A Survey of AI-Based Visual Defect Detection Techniques for Automotive Manufacturing

Abstract

Quality inspection is an essential process in automotive manufacturing, as defects in vehicle components can affect product quality, reliability, and production efficiency. Traditional inspection methods often rely on manual examination, which can be time-consuming and may produce inconsistent results under varying manufacturing conditions. Recent advancements in Artificial Intelligence (AI) and Computer Vision have created opportunities for automated visual defect inspection. This paper presents a survey of AI-based visual defect detection techniques applicable to automotive manufacturing. It examines approaches such as You Only Look Once (YOLO), image segmentation, and anomaly detection for identifying defects including scratches, dents, cracks, corrosion, and surface irregularities. The survey aims to compare these techniques based on their reported capabilities, limitations, computational requirements, and suitability for industrial applications. It also examines challenges related to dataset availability, class imbalance, changing environmental conditions, and deployment in real-world manufacturing environments. Based on the literature review, research gaps are identified to establish a foundation for Defect Lens, a proposed AI-powered visual quality inspection platform for automotive manufacturing.

Introduction

The paper presents a survey of AI-based visual defect detection techniques for automotive manufacturing and introduces Defect Lens, a proposed AI-powered quality inspection platform designed to improve product quality, inspection accuracy, and manufacturing efficiency.

Traditional quality inspection relies on manual visual examination, which is time-consuming and prone to human error, fatigue, and inconsistent judgments. Defects such as scratches, dents, cracks, corrosion, and paint irregularities can affect vehicle safety, durability, and customer satisfaction. To overcome these limitations, computer vision and artificial intelligence can automate defect identification and localisation.

The literature survey reviews classical machine vision techniques, image preprocessing methods, deep learning models, object detection algorithms, and anomaly detection approaches. Techniques such as thresholding, edge detection, Local Binary Patterns (LBP), and image enhancement are useful under controlled conditions but often struggle with varying illumination and complex surface textures. Deep learning models, particularly CNNs and YOLO-based detectors, provide improved defect detection and localisation while supporting real-time processing. However, challenges remain in detecting small or low-contrast defects, handling reflective surfaces, and working with limited labelled datasets.

The proposed Defect Lens platform integrates React.js and TypeScript for the user interface, Java and Spring Boot for backend services, Python with OpenCV for image preprocessing, YOLO for defect detection, and PostgreSQL for storing inspection records. The system processes component images, enhances their quality, identifies and localises potential defects, and presents the results through a web-based dashboard. It also maintains structured inspection records to support traceability and quality management. The platform is designed to assist human inspectors, who retain responsibility for the final acceptance or rejection of components.

The study identifies several research gaps, including the limited availability of automotive-specific defect datasets, inadequate robustness to varying lighting conditions, difficulties in detecting small defects, and insufficient integration of AI models with practical industrial inspection systems.

Conclusion

AI-based visual defect detection has emerged as a promising approach for improving quality inspection in automotive manufacturing. This survey reviewed AI-based techniques, particularly YOLO-based object detection, image preprocessing, and computer vision methods, for identifying defects such as scratches, dents, cracks, and corrosion on automotive components. The discussion highlights that detection performance depends on several factors, including image quality, lighting conditions, training data diversity, defect size, and processing requirements. Challenges such as limited labelled datasets, subtle defects, variations in surface appearance, and prediction reliability must be addressed before these systems can be adopted in manufacturing environments. The proposed Defect Lens platform aims to integrate React.js and TypeScript for the user interface, Java and Spring Boot for backend processing, Python and OpenCV for image processing, YOLO for defect detection, and PostgreSQL for storing inspection records. This integrated approach aims to identify potential defects, organise inspection information, and present results through a user-friendly dashboard. Future improvements could include defect severity assessment, enhanced detection of small defects, explainable AI, and integration with manufacturing workflows. The actual effectiveness of Defect Lens must be established through implementation, testing, and evaluation using suitable automotive defect images. Overall, AI-based visual inspection can support more consistent and efficient automotive quality-control processes when combined with reliable data, appropriate validation, and human supervision.

