Finding under vehicle undefined object - inspection system

Hi all,

There are 2 images of a under vehicle recorded with a fisheye lens. How can I detect the undefined object in the 2nd image using the main image (1st image) in Matlab ? In the 2nd image, as I have marked in red, I need to locate the undefined object.

Thanks in advance.

Answer 1

[Image Analyst]

If the images are of the same size, and of the same magnification and fairly well aligned (a doubtful scenario in a real world situation) then you can subtract the images and look for blobs of non-zero. Untested code:

% Get the difference image.
diffImage = imabsdiff(double(image1), double(image2));
% Get the max in any of the color channels
diffImage = max(diffImage, [], 3);
% Threshold to find significant differences.
mask = diffImage > 5; % or some value.
% Throw out any blobs less than 50 pixels in size (they're noise).
mask = bwareaopen(mask, 50); 
imshow(mask);
% Make measurements of area and centroid.
props = regionprops(mask, 'Area', 'Centroid');
allAreas = [props.Area]
centroids = vertcat(props.Centroid)

[mehmet irfan gedik]

The situation for the real world are a bit challenging as you said. Actually the images are not the same size (1st image- 2655x1024x3 and 2nd image- 2771x1024x3), and they're actually not exactly aligned and the last problem is that distortionCenter isn't exactly in the same location. I've been thinking about what I really need to do for a while.

[Image Analyst]

Try imresize() and imregister() to align them. If the differences are big then registration may not be necessary, though for tiny differences it would be.