高斯混合模型下建筑工人高空作业失稳检测方法

Instability detection method for construction workers working at altitude based on Gaussian mixture model

为防止施工现场高处坠落事故,实现个性化矫正管理,在考虑个体异质性对运动信号造成的差异化影响基础上,提出一种基于高斯混合模型(GMM)的实时检测方法,可及时识别建筑工人高空作业失稳状态。首先,采用姿态传感器实时采集加速度和角速度数据,以刻画建筑工人的高空作业姿态特征;然后,基于GMM算法,建立建筑工人高空作业的个性化失稳检测模型,获得个性化阈值,以判断姿势失稳状态;最后,通过试验对比基于个体数据集和公共数据集2种方式构建的模型。研究结果表明:生成的个性化检测模型在准确度P、召回率R和综合评价指标F 1值上,均远优于公共数据集模型,具有更好的个性化检测效果。该失稳检测方法能够从工人的作业姿态习惯探究个性化的高空失稳风险,促进差异化安全预控和精准化安全培训。

In order to prevent the construction site high fall accident and achieve personalized correction management,based on considering the differentiation of motion signals caused by individual heterogeneity,a real-time detection method based on GMM was proposed,which can timely identify the instability state of construction workers working at height.This method used posture sensors to collect real-time acceleration and angular velocity data to describe the posture features of construction workers working at height.Based on GMM,it established a personalized instability detection method to obtain personalized thresholds for judging the instability state of construction workers working at heights.Finally,it compared two models constructed by individual and public data sets through experiments.The results show that the personalized detection model generated is far superior to the public data set model in accuracy(P),recall rate(R)and comprehensive evaluation value(F 1).It shows the better-personalized detection effect using the personalized detection model.This study can help identify personalized risks of instability working at height from workers'habitual working postures,provide new ideas and references for preventing falling accidents,and help realize personalized correction training for workers.