2508 Object Pose Estimation Using Canonical Correlation Analysis
Details
This paper presents an appearance-based method for object pose estimation using single camera image. Canonical Correlation Analysis is introduced to build a compact appearance model and use it for pose estimation. In the approach, we first obtain a pair of training data set, i.e., object images and their pose parameters. The appearance model is given as the subspace spanned by the canonical vectors that maximize the correlation between images and poses. Pose parameters of currently observed image is predicted by finding the regression coefficient in this subspace. We also introduce the kernel methods to cope with the non-linearity lies in training data set. In the experiment, we have examined the applicability of our method for vehicle type classification using the images taken by a road monitoring camera. Pose performance of CCA and KCCA models is discussed through the experimental results.
