A RECOGNITION NETWORK MODEL-BASED APPROACH TO DYNAMIC IMAGE UNDERSTANDING

In this paper, we present definitions for a dynamic knowledge-based im age understanding system. From a sequence of grey level images, the sy stem produces a flow of image interpretations. We use a semantic netwo rk to represent the knowledge embodied in the system. Dynamic represen tation is achieved by a hypotheses network. This network is a graph in which nodes represent information and arcs relations. A control strat egy performs a continuous update of this network. The originality of o ur work lies in the control strategy: it includes a structure tracking phase, using the representation structure obtained from previous imag es to reduce the computational complexity of understanding processes. We demonstrate that in our case the computational complexity, which is exponential if we only use a purely data-driven bottom-up scheme, is polynomial when using the hypotheses tracking mechanism. This is to sa y that gain improvement in computation time is a major reason for dyna mic understanding. The proposed system is implemented; experimental re sults of road mark detection and tracking are given.