CN101557506B - Intelligent detecting device for violent behavior in elevator car based on computer vision - Google Patents
Intelligent detecting device for violent behavior in elevator car based on computer vision Download PDFInfo
- Publication number
- CN101557506B CN101557506B CN2009100989198A CN200910098919A CN101557506B CN 101557506 B CN101557506 B CN 101557506B CN 2009100989198 A CN2009100989198 A CN 2009100989198A CN 200910098919 A CN200910098919 A CN 200910098919A CN 101557506 B CN101557506 B CN 101557506B
- Authority
- CN
- China
- Prior art keywords
- sequence
- video
- crowd
- human body
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Images
Landscapes
- Indicating And Signalling Devices For Elevators (AREA)
Abstract
The invention discloses an intelligent detecting device for violent behavior in an elevator car based on computer vision, which comprises a video sensor arranged at the top part of the elevator car, an embedded system used for transmitting video data, and a monitoring center computer used for monitoring the interior of the elevator car. The video sensor is connected with the embedded system; the embedded system is wirelessly connected with the monitoring center computer; the monitoring center computer comprises a video image reading module in the car used for displaying the video data in the elevator car in real time; the video image reading module in the car is connected with a display device; the video sensor is communicated with the wireless data of the monitoring center computer; the monitoring center computer comprises a microprocessor used for the safety precaution in the elevator car; the microprocessor also comprises a background modeling and human body foreground object extracting module, a crowd behavior characteristic sequence extracting module, a modeling module of a hidden Markov model as well as an identifying module and an alarm module of violent behaviors. The intelligent detecting device for violent behavior in an elevator car based on computer vision has the advantages of intellectualization, real-time online and strong reliability.
Description
Technical field
The present invention relates to a kind of act of violence intelligent detection device in lift car that is used for based on technology such as computer vision technique, image recognitions.
Background technology
Along with building in the city covers higher and higher, the use of elevator is more and more general, and the elevator safety problem is subjected to people day by day and payes attention to.Elevator give human life bring convenient rapidly after, also the safety to the mankind challenges.Because the relative closure of elevator, it has also brought threat for human safety.In the elevator narrow space, the people also can be reduced to minimum point to the constraint of self.News media report some elevator incidents of violence through regular meeting; the victim met with in the elevator and also had been subjected to very big wound except damage to property with external psychosoma aspect after the incident of violence, and the kind of elevator incident of violence can be divided into robbing, robbing beauty and other incidents of violence in the elevator.
At present, provide the traditional video surveillance system of safety guarantee in the elevator, these massive video data are not taked any intelligentized processing, thereby can't detect automatically the different forms of violence incident; Need the security personnel in Control Room realize the elevator safety monitoring by watching video image, obviously, this monitor mode is taken very much manually, almost is difficult to realize monitoring truly, and this is because later attentiveness obviously descended at 20 minutes by artificial video image.In the Secure Application field, the target of intelligent video monitoring system is to discern personnel and their behavior under various environment in real time.Therefore exploitation has crucial meaning based on the elevator inner violence-proof intelligent detecting method of computer vision.
A kind of have dysgenic behavior, a criminal offence etc. based on act of violence intelligent detection device in the lift car of computer vision is must effective monitoring various in running process of elevator, and can make reaction pointedly, just as an elevator safety bodyguard at all times.When detecting threatening, act of violence generation, system can at first carry out phonetic warning, if behavior still continues, the control elevator is parked in nearest floor, sends alarm signal notice elevator management personnel simultaneously.
Summary of the invention
For overcome existing elevator video monitoring device need artificial video monitor, automaticity not high, almost handle the deficiency of means, poor reliability without any intellectuality, the invention provides a kind of intellectuality, real-time online, good reliability based on act of violence intelligent detection device in the lift car of computer vision.
