PERANCANGAN SISTEM EKSTRAKSI DIMENSI LINIER GAMBAR TEKNIK BERBASIS DEEP NEURAL NETWORK PADA PENYIAPAN DOKUMEN PROSES INSPEKSI
Quality inspection is a crucial aspect that must be maintained by companies, including CV CSM, a job shop metal manufacturing company in Cimahi. The current manual quality inspection process at CV CSM is deemed inefficient for a job shop with a diverse range of part complexities. This inefficienc...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/83760 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Quality inspection is a crucial aspect that must be maintained by companies,
including CV CSM, a job shop metal manufacturing company in Cimahi. The
current manual quality inspection process at CV CSM is deemed inefficient for a
job shop with a diverse range of part complexities. This inefficiency is characterized
by the extended time required to extract dimensions from engineering drawings into
inspection sheets. The higher the complexity and number of dimensions on a part,
the longer the extraction process takes.
This study focuses on designing a system for extracting dimensional text from
engineering drawings into inspection sheets using a neural network approach to
reduce the time required for the quality inspection document preparation.The study
begins with preprocessing to prepare engineering drawings as input for the next
stage, which involves developing a dimensional text extraction model. This model
development consists of five stages: creating a model for view recognition and
dimensional text area recognition using the You Only Look Once (YOLO) library,
text recognition using the OpenCV and Pytesseract libraries, processing the
drawing header with the Camelot library, and generating outputs with the
OpenPyXL library. In the final stage, a system interface will be designed, consisting
of a user interface using Streamlit and an interface with the company's existing
information system.
The outcome of this study is a design for a dimensional text extraction system using
a neural network approach, with performance metrics including an average recall,
precision, and F1-score of 84%, 87.5%, and 85%, respectively. The proposed
dimensional text extraction system significantly enhances the efficiency of
dimension extraction from engineering drawings, achieving a time saving of
92.14% compared to the manual dimension extraction process.
|
|---|