Tracking and data association
Dynamic 3D Models with Local and Global Deformations: Deformable Superquadrics
IEEE Transactions on Pattern Analysis and Machine Intelligence
Closed-Form Solutions for Physically Based Shape Modeling and Recognition
IEEE Transactions on Pattern Analysis and Machine Intelligence
A review of statistical data association for motion correspondence
International Journal of Computer Vision
Modeling a dynamic environment using a Bayesian multiple hypothesis approach
Artificial Intelligence
IEEE Transactions on Pattern Analysis and Machine Intelligence
Autonomous Exploration: Driven by Uncertainty
IEEE Transactions on Pattern Analysis and Machine Intelligence
View planning for automated three-dimensional object reconstruction and inspection
ACM Computing Surveys (CSUR)
| Hi-index | 0.00 |
The primary intent of this work is to present a method forsequentially associating three-dimensional surface measurementsacquired by an autonomous exploration agent with models that describethose surfaces. Traditional multiple-viewpoint registrationapproaches are concerned only with finding the transformation thatmaps data points to a chosen global frame. Given a parts-basedobject representation, and assuming that the view correspondence canbe found, the problem of associating the registered data with thecorrect part models still needs to be solved. While traditionalapproaches are content to group segmented data sets thatgeometrically overlap one another with the same part, there are caseswhere this causes ambiguous situations. This paper addresses themodel-data association problem as it applies to three-dimensionaldynamic object modeling. By tracking the state of part models acrosssubsequent views, we wish to identify possible events that explainmodel-data association ambiguities and represent them in a Bayesianframework. The model-data association problem is therefore relaxedto allow multiple interpretations of the object's structure, eachbeing assigned a probability. Rather than making a decision at everyiteration about an ambiguous mapping, we look to the future for theinformation needed to disambiguate it. Experimental results arepresented to illustrate the effectiveness of the approach.