Design And Implementation Of Real-time Video Stream Fall Detection System

Every year,there are a lot of person falls,and it takes a lot of money and times to process it,some person just slight abrade,some person hurt a little heavy,and some people may be die.In many cases,fall after the incident,the injured man is due to have not been found in time,lying on the ground too long failed to get effective treatment and caused serious losses,in this modern society with developed science and technology,this kind of thing can completely avoid,now even the robot can go on to win the human,moreover this kind of simple fall detection.Fall detection system,can be used in the family,shopping malls,schools,companies,construction site,such as any place,on the market at present most of the fall detection system is based on the hardware,require users to wear a bracelet or other instrument,very inconvenient and limitations,if the user is not carrying equipment,no electricity,or the device will not be able to inform other people,or the class wear up will cause the discomfort of the user,if there is no signal place such as the underground mall,such devices almost null and void.So I began to study fall detection,aims to help people in need after the fall,let they can timely rescue,don't cry because it is a waste of time and loss of the best treatment time.In study of fall detection,eliminate learned learning background,has a considerable effect on target tracking of the object,the detection of target objects in the process of how to calculate the velocity of the target object,the target object in the shape of the image analysis,the outline of the current target position,when detected falls,can have significant hints.When the background elimination,need to consider different scenarios,different light,different site conditions,to adjust the parameters of the background to eliminate,the distance of the object,etc.,all need to adjust the parameters,because when calculating the background elimination of these parameters is crucial,reasonable parameters,can make background elimination,perfect to take out the moving objects.After eliminate have a deep understanding of background,I began working on fall detection,after the fall detection is studied to the variety,found that simply from a certain Angle to judge falls,or have some misjudgment rate,if we only by examining the current tilt Angle of the characters,so may not be ready,because some people just bend over or other similar action,and if we can only through the speed to judge whether the man fell down,if the person suddenly run,so also no good way to avoid misjudgment of the,if only through the character's center of gravity,then squat down,also can cause miscarriage.We can by combining the detecting features,such as in video tracking moving objects,sudden change in detecting speed,determine if there was a sudden movement is slow,and then judge the Angle or position,whether horizontal or vertical,and then,because of Angle of view,the following need to be in a scene,arrangement of multiple taken at least two cameras,because different perspectives,the characters in the video may not be the same,so,in different angles are stalkers,in two video,as long as found fall phenomenon,can send out alarm immediately.