CN103839346B - A kind of intelligent door and window anti-intrusion device and system, intelligent access control system - Google Patents

Abstract

A kind of intelligent door and window anti-intrusion device, described anti-intrusion device includes lighting module, video acquisition module, integrated information acquisition module, warning sending module, signal processing module, lighting control module, it is characterised in that: the video acquisition that described lighting module is video acquisition module provides lighting source；Described video acquisition module starts video information around collection when lighting module is opened；Described integrated information acquisition module Real-time Collection peripheral information, and it is supplied to described signal processing module in real time；Described lighting control module determines the unlatching of lighting module, and wherein, described lighting control module determines described unlatching according to described signal processing module to the result of described peripheral information；Described signal processing module differentiates intrusion behavior according to described video information and described peripheral information, thus decides whether to start warning sending module, sends alarm signal.The invention allows for a kind of anti-intrusion system and gate control system, the anti-intrusion device of the present invention and system, information is had comprehensive analytical capacity, intrusion behavior is had Intelligent Recognition ability, significantly reduces wrong report.

Description

A kind of intelligent door and window anti-intrusion device and system, intelligent access control system

Technical field

The present invention relates to a kind of anti-illegal invasion system, be specifically related to a kind of intelligent door and window anti-intrusion device, anti-intrusion system System and intelligent access control system.

Background technology

It is known that in order to avoid theft, the anti-of photographic head etc is installed in general Hui community or outdoor corridor etc. Steal device, monitoring that a kind of indoor occupant based on intelligent video is swarmed into and warning system, mainly by photographic head The intellectual analysis of image, it is achieved indoor place is moved the analysis of target, personnel is broken in and reaches to find in time and report Alert, to ignore the motion being safe from danger simultaneously, reduce false-alarm, also have a kind of Household intelligent anti-theft system, it is at least by antitheft door, machine Tool formula lockset, window, terrace and balcony door are constituted.By being arranged on: the door net sensor on (1) antitheft door, antitheft door Magnetic Sensor, dead bolt sensor, key that (2) are arranged on lockset insert/extract sensor, pellet beats sensor, and (3) pacify The handling position sensor being contained in outer handle molectron, arranges simultaneously and enters the balcony balcony door passage from terrace Path detection district, enters the window path detection district the passage in room from window, it is achieved the owner in automatically identifying at home with The Smart Home burglary-resisting system of robber.

There is following shortcoming in existing door and window anti-intrusion system:

There is respective defect in the most single anti-intrusion detection technique, cannot normally work in some cases, once without Method normally works, and whole detecting system is in paralyzed state, the poor stability of system.As: intelligent video analysis detection may Disturbed by irrelevant motions such as ambient light change, direct reflections, caused substantial amounts of wrong report；Rpyroelectric infrared detection Yi Shouwen The interference of degree, high light, circumstance complication motion；Vibration detection is vulnerable to the interference that the external force such as wind cause door and window vibration to produce；Glass Broken detection is vulnerable to the interference of external environment noise；Door Magnetic testi needs door and window strictly to close, in the feelings that indoor needs ventilate Will be unable under condition use, such as hot summer, user generally requires ventilation of windowing.

The most traditional anti-intrusion system cannot carry out the ability of anticipation, the early warning of system to the behavior of setting foot-point before robber's crime Ability；

3. need could trigger after robber enters the room warning, and can not trigger in the very first time implementing crime and report to the police, fall The low promptness reported to the police；

4. relating to substantial amounts of various sensor, installation and debugging are extremely complex, and warning message is the most directly perceived simultaneously, it is impossible to learn existing Field situation；

The most traditional warning system cannot work, to illegal invasion person can take advantage of it in the case of cutting off external power supply Machine.

Summary of the invention

It is an object of the invention to provide a kind of intelligent door and window anti-intrusion system, solve existing anti-theft measure function list One, rate of false alarm is high, reports to the police not in time and easy out-of-work problem.

For solving above-mentioned technical problem, the present invention by the following technical solutions:

A kind of intelligent door and window anti-intrusion device, described anti-intrusion device includes lighting module, video acquisition module, comprehensively believes Breath acquisition module, warning sending module, signal processing module, lighting control module, it is characterised in that:

The video acquisition that described lighting module is video acquisition module provides lighting source；

Described video acquisition module starts video information around collection when lighting module is opened；

Described integrated information acquisition module Real-time Collection peripheral information, and it is supplied to described signal processing module in real time；

Described lighting control module determines the unlatching of lighting module, and wherein, described lighting control module is according to described signal Processing module determines described unlatching to the result of described peripheral information；

Described signal processing module differentiates intrusion behavior according to described video information and described peripheral information, thus determines Whether startup warning sending module, sends alarm signal.

Anti-intrusion device as above, it is characterised in that:

Also including power management module, be powered described anti-intrusion device, described power management module connects external Power supply and accumulator；

When external power supply normal power supply, whole product uses external power supply；

When external power supply is cut off, described accumulator is automatically enabled as described anti-intrusion device and powers.

Arbitrary described anti-intrusion device, it is characterised in that:

Described integrated information acquisition module includes audio collection module, heat release infrared detection alarm module, vibration inspection Survey the one or more modules in module, glass breaking detection module, door Magnetic testi module；

Described peripheral information includes voice messaging, infrared information, vibration information and glass breaking acoustic information, door magnetic letter Breath.

Anti-intrusion device as above, it is characterised in that:

Described glass breaking detection module profit includes glass breaking intrusion detector, and the glass of high frequency is broken by described detector Broken sound effectively detects, and the common acoustic signal of low frequency is had stronger inhibitory action.

Arbitrary described anti-intrusion device, it is characterised in that:

In the form of a ring, described heat discharges infrared detection alarm module, vibration detection mould to illuminating lamp in described lighting module Block, video acquisition module, audio collection module and warning sending module distributing installation are at the outward flange of described illuminating lamp, described letter Number processing module, power management module and switch control module are arranged on the inside of described illuminating lamp.

Arbitrary described anti-intrusion device, it is characterised in that:

Described glass breaking detection module, door Magnetic testi module will be detected by wired or wireless communication mode respectively Information is transformed into voltage signal and exports to described signal processing module.

Arbitrary described anti-intrusion device, it is characterised in that:

Described warning sending module is to use wired or wireless communication mode, the alarm signal sent by signal processing module Send.

Anti-intrusion device as above, it is characterised in that:

Described communication includes 3G, WIFI, Zigbee, Bluetooth or other communication.

Anti-intrusion device as above, it is characterised in that:

Described signal processing module collection described heat release infrared detection alarm module, vibration detection module, glass break The signal that broken detection module, door Magnetic testi module, video acquisition module and audio collection module transmit, carries out total score Analysis, thus complete described resolution intrusion behavior, and send alarm signal by described warning sending module.

Arbitrary described anti-intrusion device, it is characterised in that:

Described signal processing module includes main process chip, and described main process chip uses main core processing device to process with from core The Duo-Core Architecture pattern that device combines.

Anti-intrusion device as above, it is characterised in that:

The described video information that described main core processing device reception video acquisition module and integrated information detection module collect With described peripheral information；

Described video information and described peripheral information constitute source signal and send to from core processor；

Described from core processor resolution intrusion behavior, form final decision signal.

