Automated Detection and Recognition of Engraved Serial Numbers on Metallic Surfaces

Project Code :TCMAPY2501

Objective

The objective of this project is to develop an automated system for detecting and recognizing engraved serial numbers on metallic surfaces using deep learning techniques, specifically YOLOv8-CBAM and RT-DETR-Swin. The project aims to enhance industrial quality control by accurately identifying engraved Meter no and kWh-rating markings from metallic components captured through industrial inspection cameras. By leveraging YOLOv8-CBAM for attention-guided detection and RT-DETR-Swin for multi-scale feature extraction, the system automatically localizes and classifies engraved serial number regions. The integration of YOLOwill provide visual interpretability by highlighting regions in the image that influence the model's decisions. The goal is to create an efficient, automated solution for engraved serial number recognition on metallic surfaces, enabling rapid and accurate verification of Meter no and kWh-rating markings and enhancing industrial quality control management systems

Abstract

This paper presents a novel approach to automated detection and recognition of engraved serial numbers on metallic surfaces using YOLOv8-CBAM, RT-DETR-Swin, and YOLO. The system aims to identify and classify two target classes—Meter no and kWh-rating—engraved on metallic components by processing image datasets through advanced object detection techniques. YOLOv8-CBAM integrates the Convolutional Block Attention Module into the YOLOv8 backbone to improve feature discrimination on metallic textures with low contrast. RT-DETR-Swin employs the Swin Transformer encoder for enhanced multi-scale feature extraction, making it well-suited for detecting fine engraved characters. YOLOis utilized for visual interpretability, providing insights into the regions of interest that influence model predictions. The detection pipeline focuses on two classes, Meter no and kWh-rating, and optimizes performance by fine-tuning hyperparameters and leveraging attention-based feature extraction. The methodology improves robustness and precision in industrial inspection tasks, offering a scalable solution for automated quality control on manufacturing lines.

Keywords: YOLOv8-CBAM, RT-DETR-Swin, YOLO, CBAM, engraved serial number detection, Meter no, kWh-rating, metallic surface inspection, deep learning, computer vision

NOTE: Without the concern of our team, please don't submit to the college. This Abstract varies based on student requirements.