摘要
为预防在有限空间作业中因不安全行为引发的安全事故,着重分析了作业人员在有限空间作业中因跌倒发生的事故风险。通过安装在作业区域的视频摄像头获取的图像,对人体跌倒行为进行识别,并利用MATLAB的图像处理功能,构建了基于支持向量机(SVM)的不安全行为图像识别算法。该算法首先应用图像灰度级变换、直方图均衡化、高斯低通滤波平滑和中值滤波去噪对视频采集图像进行图像预处理;然后根据帧间差分法定位人体,并提取HOG特征和人体重心移动特征构建人体跌倒行为特征向量,应用SVM对不安全行为进行分类;最后对获取的图像资料进行识别与分析。结果表明:该图像识别算法的平均正确识别率达到了97.84%,能够对有限空间作业中的不安全行为进行有效识别,从而验证了该算法的有效性。
In order to prevent safety accidents caused by unsafe behavior in confined space,this paper focuses on the risk of accidents caused by falls in confined space operations.Through the image obtained by the video camera installed in the work area,the paper identifies the fall behavior of human body,and constructs the image recognition algorithm of unsafe behavior based on Support Vector Machine(SVM) using the image processing function of MATLAB.First of all,the paper applies image gray level transformation,histogram equalization as well as gaussian low-pass filtering smoothing to preconditioning the image acquired by video.In addition,the paper locates the human body according to the interframe difference method,and extracts the HOG feature and the moving characteristic of the human center of gravity to construct the fall behavior characteristic vector,and applies SVM to classifying unsafe behavior.Finally,the paper identifies and analyzes the obtained image data.The results show that the average correct recognition rate of the algorithm is 97.84% for the unsafe behaviors in the limited space operation,which can identify the unsafe behaviors in confined space and verify the effectiveness of the proposed algorithm.
作者
赵江平
王垚
ZHAO Jiangping;WANG Yao(College of Resources Engineering,Xi’an University of Architecture and Technology,Xi’an 710055,China)
出处
《安全与环境工程》
CAS
北大核心
2020年第1期158-165,共8页
Safety and Environmental Engineering
关键词
有限空间作业
不安全行为
图像识别技术
人体跌倒行为检测
帧间差分法
特征提取
支持向量机(SVM)
confined space operation
unsafe behavior
image recognition technology
human fall behavior detection
interframe difference method
feature extraction
Support Vector Machine(SVM)
