Improvement of feature point filtering algorithm for dynamic scenarios in substations based on fusion of YOLOv5 and ORB-SLAM

doi:10.3969/j.issn.2097-0706.2026.02.005

Abstract

Abstract:

To address the degradation in localization and mapping accuracy of intelligent inspection robots under complex and dynamic operating conditions in substations，an enhanced localization and mapping architecture integrating improved CA-YOLOv5 target detection was proposed. A multimodal attention mechanism was used to optimize the CA-YOLOv5 network and construct a real-time dynamic target recognition framework. A semantic-geometric joint constraint strategy was used to establish a dynamic region mask and a motion probability model during the feature matching stage. A dynamic feature filtering algorithm based on spatiotemporal consistency was designed to achieve precise elimination of dynamic interference sources and the effective preservation of the static scene structure in the BA optimization process. Comparative experiments on public datasets and real dynamic scenarios demonstrated that the improved system reduced the localization errors by 43.7% and improved the map reconstruction completeness by 41.5% in dynamic environments，while maintaining satisfactory real-time processing performance. The fusion framework effectively solves the problems of mismatching and map pollution caused by dynamic elements，thereby overcoming typical dynamic disturbances in substations.

Key words: ORB-SLAM3, YOLO, intelligent inspection, sensor recognition, dynamic removal, substation, attention mechanism

CLC Number:

TM63：TP242

HE Longqing, LI Xiaoyong, SHI Xin, JIANG Han, LI Yuqiang, WANG Yongjun, WANG Kai. Improvement of feature point filtering algorithm for dynamic scenarios in substations based on fusion of YOLOv5 and ORB-SLAM[J]. Integrated Intelligent Energy, 2026, 48(2): 47-58.

