What is MOT (Mutil-Object-Tracking)
MOTThe English full name is Mutil-Object-Tracking, and the Chinese name is multi-target tracking, literally 对多个目标对象进行跟踪. Specifically, we need to get the trajectory of the object we are more interested in in a video. The motion track here includes the detection frame and the category of the detected object as well as the number (ID) of the detected object
类别: In the case of pedestrian tracking and recognition, only one category is pedestrians. In some multi-category scenes, there will be pedestrians, vehicles
ID: It is basically fixed and generally does not change in the scene. If there is a change, it proves that there is an error.
In multi-target tracking tasks, everyone uses more and more practical paradigms Tracking-by-detection(detect first and then track). This paradigm is divided into two steps. 第一步First, the target detection algorithm is used to obtain the detection frame of the target. 第二步The Data associationmethod used will be different. The monitoring frames between the frames are matched to get ID.
Data association
The first frame in the video is detected first, and different ids are assigned to different detection frames. The frames of different colors in the figure represent different ids. Continue to detect the target from the second frame, but the detection process is not assigned an id, we use data asso