CVMT colloquia 2008

VBN is an increasingly important element in our daily lives. The information stored in VBN concerning publications and other activities are directly linked to the funding and other priviliges we enjoy. The Faculty is basing more and more of the distribution of funds on information we all are obligated to enter into VBN.

David Meredith has applied for our vacant assistant professorship with a focus on AI programming for game technology. As part of the evaluation of his application CVMT invited him to Aalborg to give an informal colloquium on his research work in music processing/analysis.

In celebration of publishing a chapter in a book intended for use in teaching alternative user interfaces we would like to present the contents. First we place our work within the scope of the entire book and then focus on our contribution. The focus was gesture interfaces including usage and design guidelines of such interfaces. Tehr is also a section on gesture interfaces in relation to multi-modal research. Furthermore, a case study of our own experiments is published on a website in conjunction with the book.

Title: Adaptive Mixture of Gaussians for Robot to Person Encounters

This paper introduces a new method for adaptive control of a robot approaching a person. The proposed method is based on a cost function centered in the person based on summation of four Gaussian distributions. The distributions may be adapted according to the behavior of the person by an introduced person evaluator method. The evaluator relies on three variables, the distance between the person and the robot, the area spanned by the velocity vector of the person and the vector between the person and the robot, and position of the person. The variables are used in a Case Based Reasoning (CBR) system that is trained to assess in which degree the person is interested or not in communicating with the robot. The outcome of the CBR system is used to adapt the cost function around the person, so the robot’s behavior is adapted to his or shes interested in communication. The methods is tested both in real world experiments and simulations. The results are promising and clearly indicate the potential of the proposed method.


Illustration of the cost function. a, a PersonIndicator of 1 interested in communication with a rotation angle theta = 45 and variance sigma^2 = 0.01. b, a PersonIndicator of 0.5 potential interested in communication, with a rotation angle theta = 0 and a variance sigma^2 = 0.15. c, a PersonIndicator of 0 not interested in communication, with a rotation angle theta = 0 and a variance sigma^2 of 1. The red solid line in a and b illustrates how a simulated robot would approach the person and the color scale bar the values of the cost function.

The main problem area for Augmented Reality is ensuring that the illumination of the virtual objects is continuously consistent with the illumination in the real scene. State of the art in the area typically requires the real scene illumination conditions to be captured as a High Dynamic Range environment map. The environment map is then used for shading and shadowing. Handling the real and the virtual shadows and their interaction is the single most difficult aspect. This paper presents a completely different approach to determining the illumination conditions in the real scene. Based on an assumption that the scene is outdoor we automatically detect shadows in the image and use this information to determine the ratio of sky irradiance to sun irradiance. We then present how to convert this information into radiance levels for both the sky and the sun. When combined with a computation of the Sun's position based on date, time and information about position on the Earth, we arrive at a full illumination model applicable for rendering virtual objects into real scenes.