CORROSION DETECTION AT TRANSMISSION ACCESSORIES USING COMBINATION OF OBJECT DETECTION, IMAGE CLASSIFICATION AND BACKGROUND REMOVAL
Inspection of the condition of electrical transmission accessories is an important aspect of maintaining the reliability of electricity supply. This research aims to develop an automatic corrosion detection system using a combination of object detection and image classification. The dataset used...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/87819 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Inspection of the condition of electrical transmission accessories is an important
aspect of maintaining the reliability of electricity supply. This research aims to
develop an automatic corrosion detection system using a combination of object
detection and image classification. The dataset used includes specific accessories
from the Indonesian environment such as Clevis, Dead End Compression, Shackle,
Tension Clamp, Hole, and Bolt, which have diverse backgrounds and small sizes.
The methods used involve state-of-the-art models such as YOLOv9e for object
detection and Xception for image classification. Experiments show that reducing
the number of classes and increasing the object size can improve model
performance up to 0.972 mAP@0.5 using YOLOv9e. However, background
removal has a negative impact on performance with an mAP@0.5 score of less than
0.20. The combination of object detection and image classification yielded an F1-
score of 84.05 on field test data. The results of this research indicate the potential
use of deep learning technology in the inspection of electrical transmission
accessories in Indonesia with several challenges related to dataset quality and
pipeline integration. |
|---|