Skip to main content

IC Recognition from a Printed Circuit Board

In this post, I will tell you about a project that I and my friend did last year. That was IC detection from a printed circuit board and then recognize it using optical character recognition.

This is the application of Open Computer Vision (OpenCV).

 First of all, I will show you a flowchart that will describe our methodology and then I will discuss in detail.


There are two main parts of our algorithm: Localizing each IC on the PCB, extracting and saving it. Then we use Tesseract OCR engine to read the labels of each detected IC.



A. PREPROCESSING
Before we apply any algorithms to our image we have to process it to obtain a proper image. To save processing time we will resize the image maintaining the aspect ratio. After that remove noise from the image using Gaussian Blur.

B. SEGMENTATION
First, we convert the RGB image to HSV (Hue, Saturation, Value) color space. The H channel creates a mask for all the integrated circuits on the board. We then threshold the image to get a binary image However, it is not perfect.

C. MORPHOLOGICAL OPERATIONS
As we can see the thresholding is not perfect so we performed morphological operations (first dilate and then erode the dilated image) on the threshold image.

D. EXTRACTION
To extract the integrated circuits from PCB we first have to find edges of the ICs. We used Canny Edge Detection algorithm to find edges. After that, we found contours for those edges. These contours also included small areas on PCB that were not IC so, we had to exclude them. We excluded every contour which had less area than the mean area of contours. We found bounding rectangle for remaining contours and cropped that from the original image. This gave us all the integrated circuits on the PCB.

One of the extracted IC

                                                        
E. OPTICAL CHARACTER RECOGNITION (OCR)
Once the ICs were saved on the disk, we identified the ICs by reading their labels using OCR. For OCR, we used Tesseract OCR engine. If we pass the ICs directly to Tesseract OCR engine it will fail to read the labels so we pre-processed the image by smoothing and thresholding it. We used Otsu’s algorithm to threshold the image. We also eroded the image to make edges clear. We then passed this image to Tesseract OCR  engine and got the labels in text format.

As you know the result cannot be perfect so there are some limitations also.

LIMITATIONS:
 It can only detect and identify ICs in well-lit images.
 Smaller ICs are not detected.
 The accuracy of OCR is very low.


Now I know that I have used many technical terms here and not described it fully but in future posts, I will definitely explain them one by one. However, if you want to see the code and the paper that we have written then please let me know in the comments section and I will share the link of my GitHub account where you can find all the test images and code.
Till then enjoy learning. 

Labels:

Comments

Post a Comment

Popular posts from this blog

Constructors In Dart Language

In the previous post, I have discussed Classes and Objects and continuing to that in this post I will cover Constructors. So what a constructor is? As in the previous post, we have created an object car inside the main function using the new keyword, it only creates a generic object without any properties or we can say it is just a spot in memory. And in the next few lines when we assign values to their attributes then it is completely constructed. Although, constructing directly inside the main function is a good idea but there is another way also by which we can form a constructor inside the class itself. But stop I am forgetting something. Before we go deep into constructors I want to tell you something about a very important thing this . this is used to denote the current value. Maybe it is not making sense at this time but it will be clear once we use it. Now coming back to constructors, it is a function or method like entity but without any return type. There are thr...
Post a Comment
Read more

Inheritance In Dart

So now continuing to the Blog series I will talk about the inheritance concept in this post. In previous blogs, we have seen how to create classes and objects and we can create as many classes as we want. But sometimes we observe that many classes have some properties same as another class. We should keep in mind a fact that creation of an object in memory is a very expensive task and we have to minimize it so one of the methods to do so is Inheritance.   So there are two terms: Parent Class and Child Class. The parent class has some features and the child class has similar features along with some extra specific features and the child can be inherited from the parent class so that there is no redundancy. Image source Google So talking about the above image we can see that the vehicle is the parent class and then the car, motorcycle, truck are child class of vehicle and following that again station wagon, jeep are the child class of car and so on. ...
Post a Comment
Read more