Contextual classification of Cropcam UAV high resolution images using frequency-based approach for land use/land cover mapping case study: Penang Island

Abstract

Cropcam UAV provides GPS based digital images on demand and real time data with high temporal resolution throughout the equatorial region where the sky is often covered by clouds. The images obtained by the UAV system in this research were used to overcome the problem of unclear images obtained by the satellite and manned aircraft in our study area. Conventional classification methods commonly cannot handle the complex landscape environment in the image. The result of each image has often a salt and pepper appearances which are the main characteristic of misclassification. The objective of this study is to evaluate the land use/land cover features over Penang Island using contextual classification method based on the frequency-based approach. The technique was applied to the high resolution images in three bands collected from a digital camera equipped with the platform system to extract thematic maps. Contextual classifier that utilized both spectral and spatial information could be reduce the speckle error and improve the classification performance significantly. Four classes could be classified clearly within the study area, and a high accuracy was achieved in the classification process. In order to evaluate the performance of the classifier, nine different window sizes ranging from 3 by 3 to 19 by 19 with an increment are tested. The study revealed that the frequency based-contextual classifier is effective with the images used in this research compare with the satellite images and images collected from conventional manned platforms and could be used for land use/cover mapping for the small area of coverage.