Improvement Object Tracking Process Through Filtering Methods

One of the main problems in the object tracking process is the correct estimation of the target position along the entire trajectory. However, in the tracking process, there are variations between the estimated and real positions. So, this article aims to improve the estimation of the trajectory of an object. We treat such variations as Gauss-Markov Colored Measurement Noise (CMN). To mitigate colored noise, we use measurement differentiation and employ modified filtering methods for CMN, KF, and UFIR. The tests were carried out with experimental data obtained from a dataset. We take the measurements got in bounding boxes as a reference. We evaluate the performance of the filters through the parameters estimation error and precision. The modified UFIR showed better results with high precision and low estimation error under high values of coloration factor.