Figures/Tables 13

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Fig.8

Fig.9

Fig.10

Table 1

Table 2

Table 3

References 36

[1]	姜寒, 王凯. 面向变电站无人化巡检的多源融合即时定位与建图方法综述[J]. 广东电力, 2025, 38(3): 55-68.
	JIANG Han, WANG Kai. A review of multi-source fusion SLAM methods for unmanned substation inspection[J]. Guangdong Electric Power, 2025, 38(3): 55-68.
[2]	刘涛, 李伟华, 汤熠. 综合智慧能源系统典型构架网络安全防护研究[J]. 综合智慧能源, 2024, 46(5): 81-90. doi: 10.3969/j.issn.2097-0706.2024.05.010
	LIU Tao, LI Weihua, TANG Yi. Security protection of typical networks for integrated smart energy systems[J]. Integrated Intelligent Energy, 2024, 46(5): 81-90. doi: 10.3969/j.issn.2097-0706.2024.05.010
[3]	MUR-ARTAL R, MONTIEL J M M, TARDOS J D. ORB-SLAM: A versatile and accurate monocular SLAM system[J]. IEEE Transactions on Robotics, 2015, 31(5): 1147-1163. doi: 10.1109/TRO.2015.2463671
[4]	CAI D P, LI R Q, HU Z H, et al. A comprehensive overview of core modules in visual SLAM framework[J]. Neurocomputing, 2024, 590: 127760. doi: 10.1016/j.neucom.2024.127760
[5]	王仕仪, 张学峰, 任彬, 等. 基于SLAM的楼层巡检智能车设计[J]. 科学技术创新, 2022(21): 183-186.
	WANG Shiyi, ZHANG Xuefeng, REN Bin, et al. Design of intelligent vehicle for floor inspection based on SLAM[J]. Scientific and Technological Innovation, 2022(21): 183-186.
[6]	赵康, 闫正义, 王凯. 光纤传感技术在电池荷电状态/健康状态监测中的研究进展[J/OL]. 发电技术, 2025: 1-12(2025-03-11)[2025-09-12]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=SLJX20250311001&dbname=CJFD&dbcode=CJFQ.
	ZHAO Kang, YAN Zhengyi, WANG Kai. Research advances of fiber optic sensing technology in battery SOC/SOH monitoring[J/OL]. Power Generation Technology, 2025: 1-12(2025-03-11)[2025-09-12]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=SLJX20250311001&dbname=CJFD&dbcode=CJFQ.
[7]	李彬, 张文艳, 王庆宇, 等. 支撑新型电力系统客户侧互动业务的融合通信发展及展望[J]. 综合智慧能源, 2025, 47(11): 72-86. doi: 10.3969/j.issn.2097-0706.2025.11.007
	LI Bin, ZHANG Wenyan, WANG Qingyu, et al. Development and prospects of converged communications supporting customer-side interactive services in new-type power system[J]. Integrated Intelligent Energy, 2025, 47(11): 72-86. doi: 10.3969/j.issn.2097-0706.2025.11.007
[8]	柳宁, 崔小军, 绳鹏鹏, 等. 基于SLAM的电厂巡检机器人导航技术实现[J]. 自动化与仪器仪表, 2023(7): 214-218.
	LIU Ning, CUI Xiaojun, SHENG Pengpeng, et al. Implementation of navigation technology for power plant inspection robots based on SLAM[J]. Automation & Instrumentation, 2023(7): 214-218.
[9]	TATENO K, TOMBARI F, LAINA I, et al. CNN-SLAM: Real-time dense monocular SLAM with learned depth prediction[C]// Proceedings of 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2017: 6565-6574.
[10]	蒋经纬, 吉月辉, 刘俊杰, 等. 基于轻量级CNN的视觉SLAM快速回环检测算法[J]. 计算机仿真, 2024, 41(8): 182-188.
	JIANG Jingwei, JI Yuehui, LIU Junjie, et al. Visual slam fast loop detection algorithm based on lightweight CNN[J]. Computer Simulation, 2024, 41(8): 182-188.
[11]	王伟良. 动态环境下基于Mask R-CNN的ORB-SLAM研究[J]. 现代信息科技, 2020, 4(21): 80-83.
	WANG Weiliang. Research on ORB-SLAM based on Mask R-CNN in dynamic environment[J]. Modern Information Technology, 2020, 4(21): 80-83.
[12]	ENDO Y, SATO K, YAMASHITA A, et al. Indoor positioning and obstacle detection for visually impaired navigation system based on LSD-SLAM[C]// Proceedings of 2017 International Conference on Biometrics and Kansei Engineering (ICBAKE). IEEE, 2017: 158-162.
[13]	KLAPPSTEIN J, VAUDREY T, RABE C, et al. Moving object segmentation using optical flow and depth information[C]// Advances in Image and Video Technology.Berlin, Heidelberg: Springer, 2009: 611-623.
[14]	CUI L Y, MA C W. SOF-SLAM: A semantic visual SLAM for dynamic environments[J]. IEEE Access, 2019, 7: 166528-166539. doi: 10.1109/ACCESS.2019.2952161
[15]	陈明强, 李奇峰, 冯树娟, 等. 动态环境下基于深度学习的视觉SLAM[J]. 计算机与数字工程, 2024, 52(5): 1529-1535.
	CHEN Mingqiang, LI Qifeng, FENG Shujuan, et al. Visual SLAM based on deep learning in dynamic environment[J]. Computer & Digital Engineering, 2024, 52(5): 1529-1535.
[16]	高利平, 林周勇, 柯一鸿, 等. 基于SLAM算法和动静态规划的电厂智能巡检机器控制研究[J]. 电子设计工程, 2024, 32(17): 191-195.
	GAO Liping, LIN Zhouyong, KE Yihong, et al. Research on intelligent inspection machine control for power plants based on SLAM algorithm and dynamic and static planning[J]. Electronic Design Engineering, 2024, 32(17): 191-195.
[17]	白克强, 朱亚兰, 杨秀清, 等. 融合动态目标跟踪的视觉SLAM算法[J]. 信息与控制, 2024, 53(5): 574-584.
	BAI Keqiang, ZHU Yalan, YANG Xiuqing, et al. Visual SLAM algorithm of integrating dynamic target tracking[J]. Information and Control, 2024, 53(5): 574-584.
[18]	付明磊, 卫宁伟, 金宇强, 等. 面向动态环境的视觉惯性定位方法[J]. 传感技术学报, 2024, 37(2): 268-277.
