Non-rectangular thermal image crop & matching
4 visualizaciones (últimos 30 días)
Mostrar comentarios más antiguos
How to create rectangular IR image and contour with a false color map from the measured non-rectangular image and data from an Excel sheet?
I want to transform the results of the IR image and temperature data from the Excel sheet into a rectangular format. The interrogation region from the measuring devices is non-rectangular. Procedures are attached in the PDF file. I need help matching rectangularized IR images having the original color map with Excel data.
0 comentarios
Respuesta aceptada
DGM
el 5 de Mzo. de 2024
Editada: DGM
el 5 de Mzo. de 2024
There are two CSV files. One is readable. The other is a duplicate with broken formatting. It should have simply been deleted. There's no point trying to fix it, because the underlying data is exactly identical.
Colormap estimation from a crappy JPG. Data estimation from pseudocolor images. Perspective correction. We've done this before.
% read the file
inpict = imread('thermal.jpg');
% extract a noisy estimate of the colormap
CT0 = imcrop(inpict,[568.51 49.51 30.98 379.98]);
CT0 = reshape(mean(im2double(CT0),2),[],3);
CT0 = flipud(CT0);
% the colorscale limits specified in the image
Trange = [22.82 45.42];
% estimate T from a JPG
datafromjpg = rgb2ind(inpict,CT0,'nodither');
datafromjpg = rescale(datafromjpg,Trange(1),Trange(2));
% these are the coordinates of the box corners
% you can get these using getpts() or impixelinfo() or datatips
boxm = [82 19; 510 25; 488 464; 81 426]; % [x y]
% assert that this is where they're supposed to be
% any coordinates that define a rectangle
boxf = [1 1; 512 1; 512 512; 1 512]; % [x y]
% transform/crop the image
TF = fitgeotrans(boxm,boxf,'projective');
outview = imref2d([512 512 3]);
datafromjpg = imwarp(datafromjpg,TF,'fillvalues',255,'outputview',outview);
subplot(1,2,1)
imagesc(datafromjpg)
colormap(parula(256))
cb = colorbar('southoutside');
cb.Ticks = clim();
subplot(1,2,2)
histogram(datafromjpg)
The CSV file appears to be the sensor data from the same camera, but without the annotations and destructive JPG compression. It has the same geometry, so the same transformations apply.
% read the file
datafromcsv = readmatrix('good-vip10t,single=IMG_20231219_214630_TEMP.csv');
% these are the coordinates of the box corners
% you can get these using getpts() or impixelinfo() or datatips
boxm = [82 19; 510 25; 488 464; 81 426]; % [x y]
% assert that this is where they're supposed to be
% any coordinates that define a rectangle
boxf = [1 1; 512 1; 512 512; 1 512]; % [x y]
% transform/crop the image
TF = fitgeotrans(boxm,boxf,'projective');
outview = imref2d([512 512 3]);
datafromcsv = imwarp(datafromcsv,TF,'fillvalues',255,'outputview',outview);
figure
subplot(1,2,1)
imagesc(datafromcsv)
colormap(parula(256))
cb = colorbar('southoutside');
cb.Ticks = clim();
subplot(1,2,2)
histogram(datafromcsv)
As to whether those two images are indeed taken at the same time, or whether there's some other explanation for the largest chunk of the error, nobody will ever know.
6 comentarios
DGM
el 11 de Mzo. de 2024
I actually don't have any references for thermal imaging. @Image Analyst might know of some good sources, but I'm not the resident professional image processing guy here. Everything I say is based on a casual history of image manipulation in MATLAB and a background in general engineering principles.
That said, you might want to see how much technical documentation your camera manufacturer has available to owners. You appear to have a relatively high-end camera, so with that price might come some better documentation than your typical consumer-market offering.
Más respuestas (1)
Garmit Pant
el 5 de Mzo. de 2024
Hello bk park
From what I gather, you are trying to read data from a CSV file and then converting the non-rectangular interrogation region of the temperature measured to a rectangular shape.
