Scale Adaptive Kernelized Correlation Filter Tracker with Feature Fusion

Algorithm 2

The proposed tracking algorithm.

Inputs:

 : training image patch;

 : regression target;

 : test image patch;

 : initial target position;

 : initial target scale;

Outputs:

 : detected target position;

 : detected target scale;

Training stage:

Compute the Gaussian kernel correlation of x with

itself: using (3);

Compute coefficient using (2);

Position Detection:

Compute the response using (4);

Find the target position by maximizing ;

Scale Prediction:

Sample the new patch based on size and

resize it to ;

Find the target scale by computing the response

with (6);

Model Online Update:

Update the template using (5) according to

occlusion detection strategy.