSE, GNITS, Hyderabad-8, India (e-mail: email@example.com; firstname.lastname@example.org; email@example.com).
Cite:M. Seetha, K. V. N. Sunitha, and G. Malini Devi, "Performance Assessment of Neural Network and K-Nearest Neighbour Classification with Random Subwindows," International Journal of Machine Learning and Computing vol.2, no. 6, pp. 844-847, 2012.