A Qualitative Image Descriptor QIDL+ Applied to Ambient Intelligent Systems

摘要

A model for obtaining logic descriptions of real digital images in office scenarios is presented in this paper. The QIDL+ descriptor uses qualitative features of shape, colour, topology, location and it also includes the size feature, which shows is effectivity in scenes where the point of view of the observer and the camera is the same. QIDL+ is aimed at describing the location of target objects with respect to known or unknown objects in a scene. Known objects in the scene are identified by object detectors, while other object categories are inferred using its qualitative features. Moreover, the logic description provided is shown to be useful for reasoning about spatial locations, and the qualitative features obtained can be included in a narrative description for enhancing human-machine interaction. Tests have been carried out on office top desks scenarios to illustrate the performance of the approach and promising results are obtained.