Research Article
Multifeatures Based Compressive Sensing Tracking
Algorithm 1
Multifeature based compressive sensing tracking.
Input: | (i) Current visible frame and infrared frame | (ii) Particles | (iii) Particles Measurement matrix | (iv) Template set | (v) Preset parameter | (1) Normalize each column of ; | (2) Generate particles according to state transition distribution; | (3) for do | (4) Obtain target candidates and relevant candidate center according to ; | (5) Calculate RGB histogram , LBP of candidate in and intensity histogram of candidate in ; | (6) Stack and into an observation ; | (7) Get measurements ; | (8) Solving (17) to get coefficient vector ; | (9) Calculate residual ; | (10) Obtain observation likelihood ; | (11) end | (12) Resample the particles according to ; | (13) Using Maximum A Posterior (MAP) or Mean Square Error (MSE) to | estimate the state of current frame, which characterizes the current tracking target ; | (14) Get target observation corresponding to current tracking result ; | (15) Recalculate for by solving (17); | (16) Update target template ; | Output: | (i) Tracked target | (ii) Updated target dynamic state | (iii) Updated target template set |
|