Using multi-modal 3D contours and their relations for vision and robotics

Using surfaces and surface relations in an Early Cognitive Vision system

“…From this, we can infer as the general picture of visual processing in the human cortex (1) a deep hierarchical structure in which (2) the largest part of the processing is devoted to the extraction of a generic scene representation. The work in this paper is concerned with the development of an artificial visual system (which we have called earlier 'Early Cognitive Vision' (ECV) [1,67]) following these two principles. Functional design choices of the ECV system are motivated by the primate's visual system architecture and it has been used in applications for robotic and computer vision tasks under real-time constraints.…”

“…From that, uncertainties for higher level features are derived. The uncertainty on local levels has been used for temporal disambiguation (see [33]), and as well as to improve estimation processes of higher level attributes, (see, e.g., [1]). Uncertainties of higher level entities can be used in selection processes (as in, e.g., pose estimation, see Sect.…”

“…This paper describes our work on the extension of the hierarchical Early Cognitive Vision (ECV) system [1,67] (see Fig. 2, right stream)-which previously provided a visual hierarchy in the edge domain-by a parallel hierarchy in the surface domain (see Fig.…”

“…2). The descriptors are embedded in the spatial temporal context, allowing for disambiguation, as well as semantic reasoning (see, e.g., [1,33]). …”

“…The main contribution of this paper is to first describe a richer visual hierarchy in which a hierarchy (compared to [1]) in the edge domain is supplemented by a hierarchy in the surface domain and then to discuss aspects of this visual hierarchy for the application. Individual work making use of certain aspects of this hierarchical representation for specific applications has been published already before (see, e.g., [41,43,66]).…”