plot some nodes from an .txt file

Hi! I would like to hope if there is an easy way to plot only the outermost nodes (the red nodes in the figure).

I am leaving the filename.txt file representing the nodes.

Answer 1

[Mathieu NOE]

hello Alberto

here you are ; use function boundary with shrink factor = 1

data = readmatrix('filename.txt');
x = data(:,1);
y = data(:,2);

% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 1;
k = boundary(x,y,s);
plot(x,y, 'db', x(k), y(k), '-r')

[Alberto Acri]

Thanks for the reply @Mathieu NOE! However, I only wanted the plotting of those nodes (without the connecting line between the nodes). That is, to have only the red nodes in my figure and then report on the workspace the coordinates (x,y) of only the red nodes shown in the figure.

[Mathieu NOE]

hello

It was displayed as lines because this is how I asked in the plot command '-r'

plot(x,y, 'db', x(k), y(k), '-r')

but you can that to dots if your prefer

also the x and y values that already defined as x(k) and y(k) but if you prefer we can name them x_out and y_out

data = readmatrix('filename.txt');
x = data(:,1);
y = data(:,2);

% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 1;
k = boundary(x,y,s);
x_out = x(k);
y_out = y(k);

plot(x,y, 'db', x_out, y_out, '*r')

[Alberto Acri]

In case I have such a txt?

How can I do the same thing for the two geometries?

[Mathieu NOE]

hello again

here there is one "stupid" solution as we can see the two groups are separetd by a blank zone in the y range around y = 230

so you can tell which group of data we are dealing with depending of if it's above or below that line.

for more complex cases with multiple curves you may have to use more sophisticated tools like clustering or based on Image Processing like here :

detect closed shapes formed by several arrays of points - MATLAB Answers - MATLAB Central (mathworks.com)

"stupid" code for the time being :

% data = readmatrix('filename.txt');
data = readmatrix('filename_1.txt');
x = data(:,1);
y = data(:,2);
% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 1;

%upper curve
idx = y>235;
x_temp = x(idx);
y_temp = y(idx);
k = boundary(x_temp,y_temp,s);
x_out1 = x_temp(k);
y_out1 = y_temp(k);

%lower curve
idx = y<=235;
x_temp = x(idx);
y_temp = y(idx);
k = boundary(x_temp,y_temp,s);
x_out2 = x_temp(k);
y_out2 = y_temp(k);


plot(x,y, 'db', x_out1, y_out1, '*r', x_out2, y_out2, '*c')

[Mathieu NOE]

a slight improvement, we compute the y threshold value , but that works only if the two curves have a vertical separation > 0

% data = readmatrix('filename.txt');
data = readmatrix('filename_1.txt');
x = data(:,1);
y = data(:,2);
% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 1;

% find the y threshold that separates the two curves 
% NB : it works only if there is a sufficient gap in the y direction (here
% we take 5 units separation in y distance

yy = unique(sort(y));
ind = find(diff(yy)>5);
val_low = yy(ind);
val_upper = yy(ind+1);
threshold = (val_low+val_upper)/2; % let's take the average of the two

%upper curve
idx = y>threshold;
x_temp = x(idx);
y_temp = y(idx);
k = boundary(x_temp,y_temp,s);
x_out1 = x_temp(k);
y_out1 = y_temp(k);

%lower curve
idx = y<threshold;
x_temp = x(idx);
y_temp = y(idx);
k = boundary(x_temp,y_temp,s);
x_out2 = x_temp(k);
y_out2 = y_temp(k);


plot(x,y, 'db', x_out1, y_out1, '*r', x_out2, y_out2, '*c')

[Alberto Acri]

I thank you @Mathieu NOE. You have been a great help!

I would like to ask you for one more help. If I wanted to use this code for both files (filename.txt & filename_1.txt) how can I do it?

[Mathieu NOE]

well

I don't have yet a solution for detecting automatically how many closed curves you have

the only simple solution I can suggest is to plot the raw data then ask the operator how many closed curves he sees , then use that in the code to process

or if you have a way to name your data according to how many curves they have (if you know it a priori)

[Alberto Acri]

I will take your solution into consideration.