Anti-intrusion device as above, it is characterised in that:

Described resolution intrusion behavior mainly includes running audio-video intelligent analysis algorithm, in combination with in described peripheral information Other information differentiate.

The invention allows for a kind of intelligent intrusion-proof system, described system include memory module and arbitrary described in Anti-intrusion device；Described memory module sets up communication connection with described anti-intrusion device by the way of wired or wireless communication, The information gathering described anti-intrusion device stores；

It is characterized in that:

Described resolution intrusion behavior includes the people close to anti-intrusion device is carried out identification, and described identification is logical Cross recognition of face or other mode identification completes；

While carrying out described identification, also to integrated information acquisition module described when people is close to anti-intrusion device The described peripheral information collected is analyzed；

Described signal processing module is analyzed based on described identification with to described peripheral information, forms decision-making letter Number, differentiate whether intrusion behavior is set up.

Intelligent intrusion-proof system as above, it is characterised in that:

Other mode identifications described include carrying out identification by the mobile phone carried with, described anti-intrusion device with Mobile phone communicates the identity determining people.

Anti-intrusion system as above, it is characterised in that:

Before carrying out described recognition of face, also include that face registration is to form face database；

In described recognition of face, use HAAR feature to combine adaboost algorithm and carry out Face datection and realize Face detection；

Use two-layer human eye localizer, obtain human eye by adaboost algorithm and position, utilize Face detection result to pass through Eyes are corrected to level and realize face standardization by image rotation；

Carry out feature extraction by two-dimensional Gabor filter, utilize the covariance distance between vector as matching measurement side Formula, realizes facial image to be identified by nearest neighbor classification and mates with the facial image in described face database.

Arbitrary described system, it is characterised in that:

Described resolution intrusion behavior also includes that the action in described video information carries out detection to be analyzed.

System as above, it is characterised in that:

Described action is carried out detect analysis include background modeling and motion detection, goal verification, target following, behavior divide Analysis；

Described background modeling and motion detection, be that the mode using multiple image to be averaging obtains background mean value, by background Variance be initialized as a constant, pass through formula

$f(x,y)=\left\{\begin{array}{cc}1& |f(x,y)-\mathrm{\μ}(x,y)|>N\×{\mathrm{\σ}}^2(x,y)\\ 0& else\end{array}\right.$

By foreground detection by background pixels and prospect pixel separation, wherein 1 represents prospect pixel, and 0 represents background pixels, (x y) represents that in image, (x, y) gray value of place's pixel, (x y) represents the background mean value of correspondence, σ to μ to f²(x y) represents correspondence The variance of background；

Described goal verification, filters, according to the size set, the prospect of dimensional requirement of being unsatisfactory for by target size wave filter Region；Highlighted foreground area is filtered according to setting threshold value by highlighted target filter；By foreground area with corresponding background Do gradient difference, by time domain filtering, motion continuity, motion significance and area change seriality are filtered respectively；

Described target following, according to the location estimation target of present frame and target before in the position of next frame, by public affairs Formula

$v_t^x=\frac{\underset{n=0}{\overset{N-2}{\Σ}}(x_{t-n}-x_{t-n-1})}{N-1}$

$v_t^y=\frac{\underset{n=0}{\overset{N-2}{\Σ}}(y_{t-n}-y_{t-n-1})}{N-1}$

To estimating that target velocity realizes target prodiction, whereinWithRepresent that t target is at x direction and y respectively The speed in direction, N is time window, x_t‐nAnd y_t‐nRepresent the abscissa at rectangle frame center, the t n moment target external world respectively and indulge Coordinate, in like manner x_t‐n‐1And y_t‐n‐1Represent abscissa and the vertical coordinate at rectangle frame center, the t n 1 moment target external world respectively；

Described behavior analysis, sets alarm region by user and corresponding rule realizes the judgement of specific behavior.

Arbitrary described system, it is characterised in that:

The voice messaging being analyzed including being collected described audio collection module to described peripheral information is examined Cls analysis, will invade helpful voice messaging as foreground signal from the numerous and jumbled background audio as background signal to differentiating Information extracts.

System as above, it is characterised in that:

Described voice messaging carries out detect analysis include:

Obtain the power spectral density of audio signal in described voice messaging, if the sample rate of audio signal X (t) is fs, X (t) is sequentially passed through preemphasis, framing and windowing process, before signal processing, first removes average, use in classical spectrum estimate Period map method, use fast Fourier transform FFT realize, finally give normalized power spectrum X (t) (fn), wherein n ∈ [1, N], N is sampling number, and t represents sampling instant；

Use sub belt energy ratio A i (t) to describe power spectral density, utilize formulaCalculate subband Energy ratio Ai (t) is defined as the ratio of sub belt energy and gross energy, wherein U (i) and L (i) represent subband i coboundary and under Frequency band is divided into M subband according to logarithm by border here；

Mixed Gaussian background modeling based on sub belt energy ratio and sound detection, mixed Gaussian background modeling assumes signal The probability distribution of change can use K Gauss Distribution Fitting, and the probability distribution of audio signal is expressed as

$prob\left(x\right)=\underset{i=1}{\overset{K}{\Σ}}{\mathrm{\ω}}_i{\mathrm{\η}}_i(x,{\mathrm{\μ}}_i,{\mathrm{\σ}}_i)$

In model, each Gauss distribution η i (x, μ i, σ i) gives weights omega i, and wherein μ i and σ i represents Gauss respectively The average of distribution and standard deviation, multiple Gauss distribution obtain the probability distribution of signal by linear combination, by front B Gauss distribution As background model, remaining Gauss distribution is as foreground model, wherein

Current demand signal amplitude is mated with front B Gauss distribution, and if any one Gauss distribution therein Being made into merit then this signal is background signal, is otherwise foreground signal.

Arbitrary described system, it is characterised in that:

Described system carries out described identification, is analyzed described peripheral information, forms decision signal, differentiates invasion Whether behavior sets up comprises the steps:

The first step, is collected digital video signal by described video acquisition module, is collected by described audio collection module Digital audio and video signals, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage Signal, produces deformation or the information of stress transformation by vibration detection module collection extraneous vibration and is changed into voltage signal, logical Cross glass breaking detection module and door Magnetic testi module collection glass breaking and door magnetic change information and be changed into voltage signal；

Second step, described signal processing module receives heat release infrared detection alarm module, vibration detection module, glass Glass crushes detection module and the voltage signal of door Magnetic testi module, and combines audio-video intelligent analysis, completes identification, based on The identity identified carrys out the generation of synthetic determination intrusion behavior, when there is intrusion behavior, sends alarm signal；

3rd step, warning sending module receives alarm signal, and sends warning to outside by the way of wired or wireless Information.

System as above, it is characterised in that:

Warning sending module receives alarm signal and includes to outside alert:

Warning sending module receives described digital video signal and described digital audio and video signals, and described digital video is believed Number and described digital audio and video signals send together with alarm signal.

The invention allows for a kind of intelligent access control system, described intelligent access control system includes arbitrary described intelligence Anti-intrusion system, described intelligent access control system utilizes described intelligent intrusion-proof system to complete described identification, to described week The information of enclosing is analyzed, and identifies whether to visit for guest, and automatically the information that visiting guest arrives is sent to mobile or fixing Terminal.

