Article Details

Title · Authors · Abstract · Keywords · DOI · References

Real-time Detection Algorithm for Surface Defects Based on Optical Lenses

HAN Junru, GUO Jianfeng

(College of Science and Technology Ningbo University, Ningbo 315300, China)

Section：Computer Technology · Vol. 2026, Issue 3 · Pages：127-132 · 2096-4706（2026）03-0127-06 · CLC number：TP391.4 · Document code：A · DOI：10.19850/j.cnki.2096-4706.2026.03.024

Abstract

YOLO series models have become the benchmark in the real-time detection domain due to their balanced efficiency and accuracy, yet their performance is restricted by the inherent flaws of NMS post-processing, including additional computational overhead and the erroneous elimination of valid information. Although end-to-end Transformer frameworks can break free from the reliance on NMS, their excessively high computational cost limits practical application. To address this, the study proposes a dual-dimensional innovative architecture, which constructs a dedicated Transformer network for real-time end-to-end detection and develops a hybrid Mamba-Transformer backbone network named Mamba Vision, significantly enhancing the multi-scale feature representation capability. A phased optimization strategy is adopted, with parallel acceleration achieved by decoupling intra-scale and cross-scale feature processing for speed improvement and an uncertainty-minimization query initialization strategy introduced for accuracy enhancement to boost the accuracy of target localization. On the COCO benchmark, the proposed model achieves a mean Average Precision (mAP) of 53.05% with an inference speed exceeding 100 FPS on a T4 GPU. Compared with mainstream YOLO variants, it improves detection accuracy by 4.2%～6.7%, achieves breakthroughs in both speed and accuracy and thus provides a novel paradigm for industrial real-time detection systems.

