MPhil Research

Abstract

Remote sensing offers efficient and reliable means to recognize the pattern of the real world and to provide source data for geographic information system. Supervised learning is traditionally used to extract the features from remotely sensed imagery data in order to develop land-use mapping. The classifiers generate irrelevant classes since the complexity of the real ground features and the parametric variability of the decision rules. To evade from this issue, this study is for a new approach of remotely sensed image classification. Spatial data which earned from priori knowledge can bound the signatures respective to the location of the pixels of input image. For the correlated case study for land-use mapping of tea plantation, several experiments are done through unsupervised and supervised image classifiers of ERDAS Imagine and ENVI image processing software applications for a one sample image.  Quick Bird-2008 Satellite imagery of Ganga Ihala Korale division in Kandy district, Sri Lanka is used in a subset of 5.73ha area. Although each approach generates different outputs the expected output values are not performed since inconsequent and compound classes.  In contrast Maximum Likelihood classifier shows the highest accuracy in confusion matrix, Mahalanobis Distance classifier reached the best accuracy of the manual interpretation and ground truth process. The research is focused to modify and implement a threshold scheme for the supervised learning algorithm with mahalanobis distance classifier. The hypothesis is to test that could the system consider only the signatures corresponding to the spatial boundaries and analyzing the probability of image pixels to be assigned to the classes. If the null hypothesis is rejected, pixels will be classified as the appropriate classes in the GIS layer which included spatial boundaries of land-use types.  That means GIS data can be used to increase the complexity and reliability of hyperplane. The purpose of this paper is to discuss the used supervised classification approaches in the study, methodology, results and the future work of hypothesis testing. (Last colored lines will be amended as the final output)

Keywords — Image Segmentation, Pattern Recognition, Remote Sensing, Supervised Classification.