You are here

Computational Vision

  • Computational Vision Activity

Involved researchers

Related Projects

Visual information represents the richest source of information describing our surrounding environment. Standard imagers continuously deliver sequences of images, although only a small amount of information is used to perform any kind of visual task. This aspect turns into large data redundancy, with a consequent waste of energy for data acquisition, communication and processing. Often, it is not required to get accurate and high resolved images to be able to take a decision, which mostly deals with people monitoring in domestic rooms or moving around a dangerous or an off-limit zone. Infrastructures represent a big obstacle in the diffusion of such systems, which need to be placed all around the area to be monitored. This makes battery-operated systems an attractive solution, reducing both installation and maintenance costs. Present day wireless video systems consume a factor of about 100 too much power to enable long-term battery powered operation.

Application cases: 
Many applications could take advantage from an energy-autonomous vision system: minimization of the buildings energy consumption by determining room occupacy; improvement of the quality of life for elderly people by monitoring normal activity at their homes; safeguard cities by monitoring streets, buses, cars and train stations; protection of borders, long railways lines and high voltage power lines by identifing intruders in real-time.
Objectives: 
  • Development of ultra-low power vision sensor architectures with advanced computational capabilities aimed to extract useful image features, cutting down the processing resources required by the high-level algorithm.
  • Exploiting vision technology in distributed energy-autonomous sensory systems for environmental monitoring and sensor networks applications.
Research topics: