hyperspecreleasesmatlab_hyperspectral_toolbox_v0.07.zip

clear; close all; clc; dbstop if error; %-------------------------------------------------------------------------- % This demo process the data from the RIT Target Detection Blind Test % contest which is located at: http://dirsapps.cis.rit.edu/blindtest/ % To use this file, you select a target detection algorithm and a target to % find and then the script runs the algorithm and outputs the data into the % outputDir. Two files are outputted, a .img and a .hdr. You upload these % files to the RIT website and they are automatically scored. %-------------------------------------------------------------------------- % Parameters inputFilename = 'data\blind_test\HyMap\blind_test_refl.img'; fasticaToolboxPath = '..\matlab_hyperspectral_toolbox\trunk\FastICA_25'; targetFilenames = {'data\blind_test\SPL\F5\F5_f.txt'}; outputDir = 'RIT Data Results'; % See switch statement for algorithm choices algorithm = 'ace' %algorithm = 'rmf-sum'; %algorithm = 'plmf' %algorithm = 'matchedFilter'; %algorithm = 'sam' %-------------------------------------------------------------------------- addpath('gmm'); addpath(fasticaToolboxPath); mkdir(outputDir); % Read in the data w = 280; h = 800; p = 126; M = multibandread(inputFilename, [w h p], 'int16', 0, 'bil', 'ieee-le')/1e4; lData = hyperGetHymapWavelengthsNm(); % Read in target signatures [sig1, lSig] = hyperGetEnviSignature(targetFilenames{1}); % Get signature from data for comparison fsig1 = squeeze(M(122,495,:)); %sig1 = fsig1; % Resample data to commone wavelength set desiredLambdas = lData; sig1 = squeeze(hyperResample(sig1, lSig, desiredLambdas)); figure; plot(sig1); grid on; title('Signature 1'); xlabel('Wavelength [nm]'); ylabel('Reflectance [%]'); hold on; plot(fsig1, '--'); legend('Recorded', 'From Image'); goodBands = 1:p; %[3:63 69:93 98:123]; % Image sharpening if 0 ff = fspecial('unsharp',0.2); for k=1:p M(:,:,k) = imfilter(M(:,:,k),ff,'same'); M(:,:,k) = imfilter(M(:,:,k),ff,'same'); %M(:,:,k) = imfilter(M(:,:,k),ff,'same'); end end figure; imagesc(M(:,:,40)); axis image; colormap(gray); % Try to discover in-situ to lab kernel. % TODO % sub(:,1) = M(144,515,:); % sub(:,2) = M(144,516,:); % sub(:,3) = M(144,517,:); % sub(:,4) = M(145,515,:); % sub(:,5) = M(145,516,:); % sub(:,6) = M(145,517,:); % sub(:,7) = M(146,515,:); % sub(:,8) = M(146,516,:); % sub(:,9) = M(146,517,:); % % alpha = pinv(sub)*sig1; %alpha = alpha ./ sum(alpha(:)); % err = sub*alpha - sig1; err = err - mean(err); badBands = find(abs(err)>0.02); % goodBands = setxor(1:p,badBands); % figure; plot(err); hold on; plot(sig1,'.'); plot(fsig1,'.-'); hold off; grid on; % legend({'err','lab sig','in situ sig'}) % alpha = reshape(alpha,3,3); % figure; imagesc(alpha); % % for k=1:p % %M(:,:,p) = conv2(M(:,:,p),alpha,'same'); % end % Emperical dervied goodBands = [3 4 5 6 7 8 9 10 11 12 14 15 16 17 18 ... 19 20 22 23 24 26 28 29 31 32 33 34 35 36 37 ... 38 39 40 41 42 43 44 45 46 49 51 52 53 54 55 ... 56 57 58 59 60 61 62 66 69 70 71 72 86 87 88 ... 