In the case where I have 4 curves, as in the 'filename_3.txt' case, how could I get the same results as above?

I tried with this code but there is some problem on identifying the parameters 'val_low' and 'val_upper'.

data = readmatrix('filename_3.txt');
x = data(:,1);
y = data(:,2);
s = 1;

gap = 1;
xx = unique(sort(x));
ind_x = find(diff(xx)>gap);
val_low_x = xx(ind_x);
val_upper_x = xx(ind_x + 1);
threshold_x = (val_low_x + val_upper_x)/2; % let's take the average of the two

% upper curve
idx_x = x > threshold_x;
x_temp_x = x(idx_x);
y_temp_x = y(idx_x);
k_x = boundary(x_temp_x,y_temp_x,s);
x_out1 = x_temp_x(k_x);
y_out1 = y_temp_x(k_x);

% lower curve
idx_x = x < threshold_x;
x_temp_x = x(idx_x);
y_temp_x = y(idx_x);
k_x = boundary(x_temp_x,y_temp_x,s);
x_out2 = x_temp_x(k_x);
y_out2 = y_temp_x(k_x);

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

yy = unique(sort(y));
ind_y = find(diff(yy) > gap);
val_low_y = yy(ind_y);
val_upper_y = yy(ind_y + 1);
threshold_y = (val_low_y + val_upper_y)/2; % let's take the average of the two

% upper curve
idx_y = y > threshold_y;
x_temp_y = x(idx_y);
y_temp_y = y(idx_y);
k_y = boundary(x_temp_y,y_temp_y,s);
x_out3 = x_temp_y(k_y);
y_out3 = y_temp_y(k_y);

% lower curve
idx_y = y < threshold_y;
x_temp_y = x(idx_y);
y_temp_y = y(idx_y);
k_y = boundary(x_temp_y,y_temp_y,s);
x_out4 = x_temp_y(k_y);
y_out4 = y_temp_y(k_y);


plot(x,y, 'db', x_out1, y_out1, '*r', x_out2, y_out2, '*r', x_out3, y_out3, '*r', x_out4, y_out4, '*r')

So I thought of separating the 4 curves into 4 quadrants with 4 for loops, but I don't have many ideas.

data = readmatrix('filename_3.txt');
x = data(:,1);
y = data(:,2);
s = 1;

% X-axis separation
X = 350; 

% Y-axis separation
Y = 240;

% Upper left quadrant
for x<X & y>Y
    

end

[Mathieu NOE]

This is a new code for this case

maybe with a little more work we can detect how many distinct / non overlapping curves (with always the constrain that x and y gaps > 0)

this is just a first try :

% data = readmatrix('filename.txt');
data = readmatrix('filename_3.txt');
x = data(:,1);
y = data(:,2);
% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 1;

% find the y threshold that separates the two curves 
% NB : it works only if there is a sufficient gap in the y direction (here
% we take 5 units separation in y distance

yy = unique(sort(y));
ind = find(diff(yy)>5);
val_low = yy(ind);
val_upper = yy(ind+1);
y_threshold = (val_low+val_upper)/2; % let's take the average of the two

% same approach to find x gap between left / right 
xx = unique(sort(x));
ind = find(diff(xx)>5);
val_low = yy(ind);
val_upper = yy(ind+1);
x_threshold = (val_low+val_upper)/2; % let's take the average of the two


%upper left curve
idx = x<x_threshold;
idy = y>y_threshold;
x_temp = x(idx & idy);
y_temp = y(idx & idy);
k = boundary(x_temp,y_temp,s);
x_out1 = x_temp(k);
y_out1 = y_temp(k);

%upper right curve
idx = x>x_threshold;
idy = y>y_threshold;
x_temp = x(idx & idy);
y_temp = y(idx & idy);
k = boundary(x_temp,y_temp,s);
x_out2 = x_temp(k);
y_out2 = y_temp(k);

%lower left curve
idx = x<x_threshold;
idy = y<y_threshold;
x_temp = x(idx & idy);
y_temp = y(idx & idy);
k = boundary(x_temp,y_temp,s);
x_out3 = x_temp(k);
y_out3 = y_temp(k);

%lower right curve
idx = x>x_threshold;
idy = y<y_threshold;
x_temp = x(idx & idy);
y_temp = y(idx & idy);
k = boundary(x_temp,y_temp,s);
x_out4 = x_temp(k);
y_out4 = y_temp(k);


plot(x,y, 'db', x_out1, y_out1, '*r', x_out2, y_out2, '*c', x_out3, y_out3, '*g', x_out4, y_out4, '*k')

[Alberto Acri]

Perfect!!! As always you have given me great solutions!

