CN106251544A - A kind of intrusion alarm method based on Android intelligent and alarm device - Google Patents

Abstract

The invention discloses a kind of intrusion alarm method based on Android intelligent and alarm device, device is: web camera, Moving target detection module, power module, DSP, alarm module, gsm module, Ethernet and Android intelligent client, and alarm method is: be transferred to Android intelligent client after the Video Image Processing in monitoring region；Video image in monitoring region is moved target detection, it is judged that whether monitoring region has object to invade；Characteristics of human body is identified, it may be judged whether need to report to the police；Detection is found to have human body and invades then device startup alarm module and gsm module, and to invasion, personnel frighten；Send note simultaneously and mail reminds user in time to Android intelligent client；Video image in monitoring region is carried out recording preservation and reporting to the police by user by Android intelligent client；This apparatus structure is simple, low cost, and method is simple to operate, warning accuracy rate high, is suitable to popularize.

Description

A kind of intrusion alarm method based on Android intelligent and alarm device

Technical field

The invention belongs to smart mobile phone technical field of safety protection, relate to a kind of invasion based on Android intelligent Alarm method and alarm device.

Background technology

In recent years, along with the development of mobile communication technology, mobile communication network has had coverage rate widely, intelligence hands Machine have also been obtained universal.And along with living standard constantly improves, the security protection consciousness of people is stronger, wherein family's peace Anti-concept is increasingly accepted by people.But traditional video monitoring system realizes long-range monitoring and mainly uses wired side Method, including laying the method for special line, utilizing the method etc. of telephone wire, playback terminal is PC.Although these methods can be real Existing monitoring remote video, but there are some problems, special line deployment cost is high, and supervisor is limited to position, and telephone wire passes Defeated speed is slow, is almost difficult to real-time video monitoring, and video monitoring system based on PC, its video compress and solution Pressure uses video frequency collection card based on PC so that the collection of video signal, compress with communicate complex, stability and reliably Property is the highest, and hardware investment cost is the biggest.And PC operation complexity, need personal management, also there is no the function of automatic alarm, property Valency ratio is extremely low, is not suitable for household safety-protection.It is thus desirable to relate to a kind of simple in construction, cost is relatively low, transmission is convenient, carry report The monitoring and alarming system of the functions such as police.

It addition, in terms of mobile target body detection algorithm, most monitoring technology only simple have employed currently Image and background image do difference process, consider less to the situation of background change, and fail to considering that in family, house pet is to regarding The interference in screen monitoring region so that system has the highest rate of false alarm.

Summary of the invention

The technical problem to be solved is to overcome conventional art to have that portability is poor, cost performance is low, is unfavorable for The problem that intrusion alarm is extensively applied and existing mobile target body detection method is not enough, it is provided that a kind of based on Android The intrusion alarm method of smart mobile phone and alarm device, the method and device portability is strong, cost performance is high, can be under indoor environment Effectively detect invasion human body target, and can well process owing to background light drastic change and house pet are to monitoring region detection band The interference come.

For solving above-mentioned technical problem, the present invention adopts the following technical scheme that realization:

A kind of intrusion alarm method based on Android intelligent, it is characterised in that step is as follows:

Step one, web camera are by the video image acquisition in monitoring region to DSP and right Video image in the monitoring region being collected carries out video image compression process, and the video image after compression is passed through Ethernet It is transferred to Android intelligent client；

In step 2, DSP, step one collection is monitored in region by Moving target detection module Video image moves target detection, and the change using Three image difference to calculate the video image in monitoring region judges prison Whether control region has object to invade；

Moving target detection module in step 3, DSP, according to monitoring regarding in region in step 2 Frequently the change of image, by being identified characteristics of human body, it may be judged whether needs to report to the police；If be detected that have human body to invade, then Alarm module is started by DSP；Without detecting that human body is invaded, then DSP Alarm module will not be started, and continuation is carried out video acquisition to monitoring region by web camera；

If step 4 is found to have human body by the detection of Moving target detection module and invades, then device starts alarm module And gsm module；Invasion personnel will be frightened by alarm module by audible-visual annunciator and voice broadcast module；Start warning mould While block, transmission note and mail are reminded user to Android intelligent client by gsm module in time；User receives After information and mail reminder, by the monitoring region that monitoring client web camera is gathered by Android intelligent client Video image information carry out record preserve and report to the police.

Further technical scheme includes:

Described in step one to be collected monitoring region in video image carry out video image compression process, select Video image after compression, as the compression standard of video data, is transferred to Android intelligence by Ethernet by MPEG-4 standard RTP/RTCP protocol stack can be used to realize by cell-phone customer terminal.

The detailed process whether judging described in step 2 monitors region and have object to invade is:

(1) use Three image difference that the mobile target entering monitoring region is monitored, take three two field pictures continuously, by phase Two adjacent two field pictures are the poorest, then by do the image of after the recovery respectively through gaussian filtering, maximum between-cluster variance threshold method, patrol Volume computing and morphology operations obtain the bianry image of mobile target neighbor frame difference；

If A_k-1、A_k、A_k+1It is respectively this monitoring system video image in the monitoring region collected the obtained The continuous picture of k-1, k, k+1 frame, it is thus achieved that adjacent image do difference, then:

B_k=| A_k-A_k-1|

B_k+1=| A_k+1-A_k|

For reducing noise to image B derived above_kWith image B_k+1Impact, by respectively to image B_kWith image B_k+1Adopt Gaussian filtering process is carried out by 3*3 template, i.e. by by each pixel in a 3*3 template scanogram, true by template In fixed neighborhood, the weighted average gray value of pixel goes the value of alternate template central pixel point, thus the noise that is eliminated is smooth Image C_kAnd C_k+1, the neighborhood definition of the most each pixel is a circle centered by the coordinate figure of described each pixel The set of point on inside or border；

If threshold value T_kBy image C_kBeing divided into prospect i.e. target, and background two class, wherein prospect is image C_kMiddle pixel value is big In equal to threshold value T_kPixel, background is image C_kMiddle pixel value is less than threshold value T_kPixel, even image C_kAt coordinate (x, y) the pixel C at place_k(K_x,K_y)≥T_k, then C_k(K_x,K_y) it is prospect, if image C_kAt coordinate (x, y) the pixel C at place_k (K_x,K_y)<T_k, then C_k(K_x,K_y) it is background, and threshold value T_kMake the variance between this two class of foreground and background maximum；If threshold value T_k+1By image C_k+1Being divided into prospect i.e. target, and background two class, wherein prospect is image C_k+1Middle pixel value is more than or equal to threshold value T_k+1Pixel, background is image C_k+1Middle pixel value is less than threshold value T_kPixel, even image C_k+1At coordinate (x, y) place Pixel C_k+1(K+1_x,K+1_y)≥T_k+1, then C_k+1(K+1_x,K+1_y) it is prospect, if image C_k+1At coordinate (x, y) picture at place Vegetarian refreshments C_k+1(K+1_x,K+1_y)<T_k+1, then C_k+1(K+1_x,K+1_y) it is background, and threshold value T_k+1Make this two class of foreground and background it Between variance maximum；Then T_k、T_k+1It is respectively image C_k、C_k+1Optimal threshold, the calculating process of optimal threshold is as follows:

