/normxcorr/trunk : revision 31

1

function [validx, validy, displx, disply]=RTCorrCode(grid_x,grid_y,Firstimagename,ImageFolder)

2

3

% Real time Correlation Code

4

%

5

% Written by Chris

6

% OPTIMIZE parameter is controlling which optimizations should be used.

7

% <3 - The original version, no optimizations

8

% 3 - Most of computations are performed on NVidia card

9

% 4 - Optimize FFT sizes for better performance (affects precision)

10

% 5 - Load images in C code

11

OPTIMIZE = 4;

12

CORRSIZE = 15;

13

PRECISION = 1000;

14

15

warning off Images:initSize:adjustingMag

16

17

if OPTIMIZE > 2

18

hwid = normxcorr_hw();

19

20

if hwid > 0

21

else

22

OPTIMIZE = 0;

23

end

24

else

25

hwid = 0;

26

end

27

28

if OPTIMIZE > 2

29

RTselection = menu(sprintf('End processing by end.txt or by last image?'),...

30

'Stop with end.txt','Stop with image check','Exit');

31

32

if RTselection==1

33

end

34

35

if RTselection==2

36

end

37

38

if RTselection==3

39

if hwid > 0

40

normxcorr_hw(hwid);

41

end

42

return

43

end

44

end

45

46

% Filename

47

48

if exist('Firstimagename')==0

49

[Firstimagename ImageFolder]=uigetfile('*.tif','Open First Image');

50

end

51

52

53

if ~isempty(Firstimagename)

54

% Get the number of image name

55

letters=isletter(Firstimagename);

56

Pointposition=findstr(Firstimagename,'.');

57

Firstimagenamesize=size(Firstimagename);

58

counter=Pointposition-1;

59

counterpos=1;

60

letterstest=0;

61

62

while letterstest==0

63

letterstest=letters(counter);

64

if letterstest==1

65

break

66

end

67

Numberpos(counterpos)=counter;

68

counter=counter-1;

69

counterpos=counterpos+1;

70

if counter==0

71

break

72

end

73

end

74

75

Filename_first = Firstimagename(1:min(Numberpos)-1);

76

Firstfilenumber=Firstimagename(min(Numberpos):max(Numberpos));

77

Lastname_first = Firstimagename(max(Numberpos)+1:Firstimagenamesize(1,2));

78

Firstfilenumbersize=size(Firstfilenumber);

79

onemore=10^(Firstfilenumbersize(1,2));

80

filenamelist(1,:)=Firstimagename;

81

82

83

h=figure;

84

if exist('grid_x')==0

85

fpstest=1;

86

Filelist=[Firstimagename;Firstimagename];

87

while fpstest==1

88

[grid_x,grid_y]=grid_generator(Firstimagename,ImageFolder);

89

if OPTIMIZE > 2

90

ncp = prod(size(grid_x));

91

err = normxcorr_hw(hwid, 1, ncp, CORRSIZE, PRECISION, OPTIMIZE);

92

else

93

err = 0;

94

end

95

if err == 0

96

if OPTIMIZE > 2

97

base_points_x=single(grid_x);

98

base_points_y=single(grid_y);

99

100

normxcorr_hw(hwid, 3, base_points_x, base_points_y);

101

102

if OPTIMIZE > 4

103

normxcorr_hw(hwid, 4, strcat(imagedir, Firstimagename));

104

else

105

base = uint8(mean(double(imread([ImageFolder, Firstimagename])),3));

106

normxcorr_hw(hwid, 4, base);

107

end

108

109

normxcorr_hw(hwid, 12, base_points_x, base_points_y);

110

end

111

[processingtime]=fpstestfunc(hwid,OPTIMIZE,grid_x,grid_y,Filelist,ImageFolder);

112

fpstest = menu(sprintf(['Processing the selected grid will allow ' , num2str(1/processingtime),' frames per second' ]),'Try again','Use the grid');

113

if fpstest==1

114

clear grid_x; clear grid_y;

115

end

116

else

117

ASelection = menu(sprintf('CUDA initialization failed?'),...

