can anyone guide me through an object recognition mechanism or code that recursively plots the contour of object and extract transformation invariant features so that an ANN classifier distinguishes the object through slight different projections?

I am doing a project that requires the object in a image to be recognized and the object be represented with its original pixel values. I have to use the invariant features of object to train the neural network distinguish it even with slight different view points or slight different projections of object in other image of it .

Answer 1

[Shantanu Jana]

Use connected component method to find the contour then crops the object assign particular class name to them and extract features value from them .then you can pass this feature values to different ANN model to find the recognition rate.

check this link it can be helpful for you http://in.mathworks.com/help/images/ref/bwconncomp.html