If image C_IFor the I-frame video image in the monitoring region that web camera collects, wherein C_IComprise L ash Degree level 0,1 ..., L-1}, gray value be the pixel number of i be n_i, then total pixel isGray scale is that the pixel of i goes out Existing probability isEntire image is divided into target M by thresholding t₁With background M₂Two classes, the then meter of the inter-class variance δ of this two class Calculation formula is:

${\mathrm{\&delta;}}_I^2\left(t\right)=p_0*p_1{(u1-u2)}^2$

In formula,WithIt is respectively target M₁With background M₂The probability that two classes occur； WithIt is respectively target M₁With background M₂The gray average of two classes, the optimal threshold of segmentation is that inter-class variance calculates public affairs Formula takes thresholding t value during maximum, is designated as T_I:

$T_I=Arg\underset{0\&le;t\&le;L-1}{Max}{\mathrm{\&delta;}}_I^2\left(t\right)$

Image C is tried to achieve by above-mentioned optimal threshold computational methods_k、C_k+1Optimal threshold T_k、T_k+1, then use optimal threshold Value T_k、T_k+1Respectively to image C_k、C_k+1Binaryzation obtains prospect bianry image D_k、D_k+1, it is maximum between-cluster variance threshold method, uses Formula is expressed as follows:

D_k=threshold (| C_k|)

D_k+1=threshold (| C_k+1|)

(2) the prospect bianry image D that will obtain_k、D_k+1Carry out logic "and" operation, wherein image D_kIn prospect and image D_k+1In prospect carry out AND operation after the foreground image that obtains be image A_kThe shape profile of middle moving target, formula table Reach for:

E_k=D_k And D_k+1

Due to effect of noise, image D_k、D_k+1Image E obtained after carrying out logical operations_kBorder is the most not perfectly flat Sliding, foreground area has some noise holes, and is studded with some little noise objects on background area, is opened by morphologic Computing, i.e. after excessive erosion reflation thus remove the noise hole in bianry image, obtain the image F after de-noising_k, wherein Expansion is each pixel using structural element G scanogram, and the bianry image covered with it does OR operation；Corrosion is to use Each pixel of structural element G scanogram, and the bianry image covered with it does AND-operation；Fortune is opened in the morphology of image Calculation formula is as follows:

(3) by using connection eight neighborhood rule, it is thus achieved that image F_kIn the minimum abscissa value in largest connected region, maximum Abscissa value, minimum ordinate value and maximum ordinate value, thus obtain the mobile target image of red rectangle collimation mark noteReal Existing step is as follows:

1) in two dimensional image, it is assumed that have the pixel that z is adjacent around target pixel points, wherein z≤8, if this pixel In z pixel of gray scale and this, the gray value of some pixel Q is equal, then claim this pixel and Q to have connectedness；Connection eight is adjacent Territory rule is to choose object pixel all of adjacent picture in two-dimensional space, be upper and lower, left and right, upper left, upper right, lower-left, Eight, bottom right neighbor pixel；By using connection eight neighborhood rule to image F_kIn each pixel label is set, To image F_kThe connected component labeling matrix of pixel, label initial value is all set to 0；If image F_kPixel (K_x,K_y), its connection Region labeling is label (K_x,K_y), wherein image F_kPixel (K_x,K_y) eight neighborhood coordinate be (K_x-1,K_y)、(K_x+1,K_y)、 (K_x,K_y-1)、(K_x,K_y+1)、(K_x-1,K_y-1)、(K_x-1,K_y+1)、(K_x+1,K_y-1) and (K_x+1,K_y+1)；

2) first from image F_kThe upper left corner start progressive scan, be i.e. from storage image F_k0 row 0 of the matrix of pixel arranges Starting progressive scan, from top to bottom, every a line scans from left to right；When scanning image F_kPixel (K_x,K_y) time, sweep Retouch and complete (K_x,K_y) the adjacent pixel in the top of pixel and the left side, obtain label (K_x-1,K_y)、label (K_x-1,K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1) value；

If image F_kCoordinate is (K_x,K_y) place's pixel is foreground pixel point, and the value of its pixel is adjacent pixel It is worth identical, then it is assumed that these pixels are connections, and wherein foreground pixel point is image F_kThe pixel of middle composition prospect；If picture Vegetarian refreshments (K_x,K_y) it is adjacent pixel (K_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) and (K_x,K_y-1) connection, then label (K_x,K_y) value equal to label (K_x-1,K_y)、label(K_x-1,K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1In the middle of) The label that value is minimum, and it is vertical to update the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value and maximum Coordinate figure: minimum abscissa value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberIndulge with maximum Coordinate figure maxy_number, wherein number is the label of this connected region；

If image F_kIn be numbered the connected region of number, if the scanning of this connected region is to pixel (K_x,K_y) time, The minimum abscissa value of this connected region recorded, maximum abscissa value, minimum ordinate value and maximum ordinate value: minimum Abscissa value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberWith maximum ordinate value maxy_number, and pixel (K_x,K_y) be adjacent pixel connection, then update the minimum abscissa value of this connected region, maximum Abscissa value, minimum ordinate value and maximum ordinate value, if now K_x≥maxx_number, then maxx_number=K_xAnd minx_numberValue keeps constant, if minx_number<K_x<maxx_number, then minx_numberAnd maxx_numberValue all keeps constant；If K_x≤minx_number, then minx_number=K_xAnd maxx_numberValue keeps constant；If now K_y≥maxy_number, then maxy_number =K_yAnd miny_numberValue keeps constant, if miny_number<K_y<maxy_number, then miny_numberAnd maxy_numberValue all keeps not Become；If K_y≤miny_number, then miny_number=K_yAnd maxy_numberValue keeps constant；

If image F_kPixel (K_x,K_y) and the pixel (K of its top_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) Pixel (K with the left side_x,K_y-1) do not connect, and image F_kAt pixel (K_x,K_y) place is foreground pixel point, then will increase Add a new connected region；I.e. enumerator number '=number+1 is as pixel (K_x,K_y) connection label, i.e. label(K_x,K_y)=number ', and record the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value With maximum ordinate value: minimum abscissa value minx_number′, maximum abscissa value maxx_number′, minimum ordinate value miny_number′With maximum ordinate value maxy_number′Even, minx_number′=K_x、maxx_number′=K_x、miny_number'=K_y And maxy_number'=K_y；If image F_kIt is (K at coordinate_x,K_y) place's pixel is background pixel point, then continue to scan on the next one Point, wherein background pixel point is image F_kThe pixel of middle composition background；

3) according to process 2) described in scanogram F_kIn all pixels, if scanned, then perform process 4)；

4) from left to right, from top to bottom traversing graph as F_kIn all different connected regions, according to record all differences The minimum abscissa value of connected region, maximum abscissa value, minimum ordinate value and maximum ordinate value: minx_number、 maxx_number、miny_numberAnd maxy_numberCalculate the height H of each connected region_numberWith width W_number, it may be assumed that

H_number=maxx_number-minx_number

W_number=maxy_number-miny_number

Will height H_numberWith width W_numberProduct as the similar area i.e. S of each connected region_number=H_number* W_number, by the area S of relatively each connected region_number, by area maximum MAX (S_number) connected region be designated as maximum fortune Dynamic model block, preserves the minimum abscissa value of largest motion module, maximum abscissa value, minimum ordinate value and maximum ordinate Value: MAXS (minx_number)、MAXS(maxx_number)、MAXS(miny_number) and MAXS (maxy_number)；Due to image F_kIn Largest motion module be by above-mentioned steps process image A_kObtained by, therefore image A_kIn largest motion module and figure As F_kIn largest motion module coordinate information identical, with minimum and maximum abscissa column, the minimum of largest motion module It is expert at maximum ordinate and is formed a height of MAXS (maxx_number)-MAXS(minx_number), a width of MAXS (maxy_number)- MAXS(miny_number) rectangle frame, i.e. four coordinate points of rectangle frame are (MAXS (minx_number),MAXS(miny_number)), (MAXS(minx_number),MAXS(maxy_number)), (MAXS (maxx_number),MAXS(miny_number)), (MAXS (maxx_number),MAXS(maxy_number)), the rectangle frame that order obtains is shown in red；With red rectangle frame by image A_kIn Largest motion module frame is selected, and then obtains the mobile target image with red rectangle collimation mark noteThen judge that monitoring region has Object is invaded.

Described in step 3 by characteristics of human body's recognition methods judge whether need report to the police detailed process be:

If monitoring region has object to invade, set different threshold decision conditions, it is determined that mobile target imageIn red Whether color rectangle frame meets characteristics of human body, if met, is considered human body, needs to report to the police；If not meeting characteristics of human body, then Can judge whether to report to the police according to initial setting of user；If the mobile target image with red rectangle collimation mark noteIn, mark A length of l of the red rectangle frame of note_k, a width of b_k, work as b_kMore than l_k, then it is assumed that intrusion object is upright；To people from uprightly Moving object in distinguish, characteristics of human body can be passed through, characteristics of human body's recognition methods of the present invention is described as follows:

If mobile target imageIn, a length of l of the red rectangle frame of labelling_k, a width of b_k, then l_k/b_kCan be approximated to be into Invade the length-width ratio of object, l_k*b_kCan be approximated to be the area of intrusion object；If l_k/b_kLess than θ_kOr more than β_k, or l_k*b_k Less than α_kOr more than ε_kTime, it not the most human body, and if only if θ_k≤l_k/b_k≤β_kAnd α_k≤l_k*b_k≤ε_kTime, it is judged that for human body Monitoring instruction region；Wherein, threshold θ_k、β_k、α_kAnd ε_kValue by IP Camera with monitoring region distance determine, if net The distance in network video camera and monitoring region is when H rice distance, and the monitoring range of web camera is long L_Ak, wide K_Ak；Therefore l_k/b_k Lower proportion ratio threshold θ_kIt is set asUpper limit proportion threshold value β_kIt is set as 7；l_k*b_kLower limit area threshold α_kIt is set as current Monitoring image areaUpper limit area threshold ε_kIt is set as current monitor image area

If having human body to invade by detection, then start alarm module by DSP, without inspection Measure human body invasion, then DSP will not start alarm module, and monitoring region is entered by web camera by continuation Row video acquisition.

A kind of intrusion alarm device based on Android intelligent, it is characterised in that including: web camera, movement Module of target detection, power module, DSP, alarm module, gsm module, Ethernet and Android intelligence Cell-phone customer terminal, web camera video acquisition output port is connected with DSP video acquisition input port, Power module output port of power source is connected with web camera power input port, alarm module, gsm module, power module, shifting Moving-target detection module electrically connects with DSP respectively, gsm module and Android intelligent client without Line communication connects, and the RJ-45 Ethernet interface of web camera is connected with Ethernet by netting twine, Android intelligent visitor The video information that family end can be gathered by ethernet access web camera.

Described alarm module includes the audible-visual annunciator being made up of flash lamp and speaker and can record according to user's request Enter the voice broadcast module of different phonetic；

Compared with prior art the invention has the beneficial effects as follows:

In the present invention, Android intelligent client is responsible for receiving, playing and record monitoring remote video data, compares In traditional long-distance monitoring method, it is not necessary to install independent PC or hardware decoder material, can be by Android intelligent client End unites two into one with intrusion alarm monitoring system, and smart mobile phone is more and more universal, makes average family spend 650 to 1500 yuan i.e. Can realize monitoring instruction, therefore the present invention has feature cheap and easy for installation, is mainly used in family or office etc. Report to the police in place.

Alarm method of the present invention be utilize Android intelligent client to be monitored, monitoring of a recorded programme region Video, and by audible-visual annunciator and voice broadcast module deterrence invader, this illegal invasion card contributing to preserving invader According to.

Android intelligent client is handy to carry about, and therefore user can receive from invasion anywhere or anytime The warning message of monitoring system；When warning system is started working, if having object to move in the region of network shooting machine monitoring Dynamic and system judges it is human body, then can send information and mail by designated telephone and mailbox, thus reaches to report to the police and connect in real time The purpose received, this intrusion alarm method and device has simple to operate, real-time and that warning accuracy rate is high feature.

Accompanying drawing explanation

The present invention is further illustrated below in conjunction with the accompanying drawings:

Fig. 1 is the structural schematic block diagram of a kind of intrusion alarm device based on Android intelligent of the present invention.

Fig. 2 is the flow chart of a kind of intrusion alarm method based on Android intelligent of the present invention.

Fig. 3 is the alarm module signal of a kind of intrusion alarm device based on Android intelligent of the present invention Block diagram.

In figure: 1, user, 2, Android intelligent client, 3, GSM, 4, Ethernet, 5, alarm module, 501, sound Light crossing-signal, 502, voice broadcast module, 6, gsm module, 7, power module, 8, DSP, 9, mobile target Detection module, 10, web camera, 11, invasion personnel.

Detailed description of the invention

In order to make the purpose of the present invention, summary of the invention and advantage clearer, below in conjunction with the accompanying drawings the present invention is made Detailed description.Understand refering to Fig. 1, use intrusion alarm method and device based on Android intelligent, including including net Network video camera 10, Moving target detection module 9, power module 7, DSP 8, alarm module 5, gsm module 6, Ethernet 4 and Android intelligent client 2, web camera 10 video acquisition output port and DSP Digital Signal Processing Device 8 video acquisition input port connects, and power module 7 output port of power source is connected with web camera 10 power input port, Alarm module 5, gsm module 6, power module 7, Moving target detection module 9 electrically connect with DSP 8 respectively, Gsm module 6 is connected with Android intelligent client 2 wireless telecommunications, and the RJ-45 Ethernet interface of web camera 10 leads to Crossing netting twine to be connected with Ethernet 4, Android intelligent client 2 can access what web camera 10 gathered by Ethernet 4 Video information.

