Finding Precipitate Phases in Al6061 Optical Micrographs
Contents
Import Some Little Widgets
Normalize and adjust images.
if ~exist('normalize','var') | ~exist('adjust','var') % GIST raw rawurl = 'https://gist.githubusercontent.com/tonyfast/8a2bb4752e0cfc55c99f/raw/f706ad03b824c4e17776d012eefd0ec755d133e5/adjust_normalize.m' s = urlread( rawurl ); eval( s ); clear( 'rawurl','s') end
Load in Data
Load in two reference images to find the Center of Mass of precipitates in Aged Aluminum
files = dir( fullfile( '_data', ... 'Al6061*.tif' ) ); ct = 0; A = zeros( [ [2288 2048] 2] ); clear content for file = files' ct = ct + 1; A(:,:,ct) = imresize(imread( fullfile( '_data', file.('name') ) ), 1); % Extact metadata in file name s = strsplit( file.('name') , '_'); content( ct ) = struct( ... 'local', fullfile( '_data', file.('name') ), ... 'material', s{1}, ... 'processing', s{2}, ... 'temperature', s{3}, ... 'time', s{4}, ... 'direc', s{5}, ... 'unknown',s{6} ); end
Plot raw Data
close all; for ii = 1 : 2 ax(ii) = subplot( 1, 2, ii); imshow( normalize( A(1:size(A,2),:,ii) ) ); axis equal; shading flat; axis off; axis ij; tt = sprintf( '%s %s @ %s', content( ii).('material'), ... content( ii).('processing'), ... content( ii).('temperature')); title( tt ); end linkaxes(ax) colormap gray figure(gcf);
Smooth the Aged Data
Resize by a half and resize that by two for smoothing
close all AA = imresize( imresize( normalize(A(1: size(A,2),:,2)), .5), 2); imshow( AA ); title( 'Smoothed Data', 'Fontsize', 16 ) figure(gcf)
Warning: Image is too big to fit on screen; displaying at 25%
Gradients should indicate the phases
uses : <http://www.csse.uwa.edu.au/~pk/Research/MatlabFns/Spatial/derivative5.m>>
The compute the magnitude of the gradient around the phases of interest
close all T = AA; G = cell(1,3); % Derivative7 doesnt find large enough because it is using more spatial % information. [G{1}, G{2}, G{3} ] =derivative5( T, 'x','y','xy'); GG = sqrt( G{1}.^2 +G{2}.^2 ); imshow( 1-normalize(GG) ); title( 'Zoomed Gradient Magnitude (Dark Values indicate a gradient)', 'FontSize', 16 ) xlim([ 295.5309 551.5309]) ylim([ 443 700])
Warning: Image is too big to fit on screen; displaying at 25%
Find Locally Minimal Pixels
Erode the image to find the local minima for the precipiates.
The Gradient identifies the edges of the precipitate. Erode forces the potential center to have the local minima value
E = imerode( T, ones(3));
Glue and Scotch Tape
Find Centers of Precipitates
- Invert Original Image ( Precipitates of Interest become bright )
- Invert Gradient ( Homogenous Regions are Weighted Higher )
- Multiple Inverted Image with Inverted Gradient * The center of precipitates have no gradient * Centers will emerge as bright spots in the center of a potential precipitate because the gradient lowers the weight of the edges.
% Precondition Image with Gradient Q = normalize( ( 1 - normalize( GG ) ) ... .* (1 - normalize(T) ) );
Find Bright Spots in a Precipitate
P = Find_Peaks( Q, 'neighborhood',5, 'diff', false);
Threshold potential centers
These are the two key parameters to modify to change the segmentation.
threshmult = 1.5; cutoffpix = [mean(E(:)) - std(E(:)) * threshmult];
B = P & E < cutoffpix;
Find matrix positions of centers
[pid] = find(B); [xx,yy] = find(B);
Plot Segmentation
close all imshow(T); axis equal; shading flat; hold on plot3(yy,xx,E(pid),'cd') title( sprintf( '%i precipitates found.', numel( pid ) ) ) hold off figure(gcf); snapnow; xlim([ 280.0309 792.0309]) ylim([ 305.9960 817.9960]) title( sprintf( '%i precipitates found. ZOOMED', numel( pid ) ) ) snapnow;
Warning: Image is too big to fit on screen; displaying at 25%
Export Centers as JSON
% Create export structure precipitate = struct( 'center', [xx, yy], 'cutoff', cutoffpix, 'file', content(2).('local') ) % Write data fo = fopen(fullfile( '_data', 'precipitate_center_600F_2hrs.json'), 'w') fprintf( fo, '%s\n', savejson( [], precipitate) ); fclose(fo) return
precipitate = center: [12634x2 double] cutoff: 0.3230 file: '_data/Al6061_Aged_650F_2hrs_RD_005.tif' fo = 5 ans = 0