Keywords

optical lens; real-time Object Detection; end-to-end

References

  1. [1] ZENG F G，DONG B，ZHANG Y. MOTR: End-to-end Multiple-object Tracking with Transformer [C]//European Conference on Computer Vision.Heidelberg：Springer，2022：659-675.
  2. [2] NAWARATNE R，ALAHAKOON D，DE SILVA D. Spatiotemporal Anomaly Detection Using Deep Learning For Real-Time Video Surveillance [J].IEEE Transactions on Industrial Informatics，2019，16（1）：393-402.
  3. [3] BOGDOLL D，NITSCHE M，ZÖLLNER J M. Anomaly Detection in Autonomous Driving: A Survey [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans：IEEE，2022：4488-4499.
  4. [4] BOCHKOVSKIY A，WANG C Y，LIAO H Y M. YOLOv4: Optimal Speed and Accuracy of Object Detection [J/OL].arXiv:2004.10934 [cs.CV].（2020-04-23）.https://arxiv.org/abs/2004.10934.
  5. [5] GE Z，LIU S，WANG F，et al. YOLOX: Exceeding YOLO Series in 2021 [J/OL].arXiv:2107.08430 [cs.CV].（2021-08-06）.https://arxiv.org/abs/2107.08430.
  6. [6] JOCHER G. YOLOv5 Release v7.0 [EB/OL].（2022-11-22）[2025-07-03].https://github.com/ultralytics/yolov5/tree/v7.0.
  7. [7] JOCHER G. YOLOv8 [EB/OL].（2024-07-12）[2025-07-05].https://github.com/ultralytics/ultralytics/tree/main.
  8. [8] HUANG X，WANG X X，LV W Y. PP-YOLOv2: A practical object detector [J/OL].（2021-04-21）[2025-07-16].https://arxiv.org/abs/2104.10419.
  9. [9] LI C Y，LI L L，GENG Y F，et al. YOLOv6 v3.0: A Full-Scale Reloading [J/OL].arXiv:2301.05586 [cs.CV].（2023-01-13）.https://arxiv.org/abs/2301.05586.
  10. [10] LONG X，DENG K P，WANG G Z，et al. PP-YOLO:An Effective and Efficient Implementation of Object Detector [J/OL].arXiv:2007.12099 [cs.CV].（2020-08-03）.https://arxiv.org/abs/2007.12099.
  11. [11] REDMON J，FARHADI A. YOLOv3: An Incremental Improvement [J/OL].arXiv:1804.02767 [cs.CV].（2018-04-08）.https://arxiv.org/abs/1804.02767.
  12. [12] WANG C Y，BOCHKOVSKIY A，LIAO H Y M. YOLOv7: Trainable bag-of-Freebies Sets New State-of-the-art for Real-time Object Detectors [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Vancouver：IEEE，2023：7464-7475.
  13. [13] XU S L，WANG X X，LV W Y. PP-YOLOE: An evolved version of YOLO [J/OL].arXiv:2203.16250 [cs.CV].（2022-12-12）.https://arxiv.org/abs/2203.16250.
  14. [14] ZHU X Z，SU W J，LU L W. Deformable DETR: Deformable Transformers for end-to-end Object Detection [J/OL].arXiv:2010.04159 [cs.CV].（2021-03-18）.https://arxiv.org/abs/2010.04159.
  15. [15] YAO Z Y，AI J B，LI B X. Efficient DETR: Improving End-to-end Object Detector with Dense Prior [J/OL].arXiv:2104.01318 [cs.CV].（2021-04-03）.https://arxiv.org/abs/2104.01318.
  16. [16] ZHANG H，LI F，LIU S L. DINO：DETR with Improved Denoising Anchor Boxes for End-to-End Object Detection [J/OL].arXiv:2203.03605 [cs.CV].（2022-07-11）.https://arxiv.org/abs/2203.03605.
  17. [17] GU A，DAO T. Mamba: Linear-time Sequence Modeling with Selective State Spaces [J/OL].arXiv:2312.00752v2 [cs.LG].（2024-05-31）.https://arxiv.org/html/2312.00752v2.
  18. [18] PEI X H，HUANG T，XU C. EfficientVMamba: Atrous Selective Scan for Lightweight Visual Mamba [J/OL].arXiv:2403.09977 [cs.CV].（2024-03-15）.https://arxiv.org/abs/2403.09977.
  19. [19] ZHU L H，LIAO B C，ZHANG Q. Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model [C]//Proceedings of the 41st International Conference on Machine Learning.Vienna：PMLR，2024：62429-62442.
  20. [20] REDMON J，DIVVALA S，GIRSHICK R，et al. You Only Look Once: Unified, Real-time Object Detection [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.Las Vegas：IEEE，2016：779-788.
  21. [21] CARION N，MASSA F，SYNNAEVE G，et al. End-to-end Object Detection With Transformers [C]//European Conference on Computer Vision. Heidelberg：Springer，2020：213-229.
  22. [22] ZHU X Z，SU W J，LU L W. Deformable DETR: Deformable Transformers for End-to-end Object Detection [J/OL].arXiv:2010.04159 [cs.CV].（2021-03-18）.https://arxiv.org/abs/2010.04159.
  23. [23] LIU S L，LI F，ZHANG H. DAB-DETR: Dynamic Anchor Boxes are Better Queries for DETR [J/OL].arXiv:2201.12329 [cs.CV].（2022-03-30）https://arxiv.org/abs/2201.12329.
  24. [24] LI F，ZHANG H，LIU S L. DN-DETR: Accelerate DETR Training by Introducing Query Denoising [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.New Orleans：IEEE，2022：13619-13627.
  25. [25] CHEN Q，CHEN X K，ZENG G. Group DETR: Fast Training Convergence with Decoupled one-to-many Label Assignment [J/OL].arXiv:2207.13085 [cs.CV].（2023-08-31）.https://arxiv.org/abs/2207.13085.
  26. [26] ZHU L H，LIAO B C，ZHANG Q. Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model [C]//Proceedings of the 41st International Conference on Machine Learning.Vienna：PMLR，2024：62429-62442.
  27. [27] WANG Y M，ZHANG X Y，YANG T. Anchor DETR: Query Design for Transformer-based Detector [C]//Proceedings of the AAAI Conference on Artificial Intelligence. Vancouver：AAAI，2022：2567-2575.
  28. [28] MENG D P，CHEN X K，FAN Z J. Conditional DETR for Fast Training Convergence [C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal：IEEE，2021：3651-3660.
  29. [29] GAO P，ZHENG M H，WANG X G. Fast Convergence of DETR with Spatially Modulated Co-Attention [C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal：IEEE，2021：3621-3630.

About the authors

Han Junru (1993-), female, Han ethnicity, from Hulunbuir, Inner Mongolia, teaching assistant, master's student, research direction: image processing and artificial intelligence; Guo Jianfeng (1986-), male, Han nationality, from Jinzhong, Shanxi Province, associate professor, doctoral student, research direction: image processing and artificial intelligence.