An approach towards the leaf classification

This task contains an end-term project of class Digital Image Processing & Applications taught by Prof Deboot Sheet. Leaf classification

To recognize objects may be a simple process by human visualization but for the digitally it may vary from method to method. The signature of every object has a different pattern. Even though objects are the same class will have a similar kind of shape signature pattern while statistical values are maybe a little bit different. one of the simple methods to get the signature of the object is to plot the distance from the centroid of the boundary as a function of the angle. (basically polar form).

Details of the approach are given below

1. RGB image convert into Grayscale image (but here used binary images)

2. Convert grayscale image into binary image (insure that inside boundary of the object should pixel value 1(appear as white colour) and background should 0 value(appear as black colour)

3. Apply edge detection technique('sobel') to get the boundary of the object in the images

4. Find out the centroid of the edge detected images

5. Find out the distance of the boundary pixel from centroid and angle

6. Plot shape signature for each object

% further approach towards

7. signature matching pattern technique (but I am not able to do this process)

8. Classify leaves into two categories 1. broad leaves 2. narrow leaves

9. Tested Accuracy