The technical solution adopted for the present invention to solve the technical problems is:
A kind of based on act of violence intelligent detection device in the lift car of computer vision, comprise the video sensor that is installed in elevator car roof, the embedded system that is used for transmitting video data, be used to monitor the Surveillance center's computer in the lift car, described video sensor is connected with described embedded system, described embedded system is connected with the Surveillance center computer by communication, described Surveillance center computer comprises the car inner video image read module that is used for showing in real time video data in the lift car, described car inner video image read module connects display unit, described Surveillance center computer is the center of the WLAN (wireless local area network) in building, elevator place, described video sensor and Surveillance center's computer carry out the wireless video data communication, described supervisory control comuter comprises the microprocessor that is used for safety precaution in the lift car, and described microprocessor also comprises: background modeling and human body foreground object extraction module, crowd behaviour characteristic sequence extraction module, the MBM of hidden Markov model, the identification module of act of violence and warning and phonetic warning module;
Described car inner video image read module is used for the video information in the lift car is collected and send to Surveillance center's computer, and Surveillance center's computer real-time reads the video data that sends;
Described background modeling and human body foreground object extraction module are used for the video data that is read is extracted the human body foreground object;
Described crowd behaviour characteristic sequence extraction module, be used in human body foreground object that sequence of video images extracted, and the related data that changes of the length and width that calculate the variation of area change, foreground area mass centre of the connected region of this foreground object and boundary rectangle, set up the three-dimensional feature sequence vector;
The MBM of described hidden Markov model, be used for each three-dimensional feature sequence vector is converted to a concrete symbol, it is observed value, by obtaining the characteristic vector data of the normal behaviour in a large amount of elevators, by the K-Means algorithm full feature vector data collection is carried out cluster then, obtained the code book set.At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set.These observed values are formed observation sequence, by the parameter of the best HMM model of Baum-Welch algorithm acquisition, set up the HMM model of normal behaviour;
The identification module of described act of violence, be used to distinguish normal behaviour or abnormal behaviour, go out the probability of observed value sequence by the Forward-backward algorithm computation, if its output probability of observed value sequence is very high, think then that greater than pre-set threshold it is a normal behaviour sequence.Otherwise,, then think an abnormal behaviour sequence if the output probability of an observation sequence is lower than setting threshold.
Technical conceive of the present invention is: illustrate that at first the violence intelligent video detects principle in the elevator.Under normal circumstances, the human body that rides in the elevator generally all is to be in static relatively state, produces though have of short duration action sometimes, and its movement range can be very not big, and can not have influence on other human bodies in the elevator.And when act of violence occurring, other human bodies of the side's of applying meeting active attack of act of violence have tight contact process, in case act of violence occurs between the human body, from video angle, be difficult to separate between aggrieved party's human body of the side of applying of act of violence and act of violence.The present invention is by extracting the correlated characteristic of the prospect connected region of human object in elevator, and the hidden Markov model of setting up the human body normal behaviour detects act of violence.
The detection detailed process is as follows: video data obtains in (1) elevator.Monitoring camera be installed in the lift car directly over, overlook downwards, guaranteeing in the captured lift car all human bodies all in the visual field of camera, and the situation that can not occur blocking; (2) set up the elevator background model, extract the variation of the area of the prospect connected region of human body in the elevator, the variation of foreground area mass centre and the length and width variation of boundary rectangle and set up the three-dimensional feature vector; (3) by a characteristic sequence that extracts, using the method for vector quantization to transfer to can be the observed value sequence of hidden Markov modeling use, sets up the hidden Markov model of normal behaviour; (4) for detected act of violence, the processing of being correlated with.Concrete training testing process as shown in Figure 1.
Testing process shown in 1 with reference to the accompanying drawings, the present invention will extract, cut apart from background modeling and human body foreground object, extract the crowd behaviour characteristic sequence, the modeling of hidden Markov model, aspects such as the identification of act of violence describe;
Adopt the Codebook algorithm to extract the prospect human object among the present invention, this algorithm is by observing video sequence for a long time, utilizing quantification and clustering technique to make up background model.It is set up one to each pixel and comprises one or more code books, and brightness and the intensity of variation of color in the continuous sampling process according to each pixel generates code word earlier, and the code word of all represent pixels is referred to as code book.At last, judge by the code word of record whether pixel belongs to prospect or background, but the realization list of references [1] of this algorithm.
The building process of code book is: suppose that at first χ is the training sequence to a specific pixel, it forms χ={ x by N RGB vector
1, x
2..., x
N.In addition, make C={c
1, c
2.., c
LRepresent the code book of this pixel relatively, comprise L code word.In the code book of each pixel, the quantity of code word is different, depends on the fluctuation of sample.Each code word c
i, i=1...L has comprised a RGB vector v
i=(R
i, G
i, B
i) and one 6 tuple
In order to solve the problem of illumination variation,, also used a special color model that colouring information and monochrome information are separated such as shade and high light.The color model of accompanying drawing 3 expressions can change separately assessment to change color and brightness;
Can exist noise spot by the two-value video image that obtains after the above-mentioned processing, still need further to carry out the morphology denoising, modal solution is to use a kind of method based on criterion distance, and promptly expansion-erosion operator is connected as a single entity mutual approaching foreground blocks or target; Among the present invention, filter isolated point by the corrosion operation, the cavity is then removed by expansive working, but the realization list of references [2] of this algorithm and list of references [3].
