基于多源特征融合的实验室不安全行为检测系统设计OA北大核心CSTPCD
Design of laboratory unsafe behavior detection system based on multi-source feature fusion
为降低实验室事故发生率,设计一种基于多源特征融合的实验室不安全行为检测系统.首先,构建系统基础框架,在图像采集单元中利用高清摄像机获取实验室监控视频图像,经过图像处理单元的灰度化、去噪、平滑滤波处理后,在目标识别单元中采用帧间差分法获得视频目标区域图像;然后,在特征提取单元中处理视频目标区域图像,提取HOG特征以及人体行为重心特征,通过对二者进行融合处理得到一维特征向量,将其输入到行为检测单元中,利用SVM分类器检测不安全行为.实验结果表明:所设计系统可有效检测实验室不安全行为,F1指数均值、召回率均值分别为98.94%、99.15%,错误检测数量较少.
In order to reduce the incidence of laboratory accidents,a laboratory unsafe behavior detection system based on multi-source feature fusion is designed.The basic framework of the system is constructed,and high-definition camera is used to obtain laboratory monitoring video images in the image acquisition unit.After grayscale,denoising,and smooth filtering processing in the image processing unit,the video target area image is obtained by means of the frame difference method in the target recognition unit.Then,the video target area image is processed in the feature extraction unit,and HOG features and human behavior center of gravity features are extracted.By fusing the two,a one-dimensional feature vector is obtained,which is input into the behavior detection unit.The unsafe behavior is detected by means of the SVM classifier.The experimental results show that the system can effectively detect unsafe behaviors in the laboratory,with an average F1 index and recall rate of 98.94%and 99.15%,respectively.The number of error detections is relatively small.
张国志;张晓文
山西大学,山西 太原 030000山西大学 分子科学研究所,山西 太原 030000
电子信息工程
多源特征融合实验室不安全行为检测多源特征提取HOG特征SVM分类器
multi-source feature fusionlaboratoryunsafe behavior detectionmulti-source feature exactionHOG featuresSVM classifier
《现代电子技术》 2024 (012)
52-56 / 5
评论