常用的高光谱混合像元分解算法.rar_hyperOrthorectify_streetzpi_端元_高光谱_高光谱 分解

function hyperDemo2 % HYPERDEMO2 Demonstrates new functions in the hyperspectral toolbox clear all; close all; clc; %-------------------------------------------------------------------------- % Measurements and reflectance (input) files/directory % The data files are obtained from: http://aviris.jpl.nasa.gov/data/free_data.html % (the reflectance .rar file for Cuprite) dataDir = ['~' filesep 'Downloads' filesep 'f970619t01p02r02c']; rflFile = [dataDir filesep 'f970619t01p02_r02_sc04.a.rfl']; spcFile = [dataDir filesep 'f970619t01p02_r02.a.spc']; % Results (output) directory resultsDir = '~/Dropbox/Purdue Grad/Semester 2/AAE 590/Project/Simulation/results'; if ~isdir(resultsDir) mkdir(resultsDir); end %-------------------------------------------------------------------------- % Record outputs (not including figures) diary(sprintf('%s/log.txt',resultsDir)) dnt = fix(clock); % date and time fprintf('%d/%d/%d %d:%d:%d\n\n', dnt([2,3,1,4,5,6])) fprintf(' Reading data from %s \n in the directory %s.\n', rflFile, dataDir); fprintf(' Storing results in %s directory.\n', resultsDir); %% Read in an HSI image and display one band bndnum = 120; % Band Number tmp = hyperReadAvirisSpc(spcFile); bnd = tmp(bndnum); slice = hyperReadAvirisRfl(rflFile, [1 512], [1 614], [bndnum bndnum]); figure; imagesc(slice); axis image; colormap(gray); title(['Slice of HSI, at $\lambda=$',sprintf('%5.6g nm',bnd)],... 'Interpreter', 'Latex', 'FontSize', 14); print(gcf, '-r600', '-depsc', sprintf('%s/sampleSlice', resultsDir)); %% True Color Composite rgbBands = [31,20,12]; % RGB bands (default): [665.73, 557.07, 478.17] nm tColor = hyperTruecolor(rflFile, 512, 614, 224, rgbBands, 'stretchlim'); figure; imagesc(tColor); axis image title('Cuprite, Nevada. AVIRIS 1997 data.', 'Interpreter', 'Latex', 'FontSize', 14); print(gcf, '-r600', '-depsc', sprintf('%s/truecolor', resultsDir)); %% Read part of AVIRIS reflectance data file that we will further process M = hyperReadAvirisRfl(rflFile, [1 512], [1 614], [1 224]); % Read AVIRIS .spc file lambdasNm = hyperReadAvirisSpc(spcFile); % Resample AVIRIS image [h, w, p] = size(M); M2d = hyperConvert2d(M); desiredLambdasNm = 400:(2400-400)/(224-1):2400; fprintf('Resampling M for desired wavelengths...\n'); tic; M2d = hyperResample(M2d, lambdasNm, desiredLambdasNm); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); %% Remove low SNR bands and bands affected by water absorption % Examine the spectral profile of some random pixels (targets) M = hyperConvert3d(M2d, h, w, p); ntarg = 20; % number of targets htarg = randi(h, 1, ntarg); % h position of targets wtarg = randi(w, 1, ntarg); % w position of targets targets = zeros(ntarg, p); for idx = 1:ntarg targets(idx,:) = squeeze(M(htarg(idx), wtarg(idx), :)); end figure; set(0,'Units','pixels'); scnsz = get(0,'ScreenSize'); set(gcf, 'OuterPosition',[1,1,scnsz(3)/3,scnsz(4)],'PaperPositionMode', 'auto'); subplot(211); plot(desiredLambdasNm, targets, '.'); grid on; ylim([0,0.6]); title(sprintf('(a) %d Target Signatures (All Bands)', ntarg), 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = xlabel('Wavelengths [nm]'); ax(2) = ylabel('Reflectance'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); fprintf('Removing low SNR bands...\n'); % [rows,cols,vals]=find(sum(M2d,2)==0); % goodBands = [4:104 116:149 171:224]; goodBands = [4:104 116:135 137:149 174:224]; M2d = M2d(goodBands, :); p = length(goodBands); lambdasNm = desiredLambdasNm(goodBands); % Final hyperspectral data cube that will be used throughout this demo M = hyperConvert3d(M2d, h, w, p); fprintf('...Done\n'); % Plot new target signatures (only good bands now) targets = zeros(ntarg, p); for idx = 1:ntarg targets(idx,:) = squeeze(M(htarg(idx), wtarg(idx), :)); end subplot(212); plot(lambdasNm, targets, '.'); grid on; ylim([0,0.6]); title(sprintf('(b) %d Target Signatures (Good Bands)', ntarg), 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = xlabel('Wavelengths [nm]'); ax(2) = ylabel('Reflectance'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); print(gcf, '-r600', '-depsc', sprintf('%s/targets_spectra', resultsDir)); %% -------------------------------------------------------------------------- % Perform various endmember determination algorithms % Estimate number of endmembers in image. % q = hyperHfcVd(M2d, [10^-3]); % doesn't work because corr(M') returns % some NaNs, which means that there's no variance in some of the bands... q = 6; % number of endmembers to find % M2dnorm = hyperNormalize(M2d); % normalized [0-1] reflectance %% PPI fprintf('Performing PPI for endmember determination...\n'); tic; Uppi = hyperPpi(M2d, q, 1000); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); % Plot endmember signatures figure; plot(lambdasNm, Uppi, '.'); grid on; title('PPI Recovered Endmembers', 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = ylabel('Reflectance [0-1]'); ax(2) = xlabel('Wavelength [nm]'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); l = legend(cellstr(num2str((1:q)'))', 'Location', 'EastOutside'); a = get(l, 'children'); set(a(1:3:end), 'MarkerSize', 20); print(gcf, '-r600', '-depsc', sprintf('%s/endmmbrs-ppi', resultsDir)); %% N-FINDR fprintf('Performing N-FINDR for endmember determination...\n'); tic; Unfindr = hyperNfindr(M2d, q); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); % Plot endmember signatures figure; plot(lambdasNm, Unfindr, '.'); grid on; title('N-FINDR Recovered Endmembers', 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = ylabel('Reflectance [0-1]'); ax(2) = xlabel('Wavelength [nm]'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); l = legend(cellstr(num2str((1:q)'))', 'Location', 'EastOutside'); a = get(l, 'children'); set(a(1:3:end), 'MarkerSize', 20); print(gcf, '-r600', '-depsc', sprintf('%s/endmmbrs-nfindr', resultsDir)); %% AVMAX fprintf('Performing AVMAX for endmember determination...\n'); tic; Uavmax = hyperAvmax(M2d, q); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); % Plot endmember signatures figure; plot(lambdasNm, Uavmax, '.'); grid on; title('AVMAX Recovered Endmembers', 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = ylabel('Reflectance [0-1]'); ax(2) = xlabel('Wavelength [nm]'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); l = legend(cellstr(num2str((1:q)'))', 'Location', 'EastOutside'); a = get(l, 'children'); set(a(1:3:end), 'MarkerSize', 20); print(gcf, '-r600', '-depsc', sprintf('%s/endmmbrs-avmax', resultsDir)); %% AMEE fprintf('Performing AMEE for endmember determination...\n'); tic; Uamee = hyperAmee(M, q, 5); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); % Plot endmember signatures figure; plot(lambdasNm, Uamee, '.'); grid on; ylim([0,1]); title('AMEE Recovered Endmembers', 'Interpreter', 'Latex', 'FontSize', 14); ax(1) = ylabel('Reflectance [0-1]'); ax(2) = xlabel('Wavelength [nm]'); set(ax, 'Interpreter', 'Latex', 'FontSize', 12); l = legend(cellstr(num2str((1:q)'))', 'Location', 'EastOutside'); a = get(l, 'children'); set(a(1:3:end), 'MarkerSize', 20); print(gcf, '-r600', '-depsc', sprintf('%s/endmmbrs-amee', resultsDir)); %% -------------------------------------------------------------------------- % Create abundance maps from unmixed endmembers %% From PPI results: fprintf('Creating abundance maps from PPI endmember results...\n'); tic; abundanceMaps = hyperNnls(M2d, Uppi); abundanceMaps = hyperConvert3d(abundanceMaps, h, w, q); et=toc; fprintf('...Done. (~%1.3g seconds)\n',et); fprintf('Plotting and saving PPI abundance maps...\n'); for i=1:q if i==1; figure; end; clf; imagesc(abundanceMaps(:,:,i)); colorbar; axis image; title(sprintf('Abundance Map %d/%d', i, q), 'Interpreter', 'Latex'); print(gcf, '-depsc', '-r600', sprintf('%s/abund-ppi-%d', resultsDir, i)) end close(gcf); fprintf('...Done.\n'); %% From N-FINDR results: fprintf('Creating abundance maps from N-FINDR endmember results...\n'); tic; abundanceMaps = h