CN102254395A - Article theft detecting and alarming device and method based on intelligent video analysis - Google Patents

Abstract

The invention provides an article theft detecting device and method based on intelligent video analysis. Whether articles are stolen or not is judged through comparing the similarity among all frames in a monitoring area by using a two-dimensional relevant algorithm as a detecting algorithm which has high sensitiviity and high anti-interference performance and ignores background influence after set without updating the target area pixel. The device based on the algorithm comprises a UPS (Uninterruptible Power Supply) functional power supply, a camera, an infrared lamp set, a DSP (Digital Signal Processor) algorithm analyzing circuit, an alarm circuit and video record storage equipment. The article theft detecting device and method based on intelligent video analysis can be applied to the field of theft prevention monitoring of various articles, an alarm signal is sent after the monitored article is moved away, a video within 15 seconds before and after the article is lost is stored and an investigation clue is provided for the public security bureau. The invention provides the novel device which applies an advanced electronic technology and multiple safety design and can monitor valuables in real time and prevents theft.

Description

Stolen articles detection and warning device and method based on intelligent video analysis

Technical field:

The invention belongs to the application in the monitoring field of computer vision sensor technology, image recognition technology, Computer Control Technology, is that a kind of computer vision methods of utilizing is carried out the device of intellectual analysis to judge whether monitored article have things stolen in the visual field to monitor video.

Background technology:

In tradition supervision field, supervision in the visual field generally all is that special Security Personnel 24 hours guards are in turn arranged, perhaps the special messenger in the video wall front of containing a plurality of monitoring screens monitors, the not only waste of manpower of this way, simultaneously because continuous for a long time the observation is very easy to cause visual fatigue, the details when omitting crime easily.

For solving the problem that above-mentioned manual detection recognition methods exists, present more employing be real-time video that video capture device is obtained be transferred to terminal device, on computing machine, whether be monitored by analysis has near the article moving target to occur, if there is moving target to occur, then leave the method for judging whether the monitoring article also exist when being monitored article when moving target, this method is fit to monitor the scene of a plurality of targets, but present rate of false alarm and the rate of failing to report of this method is very high.

Proposed by the invention pass through the two-dimensional correlation computing calculate similarity between the frame of video method and based on the stolen articles of the method detect and report to the police can the real time monitoring target item state, signal in time gives the alarm, improve accuracy rate, videograph after simultaneously can preserving before the case happened automatically makes things convenient for the locking suspect of owner and public security organ.These functions have well remedied the deficiency of traditional supervision properties of product, use manpower and material resources sparingly.

In recent years, along with fast development of national economy and overall national strength progressively improve, demand and the application of intelligent surveillance in early warning, safety precaution and the modern management of crucial industries such as military affairs, bank, traffic is extensive further.As the intelligence monitoring and controlling device in the video monitoring system, the present invention has good application prospects and great demand.

Summary of the invention:

The object of the present invention is to provide a kind of stolen articles detection method and device based on intelligent video analysis, the method disclosed in the present adopts the two-dimensional correlation algorithm, and this method has very strong adaptive faculty to the illumination variation of environment.Device disclosed in this invention can be finished the state of article and monitor in real time under the situation that operating personnel do not participate in, can in time report to the police to stolen incident, has the real-time height, and the advantage of the video information of crime scene is in time write down in non-stop run.

The processing procedure of stolen articles detection method based on video analysis disclosed in this invention mainly comprises: the image array in the artificial zone of choosing the target object place, self-adaptation is provided with the threshold value of detection, the similarity of objective matrix in the every frame video of real-time analysis judges according to the size of similarity and detection threshold whether article are lost.

The image array in the artificial zone of choosing the target object place is promptly chosen the position and the size of target area, sets the object that video algorithm is analyzed.Self-adaptation is provided with the threshold value of detection, according to object being arranged in the scene and not having two kinds of situations of object, calculates two similarities, then with the mean value of two similarities as threshold value, this threshold setting method has good adaptivity.In the process of monitoring in real time, calculate the similarity between every two field picture and the target image, judge according to detection threshold whether target is stolen.

