基于改进 YOLOv8 的指针式仪表读数识别算法

Pointer Meter Reading Recognition Algorithm Based on Improved YOLOv8

摘要

针对现有深度学习的仪表读数识别算法在边缘设备上存在资源消耗较大和传统图像处理算法在复杂场景下鲁棒性不足、误差累积以及难以实现端到端指针提取等问题,提出了一种轻量化改进的 YOLOv8 模型实现仪表检测,并采用 YOLOv8-pose 关键点模型提取仪表盘关键点,拟合指针与刻度线结合角度法计算仪表读数。首先,设计轻量级的 RGELAN 模块替代 C2f 模块,降低骨干和颈部网络复杂度;其次,将考虑多尺度特征贡献的 CASC-Head 检测头替代解耦头,减少检测头参数量;最后,引入 Shape-IoU 优化回归损失函数,提升检测准确性。实验结果表明:改进后的模型 YOLOv8-RSS 在 P 和 mAP@50:95 上分别为 98.5%和 90.6%,相较于原始 YOLOv8 的 P 和 mAP@50:95 仅损失了 0.3%和 0.4%,但参数量、计算量和模型大小分别减少了 48.3%,44.4%和 46%;在复杂场景下,仪表读数阶段算法的平均相对误差、平均引用误差、参数量和检测速度分别为 1.425%,0.557%,3.08M 和 78 帧/s,与其他算法相比,该算法降低了空间占用和读数误差,提升了检测速度。

Abstract

To address the issues of existing deep learning-based meter reading algorithms on edge devices, such as high resource consumption, the lack of robustness, error accumulation, and difficulty in end-to-end pointer extraction in traditional image processing methods in complex scenarios, a lightweight improved YOLOv8 model was proposed for meter detection. Meanwhile, the YOLOv8-pose keypoint model was employed to extract keypoints from the meter dial, and the reading was calculated using an angle-based method by fitting the pointer and scale lines. Firstly, a lightweight RGELAN module was designed to replace the C2f module, reducing the complexity of the backbone and neck networks. Then, the CASC-Head detection head, which considered multi-scale feature contributions, replaced the decoupled head, reducing detection parameters. Finally, the Shape-IoU optimized regression loss was introduced to improve detection accuracy. Experimental results showed that the improved YOLOv8-RSS model achieved 98.5% precision and 90.6% mAP@50:95, with only 0.3% and 0.4% losses compared with the original YOLOv8, while reducing parameters, computation, and model size by 48.3%, 44.4%, and 46%, respectively. In complex scenarios, it achieved an average relative error of 1.425%, average absolute error of 0.557%, 3.08 MB parameters, and 78 frame per second. Compared with existing methods, the proposed algorithm reduced space consumption and reading errors, and improved detection speed.

关键词

仪表读数识别 / 轻量化 YOLOv8 / SIFT 倾斜校正 / 关键点检测 / 角度法

Key words

meter reading recognition / lightweight YOLOv8 / SIFT tilt correction / key point detection / angle method

参考文献

[1] ZHANG J Q, ZHANG M, QI G W, et al. Reading recognition method of mechanical pointer meter based on machine vision[J].Engineering Advances, 2023, 3(1): 34-38.

[2] 钱玉宝,王紫涵,邱腾煌. 指针式仪表读数识别的研究现状与发展[J].电子测量技术,2024,47(8):110-119.

[3] QIAN Y B, WANG Z H, QIU T H. Research status and development of pointer meter reading recognition[J].Electronic Measurement Technology, 2024, 47(8): 110-119.

[4] 陶中涵,万熠,张桂新,等. 改进 YOLOv8 与特征匹配的单指针仪表识别方法研究[J/OL].机械科学与技术,2024:1-7.(2024—06—20)[2025—03—22].https://link.cnki.net/doi/10.13433/j.cnki.1003—8728.20240082.

[5] TAO Z H, WAN Y, ZHANG G X, et al. Research on improved YOLOv8 and feature matching for single—pointer instrument recognition method[J/OL].Mechanical Science and Technology for Aerospace Engineering, 2024: 1-7. (2024—06—20)[2025—03—22].https://link.cnki.net/doi/10.13433/j.cnki.1003—8728.20240082.

[6] 张建寰,曾亮,刘涛,等. 户外巡检机器人指针式仪表识别方法研究[J].仪表技术与传感器,2023(11):51—56,64.

[7] ZHANG J H, ZENG L, LIU T, et al. Research on pointer meter recognition algorithm of outdoor inspection robots[J].Instrument Technique and Sensor, 2023(11): 51—56, 64.

[8] 巩方超,王硕禾,周启斌,等. 基于 SURF 和 FLANN 算法的变电所仪表定位与读数识别[J].石家庄铁道大学学报(自然科学版),2020,33(1):110-115.

[9] GONG F C, WANG S H, ZHOU Q B, et al. Location and readout identification of instrument in transformer substation based on SURF and FLANN algorithm[J].Journal of Shijiazhuang Tiedao University(Natural Science Edition), 2020, 33(1): 110-115.

[10] 李巍,王鸥,刚毅凝,等. 一种自动读取指针式仪表读数的方法[J].南京大学学报(自然科学),2019,55(1):117-124.

