基于OpenPose-GLCM-Hash的配电安监数据清洗技术  

Distribution Safety Monitoring Data Cleaning Technology Based on OpenPose-GLCM-Hash

作　　者：程良伦[2] 裴求根 严宇平 余子勇 邵彦宁 严强 林嘉鑫 CHENG Lianglun;PEI Qiugen;YAN Yuping;YU Ziyong;SHAO Yanyu;YAN Qiang;LIN Jiaxin(Guangdong Power Grid Co.,Ltd.,Guangzhou 510620,Guangdong,China;Guangdong University of Technology,Guangzhou 510006,Guangdong,China;Guangdong Smooth Technology Co.,Ltd.,Jiangmen 510000,Guangdong,China)

机构地区：[1]广东电网有限责任公司,广东广州510620 [2]广东工业大学,广东广州510006 [3]广东顺畅科技有限公司,广东江门510000

出　　处：《电力大数据》2024年第12期38-46,共9页Power Systems and Big Data

基　　金：广东电网有限责任公司科技项目(037800HK23030079)。

摘　　要：随着现代配电网络规模的扩大和复杂性的日益增加,安全监控正面临严峻的挑战。持续的安全监控录像产生了大量的图像数据。由于相邻图像之间的高度相似性,如果不对这些数据进行有效地筛选和清洗,将导致数据管理效率低下且存在冗余。本文提出了一种OpenPose-GLCM-Hash数据清洗技术,结合灰度共生矩阵(grey level co-occurrence matrix,GLCM)的全局纹理特征与改进的OpenPose算法提取的人体姿态特征,通过一种加权组合策略生成图像哈希码,从而实现高效且准确的图像去重。实验结果揭示,该算法在各种环境下均展现出卓越的去重效率和低误判率,特别是在噪声环境中表现尤为突出。该研究为配电安全监控的数据管理提供了一种有效的解决方案。With the increasing scale and complexity of modern power distribution networks,security monitoring is faced with severe challenges.Continuous security surveillance video generates massive image data.Due to the high similarity of adjacent images,data management will be inefficient and redundant if it is not effectively screened and cleaned.In this paper,an OpenPose-GLCM-Hash data cleaning technique is proposed,which combines the global texture features of Grey Level Co-occurrence Matrix(GLCM)with the human pose features extracted by improved OpenPose algorithm.Image hash code is generated by weighted combination to achieve efficient image weight removal.The experimental results show that the algorithm has high weight removal rate and low error rate in different environments,especially in noisy environments.This study provides an effective solution for the data management of power distribution safety supervision.

关 键 词：配电安全监控 数据清洗 姿态估计 灰度共生矩阵 

分 类 号：TM08[电气工程—电工理论与新技术]