a class project for ESPM 271 - Advanced Remote Sensing
University of California at Berkeley
This study focused on an exploration of methods to study land cover and land cover change in a national park buffer zone area in southern Zambia. After applying standard image enhancement and registration techniques to a pair of landsat 5 and 7 images, I tested several change detection techniques including spectral unmixing with and without ground sample data, principal components analysis of difference images, differencing vegetation indices, and edge analysis. The principal components analysis of difference images provided the most satisfactory results in identifying pixels that experienced land cover change between 1989 and 2000. Future work will build upon these results to develop a change mask and classify the two images in order to generate a land cover transition matrix.
Downloadsouthern_zm_change_detection_study.pdf (3.1 Mb)
Return to: Home