Algorithm 1. MACF tracking algorithm

Input:  1: Image $I_t$.  2: Predicted target position $P{P}_t$ and scale $P{s}_t$ in previous frame.

Output:

1: Detected target position $P_t$ and scale $S_t$ in current frame.  2: Predicted target position $P{P}_{t+1}$ and scale $P{s}_{t+1}$ subsequent frame.

Loop:  1: Initialize the Translation model $A_{1,trans}$, $B_{1,trans}$ and Scale model $A_{1,scale}$, $B_{1,scale}$ in the first frame by Equations (3) and (4), and initialize the Confidence of the Squared Response Map $\mathrm{C}\mathrm{S}\mathrm{R}{\mathrm{M}}_0$ in the initial frame by Equation (19).  2: for $t\in \left[2,t_f\right]$ do.  3:  Position detection and prediction:  4:    Extract pending sample feature $Z_{t,trans}$ from $I_t$ at $P{P}_t$ and $P{s}_t$.  5:    Compute correlation scores $Y_{t,trans}$ by Equation (8).  6:    Set $P_t$ to the target position that maximizes $Y_{t,trans}$.  7:    Predict the position $P{P}_{t+1}$ of the target of subsequent frame by joint Equations (11) and (17).  8:  Scale detection and prediction:  9:    Extract pending sample feature $Z_{t,scale}$ from $I_t$ at $P_t$ and $P{s}_t$.  10:    Compute correlation scores $Y_{t,scale}$ by Equation (8).  11:    Set $S_t$ to the target scale that maximizes $Y_{t,scale}$.  12:    Predict the position $P{s}_{t+1}$ of the target of subsequent frame by joint Equations (11) and (17).  13:  Model update:  14:    Compute the Confidence of the Squared Response Map $\mathrm{C}\mathrm{S}\mathrm{R}{\mathrm{M}}_t$ in current frame by Equation (17).  15:    Compute the adaptive learning rate ${\eta }_t$ by Equation (18).  16:    Extract sample features $X_{t,trans}$ and $X_{t,scale}$ from $I_t$ at $P_t$ and $S_t$.  17:    Update motion parameters ($v_{x_t}$, $v_{y_t}$, $v_{s_t}$), ($a_{x_t}$, $a_{y_t}$, $a_{s_t}$) by Equations (9) and (10).  18:    Update Kalman filters by Equation (18).  19:    Update the translation model $A_{t,trans}$, $B_{t,trans}$ by adaptive learning rate ${\eta }_t$.  20:    Update the scale model $A_{t,scale}$, $B_{t,scale}$ by adaptive learning rate ${\eta }_t$.  21:  Return $P_t$, and $P{P}_{t+1}$, $P{s}_{t+1}$.  22: end for.