next up previous contents
Next: Building Spatial Choice Models Up: Computational Information Technology Previous: Neural Networks in Electrical

Learning Scene and Object Analysis

Researchers: Jouko Lampinen, Timo Kostiainen, Ilkka Kalliomäki, Toni Tamminen, and Aki Vehtari

The project is funded by TEKES and participating enterprises in the the USIX technology programme. The project started in June 2000 and is scheduled for three years.

The goal of the project is to develop an object recognition and scene analysis system that can locate and recognize the objects in the scene and analyze the 3D structure of the objects and the scene. The approach we study is based on combining elements from view-based and model based methods using full Bayesian inference. The objects are defined by prior models that are learned from example images. The 3D representation is based on eigen shapes, so that a linear combination of the base shapes produces the perceived shape. View based approach is used in matching the perceived image and the image due to the assumed 3D structure. There we use distortion tolerant Gabor-filter based matching, combined to prelearned aspect graphs: for each nominal view of an object, the set of vertices that define the visible shape (like the silhouette) is memorized, allowing fast matching of the pertinent feature locations. Slower rendering of the whole model will be used for learning the aspects graphs and matching the textures. Inference of the distribution of probable 3D shapes given the perceived image is carried out by Markov Chain Monte Carlo techniques (Metropolis, Gibbs and Hybrid Monte Carlo sampling).

  
Figure 4
Figure 4: Example of detail matching. Local features in the grid points in the left figure are matched to the image in the middle figure, using Gibbs sampling. In the nodes with yellow ring the probability of finding a matching detail is low. The right figure shows samples from the posterior distribution of the grid node locations, giving an idea of the accuracy of the detail matching.

  
Figure 5
Figure 5: Example of inferring the 3D shape. The right image shows a sample of probable shapes corresponding to the left figure, represented as a linear sum of the base shapes. The matching is not very accurate in the outline of the object, as the edge detectors in the system gave low confidence for the edge positions.


next up previous contents
Next: Building Spatial Choice Models Up: Computational Information Technology Previous: Neural Networks in Electrical
www@lce.hut.fi