基于傅里叶描述子的吸烟行为检测方法  

Smoking behavior detection method based on Fourier descriptor

作　　者：赵鉴福 梁金幸 史宝军 ZHAO Jianfu;LIANG Jinxing;SHI Baojun(State Key Laboratory of Reliability and Intelligence of Electrical Equipment,Hebei University of Technology,Tianjin 300401,China;School of Mechanical Engineering,Hebei University of Technology,Hebei Key Laboratory of Robot Sensing and Human-Robot Integration,Tianjin 300401,China;National Engineering Research Center for Technological Innovation Method and Tool,Tianjin 300401,China)

机构地区：[1]河北工业大学,省部共建电工装备可靠性与智能化国家重点实验室,天津300401 [2]河北工业大学机械工程学院,河北省机器人感知与人机融合重点实验室,天津300401 [3]国家技术创新方法与实施工具工程技术研究中心,天津300401

出　　处：《传感器与微系统》2023年第10期147-150,155,共5页Transducer and Microsystem Technologies

基　　金：国家重点研发计划资助项目(2019YFB1312102);国家自然科学基金资助项目(U20A20201);河北省重点研发计划资助项目(20311803D)。

摘　　要：针对烟雾传感器易受粉尘和水雾影响、视频烟雾采集易受光照和气象条件影响等问题,提出了一种基于傅里叶描述子的吸烟行为检测方法。首先,制定了手部区域获取规则,完成Azure Kinect传感器RGB相机与深度相机的标定和对齐,得到包含烟支的手部目标区域;然后,比较不同颜色模型下的烟支分割效果,选择分割效果更优的YCbCr颜色模型,并提取分割后图片基于傅里叶描述子的形状特征;最后,训练支持向量机(SVM)分类器,以交叉验证的方式选择参数。建立吸烟行为检测数据集,比较了不同的分类器。实验结果表明:本文提出的基于傅里叶描述子与SVM的吸烟检测方法准确率达86%、精确率93%、召回率达82%。本文方法检测效果优,可布置在移动巡检机器人上,为保障石化厂区的安全发挥积极作用。Aiming at the problem that smoke sensor is easily affected by dust and water mist,and the video smoke acquisition is easily affected by illumination and meteorological conditions.A smoking behavior detection method based on Fourier descriptors is proposed.Firstly,the rules for acquiring the hand area are formulated,and the calibration and alignment of the RGB camera and the depth camera of the Azure Kinect sensor are completed,and the target area of the hand containing cigarettes is obtained.Then,the segmentation effect of cigarettes under different color models are compared,and the YCbCr color model with better segmentation effect is selected,and the shape features of segmented images are extracted based on Fourier descriptors.Finally,the support vector machine(SVM)classifier is trained to select parameters in a cross-validation manner.A smoking behavior detection dataset is established and different classifiers are compared.The experimental results show that the proposed smoking detection method based on Fourier descriptor and SVM has an Accuracy of 86%,a Precision of 93%,and a Recall of 82%.The proposed method has excellent detection effect and can be deployed in mobile inspection robots to play an active role in ensuring the safety of petrochemical plants.

关 键 词：吸烟行为检测 烟支分割 傅里叶描述子 支持向量机 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]