Dynamic Identification Method of Unsafe Behaviors in Power Transmission and Transformation Construction Scene Based on UAV

Konferenz: ICMLCA 2021 - 2nd International Conference on Machine Learning and Computer Application
17.12.2021 - 19.12.2021 in Shenyang, China

Tagungsband: ICMLCA 2021

Seiten: 4Sprache: EnglischTyp: PDF

Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt

Autoren:
Zhang, Yu; Liang, Wei; Zhu, Rui; Yin, Kangyong; Jia, Mengmeng; Wang, Jingjun (Jiangsu Electric Power Research Institute Corporation Limited, State Grid Jiangsu Electric Power Co., Ltd. Research Institute, Nanjing, China)

Inhalt:
In order to improve the accuracy of identification of unsafe behaviors in power transmission and transformation scenes, a dynamic identification method based on UAV is proposed. Extraction of key elements for classifying construction unsafe behavior parallel. A Fast R-CNN module and a bottom-up attention model are constructed to extract the power transmission engineering images taken by UAVs. The LSTM network is used to compute the similarity between the image and the unsafe text and to identify the unsafe behavior dynamically. Experimental results show that the proposed algorithm can achieve 97.23% accuracy, 98.38% recall, 1.67% false alarm rate and 1.39% false alarm rate.