Described web camera 10 is by the video image acquisition in monitoring region to DSP 8 Moving target detection module 9, and the video image in the monitoring region being collected is carried out video image compression process, will compression After video image be transferred to Android intelligent client 2 by Ethernet 4；In described DSP 8 Moving target detection module 9 through to monitoring region in vedio data be identified, judge and analyze after, pass through DSP 8 controls gsm module 6 and sends warning message and mail to Android phone client 2, and triggers report Alert module 5 carries out onsite alarming, and the video image in monitoring region is passed by web camera 10 by RTP/RTCP agreement simultaneously Being passed to Ethernet 4, user 1, after receiving warning message, opens Android intelligent client 2 and plays out.Described Alarm module 5 is made up of audible-visual annunciator 501 and voice broadcast module 502, and wherein audible-visual annunciator includes flash lamp and raises one's voice Device；Described Android intelligent client (2) is carried with by user, and user (1) utilizes Ethernet (4) to connect invasion Test side web camera (10), it is achieved the function that monitoring in real time, recording evidence and assistance are reported to the police.

In the implementation case, intrusion detection end web camera model is EasyN FS-613A-M136, and external 5V is straight Stream power supply, is connected with Ethernet by RJ-45 Ethernet interface and realizes data transmission；Gsm module 7 model is Huawei GTM900-B, is inside provided with a SIM, and SIM is for realizing the communication of gsm module 7 and Android intelligent client 2 even Connect；Android intelligent model is Huawei's C199s Android mobile phone, and it is configured to 5.5 inch screen, 2G internal memory, 16G hard disk, The version selecting Android intelligent operating system is Android 4.4, and version number is V100R001C92B263；Voice is broadcast The model that report module uses Hua Bang ISD company to produce is ISD1700 voice broadcast module, running voltage 2.4V-5.5V；Numeral letter Number processor uses C667X DSP.

The video data that Android intelligent client is transmitted through Ethernet by Network Capture web camera, from And reach the purpose monitored, can record non-by touching " starting to record " button in Android intelligent client simultaneously Method invasion video information.

Refering to Fig. 2, Android intelligent client is mainly responsible for receiving, playing and monitoring of a recorded programme end web camera The video of shooting, and receive alarming short message and mail.The tool of intrusion alarm method and device based on Android intelligent It is as follows that body realizes step:

Starting stage, the initialization of equipment to be carried out, user inputs web camera in Android intelligent client IP address；At initial phase, user opens Android intelligent client, enters Android intelligent client Interface, uses the IP address needing to input this IP Camera machine at IP address field first, and the IP address of web camera can quilt Android intelligent client automatically saves, and reuses and Button Login can be selected to be directly entered Android intelligent Client.

Step one, when equipment is by after Initialize installation, and user opens intrusion alarm device, and web camera is by monitored space Video image acquisition in territory is in DSP, and carries out the video image in the monitoring region being collected Video image compression processes, and by Ethernet, the video image after compression is transferred to Android intelligent client；

The monitoring image that in step 2, DSP, step one is gathered by Moving target detection module is carried out Moving target detection, the change using Three image difference to calculate monitoring region judges to monitor the tool whether region has object to invade Body process is:

(1) use Three image difference that the mobile target entering monitoring region is monitored, take three two field pictures continuously, by phase Two adjacent two field pictures are the poorest, then by do the image of after the recovery respectively through gaussian filtering, maximum between-cluster variance threshold method, patrol Volume computing and morphology operations obtain the bianry image of mobile target neighbor frame difference；

If A_k-1、A_k、A_k+1It is respectively this monitoring system video image in the monitoring region collected the obtained The continuous picture of k-1, k, k+1 frame, it is thus achieved that adjacent image do difference, then:

B_k=| A_k-A_k-1|

B_k+1=| A_k+1-A_k|

For reducing noise to image B derived above_kWith image B_k+1Impact, by respectively to image B_kWith image B_k+1Adopt Gaussian filtering process is carried out by 3*3 template, i.e. by by each pixel in a 3*3 template scanogram, true by template In fixed neighborhood, the weighted average gray value of pixel goes the value of alternate template central pixel point, thus the noise that is eliminated is smooth Image C_kAnd C_k+1, the neighborhood definition of the most each pixel is a circle centered by the coordinate figure of described each pixel The set of point on inside or border；

If threshold value T_kBy image C_kBeing divided into prospect i.e. target, and background two class, wherein prospect is image C_kMiddle pixel value is big In equal to threshold value T_kPixel, background is image C_kMiddle pixel value is less than threshold value T_kPixel, even image C_kAt coordinate (x, y) the pixel C at place_k(K_x,K_y)≥T_k, then C_k(K_x,K_y) it is prospect, if image C_kAt coordinate (x, y) the pixel C at place_k (K_x,K_y)<T_k, then C_k(K_x,K_y) it is background, and threshold value T_kMake the variance between this two class of foreground and background maximum；If threshold value T_k+1By image C_k+1Being divided into prospect i.e. target, and background two class, wherein prospect is image C_k+1Middle pixel value is more than or equal to threshold value T_k+1Pixel, background is image C_k+1Middle pixel value is less than threshold value T_kPixel, even image C_k+1At coordinate (x, y) place Pixel C_k+1(K+1_x,K+1_y)≥T_k+1, then C_k+1(K+1_x,K+1_y) it is prospect, if image C_k+1At coordinate (x, y) picture at place Vegetarian refreshments C_k+1(K+1_x,K+1_y)<T_k+1, then C_k+1(K+1_x,K+1_y) it is background, and threshold value T_k+1Make this two class of foreground and background it Between variance maximum；Then T_k、T_k+1It is respectively image C_k、C_k+1Optimal threshold, the calculating process of optimal threshold is as follows:

If image C_IFor the I-frame video image in the monitoring region that web camera collects, wherein C_IComprise L ash Degree level 0,1 ..., L-1}, gray value be the pixel number of i be n_i, then total pixel isGray scale is that the pixel of i goes out Existing probability isEntire image is divided into target M by thresholding t₁With background M₂Two classes, the then meter of the inter-class variance δ of this two class Calculation formula is:

${\mathrm{\&delta;}}_I^2\left(t\right)=p_0*p_1{(u1-u2)}^2$

In formula,WithIt is respectively target M₁With background M₂The probability that two classes occur； WithIt is respectively target M₁With background M₂The gray average of two classes, the optimal threshold of segmentation is that inter-class variance calculates public affairs Formula takes thresholding t value during maximum, is designated as T_I:

$T_I=Arg\underset{0\&le;t\&le;L-1}{Max}{\mathrm{\&delta;}}_I^2\left(t\right)$

Image C is tried to achieve by above-mentioned optimal threshold computational methods_k、C_k+1Optimal threshold T_k、T_k+1, then use optimal threshold Value T_k、T_k+1Respectively to image C_k、C_k+1Binaryzation obtains prospect bianry image D_k、D_k+1, it is maximum between-cluster variance threshold method, uses Formula is expressed as follows:

D_k=threshold (| C_k|)

D_k+1=threshold (| C_k+1|)

(2) the prospect bianry image D that will obtain_k、D_k+1Carry out logic "and" operation, wherein image D_kIn prospect and image D_k+1In prospect carry out AND operation after obtain the image of prospect and be image A_kThe shape profile of middle moving target, formula table Reach for:

E_k=D_k And D_k+1

Due to effect of noise, image D_k、D_k+1Image E obtained after carrying out logical operations_kBorder is the most not perfectly flat Sliding, foreground area has some noise holes, and is studded with some little noise objects on background area, is opened by morphologic Computing, i.e. after excessive erosion reflation thus remove the noise hole in bianry image, obtain the image F after de-noising_k, wherein Expansion is each pixel using structural element G scanogram, and the bianry image covered with it does OR operation；Corrosion is to use Each pixel of structural element G scanogram, and the bianry image covered with it does AND-operation；Fortune is opened in the morphology of image Calculation formula is as follows:

(3) by using connection eight neighborhood rule, it is thus achieved that image F_kIn the minimum abscissa value in largest connected region, maximum Abscissa value, minimum ordinate value and maximum ordinate value, thus obtain the mobile target image of red rectangle collimation mark noteReal Existing step is as follows:

1) in two dimensional image, it is assumed that have the pixel that z is adjacent around target pixel points, wherein z≤8, if this pixel In z pixel of gray scale and this, the gray value of some pixel Q is equal, then claim this pixel and Q to have connectedness；Connection eight is adjacent Territory rule is to choose object pixel all of adjacent picture in two-dimensional space, be upper and lower, left and right, upper left, upper right, lower-left, Eight, bottom right neighbor pixel；By using connection eight neighborhood rule to image F_kIn each pixel label is set, To image F_kThe connected component labeling matrix of pixel, label initial value is all set to 0；If image F_kPixel (K_x,K_y), its connection Region labeling is label (K_x,K_y), wherein image F_kPixel (K_x,K_y) eight neighborhood coordinate be (K_x-1,K_y)、(K_x+1,K_y)、 (K_x,K_y-1)、(K_x,K_y+1)、(K_x-1,K_y-1)、(K_x-1,K_y+1)、(K_x+1,K_y-1) and (K_x+1,K_y+1)；

2) first from image F_kThe upper left corner start progressive scan, be i.e. from storage image F_k0 row 0 of the matrix of pixel arranges Starting progressive scan, from top to bottom, every a line scans from left to right；When scanning image F_kPixel (K_x,K_y) time, sweep Retouch and complete (K_x,K_y) the adjacent pixel in the top of pixel and the left side, obtain label (K_x-1,K_y)、label (K_x-1,K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1) value；

If image F_kCoordinate is (K_x,K_y) place's pixel is foreground pixel point, and the value of its pixel is adjacent pixel It is worth identical, then it is assumed that these pixels are connections, and wherein foreground pixel point is image F_kThe pixel of middle composition prospect；If picture Vegetarian refreshments (K_x,K_y) it is adjacent pixel (K_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) and (K_x,K_y-1) connection, then label (K_x,K_y) value equal to label (K_x-1,K_y)、label(K_x-1,K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1In the middle of) The label that value is minimum, and it is vertical to update the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value and maximum Coordinate figure: minimum abscissa value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberIndulge with maximum Coordinate figure maxy_number, wherein number is the label of this connected region；

If image F_kIn be numbered the connected region of number, if the scanning of this connected region is to pixel (K_x,K_y) time, The minimum abscissa value of this connected region recorded, maximum abscissa value, minimum ordinate value and maximum ordinate value: minimum Abscissa value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberWith maximum ordinate value maxy_number, and pixel (K_x,K_y) be adjacent pixel connection, then update the minimum abscissa value of this connected region, maximum Abscissa value, minimum ordinate value and maximum ordinate value, if now K_x≥maxx_number, then maxx_number=K_xAnd minx_numberValue keeps constant, if minx_number<K_x<maxx_number, then minx_numberAnd maxx_numberValue all keeps constant；If K_x≤minx_number, then minx_number=K_xAnd maxx_numberValue keeps constant；If now K_y≥maxy_number, then maxy_number =K_yAnd miny_numberValue keeps constant, if miny_number<K_y<maxy_number, then miny_numberAnd maxy_numberValue all keeps not Become；If K_y≤miny_number, then miny_number=K_yAnd maxy_numberValue keeps constant；

If image F_kPixel (K_x,K_y) and the pixel (K of its top_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) Pixel (K with the left side_x,K_y-1) do not connect, and image F_kAt pixel (K_x,K_y) place is foreground pixel point, then will increase Add a new connected region；I.e. enumerator number '=number+1 is as pixel (K_x,K_y) connection label, i.e. label(K_x,K_y)=number ', and record the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value With maximum ordinate value: minimum abscissa value minx_number′, maximum abscissa value maxx_number′, minimum ordinate value miny_number′With maximum ordinate value maxy_number′Even, minx_number′=K_x、maxx_number′=K_x、miny_number'=K_y And maxy_number′=K_y；If image F_kIt is (K at coordinate_x,K_y) place's pixel is background pixel point, then continue to scan on the next one Point, wherein background pixel point is image F_kThe pixel of middle composition background；

3) according to process 2) described in scanogram F_kIn all pixels, if scanned, then perform process 4)；

4) from left to right, from top to bottom traversing graph as F_kIn all different connected regions, according to record all differences The minimum abscissa value of connected region, maximum abscissa value, minimum ordinate value and maximum ordinate value: minx_number、 maxx_number、miny_numberAnd maxy_numberCalculate the height H of each connected region_numberWith width W_number, it may be assumed that

H_number=maxx_number-minx_number

W_number=maxy_number-miny_number

Will height H_numberWith width W_numberProduct as the similar area i.e. S of each connected region_number=H_number* W_number, by the area S of relatively each connected region_number, by area maximum MAX (S_number) connected region be designated as maximum fortune Dynamic model block, preserves the minimum abscissa value of largest motion module, maximum abscissa value, minimum ordinate value and maximum ordinate Value: MAXS (minx_number)、MAXS(maxx_number)、MAXS(miny_number) and MAXS (maxy_number)；Due to image F_kIn Largest motion module be by above-mentioned steps process image A_kObtained by, therefore image A_kIn largest motion module and figure As F_kIn largest motion module coordinate information identical, with minimum and maximum abscissa column, the minimum of largest motion module It is expert at maximum ordinate and is formed a height of MAXS (maxx_number)-MAXS(minx_number), a width of MAXS (maxy_number)- MAXS(miny_number) rectangle frame, i.e. four coordinate points of rectangle frame are (MAXS (minx_number),MAXS(miny_number)), (MAXS(minx_number),MAXS(maxy_number)), (MAXS (maxx_number),MAXS(miny_number)), (MAXS (maxx_number),MAXS(maxy_number)), the rectangle frame that order obtains is shown in red；With red rectangle frame by image A_kIn Largest motion module frame is selected, and then obtains the mobile target image with red rectangle collimation mark noteThen judge that monitoring region has Object is invaded.