	FU Minglei, WEI Ningwei, JIN Yuqiang, et al. Visual-inertial positioning method for dynamic environment[J]. Chinese Journal of Sensors and Actuators, 2024, 37(2): 268-277.
[19]	LIU J H, LI X F, LIU Y Q, et al. RGB-D inertial odometry for a resource-restricted robot in dynamic environments[J]. IEEE Robotics and Automation Letters, 2022, 7(4): 9573-9580. doi: 10.1109/LRA.2022.3191193
[20]	WEI S, LI Z. An RGB-D SLAM algorithm based on adaptive semantic segmentation in dynamic environment[J]. Journal of Real-Time Image Processing, 2023, 20(5): 85. doi: 10.1007/s11554-023-01343-2
[21]	LI F, CHEN W F, XU W F, et al. A mobile robot visual SLAM system with enhanced semantics segmentation[J]. IEEE Access, 2020, 8: 25442-25458. doi: 10.1109/Access.6287639
[22]	YIN H S, LI S M, TAO Y, et al. Dynam-SLAM: An accurate, robust stereo visual-inertial SLAM method in dynamic environments[J]. IEEE Transactions on Robotics, 2023, 39(1): 289-308. doi: 10.1109/TRO.2022.3199087
[23]	YUAN C F, XU Y L, ZHOU Q. PLDS-SLAM: Point and line features SLAM in dynamic environment[J]. Remote Sensing, 2023, 15(7): 1893. doi: 10.3390/rs15071893
[24]	柯学良, 肖玮, 曲乃铸, 等. 基于光照不变特征提取和动态特征祛除的视觉SLAM算法[J]. 测绘通报, 2025(9): 105-111. doi: 10.13474/j.cnki.11-2246.2025.0917
	KE Xueliang, XIAO Wei, QU Naizhu, et al. A visual SLAM algorithm based on illumination-robust feature extraction and dynamic feature removal[J]. Bulletin of Surveying and Mapping, 2025(9): 105-111. doi: 10.13474/j.cnki.11-2246.2025.0917
[25]	方福涛. 基于SLAM和IoT的仓储机器人的应用研究[J]. 中国设备工程, 2025(18): 98-101.
	FANG Futao. Application research of storage robot based on SLAM and IoT[J]. China Plant Engineering, 2025(18): 98-101.
[26]	王庆, 史晓琼, 严超, 等. 基于动态多目标追踪与背景提取的视觉SLAM方法[J]. 中国惯性技术学报, 2025, 33(9): 856-863.
	WANG Qing, SHI Xiaoqiong, YAN Chao, et al. Visual SLAM method based on dynamic multi-object tracking and background extraction[J]. Journal of Chinese Inertial Technology, 2025, 33(9): 856-863.
[27]	HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]// Proceedings of 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2021: 13708-13717.
[28]	尚玉朝, 刘春豪, 王凯. 基于融合优化算法的超级电容器健康状态预测模型[J/OL]. 发电技术, 2025: 1-11 (2025-01-17)[2025-09-12]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=SLJX20250117001&dbname=CJFD&dbcode=CJFQ.
	SHANG Yuzhao, LIU Chunhao, WANG Kai. A state of health prediction model for supercapacitors based on an integrated optimization algorithm[J/OL]. Power Generation Technology, 2025: 1-11(2025-01-17)[2025-09-12]. https://kns.cnki.net/KCMS/detail/detail.aspx?filename=SLJX20250117001&dbname=CJFD&dbcode=CJFQ.
[29]	时洪雷, 刘喜军, 高竟译, 等. 基于超参数优化和混合神经网络的锂硫电池健康状态估计[J/OL]. 储能科学与技术, 2025: 1-11 (2025-09-11)[2025-09-12]. https://link.cnki.net/doi/10.19799/j.cnki.2095-4239.2025.0608.
	SHI Honglei, LIU Xijun, GAO Jingyi, et al. State of health estimation of lithium-sulfur batteries based on hyperparameter optimization and hybrid neural networks[J/OL]. Energy Storage Science and Technology, 2025: 1-11(2025-09-11)[2025-09-12]. https://link.cnki.net/doi/10.19799/j.cnki.2095-4239.2025.0608.
[30]	张考, 何凯琳, 杨沛豪. 基于模糊强化学习的电力变压器故障诊断算法研究[J]. 综合智慧能源, 2024, 46(10): 48-55. doi: 10.3969/j.issn.2097-0706.2024.10.007
	ZHANG Kao, HE Kailin, YANG Peihao. Research on power transformer fault diagnosis algorithm based on fuzzy reinforcement learning[J]. Integrated Intelligent Energy, 2024, 46(10): 48-55. doi: 10.3969/j.issn.2097-0706.2024.10.007
[31]	REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection[C]// Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, 2016: 779-788.
[32]	任一鸣, 杜董生, 邓祥帅, 等. 基于Real-ESRGAN和改进YOLOv8n的输电线路绝缘子故障检测[J]. 综合智慧能源, 2024, 46(7): 29-39. doi: 10.3969/j.issn.2097-0706.2024.07.004
	REN Yiming, DU Dongsheng, DENG Xiangshuai, et al. Fault detection of transmission line insulators based on Real-ESRGAN and improved YOLOv8n[J] Integrated Intelligent Energy, 2024, 46(7): 29-39. doi: 10.3969/j.issn.2097-0706.2024.07.004
[33]	Redmon J, Farhadi A. Yolov3:An incremental improvement[J]. arXiv Preprint, 2018: 1804.02767.
[34]	WANG C Y, MARK LIAO H Y, WU Y H, et al. CSPNet: A new backbone that can enhance learning capability of CNN[C]// Proceedings of 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). IEEE, 2020: 1571-1580.
[35]	GAO S H, CHENG M M, ZHAO K, et al. Res2Net:A new multi-scale backbone architecture[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(2): 652-662. doi: 10.1109/TPAMI.34
[36]	高翔, 张涛. 视觉SLAM十四讲:从理论到实践[M]. 2版. 北京: 电子工业出版社, 2019.

