摘要
针对施工作业行为风险难以监测,无法匹配应对策略的问题,提出基于改进机器视觉的施工作业行为风险监测。采用了改进的直方图均衡算法,均衡处理了监测获取的施工场景图像;创新性地引入了方向梯度直方图(HOG)优化支持向量机(SVM),构建了基于SVM-HOG的施工安全风险识别模型,识别了图像中的禁止区域入侵行为和人员施工风险行为,发送风险行为预警信号;基于知识图谱,分析了风险行为等级,实现施工作业行为风险监测。施工现场测试结果显示:采用该方法均衡处理后的图像对比度和信息熵结果最高值分别达到7.34和37.55,提升了图像细节和质量。该方法准确识别了监测图像中的人员行为风险、机械作业不规范风险以及人员入侵安全风险,为施工安全管理提供可靠依据。
In response to the problem of difficult monitoring and inability to match response strategies for construction activity risk,a construction activity risk monitoring method based on improved machine vision is proposed.The improved histogram equalization algorithm is used to balance the construction scene images obtained from monitoring.Innovatively introduced Histogram of Oriented Gradient(HOG)to optimize Support Vector Machine(SVM),constructed a construction safety risk identification model based on SVM-HOG,identified prohibited area intrusion behavior and personnel construction risk behavior in the image,and sent risk behavior warning signal.Based on the knowledge graph,the level of risk behavior was analyzed to achieve risk monitoring of construction operations.Construction site test results show that the highest values of the contrast and information entropy of the image after equalization processing by this method reach 7.34 and 37.55,respectively,improving the image detail and quality.This method accurately identifies personnel behavior risks,non-standard mechanical operations risks,and personnel intrusion safety risks in monitoring images,providing a reliable basis for construction safety management.
作者
吴承彬
WU Chengbin(Zhonghui Construction Group Co Ltd,Fuzhou 350001,China)
出处
《土木工程与管理学报》
2023年第3期82-87,共6页
Journal of Civil Engineering and Management
关键词
机器视觉算法
安全风险监测
图像均衡处理
风险行为
machine vision algorithm
safety risk monitoring
image equalization processing
risk behavior
