An evolutionary algorithm for model-based pose estimation and tracking
Cláudio Rossi · 2005
This paper proposes an evolutionary algorithm-based procedure for the problem of pose estimation of moving 3D objects using 2D images. The procedure consists of looking for six position parameters, three for rotation and three for translation, such that the projection of a model best fits a set of points (vertices) extracted from the 2D image. The evolutionary algorithm keeps a population of candidate solutions, whose goodness is measured in terms of mean distance between model and image points, that follow the movement of the extracted points. Key features of the algorithm are speed and robustness with respect to noise on the input data. Experimental results conducted on synthetic images demonstrate the effectiveness of the proposed approach.