算法	RMSE	均值	中值	STD
ORB-SLAM3	0.622	0.573	0.483	0.241
YOLOv5-ORB-SLAM2	0.022	0.019	0.017	0.010
YOLOv5-ORB-SLAM3	0.023	0.021	0.019	0.009
YOLOv5-ORB-SLAM3 (Dense)	0.020	0.018	0.016	0.009

算法	RMSE	均值	中值	STD
ORB-SLAM3	0.027	0.021	0.016	0.016
YOLOv5-ORB-SLAM2	0.016	0.013	0.011	0.009
YOLOv5-ORB-SLAM3	0.015	0.012	0.010	0.009
YOLOv5-ORB-SLAM3(Dense)	0.014	0.011	0.010	0.008

模型	mAP@0.5/%	FPS	e_APE/m	评价
YOLOv8-ORB-SLAM3(Dense)	89.6	15.6	0.215	提升了检测精度，但结构复杂，推理速度较慢，实时性略有下降
YOLOv9-ORB-SLAM3(Dense)	90.2	16.2	0.205	精度进一步提升，FPS表现优于YOLOv8，但动态环境中的稳定性欠佳
YOLOv10-ORB-SLAM3(Dense)	91.3	16.4	0.198	轻量化优化显著，实时性更高，但在动态目标检测中的鲁棒性较差
YOLOv5-ORB-SLAM3（Dense）	92.4	16.8	0.189	精度最高，定位误差最低，鲁棒性最佳，算力消耗低，适合部署