Abstract
In disaster monitoring, an on-board autonomous system, which can process the images to extract useful images is considered necessary to speed up the decision making process. However, the current degree of automation in image processing on-board remote sensing satellites is low. Most of the remote sensing satellites operate a 'store-and-forward' mechanism and do not have intelligent imaging capabilities. This is because of the satellites being equipped with low-performance computing capabilities to save power and due to reliability considerations. Also, the existing algorithms for disaster monitoring are not fully automated and require high-performance computing. The objective of this research project is to investigate the feasibility of implementing an intelligent system on board small satellites that can perform critical image processing tasks such as image registration and change detection. Investigation of current change detection and flood detection techniques is carried out in order to evaluate their performance against state-of-the-art on-board computing systems of small satellites. Other than the limited computing capabilities and possible errors that might occur due to the radiation in space; image registration, change detection and flood detection are very challenging tasks to be implemented on board small satellites, because these tasks need to process a pair of images (i.e. comparing them): sensed and reference images. This thesis proposes a robust change detection framework to reduce the problems of pixel-to-pixel comparison. Image tiling and fuzzy inference engine are introduced in the system as a method to overcome the problems caused by direct pixel-to-pixel comparison. The performance results show that the proposed method has a better performance than pixel-to-pixel methods. The proposed new on-board flood monitoring system is simulated to evaluate its performance and to verify if it is viable for the small satellite implementation. GPS reflectometry data is proposed to be used together with existing multispectral images to reduce the problem caused by cloud cover. Recommendations and requirements for a small satellite mission for monitoring flood are presented.