摘要
为解决传统视频监控方法易受环境影响、不能实时准确监控运动目标行为的问题,提出一种基于支持向量机的手扶电梯(简称扶梯)视频监控方法。提取可变形组件模型特征,利用支持向量机检测乘客人脸;利用核相关滤波跟踪人脸目标,使用匈牙利算法更新目标的检测强度,修正人脸检测结果;基于人脸目标进行行为监控。基于16段扶梯视频的实验结果表明,该方法处理速度达到30帧/秒,人脸检测准确率为96.6%,行为监控准确率为95.1%,能够实时准确地检测乘客人脸,监控乘客行为。
To solve the problem that the traditional video monitoring method is easily affected by the environment change and it can not accurately monitor the behavior of moving targets in real time,a video monitoring method based on support vector machines for escalator was proposed.The feature of the deformable part model was extracted and the face of the passenger was detected using the support vector machine.The face target was tracked using the kernelized correlation filter,and the detection intensity of the target was updated using the Hungarian algorithm to correct the face detection result.Behaviors were monitored based on face targets.Experimental results of the 16 escalator videos indicate that the proposed monitoring method achieves a processing speed of 30 frames per second,a face detection accuracy rate of 96.6% and a behavior monitoring accuracy rate of 95.1%.It can accurately detect passenger faces and monitor passengers in real time.
引文
[1]Mabrouk AB,Zagrouba E.Abnormal behavior recognition for intelligent video surveillance systems:A review[J].Expert Systems with Applications,2017,91(1):480-491.
[2]WEI Yu,LUO Hao.Fast face recognition based on local segmentation[J].Science Technology and Engineering,2017,1(31):133-138(in Chinese).[尉宇,罗浩.基于局部分割的快速人脸识别[J].科学技术与工程,2017,1(31):133-138.]
[3]Dutta P,Bhattacharjee D.Face detection using generic eye template matching[C]//International Conference on Business and Information Management.IEEE,2017:36-40.
[4]Dahal B,Alsadoon A,Prasad PWC,et al.Incorporating skin color for improved face detection and tracking system[C]//Image Analysis and Interpretation.IEEE,2016:173-176.
[5]Li R,Li C.Adaboost face detection based on improved covariance feature[J].Journal of Computers,2014,9(5):1077-1082.
[6]Li T,Hou W,Fei L,et al.Face detection based-on depth information using HOG-LBP[C]//6th International Conference on Instrumentation and Measurement,Computer,Communication and Control.IEEE,2016:779-784.
[7]Ali AM,Alkabbany I,Farag A,et al.Facial action units detection under pose variations using deep regions learning[C]//International Conference on Affective Computing and Intelligent Interaction.IEEE,2017:395-400.
[8]Qu Jingjing,Xin Yunhong.Combined continuous frame difference with background difference method for moving object detection[J].Acta Photonica Sinica,2014,43(7):213-220.
[9]Cheng D,Gong Y,Wang J,et al.Balanced mixture of deformable part models with automatic part configurations[J].IEEE Transactions on Circuits and Systems for Video Technology,2016(99):1.
[10]Yuan Y.Canonical duality solution for alternating support vector machine[J].Journal of Industrial and Management Optimization,2017,8(3):611-621.
[11]Henriques JF,Rui C,Martins P,et al.High-speed tracking with kernelized correlation filters[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2014,37(3):583-596.
[12]Bougleux S,Gaüzère B,Brun L.A hungarian algorithm for error-correcting graph matching[C]//Graph-based Representations in Pattern Recognition.Springer,2017:118-127.
[13]Kumar S,Gupta P.Comparative analysis of intersection algorithms on queries using precision,recall and F-Score[J].International Journal of Computer Applications,2015,130(7):28-36.