基于卷积神经网络的幼儿园户外环境安全评价  

Research on Saftey Evaluation of Kindergarten Outdoor Environment Based on Convolutional Neural Network

作　　者：董潇晓 程世丹[1] DONG Xiaoxiao;CHENG Shidan(School of Urban Design,Wuhan University,Wuhan 430072,China)

机构地区：[1]武汉大学城市设计学院,湖北武汉430072

出　　处：《环境科学与技术》2021年第9期73-80,共8页Environmental Science & Technology

基　　金：住建部软科学项目(2017-K8-044)。

摘　　要：幼儿园户外环境是幼儿日常游戏的主要场所,其环境的安全与否影响到幼儿身心健康。幼儿户外环境的安全评价可以查找安全隐患,进而采取针对性的改进措施。传统的环境安全评价方法为访谈法、观察法、问卷法和案例法,具有较强的滞后性和主观性。为此,研究提出了基于卷积神经网络的幼儿园户外环境安全评价方法(SEBCNN)。该方法以幼儿在运动环境中的影像为基础,利用卷积神经网络准确生成幼儿户外运动状态的大数据,进而利用算法处理后的大数据对运动环境进行安全性分析和风险预测。经对南方某幼儿园实验验证,该方法具有数据颗粒度高、数据准确完备及鲁棒性高的特性,为幼儿园户外环境安全性分析和预测提供了新的路径。Outdoor environment in kindergarten is the main place for children’s daily games,and its safety affects children’s physical and mental health.The safety evaluation of children’s outdoor environment can find potential risks,and then take improvement measures.The traditional assessment methods of environmental safety are interview method,observation method,questionnaire method and case method,which have strong hysteresis and subjectivity.Therefore,a safety evaluation based on convolutional neural network(SEBCNN)for kindergarten outdoor environment is proposed.The SEBCNN is based on spatial image data,which can accurately record the big data of children’s outdoor behavior without contacting children.Results show that the method has high accuracy,objectivity,economy,and quantifiability,which provides a new path for evaluation and prediction of the safety of outdoor environment in kindergarten.

关 键 词：环境安全评价 卷积神经网络 大数据 幼儿园 

分 类 号：X820[环境科学与工程—环境工程]