Multifeatures Based Compressive Sensing Tracking

Algorithm 2

Template update.

Input:

 (i) Newly tracking result 𝐠 and the corresponding observation 𝐳̂

 (ii) Target coefficient vector 𝐱_𝑡

 (iii) Template set 𝐴=[𝑇,𝐼]∈𝑅^{𝑛×(𝑛+𝑁𝑡)}, where 𝑇=[𝐭₁,𝐭₂,…,𝐭_𝑁𝑡]∈𝑅^{𝑛×𝑁𝑡}

 (iv) Bhattacharyya thresholds 𝜏₁, 𝜏₂, 1>𝜏₂>𝜏₁>0

(1) 𝑚←argmax_{1≤𝑖≤𝑁𝑡}𝑥_𝑖;

(2) if  𝜏₂>𝜌(𝐳̂,𝐭_𝑚)>𝜏₁  then

(3)   𝑛←argmin_{1≤𝑖≤𝑁𝑡}𝑥_𝑖;

(4)   𝐭_𝑛←𝐳̂;

(5) end

(6) Normalize every columns of  𝑇;

Output: Updated template set 𝐴=[𝑇,𝐼]