Reference to non-existent field 'gammac'

[ubuntuslave]

After estimation, while attempting the "Calibration of two methods", I'm getting

Reference to non-existent field 'gammac'

Error in go_omni_calib_optim_iter3D (line 277) paramEst3D.gammac = paramEst.gammac;

It seems that the "Estimate Intrinsic method" is only putting gamma in the "gen_KK_est" as (1,1) and (2,2) as per the "border_estimate.m" shows (Line 121): gen_KK_est = [gamma 0 center_estimate(1);... 0 gamma center_estimate(2);... 0 0 1];

[ubuntuslave]

I added the following to "estallRT.m"

% Added by carlos
paramEst.gammac = [gen_KK_est(1,1); gen_KK_est(2,2)];
paramEst.cc = gen_KK_est(1:2,3);

This allowed me to move forward with the calibration procedure.

[zju-isee-cv]

Sorry for my careless mistake. I want to know do you have any mistake else, or do you get the estimation at last. And does the result good?

[ubuntuslave]

Hello, Dr. Xiang.

I'm wondering why from an almost perfect synthetic image of a a coaxial and symmetric configuration for a hyperbolic mirror, the center your toolbox finds is not close enough to the true value (assuming perfect synthetic optics). Mei's results are closer.

I'm using very hi-def images and patterns in 6 different positions generated with POV-ray. I was wondering if you ever did experiments with synthetic (ray traced) images? Perhaps, my problem is in the ray-tracer I'm using (POV-ray). At the moment, I don't have any other means of generating ground truth data. Also, I don't want to proceed any further with physical rigs until I evaluate my pseudo ground-truth synthetic images.

Any advice is extremely welcomed. I noticed that in your papers, you only work with analytical synthetic configurations (for simulations) and real experiments with physical rigs.

Carlos

On Aug 4, 2013, at 10:58 PM, zju-isee-cv notifications@github.com wrote:

Sorry for my careless mistake. I want to know do you have any mistake else, or do you get the estimation at last. And does the result good?

— Reply to this email directly or view it on GitHub.

[zju-isee-cv]

Hi, I have done synthetic experiments of different mirrors (hyperbolic, parabolic, sphere) with different misalignment. Also done some experiments with a real hyperbolic mirror with different misalignment. POV-ray is good enough for synthetizing here, I also use POV-ray.

I'm wondering why from an almost perfect synthetic image of a a coaxial and symmetric configuration for a hyperbolic mirror, the center your toolbox finds is not close enough to the true value (assuming perfect synthetic optics). Mei's results are closer.

Not very clear about how bad my toolbox's center is (a little?or more than a pixel). My experiments shows the center is similar with c.mei's under coaxial and symmetric configuration. And in fact our model and c.mei's model is equivalent under coaxial and symmetric configuration. Our model is better than c.mei's under misalignment.

You can try image here. mis0 is for coaxial and symmetric configuration, mis15 is for a misalignment of 15 degree.

[ccny-ros-pkg]

What's the size of the grid in mm?

[dogeyes]

50mm