Pulse Coupled Neural Networks for Automatic Change Detection at Very High Spatial Resolution

World population growth affects the environment through the swelling of the population in urban areas and by increasing the total consumption of natural resources. Monitoring these changes timely and accurately might be crucial for the implementation of effective decision-making processes. In this context, the contribution of satellite and airborne sensors might be significant for updating land-cover and land-use maps. Indeed, the recent commercial availability of very high spatial resolution visible and near-infrared satellite data has opened a wide range of new opportunities for the use of Earth-observing satellite data. In particular, new systems such as the latest WorldView-2 and WorldView1, characterized by the highest spatial resolution, now provide additional data along with very high spatial resolution platforms, such as QuickBird or IKONOS, which have already been operating for a few years. If on one side this makes available a large amount of information, on the other side, the need of completely automatic techniques able to manage big data archives is becoming extremely urgent. In fact, supervised methods risk to become unsuitable when dealing with such large amounts of data. This is even more compelling if applications dedicated to the monitoring of urban sprawl are considered. In these cases, the big potential provided by very high spatial resolution images has to be exploited for analyzing large areas, which would be unfeasible if completely automatic procedures are not taken into account. In this abstract, a novel neural network approach for the detection of changes in multi-temporal very high spatial resolution images is proposed. Pulse-coupled neural networks are a relatively new technique based on the implementation of the mechanisms underlying the visual cortex of small mammals. The visual cortex is the part of the brain that receives information from the eye. The waves generated by each iteration of the algorithm create specific signatures of the scene which are successively compared for the generation of the change map. The proposed method is completely automated since analyzes the correlation between the time signals associated to the original images. This means that no pre-processing, except for image registration, is required. Furthermore, PCNNs may be implemented to exploit at the same time both contextual and spectral information which make them suitable for processing any kind of sub-meter resolution images.