3D Object Tracking in RGB-D Images Using Particle Swarm Optimization

Abstract

Model based tracking techniques allow computing the pose of 3d objects without needing to use markers. In order to perform a precise tracking, these techniques have been using RGB-D sensors together with particle filters for evaluating several pose hypotheses of the object in a given frame from features such as points 3D coordinates, color and normal vector. This work presents a proposal to use of the particle swarm optimization method for allowing 3d object model based tracking. Experiments showed that the proposed method obtained precise results when compared to ground truth values and to state of the art techniques that perform object tracking from RGB-D images.