How can I snap my data point to the border for a colormap/heatmap?

Hey everyone!

I am trying to do a color map/heat map of Asia continent. I have my values for my color map in terms of a 70x110 double array and I want to plot at a resolution of it at every 0.5 latitude and longitude.

I have tried the following:

load coastlines
X = [0.25:0.5:34.75] ; % latitude
Y = [65.25:0.5:119.75]; %longitude
Q = pcolor(Y,X,A);
Q.FaceColor = 'interp';
set(Q, 'EdgeColor','none')
H= colormap(flipud(jet));
hold on
coast = plot(coastlon, coastlat,'Color','k');
grid on;
colorbar
xlabel('Longitude')
ylabel('Latitude')

I am able to get a decent looking map but I realised that due to the resolution the colored box is out of the coast border and it is not presentable for me. Is there any solution to this? Also is there anyway to countries borders drawn as well?

Answer 1

[darova]

Here is the idea

• create new mesh for the contour

• interpolate values

[LEE WEI JIE DANIEL]

Hello @darova!

Thanks for sharing your code! I get the concept behind it but I am still unable to understand the variables you have in your m1.m file. Do you mind explaining it to me in more details?

For example: I am using my X value to be [65:0.5:120] and Y to be [0:0.5:35]. However my Z value is a mess which is 70x110 cell array of NaN and tabulated values. How would they fit into your x1s, x2s, y1s, y2s and z1s and z2s!

Sorry to trouble you and thanks once again!

[Image Analyst]

@LEE WEI JIE DANIEL, are you using the mapping toolbox (if so I can't help). If you are, you forgot to list it in your products list.

[LEE WEI JIE DANIEL]

@Image Analyst I used coastlines only for the boundaries (is it part of mapping toolbox?), the remaining are just pure data and numbers

[darova]

• How would they fit into your x1s, x2s, y1s, y2s and z1s and z2s!

Your x1,y1,z1 should be matrices to use interp2. Maybe use meshgrid to convert your vectors into matrices

    % vectors
X = [0.25:0.5:34.75] ; % latitude
Y = [65.25:0.5:119.75]; %longitude
    % matrices (can be used with interp2)
[X1,Y1] = meshgrid(X,Y);

Here is another example using initmesh (triangulation). It's more universal

clc,clear

r = rand(1,30)+1;
t = linspace(0,2*pi-0.1,30);        % unclosed contour
[x,y] = pol2cart(t,r);

gd = [2;length(x);x(:);y(:)];       % geometry description
dl = decsg(gd);                     % decomposition
[p,e,t] = initmesh(dl);             % triangulation

    % create structure for visualization
pp1.faces = t(1:3,:)';
pp1.vertices = p;
pp1.facecolor = 'green';
patch(pp1)

Note: contour for triangulation should be unclosed (first and last points are different)

[LEE WEI JIE DANIEL]

Hello @darova thanks once again! Could I trouble you once again

I tried using the new set of codes together with the m1.m file you gave in the previous comment. I managed to process till the following: (A is my 70 x 110 matrix)

X = [0.25:0.5:34.75] ; % latitude

Y = [65.25:0.5:119.75]; %longitude

% matrices (can be used with interp2)

[X1,Y1] = meshgrid(X,Y);

load coastlines

B = coastlon;

C = coastlat;

z2 = interp2(X,Y,A',X,Y);

cla

plot(B',C')

hold on

plot(B,C)

pcolor(Y,X,A)

plot(Y,X,'.-b')

hold off

I managed to generate this plot but its kind off like my previous plot I shown above. Here the snapping to the coast borders are still not happening and I twigged pcolor(Y,X,A) from the original which is supposed to be pcolor(C,B,Z2) because my Z2 yields no color matrix and its a 9685x1 filled with NaNs while your Z2 was a 5x20 with different values that were used.

Answer 2

[Image Analyst]

pcolor() does not show all the values as pixels/tiles. It leaves off the last row and column. Or, actually, it has the pixels as edges in the image, not as the center of the little colored tiles. If you want to plot lines going to the center of the colored boxes, then don't use pcolor(), use imshow() instead. Then put hold on and call plot().

[LEE WEI JIE DANIEL]

@Image Analyst

Thanks for sharing more about this but I want to have the lines be the boundaries of the colored boxes. I do not want the color to spill over the lines. Is there a way to do so?

[Image Analyst]

I think you'd need to add a half pixel offset so that the points lie inside the boxes.

[LEE WEI JIE DANIEL]

@Image Analyst Yup that would have the points inside the boxes but I will have white spaces in between the points and line. Is there a feature in matlab that would allow us to fill up the space.

[Image Analyst]

Not sure what you're seeing. I don't have the mapping toolbox. Can you demonstrate with a small array that you synthesize?

[LEE WEI JIE DANIEL]

@Image Analyst Here! This is with my current array and plotted image. In black is my coastlines and in color is my data. This data seem to not fit the coastline and some are before the coastline while some others spill over. I would like to fill up those whitespaces and cut away the excess.

[Image Analyst]

I can't do anything with this. How can I get this into MATLAB as an array?

[LEE WEI JIE DANIEL]

@Image Analyst

X = [0.25:0.5:34.75] ; % latitude

Y = [65.25:0.5:119.75]; %longitude

For the Z value, randi(70,110)

that would generate similar matrix of the colored background

for the lines of the coast. I use load coastlines (I dont know if this is a map toolbox)

[Image Analyst]

Sorry I can't help unless you supply data. I do not have coastlines.mat.