References

[1] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once: Unified, real-time object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2016, pp. 779–788, doi: 10.1109/CVPR.2016.91. [2] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards real-time object detection with region proposal networks,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 39, no. 6, pp. 1137–1149, Jun. 2017. [3] W. Liu et al., “SSD: Single shot multibox detector,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2016, pp. 21–37. [4] [4] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal speed and accuracy of object detection,” arXiv:2004.10934, 2020. [5] C.-Y. Wang, A. Bochkovskiy, and H.-Y. M. Liao, “YOLOv7: Trainable bag-of-freebies sets new state-of-the-art for real-time object detectors,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2023. [6] J. Terven, D.-M. Córdova-Esparza, and J.-A. Romero-González, “A comprehensive review of YOLO architectures in computer vision: From YOLOv1 to YOLOv8 and YOLO-NAS,” Mach. Learn. Knowl. Extr., vol. 5, no. 4, pp. 1680–1716, 2023. [7] G. Bradski, “The OpenCV Library,” Dr. Dobb’s J. Softw. Tools, 2000. [8] K. Zuiderveld, “Contrast limited adaptive histogram equalization,” in Graphics Gems IV. San Diego, CA, USA: Academic Press, 1994, pp. 474–485. [9] J. Canny, “A computational approach to edge detection,” IEEE Trans. Pattern Anal. Mach. Intell., vol. PAMI-8, no. 6, pp. 679–698, Nov. 1986. [10] D. Czimmermann et al., “Visual-based defect detection and classification approaches for industrial applications—A survey,” Sensors, vol. 20, no. 5, Art. no. 1459, 2020. [11] P. Bergmann, M. Fauser, D. Sattlegger, and C. Steger, “MVTec AD—A comprehensive real-world dataset for unsupervised anomaly detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2019, pp. 9592–9600. [12] K. Song and Y. Yan, “A noise robust method based on completed local binary patterns for hot-rolled steel strip surface defects,” Appl. Surf. Sci., vol. 285, pp. 858–864, 2013. [13] Y. He, K. Song, Q. Meng, and Y. Yan, “An end-to-end steel surface defect detection approach via fusing multiple hierarchical features,” IEEE Trans. Instrum. Meas., vol. 69, no. 4, pp. 1493–1504, Apr. 2020. [14] K. Roth et al., “Towards total recall in industrial anomaly detection,” in Proc. IEEE/CVF Conf. Comput. Vis. Pattern Recognit. (CVPR), 2022, pp. 14318–14328. [15] C. Shorten and T. M. Khoshgoftaar, “A survey on image data augmentation for deep learning,” J. Big Data, vol. 6, Art. no. 60, 2019. [16] S. J. Pan and Q. Yang, “A survey on transfer learning,” IEEE Trans. Knowl. Data Eng., vol. 22, no. 10, pp. 1345–1359, Oct. 2010. [17] T. Ojala, M. Pietikäinen, and T. Mäenpää, “Multiresolution gray-scale and rotation invariant texture classification with local binary patterns,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 24, no. 7, pp. 971–987, Jul. 2002. [18] T.-Y. Lin et al., “Feature pyramid networks for object detection,” in Proc. IEEE Conf. Comput. Vis. Pattern Recognit. (CVPR), 2017, pp. 2117–2125. [19] O. Ronneberger, P. Fischer, and T. Brox, “U-Net: Convolutional networks for biomedical image segmentation,” in Proc. Int. Conf. Med. Image Comput. Comput.-Assist. Intervent. (MICCAI), 2015, pp. 234–241. [20] R.-G. Tulbure et al., “A review on modern defect detection models using DCNNs—Deep convolutional neural networks,” Alexandria Eng. J., vol. 61, no. 10, 2022. [21] S. B. Jha and R. F. Babiceanu, “Deep CNN-based visual defect detection: Survey of current literature,” Comput. Ind., vol. 148, Art. no. 103911, Jun. 2023, doi: 10.1016/j.compind.2023.103911. [22] G. Jocher et al., “Ultralytics YOLO,” software, 2023. [Online]. Available: https://github.com/ultralytics/ultralytics [23] R. R. Selvaraju et al., “Grad-CAM: Visual explanations from deep networks via gradient-based localization,” in Proc. IEEE Int. Conf. Comput. Vis. (ICCV), 2017, pp. 618–626. [24] S. M. Lundberg and S.-I. Lee, “A unified approach to interpreting model predictions,” in Proc. Adv. Neural Inf. Process. Syst. (NeurIPS), 2017, pp. 4765–4774. [25] Y. Chen, Y. Ding, F. Zhao, E. Zhang, Z. Wu, and L. Shao, “Surface defect detection methods for industrial products: A review,” Appl. Sci., vol. 11, no. 16, Art. no. 7657, 2021, doi: 10.3390/app11167657. [26] H. Huang and K. Zhu, “Automotive parts defect detection based on YOLOv7,” Electronics, vol. 13, no. 10, Art. no. 1817, 2024, doi: 10.3390/electronics13101817. [27] A. Garcia-Garcia, S. Orts-Escolano, S. Oprea, V. Villena-Martinez, and J. Garcia-Rodriguez, “A review on deep learning techniques applied to semantic segmentation,” arXiv:1704.06857, 2017. [28] P. Carvalho, M. Lafou, A. Durupt, A. Leblanc, and Y. Grandvalet, “Detecting visual anomalies in an industrial environment: Unsupervised methods put to the test on the AutoVI dataset,” Comput. Ind., vol. 163, Art. no. 104151, 2024, doi: 10.1016/j.compind.2024.104151. [29] M. Everingham, L. Van Gool, C. K. I. Williams, J. Winn, and A. Zisserman, “The PASCAL Visual Object Classes (VOC) challenge,” Int. J. Comput. Vis., vol. 88, no. 2, pp. 303–338, Jun. 2010. [30] T.-Y. Lin et al., “Microsoft COCO: Common objects in context,” in Proc. Eur. Conf. Comput. Vis. (ECCV), 2014, pp. 740–755. [31] Z. Zou, K. Chen, Z. Shi, Y. Guo, and J. Ye, “Object detection in 20 years: A survey,” Proc. IEEE, vol. 111, no. 3, pp. 257–276, Mar. 2023. [32] Meta Platforms, Inc., “React documentation.” [Online]. Available: https://react.dev (accessed: Oct. 13, 2026). [33] VMware/Broadcom, “Spring Boot reference documentation.” [Online]. Available: https://docs.spring.io/spring-boot/ (accessed: Oct. 13, 2026). [34] PostgreSQL Global Development Group, “PostgreSQL documentation.” [Online]. Available: https://www.postgresql.org/docs/ (accessed: Oct. 13, 2026).

Copyright

Copyright © 2026 Rutvi Landge, Sakshi Barhate, Rutuja Kashid, Ritesh Kadam, Manisha Darak. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.