Matlab 2017b

[Richard Moore]

Hi Davide,

I'm experiencing that the toolbox hangs with Matlab 2017b at the findCorners routine:

Computing center coordinates.

Iteration >>

Cpu spins but program doesn't seem to progress beyond this point. If I cntrl+c this is the call stack:

In lsqlin (line 381)

            ipqpdense(H, f, A, b, Aeq, beq, lb, ub, X0, flags,

            ...

In calibrate>omni_find_parameters_fun (line 208)

        [s_o,ob_resnorm,ob_residual,ob_exitflag,ob_output,ob_lambda]

        = lsqlin(PP,QQ,mono,mono_const,[],[],[],[],[],options);

In calibrate (line 115)

[RRfin,ss]=omni_find_parameters_fun(Xt, Yt, Xp_abs, Yp_abs, xc,

yc, RRfin, taylor_order_default, ima_proc);

In findcenter (line 60)

            [calib_data.RRfin,calib_data.ocam_model.ss]=calibrate(calib_data.Xt,

            calib_data.Yt, calib_data.Xp_abs, calib_data.Yp_abs,

            calib_data.ocam_model.xc, calib_data.ocam_model.yc,

            calib_data.taylor_order, calib_data.ima_proc);

No issues in 2016b. I'm running Win10 (1607).

cheers,

Richard

[Darren Xu]

In my test, the function lsqlin has a issue which calibrate.m uses in Matlab 2017b and later.

There has no problem in Matlab 2016a. (I didn't test in Matlab 2016b and 2017a).

You should install Matlab 2016a to run the calib program.

[leodoe...@googlemail.com]

if you convert the matrix PP to a sparse matrix, it works for me in R2019a

(still takes a minute or two, but not hours ...)

[s_o,ob_resnorm,ob_residual,ob_exitflag,ob_output,ob_lambda] = lsqlin(sparse(PP),QQ,mono,mono_const,[],[],[],[],[],options);

To understand why, read https://de.mathworks.com/matlabcentral/answers/319775-why-is-the-lsqlin-interior-point-algorithm-slower-in-r2017a-compared-to-r2016b