Intelligent access control system as above, it is characterised in that:

Described identifying whether comprises the steps: for guest's visiting

The first step, is collected digital video signal by described video acquisition module, is collected by described audio collection module Digital audio and video signals, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage Signal, produces deformation or the information of stress transformation by vibration detection module collection extraneous vibration and is changed into voltage signal, logical Cross glass breaking detection module and door Magnetic testi module collection glass breaking and door magnetic change information and be changed into voltage signal；

Second step, signal processing module receive heat release infrared detection alarm module, vibration detection module, glass break Broken detection module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis and determine whether that guest visits；

3rd step, the generation of behavior of visiting if there is guest, wirelessly by visitor at video outdoors or Photographic intelligence is sent automatically to be in the mobile or fixed terminals such as indoor master cellular phone, and activated terminals be given ring, vibrations or Voice message.

Compared with prior art, the invention has the beneficial effects as follows: intelligent door and window anti-intrusion system of the present invention uses pyroelectricity The detection of infrared detection, vibration detection, audio detection, glass breaking and door Magnetic testi technology, find door by multidimensional detection mode The illegal invasion that may be present behavior of window region, can accurately tell illegal invasion, guest's visiting or householder normal Come in and go out, the alarm signal real-time, directly perceived that audio frequency and video or picture are carrier was sent by wired or wireless communication means, the very first time Reaching such as alert receipt and the processing terminal such as householder's mobile phone, cell monitoring computer, link warning lamp the most simultaneously, and warning signal realizes incident On-the-spot sound and light alarm, or other associate device that links；Additionally native system is installed on outdoor, it is possible at utmost by illegal invasion Eliminated before it actually occurs；Compared to traditional warning system, the present invention also carries stand-by power supply, is cutting off external power supply In the case of can be power itself 2 24 hours, sensitization time section remained able to anti-intrusion detection.

Accompanying drawing explanation

Fig. 1 is the connection diagram of one embodiment of intelligent door and window anti-intrusion system of the present invention.

Fig. 2 is the hardware connection diagram of intelligent door and window anti-intrusion system of the present invention.

Fig. 3 is the software architecture diagram of the present invention bright intelligent door and window anti-intrusion system.

Fig. 4 be intelligent door and window anti-intrusion system of the present invention signal processing module in primary processor end schematic flow sheet.

Fig. 5 be intelligent door and window anti-intrusion system of the present invention signal processing module in from processor end schematic flow sheet.

Fig. 6 is face registration and the recognition of face schematic flow sheet of intelligent door and window anti-intrusion system of the present invention.

Fig. 7 is moving object detection and tracking flow process based on the video signal of intelligent door and window anti-intrusion system of the present invention Figure.

Fig. 8 is the audio detection schematic flow sheet of intelligent door and window anti-intrusion system of the present invention.

To accessing facial image analysis schematic diagram when Fig. 9 is the face standardization of intelligent door and window anti-intrusion system of the present invention.

Detailed description of the invention

In order to make the purpose of the present invention, technical scheme and advantage clearer, below in conjunction with drawings and Examples, right The present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, and It is not used in the restriction present invention.

Fig. 1 shows an embodiment of the present invention a kind of intelligent door and window anti-intrusion system: a kind of intelligent door and window anti-intrusion System, including

Video acquisition module, uses video sensor to gather video stream signal, completes the digitized of picture signal, and image is believed Number pretreatment, be met the digital video signal of signal processing module requirement and to export it to signal processing module concrete Detection mode can be to be domain mode or curtain mode；

Audio collection module, completes the digital-to-analogue conversion of voice signal, and the sample code of voice signal and Filtering Processing obtain Meet the digital audio and video signals of signal processing module requirement and export to signal processing module；

Heat release infrared detection alarm module, the difference of the temperature of inducing moving objects and background object, when human body moves Dynamic intermittent fever releases the infrared different information that can sense human body temperature and ambient temperature, is converted into voltage signal backward signal and processes Module exports；

Vibration detection module, produces extraneous vibration deformation or is changed into voltage signal by force information, to signal processing mould Block exports；

Glass breaking detection module and door Magnetic testi module, changed corresponding information by wired or wireless communication mode Voltage signal is become to export to signal processing module；

Signal processing module, collect heat release infrared detection alarm module, vibration detection module, video acquisition module and The signal that audio collection module transmits, comprehensively analyzes, and differentiates illegal invasion or guest's visiting, sends to reporting to the police respectively Module sends the information such as alarm signal or the photo sending visitor to the owner that is in；

Warning sending module, uses wired or wireless communication mode, is mainly used in receiving the report that signal processing module sends Alert signal and send warning message to householder's mobile phone or cell management center, can use wired, 3G, WIFI, Zigbee, One of Bluetooth or other communication；

Power management module, connects external power supply and accumulator, when external power supply normal power supply, outside whole product uses Connecing power supply, when extraneous power cut-off, internal standby battery enables automatically, it is ensured that above-mentioned each module is after external power interruptions Continue normal work, preferably ensure that video camera works on 2 24 hours after external power interruptions.

In figure 1 it can be seen that invention is arranged on ring like light fitting, described heat discharges infrared detection alarm module, shakes Dynamic detection module, video acquisition module, audio collection module and warning sending module are arranged on the outward flange of ring like light fitting, described The switch lamp control module of signal processing module, power management module and barrel type lamp itself is arranged in barrel type lamp Portion.Described switch lamp control module can be in conjunction with vibration, sound, infrared signal of Denging determines whether that people enters, if there being people to enter Enter, turn on lights optical illumination, provide ambient lighting for subsequent video collection.Whole system is fabricated to similar illuminating lamp by us The mutual set of appearance, system and lamp, can not only well stash native system, prevent lawless person to this intelligent door and window Anti-intrusion system destroys, moreover it is possible to the photographic head of video acquisition module in system is carried out light filling, allow gather video information Become apparent from, when ambient light illumination deficiency, to video acquisition module provide secondary light source, it is ensured that its in the environment of low-light (level) still So can collect effective view data.

Fig. 2 shows a preferred embodiment of the present invention a kind of intelligent door and window anti-intrusion system, signal processing module, master Process the Duo-Core Architecture pattern that chip uses main core processing device to be combined with from core processor.Main core processing device mainly completes numeral and regards Frequently signal and the collection of digital audio and video signals, audio/video coding, collect the signal of heat release infrared detection alarm module output With the signal of vibration detection module output, receive glass breaking detection module and the signal of door Magnetic testi module transmission, by multi-source Signal sends to from core processor together with audio, video data；Carry out the judgement of intrusion behavior from core processor and householder's identity is tested Card, wherein intrusion behavior judges mainly to include running audio-video intelligent analysis algorithm, is judged whether by video and sound Intrusion behavior, discharges the signal of infrared detection alarm module output, the signal of vibration detection module output, glass in combination with heat Glass crushes detection module and the signal of door Magnetic testi module transmission, forms final decision signal.

As it is shown on figure 3, software is mainly by signals collecting, signal analysis, reports to the police and send and coordinated signals three part composition.Letter Number gather and warning transmission operate in main core processing end, signal analysis operate in from core process end.Signals collecting interface includes Audio-video collection module, rpyroelectric infrared alarm signal acquisition module, vibration alarming signal acquisition module, glass breaking detection mould Block and door Magnetic testi module and wireless alarm signal acquisition module.Main core processing end completes the collection of switching value alarm signal and prolongs Time synchronize, together with audio-video signal be sent to from core process end.Process end from core and complete intelligent sound video analysis, multi-source alarm signal Breath merges and decision-making, is completed the confirmation of householder's identity by recognition of face mode or wireless WIFI mode, invasion row detected If for time user identification confirmation failure; to main core processing end send warning/early warning signal, main core processing end by report to the police send out Send module to send to exterior terminal to report to the police and coordinated signals information.

