Notes about rewriting calculaMediaVarianciaHSV

[mtxslv]

Aparently, from line 85 to line 183 on file calculaMediaVarianciaHSV.m we have blob_colours file.

[mtxslv]

Here we go, up to this moment I already have the frames for calibration code (calcula_media_variancia_hsv_2). I plan to suit the calibration code such that I'd only need to apply blob_colours_2 inside it and get as far as possible.

I can already see I'll need something to arrange the blobs' colours so I can averave them properly. I'll need to be careful not to average different blobs colours (like summing red blob average colour from i-th frame to what I believe is the red blob average colour from i+1-th frame).

Anyway, blob_colours_2 need the following inputs:

• frame
• boundingbox
• ndetect
• wframe_log
• value_threshold
• saturation_threshold

As I said before, I have already got all the calibration frames from the video. The threshold values are manually selected. I still need the boundingbox,ndetect and wframe_log.

[mtxslv]

to test , use this: [media, variancia] = calcula_media_variancia_hsv_2(video, tempo_inicial, tempo_final, wbackg,Vrm, nanimais, mascara, minpix, maxpix, threshold, aviobj2, criavideodiff, tipsubfundo, caixa, l, c, colorida, cor, tipfilt , INTENSO)

[mtxslv]

We will use kmeans to find the blobs' mean. Here we have some instructions:

>> load('C:\Users\mateu\Documents\zebtrack\ZebTrack\test_variables_and_results\variaveis_calculaMediaVarianciaHSV.mat')
>> [media, variancia] = calcula_media_variancia_hsv_2(video, tempo_inicial, tempo_final, wbackg,Vrm, nanimais, mascara, minpix, maxpix, threshold, aviobj2, criavideodiff, tipsubfundo, caixa, l, c, colorida, cor, tipfilt , INTENSO);
>> [idx,C] = kmeans(media, 2,'Replicates',5);

I am using 5 kmeans replicates (run 5 times and get persistent cluster centroids) because it is MATLAB example. It is very important to use kmeans replicates, to guarantee the picked centroids won't be wrong.

To plot kmeans result, run this:

figure;
plot3(media(idx==1,1),media(idx==1,2),media(idx==1,3),'r.','MarkerSize',12)
hold on
plot3(media(idx==2,1),media(idx==2,2),media(idx==2,3),'b.','MarkerSize',12)
plot3(C(:,1),C(:,2),C(:,3),'kx','MarkerSize',15,'LineWidth',3) 
legend('Cluster 1','Cluster 2','Centroids','Location','NW')
title 'Cluster Assignments and Centroids'
hold off

[mtxslv]

value_threshold = 0.15 saturation_threshold = 0.5

[mtxslv]

Logical indexing: https://blogs.mathworks.com/steve/2008/01/28/logical-indexing/ https://www.mathworks.com/company/newsletters/articles/matrix-indexing-in-matlab.html

[mtxslv]

calcula_media_variancia_hsv_2.m is out now! Here you have how to test it:

[media,variancia] = calcula_media_variancia_hsv_2(video, tempo_inicial, tempo_final, wbackg,Vrm, nanimais, mascara, minpix, maxpix, threshold, aviobj2, criavideodiff, tipsubfundo, colorida, cor, 0.15,0.5,5)

If everything is fine, the result must be:

media =

   23.3055   25.0995   57.2719
   72.8047   22.3412   30.9006

variancia =

   25.5100   34.3325  143.4295
   24.3727   22.9286   39.1767

[mtxslv]

I test swaping thresholds. This made an error appear on variance concatenation (on line 37). To this example, means were null vectors.

this test was made while using var instead of cov (on line 37). Be careful while using such test!

[mtxslv]

I just changed variance by covariance (also changed variable var from array to cell). After running the code, the result must equal:

   24.3727   -0.4904    3.7522
   -0.4904   22.9286   20.8271
    3.7522   20.8271   39.1767

   25.5100   27.2358   56.3421
   27.2358   34.3325   63.7181
   56.3421   63.7181  143.4295

Each 3x3 grid is a variance matrix. They are associated to the centroids:

media =

   72.8047   22.3412   30.9006
   23.3055   25.0995   57.2719

[mtxslv]

just changed file, function call and returned variables' names:

• media -> centroids
• variancia -> cov_matrices
• calcula_media_variancia_hsv_2 -> calcula_centroids_cov_rgb To test it, just type:

[media,variancia] = calcula_centroids_cov_rgb(video, tempo_inicial, tempo_final, wbackg,Vrm, nanimais, mascara, minpix, maxpix, threshold, aviobj2, criavideodiff, tipsubfundo, colorida, cor, 0.15,0.5,5)