This can be achieved by reading the data and first preprocessing it into a usable form. Further, you’d need to isolate the non-rectangular interrogation region using some threshold. Then, the non-rectangular region can be changed to a rectangular shape using padding with 0’s. Another method can be data interpolation. The following code snippet shows how to extract usable data, then the interrogation region and then a rectangular region using both padding and interpolation.
% Read the table from a CSV file
tbl = readtable("good-vip10t,single=IMG_20231219_214630_TEMP_01title.csv");
% Initialize an empty array to store processed temperature data
b = [];
% Loop through each row of the 'Tempeature_IR' column in the table
for i = 1:size(tbl.Tempeature_IR,1)
% Extract the temperature string for the current row, removing quotes
a1 = tbl.Tempeature_IR{i};
a1 = a1(a1~='"');
% Split the string into individual temperature values
a1 = strsplit(a1);
% Convert the temperature strings to numeric values
a1 = str2double(a1);
% Append the numeric temperature values to the array 'b'
b = [b; a1];
end
% Remove the last column from 'b'
b = b(:,1:end-1);
% EXTRACTING INTERROGATION REGION USING THRESHOLDING
% Filter columns of 'b' to keep only those where the first row value is >=
% 40. You can change the threshold.
colsToKeep = b(1, :) >= 40;
result1 = b(:, colsToKeep);
% Initialize the size of the 'result1' matrix
[rows, cols] = size(result1);
% Iterate through each row of 'result1'
for i = 1:rows
% Find the last index in the row where the value is >= 38
idx = find(result1(i, :) >= 38, 1, 'last');
% If such an index is found, set all subsequent elements in the row to 0
if ~isempty(idx)
result1(i, idx+1:end) = 0;
end
end
% PADDING THE OUTSIDE REGION WITH 0's
% Create a mask for values in 'result1' that are >= 38
mask = result1>=38;
% Initialize 'maskedresult' as a copy of 'result1'
maskedresult = result1;
% Set all values in 'maskedresult' that meet the mask condition to 0
maskedresult(mask)=0;
% INTERPOLATION
% Create a meshgrid for the coordinates of 'maskedresult'
[x, y] = meshgrid(1:size(maskedresult,2), 1:size(maskedresult,1));
% Identify indices of zero and non-zero values in 'maskedresult'
zeroInd = find(maskedresult == 0);
nonZeroInd = find(maskedresult ~= 0);
% Extract coordinates and values for interpolation
xq = x(zeroInd); % x-coordinates of zeros
yq = y(zeroInd); % y-coordinates of zeros
xv = x(nonZeroInd); % x-coordinates of non-zeros
yv = y(nonZeroInd); % y-coordinates of non-zeros
v = maskedresult(nonZeroInd); % non-zero values
% Use 'scatteredInterpolant' for interpolation with linear interpolation
% and nearest extrapolation for out-of-bound points
F = scatteredInterpolant(xv, yv, v, 'linear', 'nearest');
% Interpolate values for positions that were zero in 'maskedresult'
interpolatedValues = F(xq, yq);
% Replace zero values in 'maskedresult' with the interpolated values
maskedresult(zeroInd) = interpolatedValues;
% Display the matrix after interpolation
disp('Matrix after interpolation:');
disp(maskedresult);
The above operation give the following heatmaps:
Picture 1: Padded with 0’s
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1635081/image.png)
Picture 2: Interpolated.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1635086/image.png)
For further understanding, refer to the links to the MATLAB documentation given below:
- “scatteredInterpolant ” function- https://www.mathworks.com/help/releases/R2022b/matlab/ref/scatteredinterpolant.html
I hope you find the above explanation and suggestions useful!
Ver también
Categorías
Más información sobre Image Sequences and Batch Processing en Help Center y File Exchange.
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!