摘要
在驾驶和在线课堂这类持续时间较长的行为中,人们容易出现分心而导致事故发生或上课效率差。为检测这类分心行为,提出了一种基于低分辨率红外阵列传感器的头部运动检测方法,它在实现行为监测的同时也保护了个人隐私。首先,基于图像处理方法提取了人体的显著区域;然后设计了一种三维图像融合算法来提取时空域的变化信息;最后,设计了一个改进的残差网络来实现头部运动分类。面向驾驶和在线课堂应用场景设计了10种头部运动。实验结果表明,在50 cm到100 cm的检测范围内,平均识别率为96.76%,处理速度为9帧/s,优于现有算法。将该系统应用于车内实测,也达到了93.7%的准确率。
People easily get distracted or tired after long-duration actions such as driving and online classes,which can lead to accidents or poor efficiency.To detect such human behaviors,a head motion detection method based on low-resolution infrared array sensors is proposed with the protection of personal privacy.First,promi⁃nent areas of the human body are extracted based on image processing techniques.Then a 3D image fusion algo⁃rithm is developed to extract the change information in the spatiotemporal domain.Finally,an improved residual network is developed to achieve head motion classification.Ten head movements are designed for driving and on⁃line classroom scenarios.Experimental results show that in the detection range of 50 cm to 100 cm,our average recognition rate is 96.76%,and the processing speed is 9 frames per second,which is better than the existing state-of-the-art algorithms.The accuracy of the system is 93.7%when it is applied to the vehicle experiment.
作者
陈良琴
曾铭萱
许志猛
陈志璋
CHEN Liang-Qin;ZENG Ming-Xuan;XU Zhi-Meng;CHEN Zhi-Zhang(College of Physics and Information Engineering,Fuzhou University,Fuzhou 350108,China;Department of Electrical and Computer Engineering,Dalhousie University,Halifax,NS B3J 1Z1,Canada)
出处
《红外与毫米波学报》
SCIE
EI
CAS
CSCD
北大核心
2023年第2期276-284,共9页
Journal of Infrared and Millimeter Waves
基金
Supported by National Natural Science Foundation of China(62071125)
the Natural Science Foundation of Fujian Province(2021J01581,2018J01805)
the Scientific Research Foundation of Fuzhou University(GXRC-18083)。
