Omni-directional Visual System (OMNIVIEWS)

IST FET-Open Project No.1999-29017


PLEASE VISITE THE OFFICIAL OMNIVIEWS PAGE


Objectives

The main objective of the project is to integrate optical, hardware, and software technology for the realization of a smart visual sensor, and to demonstrate its utility in key application areas. In particular our intention is to design and realize a low-cost, miniaturized digital camera acquiring panoramic (360 deg) images and performing a useful low-level processing on the incoming stream of images. The key technologies are two: panoramic mirrors and space-variant CMOS visual sensors. By designing an appropriate matching between the curvature of the mirror and the geometry of the photosite array of the visual sensor the microcamera will deliver panoramic images immediately usable by a human observer that can be processed without previous mapping. These two technologies are an ideal match to obtain a small-size and low-cost device for application areas such as surveillance, navigation control in robotics and telepresence.

Description of the work

This project is aimed at demonstrating the feasibility of the approach by realizing an omni-directional visual camera based on the optical and electronic sensors already available. One of the crucial aspects of this project is to define the best match between the profile of the mirror and the distribution of the sensitive elements of the CMOS sensor. Another crucial aspect is the study and definition of the best match between image geometry resulting from the mirror/sensor coupling, and the visual measures required. For example, starting from a log-polar sensor, the optimal profile may be a conic mirror because, a proper match between the rate-of increase of the log-polar geometry and the angle of the cone, will produce panoramic images that: a) will demand less effort and, therefore, simpler and faster hardware structures to be processed, and b) will be directly usable by a human operator because of the proper match between the profile of the mirror and the geometry of the sensor array. On the other hand, conic mirrors seem not optimal for some classes of visual processing (e.g. stereo matching). Aspects of this kind need to be properly addressed taking into consideration technological, processing and application-derived constraints. More specifically we will build a prototype of an omnidirectional camera using already available components (i.e. mirrors and space-variant sensors) and we will test its performance in tasks such as motion detection (from a static position) and estimation of egomotion parameters from a mobile base. Experiments with direct viewing of the panoramic image by a human observer will also be performed to test the utility of the camera for telepresence applications.

Milestones and expected results

Contact Person

Prof. Giulio Sandini		tel. +39 010 3532779
DIST - University of Genova	fax  +39 010 3532148
Viale Causa, 13			e-mail: sandini@dist.unige.it
I-16145 Genova
Italy

Consortium

Participants		Role (Nation)	 	Principal Investigators
DIST			Coordinator (I) 	Giulio Sandini, Antonios Gasteratos 
CTU			Partner (CZ)		Tomas Pajdla
IST			Partner (P)		Jose' Santos-Victor