Arrivederci Roma Rosa
Chiara Solimini (UCB), Fabio Pacifici (UCB)

Logarithm of the ratios of multispectral bands
We process high resolution satellite images with neural network algorithms to extract change detection maps of urban areas. This image is a step in our classification method and represents a subarea of the city of Rome acquired by the QuickBird satellite. Each pixel of the image is a vector whose components are the logarithm of the ratios of the multispectral bands belonging to the two multitemporal images from 2002 and 2003. The image is also the input of a neural network that is trained with the above pixel values. With this method we are able to detect the changes occurring in an urban area, such as new buildings and roads constructed between the two acquisition dates, and also changed vegetated areas.

Traveling