I'll try to see if I can create a unique code that can automatically detect how many curves there are and apply the above.

Consider that I have several images like these (see attachments) and it would be faster to have the computer do everything instead of applying a particular code one at a time.

If you have any updates on how to proceed that is always welcome!

[Mathieu NOE]

hello

I am still chasing the perfect solution but maybe I need more than just the built in boundary function

have to have a look at

Computational Geometry Toolbox - File Exchange - MATLAB Central (mathworks.com)

[Mathieu NOE]

hehe

finally got some luck...

found this excellent Fex submission :

DBSCAN Clustering Algorithm - File Exchange - MATLAB Central (mathworks.com)

had to do a little bit of rework of the m functions and introduce the boundary plot in it and tada !!!

it works for all provided data files

here's one example

edit / run the main.m file and let the magic do the work !

zip attached

[Alberto Acri]

Thank you very much! You have given me a lot of help!

[Mathieu NOE]

As always, my pleasure !

[Mathieu NOE]

I have slightly changed the m files (attached) so you have the boudary points available in the workspace (' out ' array)

the data is now an output of PlotClusterinResult_mod.m

%% Load Data
% X = readmatrix('filename.txt'); % it works !!
% X = readmatrix('filename_1.txt'); % it works !!
% X = readmatrix('filename_2.txt'); % it works !!
% X = readmatrix('filename_3.txt'); % it works !!
X = readmatrix('filename_4.txt'); % it works !!


%% Run DBSCAN Clustering Algorithm

epsilon = 5;
MinPts = 2;
IDX=DBSCAN(X,epsilon,MinPts);


%% Plot Results

out = PlotClusterinResult_mod(X, IDX);
title(['DBSCAN Clustering (\epsilon = ' num2str(epsilon) ', MinPts = ' num2str(MinPts) ')']);

% double check 'out' data (boundary points) are now in the workspace
figure
plot(X(:,1),X(:,2),'d');
hold on
plot(out(:,1),out(:,2),'*k');
hold off

[Alberto Acri]

Hello again @Mathieu NOE!

May I ask how to modify the code above considering the attached text files? The coordinates of the .txt files are different from those given at the beginning of the discussion.

I wanted to know if there is a possibility to highlight only the outermost nodes as it was done for the previous cases because with the files "test_9.txt" and "test_222.txt" I get a wrong result. In figure I show you the comparison. Thank you very much.

X = readmatrix('test_9.txt');
% X = readmatrix('test_222.txt');

%% Run DBSCAN Clustering Algorithm
epsilon = 5;
MinPts = 2;
IDX=DBSCAN(X,epsilon,MinPts);

%% Plot Results
out = PlotClusterinResult_mod(X, IDX);
title(['DBSCAN Clustering (\epsilon = ' num2str(epsilon) ', MinPts = ' num2str(MinPts) ')']);

% double check 'out' data (boundary points) are now in the workspace
figure
plot(X(:,1),X(:,2),'d');
hold on
plot(out(:,1),out(:,2),'*k');
hold off

[Mathieu NOE]

hello

ok, yes I noticed with the new files that the matlab function boundary that we use here may sometims not give the expected results.

