Build model detection after features extraction

Hello,

I'm trying to code a nose detection function from a IR video.

I extracted 2 frames from the video and foud the features and compared between them.

ref_img = imread('frame_1.png');
ref_img_gray=rgb2gray(ref_img);
ref_pts=detectSURFFeatures(ref_img_gray);
[ref_features,ref_validPts]=extractFeatures(ref_img_gray,ref_pts);
figure; imshow(ref_img);
hold on; plot(ref_pts.selectStrongest(50));

image=imread('frame_50.png');
I=rgb2gray(image);
I_pts=detectSURFFeatures(I);
[I_features,I_validPts]=extractFeatures(I,I_pts);
figure;imshow(image);
hold on; plot(I_pts.selectStrongest(50));

index_pairs=matchFeatures(ref_features,I_features);
ref_matched_pts=ref_validPts(index_pairs(:,1)).Location;

I_matched_pts=I_validPts(index_pairs(:,2)).Location;
close all
figure,showMatchedFeatures(image,ref_img,I_matched_pts,ref_matched_pts);

Here the figure obtained :

What I have to do as a next step ? We can see from the figure that we got the 2 nostrils as features, so how to train a model a got a function that tracks the region for all the frames ?

thank you

Answer 1

[Manas Meena]

After SURF feature detection you can select the strongest points of interest (eg. nostrils) and the use the vision.PointTracker function to track these selected points in the video.