Egypt is a country with great cultural heritage including more than 10,000 officially listed sites that range from prehistoric monuments and pyramids to modern age buildings. In the 1980s, it became apparent that the conditions of many of these sites were worsening and that some of them were seriously affected. This is attributed to several factors such as urban growth, industrial activities resulting in damaging pollutants, agricultural reclamation and the expansion of tourism without clear site protection guidelines.

The Archaeological Site Conservation project, will develop a heritage conservation system based on deploying large number of wirelessly-connected and partially self-sufficient vision sensor nodes at the site to be monitored. The information captured by the sensors is used collaboratively in order to enable for ubiquitous observation and hence making possible a number of applications such as autonomous intelligent surveillance by detecting and profiling activities at the site and consequently reporting suspicious behaviour due to looting or breaking of imposed regulations. In addition, the proposed system will be used to quantify structural integrity of monuments at the site for early discovery of erosions and cracks. Figure 1 illustrates a typical deployment scenario for the Archaeological Site Conservation project.

Another project objective is to construct 3D model of monitored archaeological site in order to enable observing its structure over time and hence detecting any deterioration. To this end, the monument is approximately defined by the intersection of blob volumes emanating from each camera image plane where a blob of the monument, i.e. its projection in the camera image plane, is used to define a cone-like blob volume in 3D space. A shape that approximates target object is then obtained from the intersections of several blob volumes. Furthermore, using Image-Based Modelling and Rendering (IBMR) techniques, it is possible to enable interactive virtual inspection of the monitored site. Figure 2 shows multiple virtual cameras observing reconstructed model of a temple.

Archaeological Site Conservation will exploit recent machine learning methods for target classification and prediction of suspicious scenarios hence providing intelligent site monitoring based on sequence of patterns obtained from WVSN data. Target classification requires more capable sensor nodes and hence the task is usually assigned to upper tier nodes or computational hubs or both. Several techniques can be used for this task based on available computational resources, and hence division of labour between multiple upper tier nodes to allow for capable system within required budgets. For scenario analysis, the methodology to be used for predicting suspicious situations is based on applying machine learning techniques to binary image sequences and features obtained from the vision network. Due to the dynamic nature of acquired data, machine learning based on time series prediction will be used for detecting potential suspicious scenarios such as repeated human activities or vehicles outside pre-designated hours, within demarcated sensitive scene regions, or simultaneous coordinated activities. Learning paradigms, such as reinforcement learning and semi-supervised learning, will be used since they offer a lot of potential to deal with learning problems from large, non-stationary and imperfect data.

This project is being conducted by Mohamed El-Sayed, Abubakr El-Sebai, Adel Mortada, Dr Neamat El-Gayar and Dr Mohamed ElHelw. Project collaborators are Prof. Guang-Zhong Yang, Prof. Mohamed Kamel and Prof. Fathy Saleh (CultNat).