How to store SURF features of multiple images

Hello all, I'm using SURF descriptor for feature extraction from folder of images. The result i.e. surf features are of different dimension for different image i.e. N*64, N varies with each image. These features are to be classified using knn classifier,one image with all of its features as one class. But KNN classifier takes each row wise features as one class. So how to classify this varying row size features or how to store this in a format that is passed as input to KNN classifier. Below is the code where "features1{i}" contains all the surf features.

srcFiles = dir('C:\Users\nmamit\Documents\MATLAB\photos\*.jpg');  % the folder in which ur images exists

for i = 1 : length(srcFiles)

   filename = strcat('C:\Users\nmamit\Documents\MATLAB\photos\',srcFiles(i).name);

   I1 = rgb2gray(imread(filename));

      points1 = detectSURFFeatures(I1); 

   feature = zeros(1, length(srcFiles))

   [features1{i}, valid_points1{i}] = extractFeatures(I1, points1); 

   figure; imshow(I1);hold on;

   plot(valid_points1{i}.selectStrongest(10),'showOrientation',true);

end

[KS]

Hi, I am having the same problem. Can you please tell me if you came up with a solution?

Thanks.

Answer 1

[KSSV]

Store them into cells.

srcFiles = dir('C:\Users\nmamit\Documents\MATLAB\photos\*.jpg');  % the folder in which ur images exists
Features = cell(length(srcFiles),1) ;
Valid_points = cell(length(srcFiles),1) ;
for i = 1 : length(srcFiles)
    filename = strcat('C:\Users\nmamit\Documents\MATLAB\photos\',srcFiles(i).name);
    I1 = rgb2gray(imread(filename));
    points1 = detectSURFFeatures(I1);
    [features1, valid_points1] = extractFeatures(I1, points1);
    Features{i} = features1 ;
    Valid_points{i} = valid_points1 ;
    figure; imshow(I1);hold on;
    plot(valid_points1{i}.selectStrongest(10),'showOrientation',true);
end
Features
Valid_points

[KDR]

Thank you.. But this cell contents can't be classified. While passing Features to knn with a label its not accepting.

load ('Features.mat');

load('label');

mdl =ClassificationKNN.fit(Features,label);

Its giving me an error that "

Error using classreg.learning.FullClassificationRegressionModel.prepareDataCR (line 167) X must be a numeric matrix.

Error in classreg.learning.classif.FullClassificationModel.prepareData (line 473) classreg.learning.FullClassificationRegressionModel.prepareDataCR(...

Error in ClassificationKNN.prepareData (line 862) [e-mail](X,Y,varargin{:},'OrdinalIsCategorical',true);

Error in classreg.learning.FitTemplate/fit (line 206) this.PrepareData(X,Y,this.BaseFitObjectArgs{:});

Error in ClassificationKNN.fit (line 847) this = fit(temp,X,Y);

Error in k2 (line 6) mdl =ClassificationKNN.fit(Features,label); "

[KSSV]

YOu can access the feature you want using Features{i}

[lebed toufik]

@KSSV I used the features {i} but still the same problem.