As shown in Figure 4, main core processing end software flow:

After system electrification, first complete the initialization of master chip main core processing end, next complete peripheral collecting device, logical Letter equipment initializes, and starts main core processing program.Create audio-video collection thread, rpyroelectric infrared detection thread, wireless receiving Thread, communication thread, warning send thread.

Audio-video collection thread: combine a frame multi-source data after receiving audio, video data, including video data, audio frequency The mark whether data, various sensor alarm intensity and user mobile phone exist, puts in FrameBuffer and adjusts for communication thread With, complete pre-recording of audio frequency and video, in order to provide warning message simultaneously.

Alarm signal collecting thread: receive the switching value alarm signal of the alarm sensor of various wired or wireless access Ai, wherein Ai can be rpyroelectric infrared alarm signal, vibration alarming signal, glass breaking alarm signal, door magnetic alarm signal Deng.

Communication thread: receive the alarm signal sent from core processor end, notice is reported to the police and is sent thread.If FrameBuffer exists a frame multi-source data, then transmits this data to DSP end.

Reporting to the police and send thread: after receiving alarm signal, the warning message that tissue is relevant, such as the audio frequency and video pre-recorded, picture Deng, send to exterior terminal.

As it is shown in figure 5, from core processor end software flow:

First, after main core processing device end starts, multitask kernel and application program are loaded into from core processor internal memory, Complete a series of initialization from core processor, automatically create communication thread and recognition of face thread.Communication thread first determines whether The no transmission warning/early warning signal that needs, if it is to main process end transmission warning/early warning signal, otherwise judges whether to receive One frame multi-source information, wherein multi-source information includes: audio, video data, and whether various sensor alarm signals and user mobile phone exist Mark, multi-source information is put in frame buffer FrameBuffer, sends to main process end simultaneously and receive confirmation signal, show From processing, end is working properly.Recognition of face thread takes out video data from FrameBuffer and completes recognition of face, if face Identify successfully, then the mark of set identity validation.Intrusion detection thread reads frame data from FrameBuffer, passes through sound Frequency analysis, vibration information, infrared information determines whether that people will enter monitoring region, if it is opens illumination, for connecing down The video analysis come provides ambient lighting, facilitates householder to open the door simultaneously.Open illumination after setting in motion detection, due to be only concerned into Door this behavior, so needing to carry out direction of motion judgement, when be judged as entering direction time, target is tracked.Work as target Exist the time more than the setting time time, illustrate as a suspect, if the failure of WIFI mode identity validation, system can pass through language Sound prompting is cooperated with on one's own initiative and is carried out recognition of face, if authentication failure, triggers intrusion behavior and reports to the police, send report to ARM end Police/early warning signal.

Fig. 6 shows another preferred embodiment of the present invention a kind of intelligent door and window anti-intrusion system, and householder's identity judges Including recognition of face or at user mobile phone, WIFI communication software is installed, video acquisition module and user mobile phone by WIFI communicate into Row identity validation, the method for wherein said recognition of face mainly includes with lower part:

Face registration, gathers the front face photo of the different angles of same person, finds face to exist by Face detection Position in image, and standardized face's size, face angle and illumination, extract the feature of standardized face's image, typing is noted Volume customer data base；

Recognition of face, uses HAAR feature to combine adaboost algorithm and realizes Face datection and realize Face detection, have employed Two-layer human eye localizer, is all to obtain human eye by adaboost algorithm to position, utilizes Face detection result to pass through image rotation The level that is corrected to by eyes realizes face standardization, carries out feature extraction by two-dimensional Gabor filter, utilizes the association between vector Variance distance, as matching measurement mode, realizes facial image and data base's facial image of inquiry by nearest neighbor classification Coupling.Wherein Adaboost algorithm by Weak Classifier general for a large amount of classification capacities according to the mode group of training error index decreased It is combined into a strong classifier.And HAAR is characterized as that adaboost algorithm provides the weak typing feature of magnanimity, it is ensured that Adaboost algorithm totally finds the weak typing of excellent performance.In Face datection implementation process, integration histogram and cascade point The use of class device greatly reduces the process time while ensureing relatively high measurement accuracy；It is generally divided into two-layer fixed during human eye location Position, wherein ground floor is coarse positioning, and region, location have selected and includes eyes eyebrow at interior most of ocular, the second layer For being accurately positioned, region, location only comprises ocular.Coarse positioning device compared to precise localizer owing to having contained district more Domain information, the stability therefore positioned is higher, is substantially not present bigger position deviation, and precise localizer is capable of human eye Be accurately positioned, but be susceptible to eyebrow, the interference at canthus causes Wrong localization.Geometric proportion is passed through on the basis of coarse positioning Example relation determines the approximate location scope of human eye, uses precise localizer to realize being accurately positioned of human eye within the range.Pass through By slightly to the location mode of essence, reduce the impact on location such as eyebrow, canthus, improve the accuracy of location；And face mark Standardization is a step the most crucial in recognition of face, and the quality of standardization result directly affects the precision of recognition of face. Face standardization mainly completes geometric correction and the gamma correction of facial image.The result utilizing previous step human eye to position is easy to Realize the geometric correction of facial image, first pass through image rotation and eyes are corrected to level, by eyes distance d to face figure As intercepting.As it is shown in figure 9, the most finally scale the images to 80x80 pixel.

Gamma correction eliminates the uneven illumination impact on follow-up identification the most to a certain extent.Mainly include plane of illumination Matching, deducts plane of illumination, and histogram equalization and gray value normalize to zero-mean, unit variance.It is assumed here that plane of illumination is one Individual plane.Point on plane of illumination meets equation below: IP, and (x, y)=ax+by+c are write as matrix form i.e. x=Np, and wherein x represents The column vector that the picture element gray value of image is lined up, N represents the coordinate that picture element is corresponding, and first row represents abscissa, secondary series Represent vertical coordinate, the 3rd row filling 1, p=[a b c]^T.Plane parameter a, b, c can try to achieve, i.e. by the way of linear regression P=(N^TN)^-1N^Tx。

Can have selected Gabor wavelet feature, Gabor transformation has excellence in terms of analyzing image local area texture Performance.Two-dimensional Gabor filterCan be expressed as:

WhereinFor image coordinate,For the mid frequency of wave filter, k_xAnd k_yRepresent respectivelyIn throwing horizontally and vertically Shadow,For the direction of wave filter, u and v represents different values,For Gaussian envelope,It is multiple Numerical value plane wave.Two-dimensional Gabor filter modulates characteristic frequency and the sinusoidal wave Planar realization in direction by two-dimensional Gaussian function, Different scale and the analysis of different directions image texture is realized by the frequency and direction changing sinusoidal wave plane.

