(1) Configuration |
(a) Set the parameters HMS, HMCR, PAR, BWmin, BWmax and NI. |
(2) Initial population. |
(a) Build the harmony memory (HM) where each individual |
consists 4 random non-repeating indices from 1 to . |
(b) Compute homography (hypothesis ) by using the indices from . |
(c) Calculate the fitness value as the matching quality of the constructed homography |
considering the whole available data . Such fitness value is calculated by using a |
new objective function defined as: |
, (A) |
where represents the quadratic error produced by the th correspondence |
considering the hypothesis whereas is the penalty associated with the mismatch |
magnitude. Such error corresponds to the mismatch generated by the evaluation of . |
(3) Iterations . |
(a) Generate a new harmony (candidate solution) as follows: |
|
for ( = 1 to ) do |
if ( < HMCR) then |
Select randomly a number where |
= |
if ( < PAR) then |
where |
end if |
if |
|
end if |
if |
|
end if |
else |
, where |
end if |
end for |
(b) Compute homography by using the indices from . |
(d) Calculate the fitness value as the matching quality of the constructed |
homography considering the whole available data . Such fitness value is |
calculated by using the objective function described in (A). |
(e) Update the HM as if |
(4) Estimation result |
(a) The best estimation consists of the parameters computed by using the indices from |
the best element of HMin terms of its affinity, so that . |