Be to detect at crowd behaviour in the present invention, less than two man-hours in detecting car, system are automatically brought to described car inner video image read module, do not carry out follow-up calculating;
Detect about number in the car, need carry out calibration experiment according to actual conditions and obtain decision threshold, among the present invention the picture frame of single appearance in the video image experiment statistics and analysis have been done, the shared pixel quantity mean value of single foreground area is 5991 under this experimental situation, therefore should be worth as criterion, i.e. Np=5991; Ni, i=1...n are all single humanoid figure picture frame foreground pixel quantity, and experimental result shows, when 0.8Np≤Ni≤1.6Np, are judged to be single situation; When Ni＞1.6Np, be judged to be single more than.
Described crowd behaviour characteristic sequence extraction module, be used in human body foreground object that sequence of video images extracted, and the related data that changes of the length and width that calculate the variation of area change, foreground area mass centre of the connected region of this foreground object and boundary rectangle, set up the three-dimensional feature sequence vector;
In detecting car during the above situation of two people or two people; in described background modeling and the calculating of human body foreground object extraction module, extracted the sequential binary map of human body prospect; these data have obtained characteristic vector sequence after treatment, the behavior detection that can be used in the modeling of setting up hidden Markov model or pass through hidden Markov model.The characteristic vector here comprises by the length and width variation of the variation of crowd's foreground area, crowd's prospect boundary rectangle, the center of gravity set of variations of crowd's prospect becomes the three-dimensional feature vector data;
1) the variation AC of crowd's foreground area
The computational methods of the variation AC of crowd's foreground area are provided by formula (1), and the size of this value is represented the size of the variable quantity of foreground pixel quantity, can reflect the degree of human motion fierceness;
In the formula: Ap is the area of the prospect human object that obtained of former frame image, and An is the area of the prospect human object that current frame image obtained.
2) length and width of crowd's prospect boundary rectangle change WHC
The length and width of crowd's prospect boundary rectangle change WHC and are provided by formula (2), and the size of this value is represented the size of the length and width variable quantity of crowd's prospect boundary rectangle, can reflect the degree that human body attitude changed when act of violence took place;
In the formula: Hp, Wp are respectively the length of boundary rectangle of previous image frame prospect human object and wide, and Hn, Wn are the length of boundary rectangle of prospect human object of current image frame and wide.
3) center of gravity of crowd's prospect changes CC
The center of gravity of crowd's prospect changes CC and is provided by formula (3), and the size of this value is represented the size of the center of gravity variable quantity of crowd's prospect, can reflect that human body moved the degree of variation when act of violence took place;
CC＝||C
p-C
n||(3)
In the formula: Cp be previous image frame prospect human object position of centre of gravity, Cn is the position of centre of gravity of the prospect human object of current image frame.
The MBM of described hidden Markov model, be used for each three-dimensional feature sequence vector is converted to a concrete symbol, it is observed value, by obtaining the characteristic vector data of the normal behaviour in a large amount of elevators, by the K-Means algorithm full feature vector data collection is carried out cluster then, obtained the code book set.At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set.These observed values are formed observation sequence, by the parameter of the best HMM model of Baum-Welch algorithm acquisition, set up the HMM model of normal behaviour;
Because the hidden Markov model of subsequent treatment needs a concrete symbolic feature to come the representation feature vector, therefore each characteristic vector need be converted to a concrete symbol that is to say observed value.At this moment can realize by the method for using vector quantization or cluster; In fact way is by certain methods the characteristic vector training data to be divided into K classification, makes the similarity of this K classification minimum, and the central point of this K classification is formed the code book set.In case training is finished, and has obtained the code book set, just can will be arbitrarily characteristic vector transfer index value with its nearest code book in code book is gathered to, this index value conduct is an observed value.Algorithm commonly used at present has LBG vector quantization algorithm [4] and K-Means clustering algorithm [5], and what adopt among the present invention is the K-Means clustering algorithm;
The K-Means clustering algorithm is accepted input variable k; Then n data object is divided into k cluster so that make the cluster that is obtained satisfy, the object similarity in the same cluster is higher, and the object similarity in the different cluster is less.The cluster similarity is to utilize the average of object in each cluster to obtain one " center object ", and promptly center of attraction is calculated;
The roughly flow process of this algorithm is as follows:
(1) select k object as initial cluster center arbitrarily from n data object;
(2) circulation (3) to (4) till each cluster no longer changes;
(3), calculate the distance of each object and these center object, and again corresponding object is divided according to minimum range according to the average (center object) of each cluster object;
(4) recomputate the average (center object) of each (changing) cluster.
