From image sequences towards conceptual descriptions

Biological Models for Active Vision: Towards a Unified Architecture

“…There is a wealth of scene understanding systems roughly divided into four major streams: grammars [2][3][4], blackboard architectures [5,6], probabilistic models [7][8][9], and artificial intelligence methods based on ontologies and description logics [10][11][12]. Some systems perform active vision tasks by controlling cameras [13].…”

Context-Based Probabilistic Scene Interpretation

“…Compositional hierarchies have been employed for high-level scene interpretation by many researchers [1,2,3,4,5] with basically non-probabilistic (crisp) framebased representations, as commonly used in AI. Rimey [6] was the first to model compositional hierarchies with tree-shaped Bayesian Networks (BNs), requiring parts of an aggregate to be conditionally independent.…”

“…We therefore propose a probabilistic model, called Bayesian Compositional Hierarchy (BCH), where aggregates (i) are modelled by arbitrary JPDs, (ii) possess an external description abstracting from details about their parts, and (iii) depend on other aggregates only via the part-of relations of the compositional hierarchy. (1) = A (2) etc. The JPD of the complete hierarchy is given by P(A (1) .. A (N) ) = P(A (1) ) P(B 1…”

Incremental Natural Language Description of Dynamic Imagery

“…In this context, computer-based generic descriptions for complex movements become important. Those accessible in the image understanding literature have been surveyed in Nagel [1988a]. Two even more recent investigations in this direction have been published in Witkin et al [1988] (in particular Section D) and Goddard [1988].…”

“…The crucial links between the picture domain results and the natural language processing steps are provided by complex events, i.e., higher conceptual units capturing the spatio-temporal aspects of object motions. A complex event should be understood as aǹ event' in its broadest sense, comprising also notions like`episode' and`history' (see Nagel [1988a]). The recognition of intentions and plans (see Retz-Schmidt [1988]) is, however, outside the scope of this paper.…”