Moving target detection module in step 3, DSP, according to monitoring regarding in region in step 2 Frequently the change of image, by being identified characteristics of human body, it may be judged whether needs to report to the police；If be detected that have human body to invade, then Alarm module, without detecting that human body is invaded, then DSP is started by DSP Will not start alarm module, continuation is carried out video acquisition to monitoring region by web camera, is wherein identified by characteristics of human body Method judges whether that the detailed process needing to report to the police is:

If monitoring region has object to invade, set different threshold decision conditions, it is determined that mobile target imageIn red Whether color rectangle frame meets characteristics of human body, if met, is considered human body, needs to report to the police；If not meeting characteristics of human body, then Can judge whether to report to the police according to initial setting of user；If the mobile target image with red rectangle collimation mark noteIn, mark A length of l of the red rectangle frame of note_k, a width of b_k, work as b_kMore than l_k, then it is assumed that intrusion object is upright；To people from uprightly Moving object in distinguish, characteristics of human body can be passed through, characteristics of human body's recognition methods of the present invention is described as follows:

If mobile target imageIn, a length of l of the red rectangle frame of labelling_k, a width of b_k, then l_k/b_kCan be approximated to be into Invade the length-width ratio of object, l_k*b_kCan be approximated to be the area of intrusion object；If l_k/b_kLess than θ_kOr more than β_k, or l_k*b_k Less than α_kOr more than ε_kTime, it not the most human body, and if only if θ_k≤l_k/b_k≤β_kAnd α_k≤l_k*b_k≤ε_kTime, it is judged that for human body Monitoring instruction region；Wherein, the height of normal human is between 0.8-2 rice, and width is 0.3-1 rice, does not temporarily consider people here The situation of the abnormal invasion of body；Threshold θ_k、β_k、α_kAnd ε_kValue by IP Camera with monitoring region distance determine, if net The distance in network video camera and monitoring region is when M rice distance, and the monitoring range of web camera is high L_Ak, wide K_Ak；Therefore l_k/b_k Lower proportion ratio threshold θ_kIt is set asUpper limit proportion threshold value β_kIt is set as 7；l_k*b_kLower limit area threshold α_kIt is set as current Monitoring image areaUpper limit area threshold ε_kIt is set as current monitor image area

If having human body to invade by detection, then start alarm module by DSP, without inspection Measure human body invasion, then DSP will not start alarm module, and monitoring region is entered by web camera by continuation Row video acquisition；

If step 4 is found to have human body by the detection of Moving target detection module and invades, then device starts alarm module And gsm module；Invasion personnel will be frightened by alarm module by audible-visual annunciator and voice broadcast module；Wherein report to the police mould Block includes the audible-visual annunciator being made up of flash lamp and speaker and can be according to the voice broadcast of user's request typing different phonetic Module；While starting alarm module, transmission note and mail are carried in time by gsm module to Android intelligent client Wake up user；After user receives information and mail reminder, by Android intelligent client, monitoring client web camera is adopted Video image information in the monitoring region of collection carries out recording preservation and reporting to the police.User passes through Android intelligent client, Log in the IP address preset, check mobile target image in the monitoring region that web camera shootsInformation, and pass through " starting to record " button in Android intelligent client and Alarm button, record the illegal invasion video of invader And report to the police；When invader leaves monitoring region, touch " terminating to record " button and terminate to record, preserve illegal invasion video letter Cease to mobile phone.

By Android intelligent client, the video image of monitoring client web camera typing is preserved, and Display alarm button, main working process is as follows:

1) note is set by Android phone client and mail reception is reminded, when communication module sends postal to user When part and SMS, Android intelligent receive the report for police service information and mail reception prompting, by Android intelligence hands Input the IP address of map network video camera in machine client, conduct interviews；

2) by touching the recording button above Android intelligent client, start monitored area is carried out video Monitoring and recording, until selecting " terminating to record " button, return Android intelligent client；

3) by touching the alarm button above Android intelligent client, " being confirmed whether to report to the police " prompting occurs Frame asks the user whether to report to the police；If selection "Yes", report to the police the most immediately；If selection "No", then return Android intelligent Client.

Claims (6)

1. an intrusion alarm method based on Android intelligent, it is characterised in that step is as follows:

Step one, web camera (10) by the video image acquisition in monitoring region in DSP (8), And to be collected monitoring region in video image carry out video image compression process, will compression after video image by with Too net (4) is transferred to Android intelligent client (2)；

In the monitoring region that in step 2, DSP (8), step one is gathered by Moving target detection module (9) Video image move target detection, the change using Three image difference to calculate the video image in monitoring region judges Whether monitoring region has object to invade；

Moving target detection module (9) in step 3, DSP (8), monitors in region according in step 2 The change of video image, by being identified characteristics of human body, it may be judged whether needs to report to the police；If be detected that there is human body to invade, Then start alarm module (5) by DSP；Without detecting that human body is invaded, then at DSP digital signal Reason device will not start alarm module (5), and continuation is carried out video acquisition to monitoring region by web camera (10)；

If step 4 is found to have human body by Moving target detection module (9) detection and invades, then device starts alarm module And gsm module (6) (5)；Alarm module (5) will be by audible-visual annunciator (501) and voice broadcast module (502) to invasion personnel Frighten；While starting alarm module (5), gsm module (6) will send note and mail to Android intelligent visitor Family end (2) reminds user (1) in time；After user (1) receives information and mail reminder, by Android intelligent client (2) video image information in the monitoring region gathered monitoring client web camera (10) carries out recording preservation and reporting to the police.

A kind of intrusion alarm method and device based on Android intelligent the most according to claim 1, its feature exists In the video image in the monitoring region being collected being carried out video image compression process described in step one, select MPEG-4 Video image after compression, as the compression standard of video data, is transferred to Android intelligence hands by Ethernet (4) by standard Machine client (2) uses RTP/RTCP protocol stack to realize.

