Article Details

Abstract

To address the trade-off between robustness and time budget when decoding QR codes and 1D barcodes on resource-constrained edge devices, this paper proposes TAPER, a training-free, budget-aware framework. The method extracts low-overhead visual cues (contrast, tilt angle, object scale) to drive a conditional gating mechanism, transforming the conventional fixed pipeline into a multi-stage dynamic trial-based process that supports on-demand triggering of adaptive thresholding, rotation correction, and early termination. Experimental results on a self-constructed challenging dataset show that TAPER achieves a success rate of 56.44% under a 400 ms budget, representing an approximately 10% improvement over baseline methods, and reduces the average latency to 199.2 ms. It also delivers competitive performance on public benchmarks. The framework exhibits strong practicality and reproducibility, making it suitable for real-time deployment on edge devices.

Keywords

QR code decoding; 1D barcode decoding; edge computing; time budget

References

1. [1] OpenCV Contributors. OpenCV: Open Source Computer Vision Library [EB/OL].[2026-04-12].https://github.com/opencv/.
2. [2] Sauvola J，Pietikäinen M. Adaptive Document Image Binarization [J].Pattern Recognition，2000，33（2）：225-236.
3. [3] Liao M H，Zou Z S，Wan Z Y，et al. Real-time Scene Text Detection with Differentiable Binarization and Adaptive Scale Routing [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2023，45（1）：1198-1212.
4. [4] Du Y N，Li C X，Guo R Y，et al. PP-OCRv2: Bag of Tricks for Ultra Lightweight OCR System [PP/OL].https://arxiv.org/abs/2109.03144.
5. [5] Vasu P K A，Gabriel J，Feser J，et al. MobileOne: An Improved One Millisecond Mobile Backbone [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，2023：7107-7117.
6. [6] Ding X H，Zhang X Y，Ma N N，et al. RepVGG: Making VGG-style ConvNets Great Again [C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，2021：13733-13742.
7. [7] 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]//2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)，2023：7464-7475.
8. [8] Zou Z X，Chen K Y，Shi Z W，et al. Object Detection in 20 Years: A Survey [J].Proceedings of the IEEE，2023，111（3）：257-276.
9. [9] Teerapittayanon S，McDanel B，Kung H T. BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks [C]//2016 23rd International Conference on Pattern Recognition (ICPR).IEEE，2016：2464-2469.
10. [10] Han Y Z，Huang G，Song S J，et al. Dynamic Neural Networks: A Survey [J].IEEE Transactions on Pattern Analysis and Machine Intelligence，2022，44（11）：7436-7456.
11. [11] Wang Y，Shen J H，Hu T-K，et al. Dual Dynamic Inference: Enabling more Efficient, Adaptive, and Controllable Deep Inference [J].IEEE Journal of Selected Topics in Signal Processing，2020，14（4）：623-633.
12. [12] Liu S H，Zha J L，Sun J，et al. EdgeYOLO: An Edge-real-time Object Detector [PP/OL].https://arxiv.org/abs/2302.07483.

Funding

Basic Research Project of Higher Education Institutions in Liaoning Province, China (LJKQZ20222451)

About the authors

Ji Lupeng (1989-), male, Han nationality, from Heze, Shandong Province, engineer, bachelor's degree, research direction: edge computing, information security; Wu Zhanteng (2004-), male, Han nationality, born in Yingkou, Liaoning Province, is an undergraduate student. His research interests include edge computing and deep learning; Correspondence author: Wei Hao (1993-), male, Han nationality, from Jinan, Shandong Province, associate professor, master's supervisor, doctoral candidate, research direction: data mining, edge computing.