Obtained the facial image of 80x80 size by face standardization, have selected 5 filter scales, 8 filters here Ripple direction, obtains 40 different directions and the Gabor filter of frequency, obtains a facial image by after wave filter convolution Magnitude image after 40 Gabor wavelet conversion, the Gabor characteristic dimension finally obtained is 163840.Such a higher-dimension Characteristic vector can be substantially reduced the speed identifying classification, it is therefore desirable to characteristic vector is carried out dimensionality reduction.Here use 4x4 uniform Down sample realizes Feature Dimension Reduction.

Use nearest neighbor classification to realize the facial image of inquiry and mating of data base's facial image, use between vector Covariance distance, as matching measurement mode, can weigh the credibility of final coupling simultaneously by covariance distance.

According to another preferred embodiment of the present invention a kind of intelligent door and window anti-intrusion system, signal processing module

Connect and have memory module (Storage), for different applicable cases, different types of storage medium can be used, SD and TF card is controlled by SDIO, and the storage medium of this type is convenient for changing；Nand is controlled by Nandflash, this type Integrated level is high, but is not easy to change storage medium；SSD is controlled by PCI E or SATA interface, and this type memory space can The biggest to accomplish；

It is provided with Ethernet interface (RJ45)；

Connected by power management module (Power manager) and have external power supply (DC IN) and battery (Battery)；

Connect and have the internal memory (DDR) run for master cpu；

Connect and have the flash memory (Flash) for storing startup procedure, configuration parameter, log information；

Connect and have wireless module (3G, WIFI), for audio, video data and the transmission of remote control signal；

Connect and have wireless module (zigbee, Blue tooth or other wireless modules), be used for receiving distributed alarming device The alarm signal sent, such as glass breaking signal and door magnetic opening signal, it is possible to send control information to miscellaneous equipment.

Connecting has by external warning lamp, warning signal interlocking equipment, it is achieved local sound and light alarm, or controls other associate device Interlink warning module.

Fig. 7 shows another preferred embodiment of the present invention a kind of intelligent door and window anti-intrusion system, video acquisition module Detection method be broadly divided into background modeling and motion detection, goal verification, target following and behavior analysis four part, wherein

Described background modeling and motion detection, be to have employed the mode that multiple image is averaging to obtain background mean value, will the back of the body The variance of scape is initialized as a constant, passes through formula

$f(x,y)=\left\{\begin{array}{cc}1& |f(x,y)-\mathrm{\μ}(x,y)|>N\×{\mathrm{\σ}}^2(x,y)\\ 0& else\end{array}\right.$

By foreground detection by background pixels and prospect pixel separation, wherein 1 represents prospect pixel, and 0 represents background pixels.f (x y) represents that in image, (x, y) gray value of place's pixel, (x y) represents the background mean value of correspondence, σ to μ²(x y) represents the correspondence back of the body The variance of scape,

If be detected that large area prospect pixel then system thinks that scene may have occurred change, at this moment close follow-up dividing Analysis, if large area prospect pixel all being detected for a long time, system thinks that scene there occurs change really, at this moment restarts Background modeling.

The renewal of the background of native system have employed based on target aspect and the context update plan that combines based on pixel aspect Slightly.When picture element is detected as background dot, use context update based on pixel aspect, use single Gaussian Background to update here Method, including renewal and the renewal of background variance of background mean value, shown in equation below:

μ_t(x, y)=(1-α) × μ_t-1(x,y)+f_t(x,y)×α

${\mathrm{\σ}}_t^2(x,y)=max(\mathrm{\Δ},(1-\mathrm{\α})\×{\mathrm{\σ}}_{t-1}^2(x,y)+{\left(f_t\left(x,y\right)-{\mathrm{\μ}}_t\left(x,y\right)\right)}^2\×\mathrm{\α})$

Wherein subscript t represents t, and α is for updating the factor, and usually one value the least, such as 0.01.Here clamper side The minima of difference is Δ, prevents model from too restraining, causes model noise resisting ability to weaken, and the value of Δ is 36 here.

When picture element is detected as foreground point, if judging that picture element is not target pixel point by Target Segmentation, adopt With context update based on pixel aspect, if i.e. this picture element at a fixed time in be detected as prospect picture element for a long time, Picture element is disposably updated in background；If judging that picture element is target pixel point by Target Segmentation, use based on mesh The background update method of mark aspect, i.e. only judges that target, just can be by target place when not moving for a long time Area pixel point disposably updates in background.

Described goal verification, filters, according to the size set, the prospect of dimensional requirement of being unsatisfactory for by target size wave filter Region, filters highlighted foreground area by highlighted target filter according to setting threshold value, by time domain filtering to motion even Continuous property, motion significance and area change seriality are filtered respectively.The most highlighted target filter: filter highlighted prospect Region, these foreground area are probably the interference such as light source or reflective medium.The determination methods of highlighted foreground area: if prospect In region, the ratio of highlighted pixel exceedes setting threshold value, then it is assumed that this foreground area is highlighted foreground area；Time domain filtering: time Territory filtering is filtering based on target aspect, needs by target following technology.Including: motion continuity filters, and motion is notable Property filtering, area change seriality filter.Motion continuity filters: the occupied region assuming target two frames front and back is big face Long-pending overlap, i.e. will not undergo mutation in the position of target.Motion significance filters: the motion assuming target is to have certain direction Property, i.e. the vector in a period of time internal object displacement add and add with the scalar of displacement of targets and ratio should be more than certain Threshold value.Area change seriality filters: the area assuming target is metastable, will not undergo mutation.

The tracking of described target, have employed target following based on motion detection, passes through frame on the basis of motion detection Between target association, form the movement locus of target, provide foundation for succeeding target behavior analysis, according to present frame and target before Location estimation target in the position of next frame, pass through formulaWithRight Estimate that target velocity realizes target prodiction, whereinWithExpression t target is in x direction and the speed in y direction respectively, N is time window, x_t-nAnd y_t-nRepresent abscissa and the vertical coordinate at rectangle frame center, the t-n moment target external world respectively, in like manner x_t-n-1And y_t-n-1Represent abscissa and the vertical coordinate at rectangle frame center, the t-n-1 moment target external world respectively.Feature selection aspect this In have selected two features of center C of the most reliable and the most stable target area S and target boundary rectangle frame, target association just mates Former frame target can be found in the position of present frame, it is achieved the location of target.According to correlation tracking principle, as long as ensureing enough In the case of image sampling rate, same target change in location between adjacent two frames is not too large, therefore can be by target Hunting zone be limited to a less distance range, be greatly reduced the risk of error hiding simultaneously.UtilizeFormula, present frame targetWith target to be matchedFeature is just carried out when meeting following formula Joining, wherein γ is detection range threshold value, needs to be determined according to actual scene.

The criterion of coupling, shown in equation below:

$\arg min(\mathrm{\&omega;}\&times;(\frac{dist\left(O_t^j\right\{C\},O_{t-1}^i\{C\left\}\right)}{\mathrm{\&gamma;}})+(1-\mathrm{\&omega;})\&times;\frac{\left|O_t^j\left\{S\right\}-O_{t-1}^i\left\{S\right\}\right|}{O_{t-1}^i\left\{S\right\}})<T$

The weight factor that wherein ω is characterized, takes 0.5 here, is provided with error upper limit T of coupling here, prevents mistake Join, take 0.4 here.