118

'Retry another grid','Continue in software mode','Exit');

119

if Aselection==1

120

fptest = 1;

121

end

122

123

if Aselection==2

124

fptest = 0;

125

normxcorr_hw(hwid);

126

OPTIMIZE = 0;

127

end

128

129

if Aselection==3

130

return

131

end

132

end

133

end

134

end

135

136

if OPTIMIZE < 3

137

[validx, validy, displx, disply] = RTCorrCode_orig(grid_x, grid_y, Firstimagename, ImageFolder)

138

return

139

end

140

141

142

Firstfilenumber=str2num(Firstfilenumber);

143

u=1+onemore+Firstfilenumber;

144

ustr=num2str(u);

145

filenamelist(2,:)=[Filename_first ustr(2:Firstfilenumbersize(1,2)+1) Lastname_first];

146

numberofimages=2;

147

148

counter=1;

149

150

input_points_x=single(grid_x);

151

input_points_y=single(grid_y);

152

normxcorr_hw(hwid, 12, input_points_x, input_points_y);

153

154

fit_options = optimset('Display', 'off');

155

156

numberofmarkers=max(size(grid_x))*min(size(grid_x));

157

validx(:,1)=reshape(grid_x,[],1);

158

displx=zeros(numberofmarkers,1);

159

validy(:,1)=reshape(grid_y,[],1);

160

disply=zeros(numberofmarkers,1);

161

tic;

162

163

while exist([ImageFolder, 'end.txt'],'file') ==0;

164

pause(0.01);

165

166

if exist([ImageFolder, filenamelist((counter+1),:)],'file') ==2;

167

% warning(['# Processed Images: ', num2str(numberofimages-1),'; # markers:',num2str(numberofmarkers), '; Processing Image: ',filenamelist(counter+1,:)]);

168

169

if OPTIMIZE > 4

170

normxcorr_hw(hwid, 13, strcat(ImageFolder, filenamelist((counter+1),:)));

171

input_correl = normxcorr_hw(hwid, 14);

172

else

173

input = uint8(mean(double(imread([ImageFolder, filenamelist((counter+1),:)])),3));

174

normxcorr_hw(hwid, 13, input);

175

input_correl = normxcorr_hw(hwid, 14);

176

end

177

178

input_correl_x=double(input_correl(:,1));

179

input_correl_y=double(input_correl(:,2));

180

181

validx(:,counter+1)=input_correl_x; % lets save the data

182

savelinex=input_correl_x';

183

dlmwrite([ImageFolder, 'resultsimcorrx.txt'], savelinex , 'delimiter', '\t', '-append'); % Here we save the result from each image; if you are desperately want to run this function with e.g. matlab 6.5 then you should comment this line out. If you do that the data will be saved at the end of the correlation step - good luck ;-)

184

185

validy(:,counter+1)=input_correl_y;

186

saveliney=input_correl_y';

187

dlmwrite([ImageFolder, 'resultsimcorry.txt'], saveliney , 'delimiter', '\t', '-append');

188

189

subplot(2,2,1);

190

imshow([ImageFolder, filenamelist(counter+1,:)]);

191

hold on;

192

plot(grid_x,grid_y,'g+');

193

plot(input_correl_x,input_correl_y,'r+');

194

hold off;

195

drawnow;

196

197

displx(:,counter+1)=validx(:,counter+1)-validx(:,1);

198

disply(:,counter+1)=validy(:,counter+1)-validy(:,1);

199

200

subplot(2,2,2);

201

xdata=validx(:,counter+1);

202

ydata=displx(:,counter+1);

203

if counter==1

204

x(1)=0;

205

x(2)=0;

206

end

207

[x,resnormx,residual,exitflagx,output] = lsqcurvefit(@linearfit, [x(1) x(2)], xdata, ydata, [], [], fit_options);

208

plot(xdata,ydata,'.');

209

hold on;

210

ydatafit=x(1)*xdata+x(2);

211

plot(xdata,ydatafit,'r');

212

hold off;

213

xlabel('x-pos [pixel]');

214

ylabel('x-displ [pixel]');

215

title('x displ. versus x pos. in [pixel]');

216

217

slopex(counter,:)=[i x(1) x(2)];

218

219

subplot(2,2,4);

220

xdata=validy(:,counter+1);

221

ydata=disply(:,counter+1);

222

if counter==1

223

y(1)=0;

224

y(2)=0;

225

end

226

[y,resnormx,residual,exitflagx,output] = lsqcurvefit(@linearfit, [y(1) y(2)], xdata, ydata, [], [], fit_options);

