Object identification and Classification in the same image

Hi for all can evry one help me to detect ,identify and calssify the small components in the circuit boar d " inspection of printed circuit board " the output of the code implimentation is an image how all components mouted in the surface are framed for exemple the resistance or the capacitor,but not identifyed how is the resistor and how is the capacitor

then my problem is in the recognition and classification of framed Small components, by neronal network or SVM, to distiguish what are the Resistors and what are the Capacitors in the image. can every one help me to developpe this part of implimentation code ? please it is urgent

The image of circuit board:

Answer 1

[Image Analyst]

Why do you want to do this? If this is for Quality Assurance, you will already know exactly what component is supposed to be where, and all you have to do is to subtract the image from the known, perfect template image to identify regions with missing or different components. Then you just look it up. I really don't see any need to use "neronal network or SVM", unless you're just doing it because of a homework assignment or class project or something.

[Fridhi]

I need just to ditinguish between the two components and to learn and classify the components in the motherboard image into 2 classes (Resistor // Capacitor) and difenciate them by the color border of any components in the image.

I ask you have you any adea or project for Quality Assurance and how can i will already know exactly what component is supposed to be where, and all how i have to do is to subtract the image from the known, perfect template image to identify regions with missing or different components.And can you help me to develope this part please.

And i appreciate your assistance,

I will be so grateful Mr.

[Image Analyst]

1. Register (align) the test image with the reference image with imregister().
2. Subtract the image: diffImage = double(testImage) - double(refImage);
3. Threshold: binaryImage = abs(diffImage) > 5; % or whatever.
4. Get the max along the third dimension
5. Call bwlabel or bwconncomp()
6. Call regionprops and get centroids of regions.
7. Look up what component is supposed to be at that location from your reference documentation.