I first tried to tweak the so called shrink factor but that does not give a robust method as you would have to tweak it for different data sets

Instead I tried to make a better boundary function, so here we have myboundary.m !! (see attachement, you need this new function is your path)

I have updated PlotClusterinResult_mod accordingly (in attachement too)

this is the result I get with test_9.txt

[Mathieu NOE]

FYI

attached are some results / comparison between boundary and myboudary for various txt files

NB that myboudary does not require any tweaking (like the shrink factor in boundary

clc
clearvars
close all

data = readmatrix('filename.txt');
% data = readmatrix('test_9.txt');
% data = readmatrix('test_222.txt');

x = data(:,1);
y = data(:,2);

%% solution with matlab "boundary"
% you have to tweak the shrink factor to get the right result (maybe)

% k = boundary(___,s) specifies shrink factor s using any of the previous syntaxes.
% s is a scalar between 0 and 1. Setting s to 0 gives the convex hull,
% and setting s to 1 gives a compact boundary that envelops the points.
% The default shrink factor is 0.5.
s = 0.9;
k = boundary(x,y,s);
x_out = x(k);
y_out = y(k);

figure(1),
subplot(1,2,1),plot(x,y, 'db', x_out, y_out, '*r')
title(['with matlab "boundary", s = ' num2str(s)])

%% solution based on "tight boundary" but applied in both X and Y directions
%https://fr.mathworks.com/matlabcentral/answers/299796-tight-boundary-around-a-set-of-points

[x_selec,y_selec] = myboundary(x,y);

subplot(1,2,2),plot(x, y,'db',x_selec, y_selec,'*r')
title('with "myboundary"')

[Alberto Acri]

Thank you for the response @Mathieu NOE. I had seen it in the other post as well! I also write here. Can I ask you for one last help? In the case of "particular" curves in attachment, the code with shrink_factor variation always leads to some small loss. Can you always get a good result as with other curves? Thank you very much!

clc;
clear;
close all;

%% Load Data
X = readmatrix('test_1.txt');
% X = readmatrix('test_2.txt');

%% Run DBSCAN Clustering Algorithm
epsilon = 5;
MinPts = 2;
shrink_factor = 0.9;
IDX=DBSCAN(X,epsilon,MinPts);

%% Plot Results
out = PlotClusterinResult_mod(X, IDX, shrink_factor);
title(['DBSCAN Clustering (\epsilon = ' num2str(epsilon) ', MinPts = ' num2str(MinPts) ')']);

% double check 'out' data (boundary points) are now in the workspace
figure
plot(X(:,1),X(:,2),'d');
hold on
plot(out(:,1),out(:,2),'*k');
hold off

[Mathieu NOE]

hello Alberto

I did another trial based on this FEX :

Boundary extraction (identification and tracing) from point cloud data - File Exchange - MATLAB Central (mathworks.com)

and comparison of the already 2 previous solutions

this is the new code with 3 approaches :

plot for test_2.txt

code (see test3solutions.m in zip attached)

% test txt files : 

% filename.txt                  
% filename_1.txt                
% filename_2.txt                
% filename_3.txt                
% filename_4.txt                
% test_222.txt                  
% test_9.txt
% test_1.txt                    
% test_2.txt                    



data = readmatrix('test_9.txt');
x = data(:,1);
y = data(:,2);

%% test 1 : with DB cluster and  boundary (matlab native function)
shrink_factor = 1;
epsilon = 5;
MinPts = 2;
IDX=DBSCAN(data,epsilon,MinPts);
[x_out1,y_out1] = PlotClusterinResult_mod(data, IDX, shrink_factor); % myboundary


%% test 2 : withDB cluster and   myboundary (my previous suggestion)
epsilon = 5;
MinPts = 2;
out = PlotClusterinResult_mod2(data, IDX); % myboundary


%% test 3 : with find_delaunay_boundary03_fig1 
% (from Fex : https://fr.mathworks.com/matlabcentral/fileexchange/60690-boundary-extraction-identification-and-tracing-from-point-cloud-data?s_tid=ta_fx_results ) 
Fd = 1.5; %Fd = dmax (max point to point distance)
[bids, E, Ne] = find_delaunay_boundary03_fig1(data,Fd); 


%% plot
figure
subplot(1,3,1),plot(x,y, 'db', x_out1, y_out1, '*r')
title(['DB cluster w/ boundary, s = ' num2str(shrink_factor)]);
subplot(1,3,2),plot(x,y, 'db', out(:,1),out(:,2),'*r')
title('DB cluster w/ myboundary');
subplot(1,3,3),plot(x,y, 'db'),hold on
title('find-delaunay-boundary03-fig1');

for ck = 1:numel(bids)
    plot(x(bids{ck}), y(bids{ck}), '*r')
end

[Alberto Acri]

Hello and thank you and again for your help @Mathieu NOE.

