Quantitative and Qualitative Measurements of Planned and Unexpected Change using Remotely Sensed Imagery

Monitoring of planned change in objects is a fundamental responsibility of mapping agencies, which are increasingly utilising remotely-sensed imagery as a data source. In contrast, a series of recent devastating natural disasters has led to a demand for rapid detection and mapping of damage to populated areas. Change detection using remote sensors with high spatiotemporal resolutions can provide data quickly and efficiently to permit damage assessment and focus search, rescue and relief missions as well as provide a readily available data source for mapping organisations. Much change detection undertaken after unexpected events is achieved through visual interpretation of images and provides little quantitative data on building or object change. This paper proposes a novel modular framework, characterised by a pilot study of the Buncefield Oil Depot, which abstracts objects from image space into scene space. A simulated image is created from circles identified by a Hough Transform technique. The circles, representing fuel drums, are compared spectrally with a remotely-sensed image captured before the change event and an error image generated and minimised by adding bias based on initial accumulator values, creating a scene model. Future research involves updating the scene model through a forward model to effectively fit the scene to the after image and automatically store the vectors and magnitude of change. Change ontologies will provide qualitative knowledge describing the nature of change, potentially increasing accuracy and speed of damage detection after planned or unexpected events.