Perception of the real world involves several factors depending on the nature of the observed phenomena. Static or dynamic events require different acquisition modalities. Therefore, different aspects must be considered and managed: scale, resolution, acquisition speed, field of view, performance needs, accuracy, etc. Furthermore, multisensor systems are needed when different points of view are required such as in case of complex, large and cluttered environments. As a consequence, a variety of data are obtained such as: mono-dimensional signals, images, depth values, point clouds, thermographic signals, video sequences, and so on.

After acquisition, data must be processed in order to extract useful information. Raw data could drive inaccurate decisions as consistency, uniqueness, reliability, relevance, completeness are not guaranteed. Data processing or data manipulation is, therefore, a fundamental step that involves different aspects depending on the type of data, their dimension and their quality. The aim is to design and develop methodologies for registration, normalization, analysis and fusion of multisource and multimodal data. One important process in this context is the extraction of relevant features for use in model construction and in the subsequent decision making phase. Feature selection methodologies are developed in order to extract relevant information, to reduce dimensionality, to simplify models, to enhance generalization. Data mining problems are also considered to manage large amount of data and discover interesting patterns that can be used for further analysis and exploration.

The final stage includes the design and development of methodologies using Machine Learning and Artificial Intelligence algorithms, to generate data interpretation or to support decisions. ISP systems can operate autonomously or in cooperation with operators to drive better decisions and to control problems.