Stolen articles based on intelligent video analysis disclosed in this invention detects and warning device, employing has the power supply of UPS function, can eliminate " power pollutions " such as surge on the civil power, instant high-voltage, moment low-voltage, electric wire noise and frequency shift (FS)s, improve power quality, signal simultaneously simultaneously can operate as normal when system cut-off gives the alarm.

Video camera and infrared lamp fit together, and the place ahead that is installed in target to be monitored can obtain target larger area position in the visual field, in 10 meters scopes of distance.

DSP video acquisition and treatment circuit comprise sheet SDRAM (MT48LC4M32B2) chip as processor memory, be provided with the JTAG debug port, pci interface, the FLASH chip AM29LV033C of stored programme and data, video decoding chip TVP5150, video coding chip ADV7174 etc. constitute the platform that a video analysis is handled.

Warning circuit comprises the dry contact output of two-way relay, and as the switch and the chimes of doom switch of blinking light, after article were removed, on-the-spot flashing lamp was opened respectively, and loudspeaker sends alarm sound.

The videograph memory device: use industrial computer and hard-disk system as the video record memory device, be connected with treatment circuit with the DSP video acquisition by PCI, can be continuous carry out video storage, the perhaps video information of a store alarms front and back 15s.

Description of drawings:

Fig. 1 is an algorithm block diagram of the present invention.

Fig. 2 is the structured flowchart of apparatus of the present invention.

Fig. 3 is the DSP video acquisition and the treatment circuit of apparatus of the present invention

Embodiment:

Present technique field professional can methods of this invention will be better understood and device in order to make, and the present invention is described in further detail below in conjunction with drawings and embodiments.

Fig. 1 is an algorithm flow chart of the present invention, needs initialization module 101 and module 102 when setting monitoring objective, and more fresh target all needs to reinitialize this two modules each time.At first record one section aimless background video when reinitializing, put target item then and recording one section video, after two sections videos are finished, manually on the main control computer display, extract the pixel region of target item then, next system can carry out the computing of module 102 automatically, enters duty at last automatically.After entering duty, system can cycle criterion in 103,103,105 modules, and when 106 modules or 105 modules were judged as not, algorithm can enter into output 106 modules of reporting to the police and go, and continues the signal that gives the alarm, till managerial personnel's resetting system.

Fig. 2 is the structured flowchart of apparatus of the present invention, is the power module of UPS function as Figure 20 1 module, for total system provides the power supply of stability and safety, has 3 minutes continued power ability when power supply is cut off by the offender, the simultaneity factor signal that can give the alarm.For target image 202 modules of shooting clear are installed within 10 meters scopes of monitored object.203 modules are directly analyzed the vision signal from 202, and continuously with video signal transmission to 205 devices, send alerting signal when stolen and give warning circuit when 203 modules detect article.205 can be provided with continuously or the fragmented storage video information for industrial computer equipment.

Fig. 3 is the structure of DSP video acquisition and treatment circuit, except that comprising basic chips such as SDRAM, FLASH, JTAG, also has 301 modules as video decoding chip in the circuit, converts the vision signal of simulating to digital signal.The DSP video acquisition is connected by pci interface with main control computer with treatment circuit, can realize high-speed data transmission.

Claims (10)

1. intelligent video stolen articles detection method is characterized in that this method may further comprise the steps:

Step 1, extraction target item area pixel; The similarity that the frame that step 2, the frame that is removed according to target and target exist is surveyed is set monitoring threshold; Step 3, the object pixel of present frame and extraction is carried out related operation extract similarity; The similarity of step 4, comparison step three gained and detection threshold are relatively reported to the police when similarity is lower than threshold value.

