Research On Key Technologies Of State Monitoring Of Human Body Based On Machine Vision

The incidence of juvenile myopia in China is increasing year by year,and one of the reasons for the cause of juvenile myopia is bad sitting posture.In road traffic accidents,many instances are caused by drivers being distracted and fatigued while they are driving.In view of the above two problems,this paper studies the sitting posture and fatigue state monitoring based on machine vision.Based on face detection method,two different monitoring methods for each application in the two scene of monitoring sitting posture and fatigue driving are proposed.These effective methods include face color statistical method and regional discriminant feature point matching for the sitting posture monitoring,mouth moving area fusion edge statistical method for the yawning detection,human eye and pupil detection based method for closing eye detection.And a"bad sitting posture behavior monitoring system" and a "sitting posture and fatigue monitoring simulation system" are designed.Firstly,in the bad sitting posture behavior monitoring system,the face detection method based on RGB color video is improved:employing the skin color based face detection method can effectively reduce the false face recognition rate and the maximum single target face detection method can detect the face target.And then a single target face detection method of self-adaptive detection window scale is proposed.This method utilizes the size of a single subject target detected in the previous frame,to self-adaptively adjust the scope of the detection window,making the detection speed greatly improved.Secondly,a sitting posture monitoring method based on face color statistics is proposed.The method according to the detected face frame plans out the left,the right and the center these three facial color regions and then by comparing the these three skin color regions with the ones in the correct posture case,to judge the position is to the left or to the right and by the statistics comparison of face color area under current and correct conditions respectively within the face detection box to judge the position is on the front or rear.The experimental results show that the left-right detection accuracy of the proposed method is 100%,and the front-rear detection accuracy is 97.3%under the circumstance avoiding the background color is similar to or the same as the face color.In view of the monitoring of driver fatigue and bad sitting posture,three improved methods are proposed:(1)The regional discriminant feature point matching based sitting posture judgment method is proposed.By analyzing templates of the correct posture and three pairs of the best matching of SURF feature points in the real-time monitoring region,it can determine the driver's current sitting posture is correct or not.(2)A mouth moving area fusion edge statistical method for the yawning judgment is proposed.The experimental results show that the activity in the mouth almost occurs in the lower region of a face detection box,so distinguishing yawning mouth is mainly about drawing out the mouth activity region on a face box,then doing the statistics of moth edge(the edge detected by the Prewitt and Canny detection operators with an infrared camera)longitudinal projection ratio in this area to gauge the mouth opening degree and through the degree to determine the state of yawning.(3)A human eye and pupil detection based method for closing eye judgment is proposed.According to the face box,the general area of human eyes is drawn.on this basis,the position of the eyes can be detected effectively.Thus it greatly reduces the error brought by overall detection,and improves the detection efficiency and direction as well.And then the detected eyes are zoomed in properly and used for the Hough circle detection.Through the presence or absence of the Hough circle,eye opening or closing state is detected.The experimental results show that the accuracy of the head posture,yawning and eye fatigue discriminant modules of the designed "Auxiliary driving posture and fatigue monitoring simulation system" are 98.9%,100%,97.8%respectively.