The mode that the renewal of target characteristic all uses single order smooth updates, shown in equation below:

$\begin{array}{cc}{O}_t^i\left\{F\right\}=\mathrm{\&alpha;}\&times;O_{t-1}^i\left\{F\right\}+(1-\mathrm{\&alpha;})\&times;O_t^i\left\{F\right\}& F=\{S,C\}\end{array}$

Wherein α is for updating the factor, takes 0.2 here.

Described behavior analysis, sets alarm region by user and corresponding rule realizes the judgement of specific behavior.Such as Alarm region can be set to a region, if the people entering this region exceedes setting value at the time of staying, being considered can The intrusion behavior of energy, triggers and reports to the police.The large area at door place can also be set to detection region of hovering, enter this region If people exceedes setting value at the time of staying, it is considered possible behavior of setting foot-point, triggers early warning.

Fig. 8 shows another preferred embodiment of the present invention a kind of intelligent door and window anti-intrusion system, audio collection module Detection method,

Intrusion detection flow process based on sound is as shown below, the audio signal obtained by audio collecting device, first Carry out pretreatment, mainly include preemphasis, framing and windowing, meet the needs of succeeding spectral analysis and identification, next use Mixed Gauss model sets up background sound model, starts sound detection, occur being different from when detecting in scene after having modeled During the abnormal noise of background sound, warning level starts to add up, if warning level is more than setting threshold value, illustrate that there is invasion goes For, trigger and report to the police, specific as follows,

Obtain the power spectral density of audio signal, if the sample rate of audio signal X (t) is f_s, (such as taking 48kHz), by X T () sequentially passes through preemphasis, framing and windowing process, (frame length is 48K sampled point, and frame moves as 24K sampled point, window function Select Hanning window), before signal processing, first remove average, it is to avoid DC component produces impact to the spectral line near at ω=0, Use the period map method in classical spectrum estimate, use fast Fourier transform FFT to realize, finally give normalized power spectrum X^(t)(f_n), wherein n ∈ [1, N], N is sampling number, and t represents sampling instant, (when sample rate is f_sThe value of=48kHz, N is 48000, for subsequent analysis, amount of calculation is bigger), use sub belt energy ratioDescribe power spectral density, utilize formulaCalculate sub belt energy ratioIt is defined as the ratio of sub belt energy and gross energy, wherein U (i) and L I () represents that frequency band is divided into M subband according to logarithm and (works as M=10, f by the coboundary of subband i and lower boundary here_s=48kHz Time corresponding subband border be 0.023kHz, 0.046kHz, 0.093kHz, 0.187kHz, 0.375kHz, 0.750kHz, 1.5kHz, 3kHz, 6kHz, 12kHz and 24kHz)；

Mixed Gaussian background modeling based on sub belt energy ratio and sound detection, mixed Gaussian background modeling assumes signal intensity Probability distribution can use K Gauss Distribution Fitting, the probability distribution of audio signal is expressed as Each Gauss distribution η in model_i(x,μ_i,σ_i) all give a weights omega_i, wherein μ_iAnd σ_iRepresent the average of Gauss distribution respectively And standard deviation, multiple Gauss distribution obtain the probability distribution of signal, ω by linear combination_iEmbody model high for i-th The attention degree of this distribution, when in a period of time, the amplitude of signal falls in this Gauss distribution, ω_iValue can be gradually increased, Otherwise ω_iCan be gradually reduced.σ reflects the convergence of Gauss distribution, for a Gauss distribution tending to convergence, and amplitude Fluctuation around average should be less, and σ is the least, and such Gauss distribution is just suitable for describing the change of background.K Gauss distribution According to the descending of ω/σ, arrange forward Gauss distribution and be best able to represent the distribution of background, using front B Gauss distribution as Background model, remaining Gauss distribution is as foreground model, whereinBy current demand signal amplitude Mating with front B Gauss distribution, if the match is successful with any one Gauss distribution therein, this signal is background, no It is then foreground signal, utilizes | I (x, t)-μ_i(x, t-1) | ＜ c × σ_i(x, t-1) mates.

According to another embodiment of the present invention a kind of intelligent door and window anti-intrusion system, preventing of intelligent door and window anti-intrusion system Inbreak method, step is as follows:

The first step, collects digital video signal by video acquisition module, collects DAB by audio collection module Signal, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage signal, logical Cross vibration detection module collection extraneous vibration produce deformation or the information of stress transformation and be changed into voltage signal, broken by glass Broken detection module and door Magnetic testi module collection glass breaking and door magnetic change information close and are changed into voltage signal；

Second step, signal processing module receive heat release infrared detection alarm module, vibration detection module, glass break Broken detection module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis, mastership identification checking, comprehensively Judging the generation of intrusion behavior, when there is intrusion behavior, sending alarm signal；

3rd step, warning sending module receives alarm signal, receives video acquisition module simultaneously and collects digital video signal With by audio collection module collect digital audio and video signals, and be transferred to by the way of wired or wireless external management terminal or Handheld terminal.

According to another embodiment of the present invention a kind of intelligent door and window anti-intrusion system, its use of intelligent door and window anti-intrusion system The method making visual entrance guard is as follows:

The first step, collects digital video signal by video acquisition module, collects DAB by audio collection module Signal, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage signal, logical Cross vibration detection module collection extraneous vibration produce deformation or the information of stress transformation and be changed into voltage signal, broken by glass Broken detection module and door Magnetic testi module collection glass breaking and door magnetic change information close and are changed into voltage signal；

Second step, signal processing module receive heat release infrared detection alarm module, vibration detection module, glass break Broken detection module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis and determine whether that guest visits, as There is guest and visit the generation of behavior in fruit, the most wirelessly visitor is automatically sent at video outdoors or photographic intelligence To the mobile or fixed terminals such as indoor master cellular phone of being in, and activated terminals provides ring, vibrations or voice message.

In the present invention:

Video acquisition module utilize intelligent video analysis technology, i.e. use computer, Embedded Double core processor etc. from By computing and analysis in video, extract a technology of useful information in video.This technology is powerful by various processors Data-handling capacity, carries out high speed analysis to the mass data in video pictures, filters out the unconcerned information of user, is only The key message that user is provided with.

The rpyroelectric infrared detection technique that infrared detection alarm module utilizes, i.e. the one of passive infrared detection, sensing The difference of the temperature of mobile object and background object.In monitoring region, during without human motion, pyroelectric infrared sensor senses Simply ambient temperature, when human body enters region, the infrared ray of human-body emitting gathers heat after being strengthened by Fresnel Lenses Release on electricity infrared inductor, the difference of the pyroelectricity element senses in rpyroelectric infrared to human body temperature Yu ambient temperature, temperature Changing, lose charge balance, outwards discharge electric charge, subsequent conditioning circuit detection produces alarm signal after processing.

The vibrating burglar alarm technology that vibration detection module utilizes: employing piezoelectric ceramic sheet is as sensor, by the external world Vibration produces deformation or stress is changed into voltage signal, is completed transmission and the stopping of alarm signal by follow-up auxiliary circuit.

The glass breaking detection technique that glass breaking detection detection module is utilized: utilize the piezoelectricity of piezoelectric ceramic piece to imitate Should, glass breaking intrusion detector can be made.The glass breaking sound (10k～15kHZ) of high frequency is effectively detected, and The acoustical signal of below 10kHZ (such as speak, sound of walking) there is stronger inhibitory action.The most just acoustic control and vibrations detection two Plant technical combinations together, the shake that the high-frequency sound signal sent when being only concurrently detected glass breaking and percussion glass cause Dynamic, just output alarm signal.

