GDEST 2008 Conference Sessions
Theme 3 - The African Data Stream
Multiscale remote sensing data segmentation and post-segmentation change detection based on logical modeling: results for cost-effective forestland cover change quantification with case of Mau Forest Complex in Kenya
Yashon Ombada Ouma, Moi University, Kenya
Quantification of forestland cover extents, changes and causes thereof are currently of global research priority. Remote sensing data plays a significant role in this exercise. However, in-situ and supervised-based forest dynamics mapping from remotely sensed data are limited by lack of ground-truth data collection logistics and spectral-only based methods respectively. In this paper, first results of a methodology to detect change/no-change based on unsupervised multiresolution image transformation are presented, as a cost-effective strategy for forest inventory data acquisition in the context of sustainable development are presented. The technique combines directional wavelet transformation texture and multispectral imagery in an anisotropic diffusion aggregation algorithm, using unsupervised self-organizing feature map neural network implementation. Using Landsat TM (1986) and ETM+(2001), logical-operations based change detection results for part of the Complex Mau forests in Kenya are presented. An overall accuracy for change detection of 88.4% and kappa of 0.8265 was obtained. The methodology is able to predict the change information a-posteriori as opposed to the conventional methods that require land cover classes a-priori, for change detection. Most importantly, the approach can be used to predict the existence, location and extent of disturbances within natural environmental systems. This system is to be used in the National Forestry Inventory Systems Project for applications in modeling of: climate change; national water resources budget; desertification and agricultural yield systems for food security.