Intelligent Vision Systems
DIT-PRJ-06-065
Status NOT active project
DISI role Coordinator
Project type Industrial Project
Dimension Trentino
Acquisition date 2006-07-13
Start date 2006-07-13
End date 2006-10-30
Project details
Project astract The project involves the study and the validation of techniques for the detection and evaluation of two and three dimensional image data obtained from emebedded "smart" imagers.<br/>In particular, an area image sensor with processing capabilities will be used to acquire the two dimensional data. By taking advantage of time-of-flight computation, or by using stereo vision techniques, a map of distances can also be obtained.<br/>These two kinds of information are used to detect the presence of specific items in the field of view of the smart camera; specifically, for this research, the goal is to count and possibly classify the people that are seen by the smart camera.<br/>The emebdded system that is at the core of the applications envisioned here has aldready been developed by NeuriCam as an embedded system running Linux.<br/><br/>One of the questions that will be addressed in this research is weather the kind of task at hand needs a real-time operating system to guarantee that no events are missed. Commercial real time operating systems are available and will be evaluated for their suitability and cost.<br/><br/>A second area of research is related to the possibility of "fusing" the 2D and 3D data to facilitate the people-counting task. Intuitively, by having access to the three-dimensional values of what is "seen" by the camera ought to help discriminate cases where multiple people are in field of view of the camera, a situation that can cause problems when only a two-dimensional image is available. <br/><br/>Finally, the problems stemming from the power consumption of the devices envisioned for the people counting tasks will be examined. It has been observed that many applications of people counting would benefit from a stand alone operation, where each smart camera is part of a "network of sensors" so that globally they provide an overall surveillance of a vast area. <br/>In these situations, it would be desirable to operate with limited power (for instance, from a battery) and thus one would like to maintain the sensor in a low power mode (stand-by) while there is nothing "interesting" happening in the field of view. The possibilities offered by a hierarchical approach to activation of higher levels of data acquisition and elaboration will be investigated.<br/><br/>
Keywords Image vision, people counting, emebedded system
Fundings 20000 €