89 90 91 92 93 96 97 98 99 100 101 102 103 104 105 ... 106 107 108 109 110 111 112 113 115 116 117 119 120 121 122]; %sig1 = squeeze(hyperResample(sig1, lSig, desiredLambdas)); figure; plot(sig1(goodBands)); grid on; title('Signature 1 - good bands only'); xlabel('Wavelength [nm]'); ylabel('Reflectance [%]'); hold on; plot(fsig1(goodBands), '--'); legend('Recorded', 'From Image'); % Display data M = hyperConvert2d(M); %[M, H, snr] = hyperMnf(M, w, h); M_pct = hyperPct(M, 3); M_pct = hyperNormalize(hyperConvert3d(M_pct,w,h,3)); figure; imagesc(M_pct); axis image; title('Scene'); % Data conditioning M = M(goodBands, :); sig1 = sig1(goodBands); %fsig1 = fsig1(goodBands); %sig1 = fsig1; %q = hyperHfcVd(M); % Do PCT if 0 M = [M sig1]; %[M,V] = hyperPct(M,size(M,1)); [M,V] = hyperPct(M,55); sig1 = M(:,end); M = M(:,1:end-1); p = size(M,1); goodBands = 1:p; end %q = hyperHfcVd(M); q = 39; algorithm = lower(algorithm); tic switch algorithm case 'ica-eea' [U, X] = hyperIcaEea(M, 50, sig1); r = X(1,:); r = hyperConvert3d(r, w, h, 1); case 'rx' r = hyperConvert3d(hyperRxDetector(M), w, h, 1); case 'matchedfilter' r = hyperConvert3d(hyperMatchedFilter(M, sig1), w, h, 1); case 'ace' r = hyperConvert3d(hyperAce(M, sig1), w, h, 1); case 'mace' r = hyperConvert3d(hyperMace(M, sig1), w, h, 1); case 'sid' r = hyperConvert3d(hyperSid(M, sig1), w, h, 1); case 'cem' r = hyperConvert3d(hyperCem(M, sig1), w, h, 1); case 'plmf' r = hyperPlmf(hyperConvert3d(M,w,h,p),sig1,9); case 'rmf-sum' r = hyperRmf(hyperConvert3d(M,w,h,p),sig1,11,'sum'); case 'rmf-meanlocal' r = hyperRmf(hyperConvert3d(M,w,h,p),sig1,11,'meanLocal'); case 'rmf-meangloballocal' r = hyperRmf(hyperConvert3d(M,w,h,p),sig1,11,'meanGlobalLocal'); case 'glrt' r = hyperConvert3d(hyperGlrt(M, sig1), w, h, 1); case 'osp' U = hyperAtgp(M, q, sig1); r = hyperConvert3d(hyperOsp(M, U, sig1), w, h, 1); case 'amsd' r = hyperConvert3d(hyperAmsd(M, U, sig1), w, h, 1); case 'hud' U = hyperAtgp(M, q, sig1); r = hyperConvert3d(hyperHud(M, U, sig1), w, h, 1); case 'nnls' U = hyperAtgp(M, q, sig1); r = hyperConvert3d(hyperNnls(M,U),w,h,q); r = r(:,:,1); case 'fcls' U = hyperAtgp(M, q, sig1); r = hyperConvert3d(hyperFcls(M,U),w,h,q); r = r(:,:,1); case 'ucls' U = hyperAtgp(M, q, sig1); r = hyperConvert3d(hyperUcls(M,U),w,h,q); r = r(:,:,1); case 'sam' r = (1./(eps+hyperConvert3d(hyperSam(M, sig1), w, h, 1))); otherwise error('Incorrect algorithm name specified!\n'); end toc % Display results and write to file figure; imagesc(r); axis image; colorbar; title(algorithm); [a,b]=sort(r(:),'descend'); tmp = a(1:20); figure; plot(tmp./tmp(1)); grid on; [x, y, val] = hyperMax2d(r); % d1 = r(122,494) % d2 = r(127,490) % N = prod(size(r)); % [v] = sort(r(:),'ascend'); % idx = find(v==d2); % N-idx % figure; hist(r(:),100); tmp = (hyperNormalize(r)*2^10); multibandwrite(tmp, sprintf('%s\\results.img', outputDir), 'bil', 'PRECISION', 'int16', 'MACHFMT', 'ieee-le'); [pd,fa] = hyperRoc(r); figure; plot(fa,pd,'.'); grid on; title(sprintf('%s\n%s',algorithm, targetFilenames{1}));