The door Magnetic testi technology that door Magnetic testi module utilizes: be made up of two parts: permanent magnet and the tongue tube of open type, Permanent magnet is used for producing constant magnetic field, and when permanent magnet and tongue tube are in close proximity, door status sensor is in work and waits for State, after permanent magnet leaves tongue tube certain distance, is in normally open, triggers and report to the police.

Native system workflow is:

User can be turned on and off warning function as required, and the mode of detection can select Video Detection, audio frequency to examine Survey, rpyroelectric infrared detects, vibration detection, and glass breaking detects, one or more in door Magnetic testi, as window area uses Rpyroelectric infrared detection, vibration detection, glass breaking detection, door Magnetic testi, door region uses the inspection of Video Detection, audio frequency Survey, rpyroelectric infrared detection, vibration detection, door Magnetic testi.Warning message can be the one in video, audio frequency or picture or several Kind.

(1) system electrification or reset after, signal processing module loads operating system and application program from FLASH, completes Initialization and the configuration of peripheral hardware to main process chip, next completes the initialization to each subsystem, just finally enters Often duty.

(2) the continuous acquisition monitoring of primary processor end of main process chip is on-the-spot audio-video signal and heat release infrared detection Signal, receives the alarm signal of other distributed alarming sensor simultaneously by wireless receiving module, by multi-source data send into from Processor end is analyzed, and carries out audio frequency and video simultaneously and pre-records.If receiving the warning from processor end or early warning signal, then will be pre- The audio frequency and video of record, are sent to the mobile phone of cell monitoring center or householder together with the photo captured by alarm signal sending module On.

(3) the running intelligent sound video analysis algorithm and intrusion behavior from processor end and judge of main process chip.Pass through sound Frequency analysis, vibration information, infrared information determines whether that people will enter monitoring region, if it is opens illumination, for connecing down The video analysis come provides ambient lighting, facilitates householder to open the door simultaneously.Open illumination after setting in motion detection, due to be only concerned into Door this behavior, so needing to carry out direction of motion judgement, when be judged as entering direction time, target is tracked.Work as target Exist the time more than the setting time time, illustrate as a suspect, if the failure of WIFI mode identity validation, system can pass through language Sound prompting is cooperated with on one's own initiative and is carried out recognition of face, if authentication failure, triggers intrusion behavior and reports to the police, send out to primary processor end Deliver newspaper police/early warning signal.

(4), after Surveillance center and householder receive early warning or report to the police, can be confirmed by the warning message sent back, and Take counter-measure.

Experiment:

The installation of native system: ceiling is installed, it is ensured that monitoring range can cover whole door region and part in front of the door Aisle region, adjusts in shooting angle ensures monitoring visual field the portrait that can obtain approximating front, and lighting condition is good, it is possible to To facial image clearly, face elemental area is more than 80*80 pixel.

The present invention judges that two foundations that entrance behavior occurs are:

Monitoring region occurs in that a suspect: realize a suspect's by the way of multiple alarm sensor combines

Detection.

A suspect's authentication failure: householder can complete authentication by two ways: video human face is known

Other mode and WIFI communication.Video human face recognition method: user cooperates with shooting on one's own initiative after entering region Machine completes recognition of face, by comparing with the face database of registration, completes the checking of identity.WIFI communication: take the photograph Camera and user mobile phone have been communicated authentication by WIFI.

Therefore experiment is divided into two parts: region enters detection and authentication.

Region enters test experience

The present invention first passes through audio analysis, vibration detection, and infrared detection carries out preliminary examination to the behavior entering region, connects down By the way of Video Motion Detection is checked, feasible region enters detection, in order to verify the actual performance of product, carries out 5 tests, test is 7*24 hour every time, compares, for single detection side with the performance of single-sensor simultaneously As long as triggering warning for formula, then it is assumed that being to there occurs that region enters behavior, test result is as follows:

It is experimentally confirmed no matter effectively combining compared to single alarm in detection essence of multiple alarm sensor Degree aspect, or all improve a lot in terms of reducing wrong report.

Authentication is tested

When intrusion behavior being detected, system open authentication, sets verify time-out time as 10 seconds, in 10 seconds such as Fruit does not confirm user identity, then it is assumed that be illegal invasion.Both verification modes that communicates recognition of face with WIFI respectively are entered Having gone test, test result is as follows:

Identity validation mode	Experiment number	Number of success	The frequency of failure	Errors number
					WIFI communicates	1000	991	9	0
Recognition of face	1000	947	53	17
					Both combine	1000	1000	2	0

Finding in an experiment, recognition of face is to human face posture, and the requirement of photoenvironment is the highest, if environment sidelight is tight Weight or personnel's attitude coordinate the most tacit meeting to cause None-identified or wrong identification, and join without user by the way of WIFI communicates Close, but connection may be caused unsuccessful by the interference of other WIFI equipment, by description of test, use WIFI communicate into Row identity validation is better than recognition of face and realizes identity validation, and both combinations can improve the accuracy rate of identity validation further.

Although reference be made herein to invention has been described for the multiple explanatory embodiment of the present invention, however, it is to be understood that Those skilled in the art can be designed that a lot of other amendments and embodiment, and these amendments and embodiment will fall in this Shen Within spirit disclosed in please and spirit.More specifically, in the range of disclosure, drawings and claims, can So that the building block of theme composite configuration and/or layout are carried out multiple modification and improvement.Except to building block and/or layout Outside the modification carried out and improvement, to those skilled in the art, other purposes also will be apparent from.

Claims (8)

1. an intelligent intrusion-proof system, described system includes memory module and anti-intrusion device；Described memory module is by having The mode of line or radio communication sets up communication connection with described anti-intrusion device, and the information gathering described anti-intrusion device is carried out Storage；

Described anti-intrusion device includes lighting module, video acquisition module, integrated information acquisition module, warning sending module, letter Number processing module, lighting control module；

It is characterized in that:

The video acquisition that described lighting module is video acquisition module provides lighting source；

Described video acquisition module starts video information around collection when lighting module is opened；

Described integrated information acquisition module Real-time Collection peripheral information, and it is supplied to described signal processing module in real time；

Described lighting control module determines the unlatching of lighting module, and wherein, described lighting control module is according to described signal processing Module determines described unlatching to the result of described peripheral information；

Described signal processing module differentiates intrusion behavior according to described video information and described peripheral information, thus decides whether Start warning sending module, send alarm signal；

Described resolution intrusion behavior includes the people close to anti-intrusion device is carried out identification, and described identification is to pass through people Face identification or other mode identification complete；

While carrying out described identification, also the described integrated information acquisition module when people is close to anti-intrusion device is gathered To described peripheral information be analyzed；

Described signal processing module is analyzed based on described identification with to described peripheral information, forms decision signal, point Distinguish whether intrusion behavior is set up；

Described resolution intrusion behavior also includes that the action in described video information carries out detection to be analyzed, including background modeling and fortune Dynamic detection, goal verification, target following, behavior analysis；

Described background modeling and motion detection, be that the mode using multiple image to be averaging obtains background mean value, by the side of background Difference is initialized as a constant, passes through formula