A kind of intrusion alarm method and device based on Android intelligent the most according to claim 2, its feature exists In the judgement detailed process whether monitoring region has object to invade described in step 2 it is:

(1) use Three image difference that the mobile target entering monitoring region is monitored, take three two field pictures continuously, by adjacent Two two field pictures are the poorest, then will do the image of after the recovery respectively through gaussian filtering, maximum between-cluster variance threshold method, logic fortune Calculation and morphology operations obtain the bianry image of mobile target neighbor frame difference；

If A_k-1、A_k、A_k+1Be respectively this monitoring system from collect monitoring region in video image obtain kth-1, k, The continuous picture of k+1 frame, it is thus achieved that adjacent image do difference, then:

B_k=| A_k-A_k-1|

B_k+1=| A_k+1-A_k|

For reducing noise to image B derived above_kWith image B_k+1Impact, by respectively to image B_kWith image B_k+1Use 3*3 Template carries out gaussian filtering process, i.e. by with each pixel in a 3*3 template scanogram, the neighbour determined by template In territory, the weighted average gray value of pixel goes the value of alternate template central pixel point, thus the smoothed image C of the noise that is eliminated_k And C_k+1, the neighborhood definition of the most each pixel be centered by the coordinate figure of described each pixel one circle internal or The set of point on border；

If threshold value T_kBy image C_kBeing divided into prospect i.e. target, and background two class, wherein prospect is image C_kMiddle pixel value is more than In threshold value T_kPixel, background is image C_kMiddle pixel value is less than threshold value T_kPixel, even image C_kCoordinate (x, y) The pixel C at place_k(K_x,K_y)≥T_k, then C_k(K_x,K_y) it is prospect, if image C_kAt coordinate (x, y) the pixel C at place_k(K_x,K_y)< T_k, then C_k(K_x,K_y) it is background, and threshold value T_kMake the variance between this two class of foreground and background maximum；If threshold value T_k+1By image C_k+1Being divided into prospect i.e. target, and background two class, wherein prospect is image C_k+1Middle pixel value is more than or equal to threshold value T_k+1Pixel Point, background is image C_k+1Middle pixel value is less than threshold value T_kPixel, even image C_k+1At coordinate (x, y) pixel at place C_k+1(K+1_x,K+1_y)≥T_k+1, then C_k+1(K+1_x,K+1_y) it is prospect, if image C_k+1At coordinate (x, y) the pixel C at place_k+1(K +1_x,K+1_y)<T_k+1, then C_k+1(K+1_x,K+1_y) it is background, and threshold value T_k+1Make the variance between this two class of foreground and background Greatly；Then T_k、T_k+1It is respectively image C_k、C_k+1Optimal threshold, the calculating process of optimal threshold is as follows:

If image C_IFor the I-frame video image in the monitoring region that web camera collects, wherein C_IComprise L gray level 0,1 ..., L-1}, gray value be the pixel number of i be n_i, then total pixel isGray scale is the pixel probability of occurrence of i ForEntire image is divided into target M by thresholding t₁With background M₂Two classes, the then computing formula of the inter-class variance δ of this two class For:

In formula,WithIt is respectively target M₁With background M₂The probability that two classes occur；WithIt is respectively target M₁With background M₂The gray average of two classes, the optimal threshold of segmentation is inter-class variance computing formula In thresholding t value when taking maximum, be designated as T_I:

Image C is tried to achieve by above-mentioned optimal threshold computational methods_k、C_k+1Optimal threshold T_k、T_k+1, then use optimal threshold T_k、 T_k+1Respectively to image C_k、C_k+1Binaryzation obtains prospect bianry image D_k、D_k+1, it is maximum between-cluster variance threshold method, uses formula It is expressed as follows:

D_k=threshold (| C_k|)

D_k+1=threshold (| C_k+1|)

(2) the prospect bianry image D that will obtain_k、D_k+1Carry out logic "and" operation, wherein image D_kIn prospect and image D_k+1 In prospect carry out AND operation after the prospect that obtains be image A_kThe shape profile of middle moving target, formula is expressed as:

E_k=D_k And D_k+1

Due to effect of noise, image D_k、D_k+1Image E obtained after carrying out logical operations_kBorder generally the most not perfectly flat cunning, front Scene area has some noise holes, and is studded with some little noise objects on background area, by morphologic opening operation, I.e. after excessive erosion reflation thus remove the noise hole in bianry image, obtain the image F after de-noising_k, wherein expand It is each pixel using structural element G scanogram, and the bianry image covered with it does OR operation；Corrosion is to use structure Each pixel of element G scanogram, and the bianry image covered with it does AND-operation；The morphology opening operation of image is public Formula is as follows:

(3) by using connection eight neighborhood rule, it is thus achieved that image F_kIn the minimum abscissa value in largest connected region, maximum horizontal seat Scale value, minimum ordinate value and maximum ordinate value, thus obtain the mobile target image of red rectangle collimation mark noteRealize step Rapid as follows:

1) in two dimensional image, it is assumed that have the pixel that z is adjacent around target pixel points, wherein z≤8, if this pixel grey scale Equal with the gray value of some pixel Q in this z pixel, then to claim this pixel and Q to have connectedness；Connection eight neighborhood rule It is then to choose object pixel all of adjacent picture in two-dimensional space, is upper and lower, left and right, upper left, upper right, lower-left, bottom right Eight neighbor pixels；By using connection eight neighborhood rule to image F_kIn each pixel label is set, obtain figure As F_kThe connected component labeling matrix of pixel, label initial value is all set to 0；If image F_kPixel (K_x,K_y), its connected region It is numbered label (K_x,K_y), wherein image F_kPixel (K_x,K_y) eight neighborhood coordinate be (K_x-1,K_y)、(K_x+1,K_y)、(K_x, K_y-1)、(K_x,K_y+1)、(K_x-1,K_y-1)、(K_x-1,K_y+1)、(K_x+1,K_y-1) and (K_x+1,K_y+1)；

2) first from image F_kThe upper left corner start progressive scan, be i.e. from storage image F_k0 row 0 row of the matrix of pixel start Progressive scan, from top to bottom, every a line scans from left to right；When scanning image F_kPixel (K_x,K_y) time, scan through Become (K_x,K_y) the adjacent pixel in the top of pixel and the left side, obtain label (K_x-1,K_y)、label(K_x-1, K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1) value；

If image F_kCoordinate is (K_x,K_y) place's pixel is foreground pixel point, and the value of its pixel is adjacent pixel point value phase With, then it is assumed that these pixels are connections, and wherein foreground pixel point is image F_kThe pixel of middle composition prospect；If pixel (K_x,K_y) it is adjacent pixel (K_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) and (K_x,K_y-1) connection, then label (K_x, K_y) value equal to label (K_x-1,K_y)、label(K_x-1,K_y-1)、label(K_x-1,K_y+1) and label (K_x,K_y-1) when intermediate value Little label, and update the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value and maximum ordinate Value: minimum abscissa value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberAnd maximum ordinate Value maxy_number, wherein number is the label of this connected region；

