pcolor issue with irregular grid

I have been spending way too much time and would appreciate your insight and comments.

I am using “pcolor” to plot my data on an irregular grid (regularizing it is not an option).

-when I plot the even rows of my X, Y, and data the figure is what I expect (figure 1).

-when I plot the odd rows of my X, Y, and data the figure is what I expect (figure 2).

-when I plot the all rows of my X, Y, and data the figure is NOT what I expect (figure 3).

**figure 3 should be similar to figures 1&2 (basically those 2 superimposed).

the data, X, and Y are attched.

Thanks again.

clear all; close all; clc

load('C:\Desktop\Test\X.mat');
load('C:\Desktop\Test\Y.mat');
load('C:\Desktop\Test\data.mat');

% replaceing zeros with NAN
X(X == 0) = NaN;
Y(Y == 0) = NaN;
data(data == 0) = NaN;


% choosing even rows
X1 = X(2:2:end,:);
Y1 = Y(2:2:end,:);
data1 = data(2:2:end,:);

% choosing odd rows
X2 = X(1:2:end,:);
Y2 = Y(1:2:end,:);
data2 = data(1:2:end,:);

figure(1)
h1 = pcolor(X1,Y1,data1);
set(h1,'edgecolor','none');

figure(2);
h2 = pcolor(X2,Y2,data2);
set(h2,'edgecolor','none');

figure(3);
h3 = pcolor(X,Y,data);
set(h3,'edgecolor','none');

[Mathieu NOE]

hi

have to figure out why pcolor is doing that

with imagesc I don't have the problem

clear all; close all; clc
load('X.mat');
load('Y.mat');
load('data.mat');
% replaceing zeros with NAN
X(X == 0) = NaN;
Y(Y == 0) = NaN;
data(data == 0) = NaN;
% choosing even rows
X1 = X(2:2:end,:);
Y1 = Y(2:2:end,:);
data1 = data(2:2:end,:);
% choosing odd rows
X2 = X(1:2:end,:);
Y2 = Y(1:2:end,:);
data2 = data(1:2:end,:);
figure(1)
h1 = imagesc(data1);
% set(h1,'edgecolor','none');
figure(2);
h2 = imagesc(data2);
% set(h2,'edgecolor','none');
figure(3);
h3 = imagesc(data);
% set(h3,'edgecolor','none');

[Amin]

Thanks for giving it a try. unfortunately, imagesc is out of question. pcolor(data) just like imagesc(data) will work if true coordinates are not included (they produce the image base on indices but the image will be much noisier because of duplicates). The problem is that imagesc cannot handle irregular X and Y and pcolor cannot handle all X and Y together when in principle it should (can only do even ones together or add ones together using the true X,Y).

cheers,

Answer 1

[Walter Roberson]

surf(X(1:2:end,:), 'edgecolor', 'none')

and rotate it a bit, and you can see that the bulk of it is a single titled plane.

surf(X(2:2:end,:), 'edgecolor', 'none')

is likewise a single titled plane.

Now

surf(X, 'edgecolor', 'none')

and you will observe a thick wedge.

So although even X and odd X individually form consistent planes, when you take both together they do not form a single plane. When used as coordinates, pcolor would have to try to compensate.

Note: pcolor is surf() followed by view(2), together with some logic that sets the Z coordinates of the surf to all 0.

[Amin]

Thank you for looking into it. I have also noticed the irregularities between two consecutive X-matrix rows (row 1 and 2 for example) but there is not much that can be done about it as the coordinate detection system has some limitations. for example, its collection speed changes as it is collecting coordinate data (X, Y). I was under the impression that pcolor could handle such irregularities in grid but it seems not. Any suggestion of how it could be dealt with?

Cheers,

[Walter Roberson]

load X
load Y
load data
X(X == 0) = NaN;
Y(Y == 0) = NaN;
data(data == 0) = NaN;
mask = ~(isnan(X) | isnan(Y) | isnan(data));

F = scatteredInterpolant(X(mask), Y(mask), data(mask));

%trim down to useful part of data
[~, idx] = max(X(:));
[R, C] = ind2sub(size(X), idx);

lastC = C-2;    %some of the last couple of columns dip down
Xsub = X(:,1:lastC);
minx = min(Xsub, [], 'all');
maxx = max(Xsub, [], 'all');
Ysub = Y(:,1:lastC);
miny = min(Ysub, [], 'all');
maxy = max(Ysub, [], 'all');

xv = linspace(minx, maxx, lastC);
yv = linspace(miny, maxy, size(X,1));
[XG, YG] = ndgrid(xv, yv);

Z = F(XG, YG);

h = pcolor(XG, YG, Z);
h.EdgeColor = 'none';

In theory this takes all of the data into account. However, it interpolates data coordinates, and in theory it would work out something like taking the bottom (even-numbered) layer and shifting it to the left, and then merging the two sheets. The results look okay, but before using the [XG, YG, Z] data for serious analysis, you should think more about how the triangulation process works to make sure that you really are getting contributions from both half.

[Amin]

This seems to be working but like you said I need to look into if full contribution is there or not. Thanks again.