High-level and low-level computer Vision: Towards an integrated approach

Abstract

The term perception refers to the means by which information acquired from the environment via the sense organs is transformed into experiences of objects, events, sounds, tastes, etc. In this paper, we shall focus on the problem of object perception dealing exclusively with the visual modality. More precisely we will present a system, SVL — Symbolic Vision Lab, which is a development environment for object perception algorithms. SVL is mainly devoted to symbolic computation and can exploit for low level tasks a massively parallel computer, such as the general purpose Connection Machine, or an ad hoc specific VLSI architectures whose efficiency can be simulated in advance on the Connection Machine itself.

This work was supported by CNR-Progetto Finalizzato Trasporti-Prometheus under contract n.89.01458.93 and n.89.01446.93

This work has been done in the frame of the PROMETHEUS project, a European effort set up to increase the safety of road traffic. In this case the medium-term goal of the activity is to prove the feasibility of a computer vision approach to help the driver task. As always in the case of automotive electronics, cost factors and real-time constraints are key points for the adoption of a solution. The project described in this paper is a contribution in this direction: SVL is devoted to study and verify, using AI technologies, algorithms which could be subsequently implemented on specialized low-cost hardware. The Connection Machine CM-2, connected via Ethernet with the SVL host, is devoted to simulate specialized hardware before its actual implementation.