| Resumo: | This report contemplates feature tracking on image sequences using Computer Vision. To do so a stochastic method is used, the Kalman filter, allied to a global optimization method to obtain the best set of correspondences, and so to guarantee the global optimization of the tracking results. A management model is also included in this methodology to evaluate if each feature's tracking should be continued or not. By doing so, the proposed approach deals with temporary occlusion, permanent disappearance, appearance or reappearance of features, maintaining a controlled number of tracked features in each time instant, which reduces the computer cost to what is strictly necessary.The proposed methodology is here tested and validated with feature's tracking along synthetic and real image sequences.
|
|---|