[11] LI W, WANG O, GANG Y N, et al. An automatic reading method for pointer meter[J].Journal of Nanjing University(Natural Science), 2019, 55(1): 117-124.

[12] ZHANG C L, SHI L, ZHANG D D, et al. Pointer meter recognition method based on Yolov7 and hough transform[J].Applied Sciences—basel, 2023, 13(15): 8722.

[13] WU H, QI Z Y, TIAN H P, et al. SF₆ pointer pressure meter reading method based on fusion attention feature UNet[J].IEEE Access, 2023, 11: 107451-107462.

[14] 宫倩,别必龙,范新南,等. 基于关键点检测的指针仪表读数算法[J].电子测量与仪器学报,2023,37(3):66-73.

[15] GONG Q, BIE B L, FAN X N, et al. Pointer meter reading algorithm based on key point detection[J].Journal of Electronic Measurement and Instrumentation, 2023, 37(3): 66-73.

[16] JIAO W H, ZHAO D, MEI X, et al. Multiresolution deep feature learning for pointer meters reading recognition[J].Journal of Manufacturing Processes, 2024, 114: 168-177.

[17] TAN M, LE Q. EfficientNet: Rethinking model scaling for convolutional neural networks[EB/OL]. (2019—05—28)[2025—03—22].https://arxiv.org/abs/1905.11946.

[18] HAN K, WANG Y H, TIAN Q, et al. GhostNet: more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway: IEEE, 2020: 1577-1586.

[19] 张震,陈可鑫,陈云飞. 优化聚类和引入 CBAM 的 YOLOv5 管制刀具检测[J].郑州大学学报(工学版),2023,44(5):40—45,61.

[20] ZHANG Z, CHEN K X, CHEN Y F. YOLOv5 with optimized clustering and CBAM for controlled knife detection[J].Journal of Zhengzhou University(Engineering Science), 2023, 44(5): 40-45, 61.

[21] Ultralytics. YOLOv8[EB/OL]. (2023—01—10)[2025—03—23].https://github.com/ultralytics/ultralytics.

[22] Ultralytics. YOLOv5[EB/OL]. (2020—05—18)[2025—03—23].https://github.com/ultralytics/yolov5.

[23] WANG C Y, LIAO H M, YEH I H. Designing network design strategies through gradient path analysis[EB/OL]. (2022—11—09)[2025—03—23].https://arxiv.org/abs/2211.04800v1.

[24] DING X H, ZHANG X Y, MA N N, et al. RepVGG: making VGG—style ConvNets great again[C]//2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway: IEEE, 2021: 13733-13742.

[25] TIAN Z, SHEN C H, CHEN H, et al. FCOS: fully convolutional one—stage object detection[C]//2019 IEEE/CVF International Conference on Computer Vision(ICCV). Piscataway: IEEE, 2019: 9626-9635.

[26] ZHANG H, ZHANG S J. Shape—IoU: more accurate metric considering bounding box shape and scale[EB/OL]. (2024—01—12)[2025—03—23].https://arxiv.org/abs/2312.17663.

[27] LOWE D G. Distinctive image features from scale—invariant keypoints[J].International Journal of Computer Vision, 2024, 60(2): 91-110.

[28] 张香怡. 基于机器视觉的指针式仪表智能读数方法研究[D].北京:中国石油大学(北京),2021.

[29] ZHANG X Y. Research on intelligent reading method of pointer instrument based on machine vision[D]. Beijing: China University of Petroleum(Beijing),2021.

[30] 夏臻康,李维刚,田志强. 基于 YOLOv5 的指针式仪表自动读数方法研究[J].仪表技术与传感器,2023(6):44-51.

[31] XIA Z K, LI W G, TIAN Z Q. Research on automatic reading method of pointer meter based on YOLOv5[J].Instrument Technique and Sensor, 2023(6): 44-51.

[32] ZHOU D K, YANG Y, ZHU J, et al. Intelligent reading recognition method of a pointer meter based on deep learning in a real environment[J].Measurement Science and Technology, 2022, 33(5): 055021.

[33] NI T, MIAO H F, WANG L L, et al. Multi—meter intelligent detection and recognition method under complex background[C]//2020 39th Chinese Control Conference(CCC). Piscataway: IEEE, 2020: 7135-7141.

[34] REDMON J, FARHADI A. YOLOV3: an incremental improvement[EB/OL]. (2018—04—08)[2025—03—22].https://arxiv.org/abs/1804.02767.

[35] LI C Y, LI L L, GENG Y F, et al. YOLOv6 v3.0: a full—scale reloading[EB/OL]. (2023—01—13)[2025—03—22].https://arxiv.org/abs/2301.05586v1.

[36] WANG C Y, BOCHKOVSKIY A, LIAO H 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). Piscataway: IEEE, 2023: 7464-7475.

[37] WANG C Y, YEH I H, LIAO H M. YOLOv9: learning what you want to learn using programmable gradient information[EB/OL]. (2024—02—29)[2025—03—28].https://arxiv.org/abs/2402.13616v2.

[38] SUN K, XIAO B, LIU D, et al. Deep high—resolution representation learning for human pose estimation[C]//2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition(CVPR). Piscataway: IEEE, 2019: 5693-5703.

基金资助

河南省重点研发专项项目(231111211600)