SKT-MOT and DyTracker: A Multiobject Tracking Dataset and a Dynamic Tracker for Speed Skating Video

Algorithm 2

KFDU (state update step at sate k).

Input: Observation
Observation noise covariance
Measurement occlusion degree
Predicted state
Predicted state covariance
The observation model
Output: Updated state
Updated state covariance
Step:
1
//Updating dynamically observation noise covariance
2
//Calculating corrected Kalman gain
3
//Based on K, fusing observation and predicted state
4
//Updating state covariance