Comparative model for classification of forest degradation
Loading...
Date
2015-03
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universiti Teknologi Malaysia
Abstract
The challenges of forest degradation together with its related effects have attracted research from diverse disciplines, resulting in different definitions of the concept. However, according to a number of researchers, the central element of this issue is human intrusion that destroys the state of the environment. Therefore, the focus of this research is to develop a comparative model using a large amount of multi-spectral remote sensing data, such as IKONOS, QUICKBIRD, SPOT, WORLDVIEW-1, Terra-SARX, and fused data to detect forest degradation in Cameron Highlands. The output of this method in line with the performance measurement model. In order to identify the best data, fused data and technique to be employed. Eleven techniques have been used to develop a Comparative technique by applying them on fifteen sets of data. The output of the Comparative technique was used to feed the performance measurement model in order to enhance the accuracy of each classification technique. Moreover, a Performance Measurement Model has been used to verify the results of the Comparative technique; and, these outputs have been validated using the reflectance library. In addition, the conceptual hybrid model proposed in this research will give the opportunity for researchers to establish a fully automatic intelligent model for future work. The results of this research have demonstrated the Neural Network (NN) to be the best Intelligent Technique (IT) with a 0.912 of the Kappa coefficient and 96% of the overall accuracy, Mahalanobis had a 0.795 of the Kappa coefficient and 88% of the overall accuracy and the Maximum likelihood (ML) had a 0.598 of the Kappa coefficient and 72% of the overall accuracy from the best fused image used in this research, which was represented by fusing the IKONOS image with the QUICKBIRD image as finally employed in the Comparative technique for improving the detectability of forest change
Description
Thesis (Ph.D (Remote Sensing))
Keywords
Geoinformation and real estate, Environment