227

plot(xdata,ydata,'.g');

228

hold on;

229

ydatafit=y(1)*xdata+y(2);

230

plot(xdata,ydatafit,'r');

231

hold off

232

xlabel('y-pos [pixel]')

233

ylabel('y-displ [pixel]')

234

title('y displ. versus y pos. in [pixel]');

235

236

slopey(counter,:)=[i y(1) y(2)];

237

238

subplot(2,2,3);

239

plot(slopex(:,2),'-b');

240

hold on;

241

plot(slopey(:,2),'-g');

242

hold off;

243

xlabel('Image # [ ]');

244

ylabel('x- and y-strain [ ]');

245

title('Strain in x and y direction versus Image #');

246

247

counter=counter+1;

248

249

u=1+u;

250

ustr=num2str(u);

251

filenamelist(counter+1,:)=[Filename_first ustr(2:Firstfilenumbersize(1,2)+1) Lastname_first];

252

[numberofmarkers numberofimages]=size(validx);

253

254

if RTselection==2

255

if exist([ImageFolder, filenamelist((counter+1),:)],'file') ==0;

256

save ([ImageFolder, 'validx.dat'], 'validx', '-ascii', '-tabs');

257

save ([ImageFolder, 'validy.dat'], 'validy', '-ascii', '-tabs');

258

%warning('Last image detected, RTCorrCode stopped');

259

normxcorr_hw(hwid);

260

return

261

end

262

end

263

264

265

subplot(2,2,1),title(['# Processed Images: ', num2str(numberofimages-1),'; fps: ', num2str((numberofimages-1)/toc),'; # markers:',num2str(numberofmarkers), '; Waiting for Image: ',filenamelist(counter+1,:)]);

266

267

end

268

end

269

270

normxcorr_hw(hwid);

271

272

save ([ImageFolder, 'validx.dat'], 'validx', '-ascii', '-tabs');

273

save ([ImageFolder, 'validy.dat'], 'validy', '-ascii', '-tabs');

274

msgboxwicon=msgbox('end.txt file detected, RTCorrCode stopped','Processing stopped!');

275

%warning('end.txt file detected, RTCorrCode stopped');

276

end

277

278

%----------------------------------

279

%

280

281

function [processingtime]=fpstestfunc(hwid, OPTIMIZE, grid_x, grid_y, filenamelist, ImageFolder)

282

283

tic;

284

285

if hwid > 0

286

if OPTIMIZE > 4

287

normxcorr_hw(hwid, 13, strcat(ImageFolder, filenamelist(2,:)));

288

input_correl = normxcorr_hw(hwid, 14);

289

else

290

input = uint8(mean(double(imread([ImageFolder, filenamelist(2,:)])),3));

291

normxcorr_hw(hwid, 13, input)

292

input_correl = normxcorr_hw(hwid, 14);

293

end

294

input_correl_x=double(input_correl(:,1));

295

input_correl_y=double(input_correl(:,2));

296

else

297

input_points_x=grid_x;

298

base_points_x=grid_x;

299

300

input_points_y=grid_y;

301

base_points_y=grid_y;

302

303

base = uint8(mean(double(imread([ImageFolder, filenamelist(1,:)])),3)); % read in the base image ( which is always image number one. You might want to change that to improve correlation results in case the light conditions are changing during the experiment

304

input = uint8(mean(double(imread([ImageFolder, filenamelist(2,:)])),3)); % read in the image which has to be correlated

305

306

input_points_for(:,1)=reshape(input_points_x,[],1); % we reshape the input points to one row of values since this is the shape cpcorr will accept

307

input_points_for(:,2)=reshape(input_points_y,[],1);

308

base_points_for(:,1)=reshape(base_points_x,[],1);

309

base_points_for(:,2)=reshape(base_points_y,[],1);

310

input_correl(:,:)=cpcorr(input_points_for, base_points_for, input, base); % here we go and give all the markers and images to process to cpcorr.m which ic a function provided by the matlab image processing toolbox

311

input_correl_x=input_correl(:,1); % the results we get from cpcorr for the x-direction

312

input_correl_y=input_correl(:,2); % the results we get from cpcorr for the y-direction

313

end

314

315

processingtime=toc;

b'\\ No newline at end of file'