Cognitive vision research in ambient intelligence — i.e. in electronic environments that respond to the presence of people — is pivotal for advancing many fields, from robotics and security through to medicine and linguistics. The EU-funded COGNITIVE-AMI (Semantic and cognitive descriptions of scenes for reasoning and learning in ambient intelligence) project worked on furthering such research by extracting qualitative information from images and videos taken indoors. More specifically, the project team used computer vision to recognise objects, regions or movements, building on the data obtained to define concepts that preserve the properties of space. The team aimed to describe scenes, such as where a table ends and where a wall begins, in a more cognitive manner. To achieve its aims, the project team captured images and videos located on a robot inside a university building. It developed a model that extracts a logic and a narrative description of spaces using qualitative features such as shape, colour, topology, location and size. In doing so, it aimed to describe the location of the objects needed for a task, as well as to describe unknown objects through colour, shape or location in order to identify and name specific objects. A key project result in this respect was the development of a qualitative model for describing 3D objects based on depth and different perspectives. This involved defining, testing and developing related logic descriptions. The project team also articulated an approach to cognitively describe real 3D scenes in natural language. Another important project result included cognitive tests about creativity and its relationship with associations people have between linguistic and visual concepts. This led to the development of a computational method that can help measure creativity. Overall, the results will be useful in taking cognitive vision research in ambient intelligence to the next level. The project’s outcomes, along with illustrative graphs and renditions, have been published on the project website and are set to prove useful for other researchers in the field.
Cognitive vision, ambient intelligence, robotics, COGNITIVE-AMI, videos, 3D