2. method according to claim 1 is characterized in that the extraction of object pixel has operating personnel to set up on their own, chooses the interior pixel of the minimum rectangle frame that can surround target as object pixel, determines the size and the positional information of rectangle frame.

3. method according to claim 1, the setting that it is characterized in that detection threshold are to ask similarity R between the frame of frame that target exists and target existence by the two-dimensional correlation computing _Sim1And the frame of target existence and the similarity R between the non-existent frame of target _Sim2, and ask for similarity R _Sim1With similarity R _Sim2Mean value as detection threshold.

The detection threshold computing formula is: $R_{\mathrm{sim}1}=\mathrm{sum}\{(A_1-\mathrm{mean}\left(A_1\right))\⊗(A_1-\mathrm{mean}\left(A_1\right))\}$

$R_{\mathrm{sim}2}=\mathrm{sum}\{(A_1-\mathrm{mean}\left(A_1\right))\⊗(A_n-\mathrm{mean}\left(A_n\right))\}$

G _T＝(R _sim1+R _simm2)/2

A wherein ₁The picture element matrix of target area when the expression target exists, A _nThe picture element matrix of target area when the expression target does not exist, two identical m * n that are of picture element matrix size, the mean value of all elements among mean (A) the expression A, R _SimRepresent similarity, Represent the matrix multiplication cross, the summation of sum{} representing matrix element, G _TThe expression monitoring threshold.

4. method according to claim 1 is characterized in that the similarity of current video frame objective matrix is:

$R_{\mathrm{simc}}=\mathrm{sum}\{(A_1-\mathrm{mean}\left(A_1\right))\⊗(A_c-\mathrm{mean}\left(A_c\right))\}$

Compare R _SimcWith G _TIf, R _SimcGreater than G _TThen object judges that target exists; If R _SimcLess than G _TThen object judges that target is stolen.

5. the intelligent video stolen articles according to the described method of claim 1 detects and warning device, it is characterized in that being made up of six parts such as UPS function power supply, video camera, infrared lamp group, DSP video acquisition and treatment circuit, warning circuit, videograph memory device.Power unit has the UPS circuit, device is worked on when abnormity of power supply and sends alerting signal.Video camera is taken in real time to the supervision scene in the visual field, the infrared lamp group is regulated the brightness that exposure intensity makes the visual field automatically and is kept within limits when dark in the visual field, the vision signal of DSP video acquisition and treatment circuit capture camera, carry out real-time targeted surveillance and camera status analysis, when monitored object is lost, video camera is blocked, and sends alerting signal by warning circuit when losing the interruption of frame or mains-supplied, and stores 15s and the videograph of 15s before afterwards.

6. intelligent video stolen articles according to claim 5 detects and warning device, it is characterized in that power supply is the combination of commercial power rectification filter regulator circuit or battery and ups power, effectively reduce power supply ripple, isolate surge, after being cut off by the offender, power supply can work alone 3 minutes the simultaneity factor signal that can give the alarm.

7. intelligent video stolen articles according to claim 5 detects and warning device, it is characterized in that video camera is the video format output of PAL or NTCS, and makes up the dynamic adjustment of carrying out field luminance with infrared lamp.

8. intelligent video article anti-theft supervising device according to claim 5, it is characterized in that DSP video acquisition and treatment circuit be made up of DSP master's process chip, video decoding chip, FLASH program storage, SDRAM data-carrier store, main control computer telecommunication circuit and peripheral circuit, wherein the intelligent video stolen articles detection method described in DSP master's process chip operation claim 1.

9. intelligent video stolen articles according to claim 5 detects and warning device, it is characterized in that warning circuit is made up of the LED lamp and the high-power alarm speaker of violent flicker.

10. intelligent video stolen articles according to claim 5 detects and warning device, it is characterized in that the videograph memory device is made up of industrial computer and hard-disk system, can according to set the real-time continuous stored video signal or only deposit report to the police before and after the surveillance video of 15s.

Images