Two questions:

1) How can I save the matrix (rx2) of coordinates with the red *? I saw the cell "bids" but it contains a different number of rows (230x1 double and 210x1 double).

2) Is it possible to recover the coordinates that exist in the left figure but disappear in the right figure? You can see in the attached figure some nodes that disappear.

[Mathieu NOE]

hello Alberto

answers are in the code below

yes we can combine the two last options to get best of both solutions

hope it helps

clc
clearvars
close all

% test txt files : 

% filename.txt                  
% filename_1.txt                
% filename_2.txt                
% filename_3.txt                
% filename_4.txt                
% test_222.txt                  
% test_9.txt
% test_1.txt                    
% test_2.txt                    

data = readmatrix('test_1.txt');
x = data(:,1);
y = data(:,2);


%% test 2 : with DB cluster and   myboundary (my previous suggestion)
epsilon = 5;
MinPts = 2;
IDX=DBSCAN(data,epsilon,MinPts);
out = PlotClusterinResult_mod2(data, IDX); % myboundary


%% test 3 : with find_delaunay_boundary03_fig1 
% (from Fex : https://fr.mathworks.com/matlabcentral/fileexchange/60690-boundary-extraction-identification-and-tracing-from-point-cloud-data?s_tid=ta_fx_results ) 
Fd = 1.5; %Fd = dmax (max point to point distance)
[bids, E, Ne] = find_delaunay_boundary03_fig1(data,Fd); 
x3 = [];
y3 = [];
for ck = 1:numel(bids)
    x3 = [x3; x(bids{ck})];
    y3 = [y3; y(bids{ck})];
end

%% combine both datasets and keep unique pairs
out_both = [out; [x3 y3]];
out_both = unique(out_both, 'rows');

%% plot 1 (zoom)
figure
subplot(1,3,1),plot(x,y, 'db', out(:,1),out(:,2),'*r')
xlim([100 160])
ylim([240 280])
title('DB cluster w/ myboundary');
subplot(1,3,2),plot(x,y, 'db', x3,y3,'*r')
xlim([100 160])
ylim([240 280])
title('find-delaunay-boundary03-fig1');
subplot(1,3,3),plot(x,y,'db', out_both(:,1),out_both(:,2),'*r')
xlim([100 160])
ylim([240 280])
title('combined sets');


%% plot 2 (full range)
figure
subplot(1,3,1),plot(x,y, 'db', out(:,1),out(:,2),'*r')
title('DB cluster w/ myboundary');
subplot(1,3,2),plot(x,y, 'db', x3,y3,'*r')
title('find-delaunay-boundary03-fig1');
subplot(1,3,3),plot(x,y,'db', out_both(:,1),out_both(:,2),'*r')
title('combined sets');

[Alberto Acri]

Thank you @Mathieu NOE! The code works as I wished.

I have created a new question at this link about correct positioning of some coordinates. Since you gave me a lot of help with this question, maybe you can help me. :)

[Alberto Acri]

Hi @Mathieu NOE! Can I ask you again for your help? Using the attached files (the files you had attached previously), I noticed a "problem" with the file test_114_check.txt of which I also attach an image.

How can I do to extrapolate, even for this curve, only the outer coordinates?

I thank you for your help in advance!

[Mathieu NOE]

hello Alberto

happy new year !!

for your problem above, simply increase Fd for the second method until you get the expected results

file : code_v3.m

%% test 3 : with find_delaunay_boundary03_fig1 
% (from Fex : https://fr.mathworks.com/matlabcentral/fileexchange/60690-boundary-extraction-identification-and-tracing-from-point-cloud-data?s_tid=ta_fx_results ) 
% Fd = 1.5; %Fd = dmax (max point to point distance)
Fd = 3; %Fd = dmax (max point to point distance)