What would be the best method of image recognition for components on an image of a PCB?

I have an image of a PCB and want matlab to recognise the the PIC18 in the middle of the picture. The end goal is to get matlab to recognise resistors, capacitors and other components but my first step would be to get it to identify the PIC18 before i can go any further as i would need to understand how the process is done. I have an image and have applied some edge detection techniques so far to get matlab to detect the edges. From my understanding, i would have to create an algorithm next but that is where i am unsure of what and how to do it. The image below is of the edge detection applied to the image

It shows that all the edges are indeed shown clearly but what would be the next step to get matlab to detect the chip in the middle?

[John BG]

automating electronic circuit board components detection/location/classification is a dodgy task for the following reasons.

1.- big resistors, big capacitors .. the bit chip, just one big chip .. all this is relatively simple, provided you have enough light on the board, that it's obvious by the photo you have attached to the question that you don't have (enough light on the board i mean)

2.- you need colour and resolution

3.- The only way to id SMD components is measuring them with jigs, or manually. Visual on SMD resistor 1k2 or SMD resistor 5k6 is the same, yet you need to measure. Factories have expensive jigs many needles that measure single components. Same applies to arrayed capacitors/resistors inside chips.

3.- Bear in mind that some manufacturers literally delete chip codes, or stick labels on them. So you have spotted the micro-controller, yet do you know the exact version? part number? variant? again it's faster to have the right tools, plug, and ask (the micro, to id)

Having said all this, your rookie reverse engineering effort is laudable, who wouldn't like to have a program that spots all the components on a board? factories have similar systems, to place components on the right board coordinates.

What you can do, is to build a list of component images, and tell y our application to spot them on the board one by one, which means you need to know the BOM (Bill of Materials) in advance, and take a photo of each component, and hope the components will not have things like radiators, labels, glue .. on them.

Good luck

John

[e-mail]

[Hashil Patel]

Thank you for your Feedback John. i have made this library that you have advised me to make but the next step is the toughest part and the most difficult. Thats where i am. Getting MATLAB to spot the component. any particular function i should be looking into that would help me get the job done??

[Walter Roberson]

That was good feedback, John. I would have voted for it.

[KIVANÇ mutlu]

Ben pcb üzerinde fiducial noktasını öğretmek ve daha sonra bu öğretilen noktalara farklı yerlerde gelen pcbs karşılaştırmak istiyorum. nasıl yapabilirim.

[KIVANÇ mutlu]

I made comparison of fixed pcb images but XOR comparison does not work on samples from different locations. I need to first identify the PCB location using the fiducial reference.

Answer 1

[ATTAR Mehenni]

Hi, if you found the solution, share the code matlab, thanks

[Image Analyst]

I don't think you can identify random little specks or boxes on a PCB - they could be anything. But what you can do is to compare a test board against a reference board and detect differences. Is that what you really need to do?

[ATTAR Mehenni]

I want to detect the lack of components on a pcb and what it is: a resistor, a capacitor ...

[Image Analyst]

Why do you need to do this? Don't you know what is supposed to be in each location? Many components look the same, just like a tiny box.