施工现场事故频发,大多数情况都是因为建设工程工人的不安全行为造成的.构建科学合理的识别模型有助于及时识别工人的不安全行为.为了降低事故发生的概率,保障工人的生命安全,此次研究在多模态融合算法的基础上,结合建筑信息模型(Building Information Modeling,BIM)及相关信息技术构建工人不安全行为的识别模型.实验结果显示,训练集的后验概率误差为0.038,测试后验概率误差为0.042;在测试实验中,研究算法的准确率与召回率均较高,平均准确率为90.09％,平均召回率为89.77％;在与其他两种算法的对比中,研究算法的准确率与召回率均更高,研究算法的平均准确率为96.42％,平均召回率为95.97％.验证了研究方法的优越性,说明了研究构建模型能够为建设工程安全管理提供新的思路和方法.
A Model for Identifying Unsafe Behaviors of Construction Workers Based on Multimodal Fusion
Accidents occur frequently on construction sites,mostly due to unsafe behavior of con-struction workers.Building a scientific and reasonable identification model helps to timely identify un-safe behaviors of workers,prevent them in advance,and avoids disasters.In order to reduce the proba-bility of accidents and ensure the safety of workers'lives,this research builds a recognition model of un-safe behaviors of workers based on multimodal fusion algorithm,combined with Building information modeling(BIM)and relevant information technology.The experimental results show that the Posterior probability error of the training set is 0.038,and the Posterior probability error of the test set is 0.042.In the testing experiment,the accuracy and recall rate of the research algorithm were both high,with an average accuracy of 90.09％ and an average recall rate of 89.77％.In comparison with the other two al-gorithms,the accuracy and recall rate of the research algorithm are higher,with an average accuracy of 96.42％ and an average recall rate of 95.97％.The superiority of the research method has been verified,indicating that the research and construction model can provide new ideas and methods for safety man-agement of construction projects.
construction engineeringunsafe behavioridentification modelBIMinformation technology
万方数据
维普
