Deyun Society: What are the target tracking algorithms of the extended module?

1.KCF: The tracker KCF uses the circular matrix around the target to collect positive and negative samples, and uses ridge regression to train the target detector. The matrix operation is successfully converted into Hadamad product of vectors, that is, point multiplication of elements, which greatly reduces the amount of operation, improves the operation speed, and makes the algorithm meet the real-time requirements.

2.2.MIL:TrackerMIL trains the classifier online to separate objects from the background; Multi-instance learning avoids the drift problem of robust tracking

OLB: Tracker Boosting Online Real-time Target Tracking Based on AdaBoost Algorithm. The classifier uses the surrounding background as a counterexample in the update step to avoid the drift problem.

4.MediaFlow: Tracker MediaFlow Tracker is suitable for very smooth and predictable motion, and objects can be seen in the whole sequence.

5.5.TLD:TrackerTLD decomposes long-term tracking tasks into tracking, learning and detection. A tracker tracks objects between frames. The detector locates all observed phenomena and corrects the tracker if necessary. The tracker learns to estimate the errors of the detector and updates them to avoid the repetition of these errors. The tracker can handle fast moving, partially occluded and lost objects.