Comparison of mapping trees on farmlands using markov random field super-resolution mapping and maximum likelihood classifier
| dc.contributor.author | Gaisie-Essilfie, Franz Alex | |
| dc.date.accessioned | 2014-10-17T12:18:41Z | |
| dc.date.accessioned | 2023-04-21T07:37:27Z | |
| dc.date.available | 2014-10-17T12:18:41Z | |
| dc.date.available | 2023-04-21T07:37:27Z | |
| dc.date.issued | 2014-10-17 | |
| dc.description | A thesis submitted to the Department of Wildlife and Range Management, Kwame Nkrumah University of Science and Technology in partial fulfilment of the requirements for the degree of Master of Science in Geo-Information Science, 2014 | en_US |
| dc.description.abstract | Mapping trees is of prime importance due to their notable ability to absorb carbon dioxide from the atmosphere and store it away for extended periods. In view of this, there have been a number of efforts aimed at maintaining and improving tree stands, notable among them is the REDD+ programme. For such programmes to be effective, there is the need to monitor the rate of increase or loss of forest cover. This is traditionally done by forest inventory, a slow and rather costly process. With the recent advancements in satellite imaging, it is becoming increasingly possible to map features on the ground using remotely sensed data captured by imaging satellites in orbit around the Earth. Due to limitations of small image extent, data availability and in a number of cases, cost, it is not practical to use high resolution satellite images to map trees. Consequently, users often resort to coarser medium resolution images. In view of this challenge and the need to extract more and more information from images, there have been a number of techniques, broadly termed super-resolution mapping techniques, developed to map image pixels at a sub-pixel level. This study compared the relative accuracy of mapping using one such technique, Markov Random Field Super-Resolution Mapping (MRF-SRM), with the Maximum Likelihood Classifier (MLC), a pixel-based classification algorithm in mapping trees on farmlands in two tests; image classification accuracy and tree canopy identification accuracy. An ASTER image of a community in the Ejisu-Juabeng district of Ghana was classified using the MRF-SRM and the MLC techniques after which tests were conducted to compare the image classification accuracy and tree canopy identification accuracy using field data as reference. It was observed that the MRF-SRM technique yielded higher image classification accuracy (74.29%) than the MLC (65.71%). The test of the tree canopy identification accuracy yielded 17.96% and 34.88% for the MRF-SRM and MLC techniques respectively. The results obtained indicate that the MRF-SRM technique is more suitable for mapping large areas than MLC. MRF-SRM, however, less suitable for identifying small targets as compared to MLC. | en_US |
| dc.description.sponsorship | KNUST | en_US |
| dc.identifier.uri | https://ir.knust.edu.gh/handle/123456789/6618 | |
| dc.language.iso | en | en_US |
| dc.subject | Markov Random Field, | en_US |
| dc.subject | super resolution mapping | en_US |
| dc.subject | trees, | en_US |
| dc.subject | farmlands | en_US |
| dc.title | Comparison of mapping trees on farmlands using markov random field super-resolution mapping and maximum likelihood classifier | en_US |
| dc.type | Thesis | en_US |