CN102737474A

CN102737474A - Monitoring and alarming for abnormal behavior of indoor personnel based on intelligent video

Info

Publication number: CN102737474A
Application number: CN2011100868608A
Authority: CN
Inventors: 胡茂林
Original assignee: Individual
Current assignee: Individual
Priority date: 2011-04-07
Filing date: 2011-04-07
Publication date: 2012-10-17

Abstract

The invention discloses an automatic monitoring system for indoor personnel (especially the elderly living on their own and the disabled) based on intelligent video. A behavior of a moving object (referring to a human) in a video monitoring area is analyzed, to realize identification of body characteristics including standing, sitting, falling down, etc; and if a judging result is falling down, an alarming message is sent to terminal equipment of other family members, community managing staff or civil administration staff in a wired or wireless manner, wherein the terminal equipment comprises a mobile, a computer screen, a tablet pc and the like. The method of the invention can be used in indoor places including homes, nursing homes for the elderly, activity centers, hospitals, etc.

Description

Supervision and warning based on the indoor occupant abnormal behaviour of intelligent video

Technical field

The invention belongs to the intelligent video field.Through intellectual analysis to camera image, realize household, health care for the aged institute, indoor place human behaviors such as activity centre and hospital are discerned, and reach abnormal behaviour is found as early as possible and reported to the police.

Background technology

Along with the video camera industry is changed on a large scale, the application of camera is popularized; At public safety, all used a large amount of cameras in the monitoring of Smart Home widely.Artificial and to manage these numerous cameras constantly obviously impossible, in addition in commercial markets such as Smart Homes, traditional monitoring means that utilize the pulpit giant-screen are not suitable for general user's demand, and following monitor terminal should be mobile devices such as mobile phone.Therefore need camera can judge intelligently that interested target occurs, and only send needed information, reduce cost of labor, simultaneously, save communication network bandwidth and secret protection.

In field of video monitoring, most watch-dog all is the content that directly shows shot by camera at present.There are certain applications the picture that relates to private residence to be blocked and mosaic processing, but the behavior of indoor environments such as being applied in the family is judged do not have in providing of public safety field.

Summary of the invention

To household, health care for the aged institute, the indoor special messengers of needs such as activity centre and hospital keep watch on the place, and the present invention's research and development are based on the automatic monitor system of intelligent video.Behaviors such as the personnel through to the video camera shooting area stand, walk, sit, fall identification reaches abnormal behaviour is found and warning in time as early as possible.At first input picture is carried out moving object detection, detected moving-target image is carried out characteristic (such as length breadth ratio) analysis, judge whether it is abnormal behaviour; If, and the lasting regular hour, just through mobile phone, panel computer wireless or that wireline communication network connects, computer etc. are reported to the police to the relevant personnel.

Description of drawings

Fig. 1 system flowchart

Embodiment

Video camera among the present invention sets up in the zone of required monitoring, and video camera is connected with main control computer through data line.Total system mainly contains three parts and forms: (1) motion target detection; (2) behavioural analysis; (3) moving-target based on different platform shows and warning.The system works quick-reading flow sheets is as shown in Figure 1.

Suppose to have only in the scene people in motion, therefore detected moving region is the human body region.Usual way has: (1) background subtraction method, and utilize the difference of present image and background image to detect the moving region; (2) time differencing method adopts the time difference based on pixel between two or three consecutive frames in the continuous images sequence, obtain frame difference image, and it is carried out aftertreatment extracts the moving region in the image.(3) optical flow method has adopted the time dependent light stream characteristic of moving target.

The key of motion detection is that sequence image is extracted region of variation from background image, because the people who moves in the common region-of-interest can not represent that template matching method is inapplicable with a fixing template; The optical flow approach calculated amount is bigger, and noiseproof feature is poor, and is also inapplicable under complex background, and what often adopt is the method for background subtraction.On the other hand, because the subtle change of some backgrounds often appears in the complex environment, like the swing of curtain, the flicker of computer screen etc., simple background subtraction effect will be affected.Here used adaptive background modeling method, promptly in system, set up the background statistical model of ambient image, extracted foreground area with adaptive background subduction algorithm, and the statistical model of online updating background.Test findings proves that this method can be applicable to indoor environments such as house preferably.

Before the detection beginning, system continues observation a period of time to nobody's environment earlier, sets up the statistical model of initial environment background.After, to each image that collects constantly, at first carry out denoising, pre-service such as level and smooth.The detection of moving region utilizes the front to set up good background statistical model, draws foreground area by adaptive dynamic background subduction algorithm, and upgrades original background statistical model.Moving-target is judged whether it is because illumination or curtain throw open the background change that causes, if, carry out context update; Otherwise just explain that guarded region has moving target, passing thresholdization obtains a bianry image, carries out binary conversion treatment then.

Binary conversion treatment is for the little target (noise) in the removal of images and fills the cavity in the interesting target, and this binary image and input picture are merged, and produces the image of moving target; The parameter of statistical study moving-target: length breadth ratio, distribution of color, movement velocity etc.If these parameter generation ANOMALOUS VARIATIONS just send to this moving-target image on the terminal device through wired or wireless mode and report to the police.

Claims

1. no matter adopt which kind of behavioural analysis algorithm.

2. no matter show on the different terminal equipment, such as mobile phone, display, giant-screen, panel computer etc.;

3. no matter at terminal device epigraph display format, demonstration be entire image or parts of images, or one section video;

4. no matter whether system combines other panalarms or acousto-optic electricity apartment warning.

All in the protection of this patent.