Tracking Features in Image Sequences with Kalman Filtering, Global Optimization, Mahalanobis Distance and a Management Model

This work addresses the problem of tracking feature points along image sequences. In order to analyze the undergoing movement, an approach based on the Kalman filtering technique has been used, which basically carries out the estimation and correction of the features' movement in every image frame. So as to integrate the measurements obtained from each image into the Kalman filter, a data optimization process has been adopted to achieve the best global correspondence set. The proposed criterion minimizes the cost of global matching, which is based on the Mahalanobis distance. A management model is employed to manage the features being tracked. This model adequately deals with problems related to the occlusion of the tracked features, the appearance of new features, as well as optimizing the computational resources used. Experimental results obtained through the use of the proposed tracking framework are presented.