An Uncivilized Behavior Detection Method based on Improved ECO Algorithm

• Published in: 2020 35th Youth Academic Annual Conference of Chinese Association of Automation (YAC) pp. 843 - 847
• Main Authors: Wu, Hao, Liang, Shasha, Niu, Dan, Ding, Li, Hu, Yaocong, Zhu, Xiaoci, Xu, Ruohan, Chen, Xisong
• Format: Conference Proceeding
• Language: English
• Published: IEEE 16.10.2020
• Subjects: Behavior detection; Buildings; Computational efficiency; ECO algorithm; elevator video surveillance; Elevators; Monitoring; Non-local attention mechanism; Personnel; Safety; Video surveillance
• DOI: 10.1109/YAC51587.2020.9337598

In recent years, with the increase of high-rise buildings, elevators have become more and more common. The subsequent problem of safe riding has entered the public's attention. In daily life, uncivilized elevator riding behaviors are often seen such as picking doors, blocking doors, and elevator doors that are always open in the elevator. They not only endanger the safety of the elevator, but also easily cause the elevator to malfunction. Therefore, it is necessary to effectively monitor and manage the uncivilized behavior in the elevator. Aiming at the most common undesirable behaviors such as blocking the elevator doors in the elevator, this paper proposes and implements an improved ECO algorithm-based elevator behavior detection method. Considering the large span of elevator monitoring video series, a Non-local attention mechanism is added to the ECO network to improve the network's ability to capture long-range information and effectively recognize uncivilized behaviors in elevators. The test results show that the proposed improved-ECO method combing with the Non-local attention can achieve higher detection accuracy for uncivilized behaviors and high computational efficiency. It is quite important for the management personnel to monitor and manage so many elevators in parallel.