Detecting head in pillow defect (HIP) by using deep learning and image processing technique
Deep learning is an Artificial Intelligence (AI) method that mimics the ways human brain processing data and recognizes the data or objects. It is a subset of machine learning which utilizes the hierarchical level of artificial neural networks (ANN) to perform the process of machine learning (Hargra...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Year Project / Dissertation / Thesis |
| Published: |
2021
|
| Subjects: | |
| Online Access: | http://eprints.utar.edu.my/4272/2/17ACB02302_FYP.pdf http://eprints.utar.edu.my/4272/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Deep learning is an Artificial Intelligence (AI) method that mimics the ways human brain processing data and recognizes the data or objects. It is a subset of machine learning which utilizes the hierarchical level of artificial neural networks (ANN) to perform the process of machine learning (Hargrave, 2019). Deep learning has very great potential of wide adoption in various industries. In fact, deep learning has already been used by corporations and start-ups such as Google, Facebook, Amazon, Tesla etc for several different tasks such as filtering fake news, analysing shopping trends and developing self-driving cars. In the manufacturing sectors, deep learning techniques were usually used to aid the engineers or inspectors in making decisions in the production line or the quality inspection phase. However, there are still various reasons why deep learning was not largely implemented in the manufacturing sector especially in the detection of Head in Pillow (HIP) defects that occurred in the Ball Grid Array (BGA) of a printed circuit board (PCB). This project aims to design a robust deep learning model that could be implemented to speed up and ease the process of detecting the HIP defects. The 3 Dimensional (3D) Convolutional Neural Network (CNN) will be the foundation of the deep learning model which will deal with the grayscale BGA slice images that were stacked together. The outcome of the project will be a robust deep learning model that could classify the HIP defects on BGA joints in greyscale which have not more than 9 slices. Over 200 of 3D CNN models with different hyperparameters and architecture are created in this project to achieve the objectives of the project. |
|---|