AI-Driven Multi-Stage Computer Vision System for Defect Detection in Laser-Engraved Industrial Nameplates

TL;DR

This work tackles automatic defect detection for laser-engraved nameplates in air-disc brake manufacturing by engineering a multi-stage AI pipeline. It combines YOLOv7-tiny for string/logo detection, Tesseract OCR for character recognition, a traditional image-diff logo defect detector, and a Residual Variational Autoencoder (ResVAE) for character-level anomaly detection, all validated against MES cross-checks. The PoC demonstrates strong recall across stages (notably 100% in several modules) and competitive accuracy, though precision is impacted by lighting and processing artifacts, highlighting opportunities for adaptive thresholding, improved binarization, and enhanced preprocessing. Overall, the approach advances automated, end-to-end nameplate inspection by localizing defects at the character and logo level, potentially reducing manual inspection and improving manufacturing traceability. The results underscore a viable path toward robust, scalable quality control in laser-engraved nameplates with implications for production efficiency and safety.

Abstract

Automated defect detection in industrial manufacturing is essential for maintaining product quality and minimizing production errors. In air disc brake manufacturing, ensuring the precision of laser-engraved nameplates is crucial for accurate product identification and quality control. Engraving errors, such as misprints or missing characters, can compromise both aesthetics and functionality, leading to material waste and production delays. This paper presents a proof of concept for an AI-driven computer vision system that inspects and verifies laser-engraved nameplates, detecting defects in logos and alphanumeric strings. The system integrates object detection using YOLOv7, optical character recognition (OCR) with Tesseract, and anomaly detection through a residual variational autoencoder (ResVAE) along with other computer vision methods to enable comprehensive inspections at multiple stages. Experimental results demonstrate the system's effectiveness, achieving 91.33% accuracy and 100% recall, ensuring that defective nameplates are consistently detected and addressed. This solution highlights the potential of AI-driven visual inspection to enhance quality control, reduce manual inspection efforts, and improve overall manufacturing efficiency.

Figures (13)

• Figure 1: (a) Annotated defective laser-engraved result from YOLOv7 wang2023yolov7, with a blue bounding box for the logo and yellow bounding boxes for strings, along with confidence values, (b) Final pipeline output with defects highlighted in red boxes.
• Figure 2: The complete pipeline for nameplate defect detection and string verification. (a) Image alignment, (b) YOLO-based string and logo detection, (c) Logo defect detection through image differencing, (d) OCR for character recognition/detection, and (e) Anomaly detection using ResVAE.
• Figure 3: Logo defect detection. (a) Extracted reference (top) and captured images (bottom). (b) Pre-processing with noise reduction and smoothing. (c) Pixel-wise difference. (d) Binary thresholding and morphological operations. (e) Connected component analysis. (f) Defect localization and visualization through contouring.
• Figure 4: Character Detection and Recognition. (a) Gaussian Blur, (b) Background subtraction, (c) Bitwise NOT, (d) Binary thresholding and morphological operations, (e) Tesseract OCR detection/recognition, (f) OCR output, (g) Bounding Box Adjustment
• Figure 5: Block architecture of the proposed Residual Variational Autoencoder.