|
Common video tracking algorithms mainly include: centroid tracking algorithm (Centroid), multi-target tracking algorithm (MTT), correlation tracking algorithm (Correlation), edge tracking algorithm (Edge), phase correlation tracking algorithm (Phase Correlation), scene locking algorithm ( SceneLock) and Combined tracking algorithms.
|
| Centroid Tracking Algorithm (Centroid) |
|
This tracking method is used to track bounded targets such as airplanes, and the target is completely included in the field of view of the camera. Some preprocessing algorithms can be used for this tracking method: such as white heat (positive contrast) enhancement, black heat (negative contrast) ) enhancement, and histogram-based statistical (bipolar) enhancement.
|
 |
Correlation tracking algorithm (Correlation) |
|
Correlation can be used to track many types of targets and is very effective when the target is unbounded and the dynamics are not very strong. Typical application: The target is in the range of close range, and the target extends beyond the field of view of the camera, such as a ship.
|
 |
Edge Tracking Algorithm (Edge) |
|
Edge tracking is the most effective algorithm when the tracked object has one or more definite edges and uncertain edges at the same time. Typically a rocket launch has a well-defined leading edge, but the trailing edge is indeterminate due to the jet.
|
 |
Multi-target tracking algorithm (MMT) |
|
Multi-target tracking is used for bounded targets such as airplanes, ground vehicles, etc. They are completely within the tracking window. Small target tracking MMT in complex environment can give a better performance.
|
Phase Correlation Tracking Algorithm (Phase Correlation) |
|
The phase correlation algorithm is a very general algorithm and can be used to track both unbounded and bounded targets. It can give a good effect in complex environments (such as cars on the ground).
|
 |
Scene Lock Tracking Algorithm (SceneLock) |
|
This algorithm is specialized for tracking of complex scenes. Suitable for air-to-ground and ground-to-ground scenarios. This algorithm tracks multiple targets in the scene, and then estimates the global motion of the entire scene based on the motion of each point. Targets and positioning in the scene are automatically selected. When an existing tracker moves outside the camera's field of view, a new tracker can be automatically identified. The aiming point is initialized to a point in the scene, tracking starts, and the aiming line is positioned. In this mode, the position of objects in the scene can be continuously tracked and reported.
|
 |
Combined Tracking Algorithm (Combined) |
|
As the name implies, this tracking method is a combination of two tracking algorithms with complementary characteristics: correlation class algorithm + centroid class algorithm. It is suitable for scenarios where the target size, surface, and characteristics vary greatly (such as a boat traveling in rough seas).
|