Best Combination of Parameters via Mathematical Optimization

Greetings MATLAB Community,

The equation shown below is intended to be used for the detection of white pixels of the tennis court lines. Reference to this equation can be found in http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.59.3604

There are 3 parameters in the equations which are sigma_l, sigma_d and r. Based on this, I would like to inquire a few questions:

• Do I need to perform a certain manual computation on the equation before feeding it into the MATLAB optimization tool?
• Which optimization tool in MATLAB is most suitable for the computation/optimization of the combination of the parameters?
• Any reference to such equation and solution provided?

Thanks.

Ronald

[Walter Roberson]

I think that "r" is a "tau".

[rondesu]

Yes, it is a "tau". This variable denotes the width of the court lines.

Answer 1

[Alan Weiss]

Your objective function looks discontinuous. So according to the Optimization Decision Table you should use a Global Optimization Toolbox solver.

According to the Table for Choosing a Solver and Choosing Between Solvers for Nonsmooth Problems, for a nonsmooth problem you should probably try patternsearch first.

That said, if your objective function is really just 1 or 0, I don't understand what you are doing, why you want to optimize anything.

Good luck,

Alan Weiss

MATLAB mathematical toolbox documentation