基于电梯视频的乘客倒地检测算法研究  

Research on Passenger Landing Detection Elevator Based on Video

作　　者：马志伟 章国宝[1] 

机构地区：[1]东南大学自动化学院,江苏南京210096

出　　处：《工业控制计算机》2018年第4期71-73,76,共4页Industrial Control Computer

基　　金：国家自然科学基金青年项目(61503081);江苏省质量技术监督局科技项目:电梯运行环境智能管理系统的研究及应用(KJ168355);江苏省科技支撑计划资助项目(BE2014728)支持

摘　　要：提出了一种基于电梯监控视频的用以检测轿厢内是否有乘客倒地的算法。首先分析了一些常用前景提取法在电梯轿厢环境进行前景提取的不足,从而提出了一种改进的带有更新法则的背景差法。接着介绍Canny边缘检测的原理,用来提取边缘前景特征。主要考虑电梯只有一位乘客且乘客倒地的情况,所以结合前景主要集中在轿厢下部的特点,给定了倒地判定规则,采用不对称的限幅积分器对判定结果进行滤波。实验结果表明,算法在快速计算的同时,能平稳有效地对单乘客倒地行为进行准实时报警。This paper presents an algorithm to detect whether there is a passenger in the car falling to the ground based on elevator monitoring video.First of all,this paper analyzes the shortcomings of some common foreground extraction methods which are used to extract the foreground in the elevator car environment.An improved background subtraction method with updating rule is proposed in this paper.Then it introduces the principle of Canny edge detection,and used to extract the edge of the foreground features.This paper mainly considers the situation that there is only one passenger in the elevator and the passenger fell to the ground.

关 键 词：轿厢环境 背景差法 CANNY边缘检测 限幅积分器 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]