$f(x,y)=\left\{\begin{array}{cc}1& |f(x,y)-\mathrm{\&mu;}(x,y)|>N\&times;{\mathrm{\&sigma;}}^2(x,y)\\ 0& else\end{array}\right.$

By background pixels and prospect pixel separation, wherein 1 represents prospect pixel, and 0 represents background pixels, and (x y) represents in image f (x, y) gray value of place's pixel, (x y) represents the background mean value of correspondence, σ to μ²(x, y) represents the variance of corresponding background, and N is for adjusting Joint coefficient；

Described goal verification, is filtered according to the size set by target size wave filter and is unsatisfactory for the foreground zone of dimensional requirement Territory；Highlighted foreground area is filtered according to setting threshold value by highlighted target filter；Foreground area is done with corresponding background Gradient difference, is filtered motion continuity, motion significance and area change seriality respectively by time domain filtering；

Described target following, according to the location estimation target of present frame and target before in the position of next frame, passes through formula

$v_t^x=\frac{\underset{n=0}{\overset{N-2}{\&Sigma;}}(x_{t-n}-x_{t-n-1})}{N-1}$

$v_t^y=\frac{\underset{n=0}{\overset{N-2}{\&Sigma;}}(y_{t-n}-y_{t-n-1})}{N-1}$

To estimating that target velocity realizes target prodiction, whereinWithRepresent that t target is in x direction and y direction respectively Speed, N is time window, x_t-nAnd y_t-nRepresent abscissa and the vertical coordinate at rectangle frame center, the t-n moment target external world respectively, In like manner x_t-n-1And y_t-n-1Represent abscissa and the vertical coordinate at rectangle frame center, the t-n-1 moment target external world respectively；

Described behavior analysis, sets alarm region by user and corresponding rule realizes the judgement of specific behavior.

2. anti-intrusion system as claimed in claim 1, it is characterised in that:

Before carrying out described recognition of face, also include that face registration is to form face database；

In described recognition of face, use HAAR feature to combine adaboost algorithm and carry out Face datection and realize Face detection；

Use two-layer human eye localizer, obtain human eye by adaboost algorithm and position, utilize Face detection result to pass through image Eyes are corrected to level and realize face standardization by rotation；

Carry out feature extraction by two-dimensional Gabor filter, utilize the covariance distance between vector as matching measurement mode, lead to Cross nearest neighbor classification to realize facial image to be identified and mate with the facial image in described face database.

3. the system as described in any one of claim 1-2, it is characterised in that:

The voice messaging being analyzed described peripheral information including being collected audio collection module carries out detection analysis, will Carry from the numerous and jumbled background audio information as background signal as foreground signal differentiating the helpful voice messaging of invasion Take out.

4. system as claimed in claim 3, it is characterised in that:

Described voice messaging carries out detect analysis include:

Obtain the power spectral density of audio signal in described voice messaging, if the sample rate of audio signal X (t) is fs, by X T () sequentially passes through preemphasis, framing and windowing process, first remove average before signal processing, uses in classical spectrum estimate Period map method, uses fast Fourier transform FFT to realize, finally gives normalized power spectrum X^(t)(f_n), wherein n ∈ [1, N], N is sampling number, and t represents sampling instant；

Use sub belt energy ratio A_i ^(t)Describe power spectral density, utilize formulaCalculate sub belt energy Ratio, A_i ^(t)Being defined as the ratio of sub belt energy and gross energy, wherein U (i) and L (i) represents that the coboundary of subband i is with following Boundary, i represents the sequence number that frequency band divides subband according to logarithm；

Mixed Gaussian background modeling based on sub belt energy ratio and sound detection, mixed Gaussian background modeling assumes signal intensity Probability distribution can use K Gauss Distribution Fitting, and the probability distribution of audio signal is expressed as Each Gauss distribution η in model_i(x,μ_i,σ_i) all give a weights omega_i, wherein μ_iAnd σ_iRepresent the average of Gauss distribution respectively And standard deviation, multiple Gauss distribution obtain the probability distribution of signal by linear combination, using front B Gauss distribution as background mould Type, remaining Gauss distribution as foreground model,Wherein T is the error upper limit, and b is 1 to K Between positive integer；Current demand signal amplitude is mated with front B Gauss distribution, if divided with any one Gauss therein Cloth the match is successful then this signal be background signal, be otherwise foreground signal.

5. the system as described in claim 1-2,4 any one, it is characterised in that:

Described system carries out described identification, is analyzed described peripheral information, forms decision signal, differentiates intrusion behavior Whether establishment comprises the steps:

The first step, collects digital video signal by described video acquisition module, collects DAB by audio collection module Signal, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage signal, logical Cross vibration detection module collection extraneous vibration produce deformation or the information of stress transformation and be changed into voltage signal, broken by glass Broken detection module and door Magnetic testi module collection glass breaking and door magnetic change information are also changed into voltage signal；

Second step, described signal processing module receive heat release infrared detection alarm module, vibration detection module, glass break Broken detection module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis, complete identification, based on identification Identity carry out the generation of synthetic determination intrusion behavior, when there is intrusion behavior, send alarm signal；

3rd step, warning sending module receives alarm signal, and to outside alert by the way of wired or wireless.

6. system as claimed in claim 3, it is characterised in that:

Described system carries out described identification, is analyzed described peripheral information, forms decision signal, differentiates intrusion behavior Whether establishment comprises the steps:

The first step, collects digital video signal by described video acquisition module, collects numeral by described audio collection module Audio signal, collects external heat temperature information by heat release infrared detection alarm module and is changed into voltage letter Number, produce deformation or the information of stress transformation by vibration detection module collection extraneous vibration and be changed into voltage signal, passing through Glass breaking detection module and door Magnetic testi module collection glass breaking and door magnetic change information are also changed into voltage signal；

Second step, described signal processing module receive heat release infrared detection alarm module, vibration detection module, glass break Broken detection module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis, complete identification, based on identification Identity carry out the generation of synthetic determination intrusion behavior, when there is intrusion behavior, send alarm signal；

3rd step, warning sending module receives alarm signal, and to outside alert by the way of wired or wireless.

7. an intelligent access control system, described intelligent access control system include the intelligence as described in any one of claim 1-6 prevent into Invading system, described intelligent access control system utilizes described intelligent intrusion-proof system to complete described identification, to believe described surrounding Breath is analyzed, and identifies whether to visit for guest, and automatically the information that visiting guest arrives is sent to mobile or fixed terminal.

8. intelligent access control system as claimed in claim 7, it is characterised in that:

Described identifying whether comprises the steps: for guest's visiting

The first step, collects digital video signal by described video acquisition module, collects DAB by audio collection module Signal, is collected external heat temperature information by heat release infrared detection alarm module and is changed into voltage signal, logical Cross vibration detection module collection extraneous vibration produce deformation or the information of stress transformation and be changed into voltage signal, broken by glass Broken detection module and door Magnetic testi module collection glass breaking and door magnetic change information are also changed into voltage signal；

Second step, signal processing module receives heat release infrared detection alarm module, vibration detection module, glass breaking inspection Survey module and the voltage signal of door Magnetic testi module, and combine audio-video intelligent analysis and determine whether that guest visits；

3rd step, the generation of behavior of visiting if there is guest, wirelessly by visitor at video outdoors or photo Information is sent automatically to be in the movement of indoor owner or fixed terminal, and activated terminals provides ring, vibrations or voice message.