To address the problems of low efficiency, high labor intensity, and significant safety risks associated with traditional manual inspection methods in railway signal equipment rooms, a multi-task visual collaborative recognition framework based on a “backbone detection+task-oriented” strategy is proposed, enabling high-precision identification and condition monitoring of key equipment in signal rooms. Firstly, by analyzing the inspection environment and functional requirements of signal equipment rooms, a unified object detection model based on YOLOv12 is constructed to achieve the classification and localization of main inspection targets, including digital meters, pointer meters and indicator lights. Subsequently, differentiated task-level recognition strategies are designed according to the characteristics of different devices: a fusion recognition method combining the wire-crossing technique and PaddleOCR is adopted for digital meters; a reading method based on image processing and PaddleOCR is applied for pointer meters; and a state recognition method integrating the Hough transform, HSV color space analysis, and PaddleOCR is used for indicator lights. Finally, experimental results on a self-constructed dataset demonstrate that the proposed method achieves accuracies of 94.0%, 92.4%, 91.5% and 94.2% in meter classification, digital display recognition, pointer reading, and indicator light state identification, respectively. Compared with traditional methods, the proposed framework significantly improves recognition accuracy and detection efficiency, exhibiting strong engineering applicability.
WangShizhong. Research on intelligent inspection system of high speed railway signal equipment room[D]. Beijing: China Academy of Railway Sciences, 2022:45. (in Chinese)
[3]
WeiX, ReyW P. Advancements in substation inspection robots: a review of research and development[J]. E3S Web of Conferences, 2024, 528: 02017.
[4]
QinZ, XuZ, SunQ, et al. Investigation of intelligent substation inspection robot by using mobile data[J]. International Journal of Humanoid Robotics, 2023, 20(3): 2240003.
[5]
梁松伟. 基于多源异构视觉的变电站机器人巡检技术研究[D]. 杭州: 浙江大学, 2021: 3.
[6]
LiangSongwei. Research on inspection technology of substation robot based on multi-source heterogeneous vision[D]. Hangzhou: Zhejiang University, 2021: 3. (in Chinese)
LiuMingfeng, LiuSunjun, GuoShunsen, et al. Study and design of computer room inspection system based on DBN image recognition[J]. Measurement & Control Technology, 2018, 37(11): 45-49. (in Chinese)
ChenZhuoxiong, YuXiaoli. Research on the application of intelligent inspection robot in subway communication and signal room[J]. China New Telecommunications, 2019, 21(11): 7-8. (in Chinese)
HuangJiadong, ChenHaiping. Design and application of intelligent inspection robot in data center computer room[J]. Computer & Network, 2017, 43(22): 46-47. (in Chinese)
[15]
MaY, JiangQ. A robust and high-precision automatic reading algorithm of pointer meters based on machine vision[J]. Measurement Science and Technology, 2019, 30(1): 015401.
GuoNini, QiaoGangzhu, ZhangGuanghua, et al. Classification of diabetic retinopathy using generative adversarial networks and dual attention network[J]. Journal of North University of China (Natural Science Edition), 2023, 44(1): 39-47. (in Chinese)
[18]
MustafaT, KarabatakM. Real time car model and plate detection system by using deep learning architectures[J]. IEEE Access, 2024, 12: 107616-107630.
[19]
刘文亮. 七段式数显仪表中数字识别的研究与实现[D]. 大连: 大连理工大学, 2013: 16.
[20]
LiuWenliang. Research and implementation of number recognitionin seven-segment numeric instrument[D]. Dalian: Dalian University of Technology, 2013: 16. (in Chinese)
ZHANGXun, HUANGJunkai, ZHAOChao, et al. High voltage circuit breaker mechanical fault diagnosis technology based on KPCA-SVM[J]. Journal of Test and Measurement Technology, 2023, 37(2): 158-164. (in Chinese)
[23]
KharinaN, ChernyadyevS. Software for car license plates recognition with minimal computing resources[C]//2022 24th International Conference on Digital Signal Processing and its Applications (DSPA), 2022: 1-4.
[24]
KuindersmaS, DeitsR, FallonM, et al. Optimization-based locomotion planning, estimation, and control design for the atlas humanoid robot[J]. Autonomous Robots, 2016, 40(3): 429-455.
[25]
WangS, ZhouC, ShenY. Design of a new single portable six foot robot based on curved foot[C]//4th International Conference on Artificial Intelligence and Advanced Manufacturing (AIAM), 2023: 283-287.
[26]
WangH, HuY, YinH, et al. Research on the recognition method of dial scales for arrester pointer instruments based on deep learning[J]. Applied Sciences, 2024, 14(5): 2049.
[27]
HuangZ Y. Research on single image super-resolution based on very deep super-resolution convolutional neural network[J]. Journal of Measurement Science and Instrumentation, 2022, 13(3): 276-283.
YuYangbo, DongZhaoyang, LiJun, et al. Algorithm for recognizing the status of plates and indicator lights based on machine vision technology[J]. Power Systems and Big Data, 2024, 27(6): 81-92. (in Chinese)
SuMo, ZhaoJiaxu, GuoXiangkun, et al. Device state recognition method based on augmented reality[J]. Journal of Chinese Computer Systems, 2025, 46(7): 1684-1691. (in Chinese)
ChenHaotian. Research on substation indicator status recognition method based on text positioning[D]. Beijing: North China Electric Power University, 2024: 37. (in Chinese)
ChenTan. Research on inspection system and recognition algorithm for data center[D]. Hangzhou: Zhejiang Sci-Tech University, 2023: 26. (in Chinese)
[37]
SarkarO, SinhaS, JenaA K, et al. Automatic number plate character recognition using paddle-OCR[C]//2024 International Conference on Innovations and Challenges in Emerging Technologies (ICICET), 2024: 1-7.
[38]
BaiW, ZhaoJ, DaiC, et al. Two novel models for traffic sign detection based on YOLOv5s[J]. Axioms, 2023, 12(2): 160.
[39]
ZhengX, BiJ, LiK, et al. SMN-YOLO: lightweight YOLOv8-based model for small object detection in remote sensing images[J]. IEEE Geoscience and Remote Sensing Letters, 2025, 22: 8001305.
[40]
WeiZ, WeiY. YOLOv10n-based defect detection in power insulators: attention enhancement and feature fusion optimization[J]. IEEE Access, 2025, 13: 115888-115897.