Appearance-Based Statistical Object Recognition by Heterogeneous Background and Occlusions

In this paper we present a new approach for the localization and classification of 2-D objects that are situated in heterogeneous background or are partially occluded. We use an appearance-based approach and model the local features derived from wavelet multiresolution analysis by statistical density functions. In addition to the object model we define a new model for the background and a function that assigns the single feature vectors either to the object or to the background. Here, the background is modelled as uniform distribution, therefore we need for all possible backgrounds only one density function. Experimental results show that this model is well suited for this recognition task.

[1]  Heinrich Niemann,et al.  Object Localization with Mixture Densities of Wavelet Features , 1998 .

[2]  Hermann Ney,et al.  Discriminative Training of Gaussian Mixtures for Image Object Recognition , 1999, DAGM-Symposium.

[3]  Joachim Hornegger,et al.  Statistische Modellierung, Klassifikation und Lokalisation von Objekten , 1996, Ausgezeichnete Informatikdissertationen.

[4]  Horst Bischof,et al.  Dealing with occlusions in the eigenspace approach , 1996, Proceedings CVPR IEEE Computer Society Conference on Computer Vision and Pattern Recognition.

[5]  Josef Pösl,et al.  Erscheinungsbasierte statistische Objekterkennung , 2002, Informatik Forschung und Entwicklung.

[6]  Hermann Ney,et al.  An Automatic Approach to Invariant Radiograph Classification , 2001, Bildverarbeitung für die Medizin.

[7]  Bernt Schiele,et al.  The Robustness of Object Recognition to View Point Changes Using Multidimensional Receptive Field Histograms , 1996 .