By seamlessly integrating advanced machine vision and deep learning, MVTec is driving a new era in automotive quality control.
Precision Meets Efficiency on the Assembly Line
Consistent, defect free welds are critical to certified safety and structural integrity in modern car assembly. To meet these high standards, Spanish automation firm DGH, now part of Groupe ADF, deployed an AI powered solution on behalf of a major French automaker. The system automatically captures images of MIG and laser welded body in white joints and flags any anomalies in real time.
Utilising MVTec’s HALCON software and Deep Learning Tool, the system applies Instance Segmentation to localise weld seams and Anomaly Detection to reliably separate acceptable from defective welds. This yields high detection accuracy and robust performance even under fluctuating lighting and reflective metal surfaces.
From Minimal Data to Maximum Insight
Unlike traditional AI systems that depend on extensive datasets of flawed examples, the MVTec approach thrives on defective free “good” images. According to Guillermo Martín, Innovation and Technology Director at DGH, only a few defect bearing samples are needed to fine tune thresholds, making the system practical even when defect data is scarce.
Labelling these samples is simplified through the Smart Label Tool in MVTec’s Deep Learning Tool, which automatically identifies weld seam areas with minimal user input. The result is faster, more reliable training with significantly reduced effort.
Deployment and Impact
The first installation went live in early 2024 at the OEM plant, delivering clear benefits in throughput and consistency. Following that success, a second system commenced operation in April 2024. According to DGH, the automated system has markedly decreased dependency on skilled labor, accelerated inspection cycles, and enhanced overall weld quality.
Why It Matters
This inspection system exemplifies a broader shift across industries. AI enabled machine vision is replacing manual inspections with automated, objective, and reproducible processes. Deep learning combined with traditional vision tools is raising defect detection rates while lowering false positives across manufacturing sectors.
Bottom Line: The Future Is Now
MVTec’s HALCON remains a cornerstone in industrial vision technology. Its flexible architecture, AI integration, and support for edge computing hardware enable real time, high precision inspection, driving consistent quality control across demanding production environments.
As the automotive sector accelerates toward fully automated production lines, solutions like this project illustrate how AI in machine vision is rapidly reducing costs, boosting accuracy, and reinforcing safety.
