基于深度传感器的监控技术与算法
|
摘要
针对现有的视频监控技术对关键设施的监控实时性低,以及在夜晚实施监控时效果差的问题,提出基于深度传感器的关键设施监控技术,并通过支持向量机(SVM)算法对其可行性进行验证。利用不同人体的骨架长度存在差异的特性,采用深度传感器作为数据采集端,瞬时采集人体的8个骨架长度作为特征值,作为人体骨架信息,由于深度传感器以红外结构光检测为核心,故在夜晚也可无损地采集数据。基于SVM进行分类器的设计,得到最佳人体识别算法。实验表明:所提模型及算法的识别成功率可以达到80%以上,具有较好的识别效果,故基于深度传感器的监控技术可以为关键设施监控提供一种新思路。
In view of the low real-time performance of the existing video surveillance technology for the monitoring of key facilities and the poor effect of monitoring during the night,a key facility monitoring technology based on depth sensor is proposed,and its feasibility is verified by SVM algorithm. Taking characteristics of the difference in skeleton length of different human bodies,the depth sensor is used as data collection terminal,and the eight skeleton lengths of the human body are instantaneously collected as the eigenvalues,and the values are used as the human skeleton information,the depth sensor takes the infrared structure light detection as core,so data can also be collected lossless at night. The best human recognition algorithm can be obtained based on design of SVM classifier. Experimental results show that the recognition success rate of the proposed model and algorithm can reach above 80 %,and has good recognition effect. Therefore,the monitoring technology based on depth sensor can provide a new idea for monitoring of key facilities.
In view of the low real-time performance of the existing video surveillance technology for the monitoring of key facilities and the poor effect of monitoring during the night,a key facility monitoring technology based on depth sensor is proposed,and its feasibility is verified by SVM algorithm. Taking characteristics of the difference in skeleton length of different human bodies,the depth sensor is used as data collection terminal,and the eight skeleton lengths of the human body are instantaneously collected as the eigenvalues,and the values are used as the human skeleton information,the depth sensor takes the infrared structure light detection as core,so data can also be collected lossless at night. The best human recognition algorithm can be obtained based on design of SVM classifier. Experimental results show that the recognition success rate of the proposed model and algorithm can reach above 80 %,and has good recognition effect. Therefore,the monitoring technology based on depth sensor can provide a new idea for monitoring of key facilities.
引文
[1]赵丹,肖继学,刘一.智能传感器技术综述[J].传感器与微系统,2014,33(9):4-7.
[2]赵小川,罗庆生,韩宝玲.机器人多传感器信息融合研究综述[J].传感器与微系统,2008,27(8):1-4,11.
[3]陈玲.生物传感器的研究进展综述[J].传感器与微系统,2006,25(9):4-7.
[4] Wen X B,Man J F,Li Q Q,et al. Human segmentation andtracking for crowded people in video surveillance[J]. Journal ofChinese Computer Systems,2012,33(4):891-895.
[5] Choi W,Pantofaru C,Savarese S. A general framework fortracking multiple people from a moving camera[J]. IEEE Transon Pattern Analysis and Machine Intelligence,2013,35(7):1577-1591.
[6] Nakashima Y,Babaguchi N,Fan J P. Intended human objectdetection for automatically protecting privacy in mobile videosurveillance[J]. Multimedia Systems,2012,18(2):157-173.
[7]王春兰.智能视频监控系统中运动目标检测方法综述[J].自动化与仪器仪表,2017(3):1-3.
[8]张海山.基于云存储视频监控系统的研究[J].电子设计工程,2015,23(10):96-101.
[9]张军,杨子晨.多传感器数据采集系统中的数据融合研究[J].传感器与微系统,2014,33(3):52-54,57.
[10]张诚,马华东,傅慧源.基于时空关联图模型的视频监控目标跟踪[J].北京航空航天大学学报,2015,41(4):713-720.
[11]李忠海,邢晓红,金海洋.基于视频监控的人的异常行为系统设计[J].火力与指挥控制,2017,42(4):165-170.
[12]石曼银. Kinect技术与工作原理的研究[J].哈尔滨师范大学自然科学学报,2013,29(3):83-86.
[2]赵小川,罗庆生,韩宝玲.机器人多传感器信息融合研究综述[J].传感器与微系统,2008,27(8):1-4,11.
[3]陈玲.生物传感器的研究进展综述[J].传感器与微系统,2006,25(9):4-7.
[4] Wen X B,Man J F,Li Q Q,et al. Human segmentation andtracking for crowded people in video surveillance[J]. Journal ofChinese Computer Systems,2012,33(4):891-895.
[5] Choi W,Pantofaru C,Savarese S. A general framework fortracking multiple people from a moving camera[J]. IEEE Transon Pattern Analysis and Machine Intelligence,2013,35(7):1577-1591.
[6] Nakashima Y,Babaguchi N,Fan J P. Intended human objectdetection for automatically protecting privacy in mobile videosurveillance[J]. Multimedia Systems,2012,18(2):157-173.
[7]王春兰.智能视频监控系统中运动目标检测方法综述[J].自动化与仪器仪表,2017(3):1-3.
[8]张海山.基于云存储视频监控系统的研究[J].电子设计工程,2015,23(10):96-101.
[9]张军,杨子晨.多传感器数据采集系统中的数据融合研究[J].传感器与微系统,2014,33(3):52-54,57.
[10]张诚,马华东,傅慧源.基于时空关联图模型的视频监控目标跟踪[J].北京航空航天大学学报,2015,41(4):713-720.
[11]李忠海,邢晓红,金海洋.基于视频监控的人的异常行为系统设计[J].火力与指挥控制,2017,42(4):165-170.
[12]石曼银. Kinect技术与工作原理的研究[J].哈尔滨师范大学自然科学学报,2013,29(3):83-86.