If image F_kIn be numbered the connected region of number, if the scanning of this connected region is to pixel (K_x,K_y) time, remember The minimum abscissa value of this connected region of record, maximum abscissa value, minimum ordinate value and maximum ordinate value: minimum horizontal seat Scale value minx_number, maximum abscissa value maxx_number, minimum ordinate value miny_numberWith maximum ordinate value maxy_number, And pixel (K_x,K_y) be adjacent pixel connection, then update the minimum abscissa value of this connected region, maximum abscissa value, Minimum ordinate value and maximum ordinate value, if now K_x≥maxx_number, then maxx_number=K_xAnd minx_numberValue keeps not Become, if minx_number<K_x<maxx_number, then minx_numberAnd maxx_numberValue all keeps constant；If K_x≤minx_number, then minx_number=K_xAnd maxx_numberValue keeps constant；If now K_y≥maxy_number, then maxy_number=K_yAnd miny_numberValue Keep constant, if miny_number<K_y<maxy_number, then miny_numberAnd maxy_numberValue all keeps constant；If K_y≤ miny_number, then miny_number=K_yAnd maxy_numberValue keeps constant；

If image F_kPixel (K_x,K_y) and the pixel (K of its top_x-1,K_y)、(K_x-1,K_y-1)、(K_x-1,K_y+1) and the left side Pixel (K_x,K_y-1) do not connect, and image F_kAt pixel (K_x,K_y) place is foreground pixel point, then will increase by one New connected region；I.e. enumerator number '=number+1 is as pixel (K_x,K_y) connection label, i.e. label (K_x, K_y)=number ', and it is vertical to record the minimum abscissa value of this connected region, maximum abscissa value, minimum ordinate value and maximum Coordinate figure: minimum abscissa value minx_number′, maximum abscissa value maxx_number′, minimum ordinate value miny_number′And maximum Ordinate value maxy_number′Even, minx_number′=K_x、maxx_number′=K_x、miny_number′=K_yAnd maxy_number′=K_y； If image F_kIt is (K at coordinate_x,K_y) place's pixel is background pixel point, then continue to scan on next point, wherein background pixel Point is image F_kThe pixel of middle composition background；

3) according to process 2) described in scanogram F_kIn all pixels, if scanned, then perform process 4)；

4) from left to right, from top to bottom traversing graph as F_kIn all different connected regions, according to record all different connections The minimum abscissa value in region, maximum abscissa value, minimum ordinate value and maximum ordinate value: minx_number、maxx_number、 miny_numberAnd maxy_numberCalculate the height H of each connected region_numberWith width W_number, it may be assumed that

H_number=maxx_number-minx_number

W_number=maxy_number-miny_number

Will height H_numberWith width W_numberProduct as the similar area i.e. S of each connected region_number=H_number*W_number, Area S by relatively each connected region_number, by area maximum MAX (S_number) connected region be designated as largest motion module, Preserve the minimum abscissa value of largest motion module, maximum abscissa value, minimum ordinate value and maximum ordinate value: MAXS (minx_number)、MAXS(maxx_number)、MAXS(miny_number) and MAXS (maxy_number)；Due to image F_kIn maximum fortune Dynamic model block is to process image A by above-mentioned steps_kObtained by, therefore image A_kIn largest motion module and image F_kIn Largest motion module coordinate information is identical, indulges with minimum and maximum abscissa column, the minimum and maximum of largest motion module Coordinate is expert at and is formed a height of MAXS (maxx_number)-MAXS(minx_number), a width of MAXS (maxy_number)-MAXS (miny_number) rectangle frame, i.e. four coordinate points of rectangle frame are (MAXS (minx_number),MAXS(miny_number)), (MAXS(minx_number),MAXS(maxy_number)), (MAXS (maxx_number),MAXS(miny_number)), (MAXS (maxx_number),MAXS(maxy_number)), the rectangle frame that order obtains is shown in red；With red rectangle frame by image A_kIn Largest motion module frame is selected, and then obtains the mobile target image with red rectangle collimation mark noteThen judge that monitoring region has Object is invaded.

A kind of intrusion alarm method and device based on Android intelligent the most according to claim 3, its feature exists In described in step 3 by characteristics of human body's recognition methods judge whether need report to the police detailed process be:

If monitoring region has object to invade, set different threshold decision conditions, it is determined that mobile target imageMiddle red square Whether shape frame meets characteristics of human body, if met, is considered human body, needs to report to the police；If not meeting characteristics of human body, the most permissible Initial setting according to user judges whether to report to the police；If the mobile target image with red rectangle collimation mark noteIn, labelling A length of l of red rectangle frame_k, a width of b_k, work as b_kMore than l_k, then it is assumed that intrusion object is upright；To people from upright fortune Distinguishing in animal body, can pass through characteristics of human body, characteristics of human body's recognition methods of the present invention is described as follows:

If mobile target imageIn, a length of l of the red rectangle frame of labelling_k, a width of b_k, then l_k/b_kCan be approximated to be invader The length-width ratio of body, l_k*b_kCan be approximated to be the area of intrusion object；If l_k/b_kLess than θ_kOr more than β_k, or l_k*b_kIt is less than α_kOr more than ε_kTime, it not the most human body, and if only if θ_k≤l_k/b_k≤β_kAnd α_k≤l_k*b_k≤ε_kTime, it is judged that invade for human body Monitoring region；Wherein, threshold θ_k、β_k、α_kAnd ε_kValue by IP Camera with monitoring region distance determine, if network is taken the photograph The distance in camera and monitoring region is when H rice distance, and the monitoring range of web camera is long L_Ak, wide K_Ak；Therefore l_k/b_kUnder Limit proportion threshold value θ_kIt is set asUpper limit proportion threshold value β_kIt is set as 7；l_k*b_kLower limit area threshold α_kIt is set as current monitor Image areaUpper limit area threshold ε_kIt is set as current monitor image area

If having human body to invade by detection, then start alarm module (5) by DSP (8), without Detect that human body is invaded, then DSP (8) will not start alarm module, and web camera (10) is right by continuation Monitoring region carries out video acquisition.

5. an intrusion alarm device based on Android intelligent, it is characterised in that including: web camera (10), shifting Moving-target detection module (9), power module (7), DSP (8), alarm module (5), gsm module (6), with Too net (4) and Android intelligent client (2), web camera (10) video acquisition output port and DSP digital signal Processor (8) video acquisition input port connects, and power module (7) output port of power source inputs with web camera (10) power supply Port connects, and alarm module (5), gsm module (6), power module (7), Moving target detection module (9) are digital with DSP respectively Signal processor (8) electrically connects, and gsm module (6) is connected with Android intelligent client (2) wireless telecommunications, network shooting The RJ-45 Ethernet interface of machine (10) is connected with Ethernet (4) by netting twine, and Android intelligent client (2) can be passed through Ethernet (4) accesses the video information that web camera (10) gathers.

A kind of intrusion alarm device based on Android intelligent the most according to claim 5, it is characterised in that institute The alarm module (5) stated includes the audible-visual annunciator (501) being made up of flash lamp and speaker and can be according to user's request typing The voice broadcast module (502) of different phonetic.