Hidden Markov model below is represented with HMM, it be a kind of with parametric representation, be used to describe the probabilistic model of statistics of random processes characteristic.A typical hidden Markov model as shown in Figure 4;
HMM can be designated as λ=(N, M, π, A, B),
The HMM model specifically can be described by following parameters:
(1) N: the state number in the model.Remember N state S
1..., S
N
(2) M: the possible observed value number of each state correspondence.Remember that M observed value is O
1..., O
M
(3) π: initial condition probability vector, π=(π
1..., π
N) wherein, π
i=P (q
1=S
i), 1≤i≤N, and satisfy π
i〉=0,
It is used to describe observation sequence when initial belongs to different state when t=1 probability.
(4) A is a state transition probability matrix, A=(a
Ij)
N * NWherein A is a state transition probability matrix, A=(a
Ij)
N * NWherein, a
Ij=p (q
T+1=S
j| q
t=S
i), 1≤i, j≤N, and,
It is used to describe the probability distribution that is transformed to the state j of time t+1 from the state i at time t.
(5) B is the observed value probability matrix, B=(b
Ji)
N * MWherein, b
j(i)=p (o
t=O
i| q
t=S
j)=p (Q
i| S
j) 1≤j≤N, 1≤i≤M, and
It is used to be described in time t state S
jSituation under to produce observed value be O
iProbability.
Realize that by HMM the act of violence detection in the elevator must solve three basic problems.
(a) computational problem.For given observed value sequence o=O
1O
2... O
NAnd specific HMM model parameter λ=(N, M, π, how A B), calculates the probability that this specific HMM model produces characteristic sequence o.This problem is very important, can pass through Forward-backward[6] the algorithm solution.
(b) identification problem.For given observed value sequence o=O
1O
2... O
N, how to select best status switch S so that this sequence of best explanation.This can pass through Viberti-Decoding[6] the algorithm solution.
(c) training problem, how the adjustment model parameter lambda makes P (o| λ) probability maximum.This can lead to Baum-Welch algorithm [6] and solve.
But realization list of references [6] about Forward-backward algorithm, Viberti-Decoding algorithm and Baum-Welch algorithm; After having solved above-mentioned three problems, just can come act of violence is detected by the HMM modeling;
It is simple relatively, easy for normal behavioral data is obtained, and abnormal behaviours such as violence have the consideration of the infinite property enumerated, the present invention comes act of violence is detected by the normal behaviour in the elevator is carried out the HMM modeling, and what select for use is that from left to right two condition shifts HMM, as shown in Figure 4;
The use of HMM model needed through training and two stages of detection.Promptly set up the HMM model stage in the training stage.At first, obtain the characteristic vector data of the normal behaviour in a large amount of elevators, by the K-Means algorithm full feature vector data collection is carried out cluster then, obtained the code book set.At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set.These observed values are formed observation sequence, obtain the parameter of best HMM model by the Baum-Welch algorithm, thereby set up the HMM model of normal behaviour.Concrete modeling process as shown in Figure 5;
The identification module of described act of violence, be used to distinguish normal behaviour or abnormal behaviour, go out the probability of observed value sequence by the Forward-backward algorithm computation, if its output probability of observed value sequence is very high, think then that greater than pre-set threshold it is a normal behaviour sequence.Otherwise,, then think an abnormal behaviour sequence if the output probability of an observation sequence is lower than setting threshold;
At cognitive phase, the characteristic vector sequence to input is converted to the observed value sequence by the code book collection that obtains in the training stage.Go out the probability of observed value sequence by the Forward-backward algorithm computation,, think then that greater than pre-set threshold it is a normal behaviour sequence if its output probability of observed value sequence is very high.Otherwise,, then think an abnormal behaviour sequence if the output probability of an observation sequence is lower than setting threshold.Concrete detection identifying as shown in Figure 6;
As shown in Figure 6, the probability of observation sequence Model Calculation thus obtains, and the process of identification behavior is exactly by whether judging the output probability value less than a minimum threshold, if just this behavior sequence is classified as abnormal behaviour less than a minimum threshold;
Described minimum threshold generally can obtain by the training study of normal sequence, usually by observing the probable value of exporting in the training stage, the establishing method of minimum threshold is to deduct and organize 5% of probit range δ that normal sequence exports more one fixing on minimum value in the probable value that many group normal sequence export;
Described warning and phonetic warning module, be used for the abnormal behaviour sequence is carried out phonetic warning, system sends in various degree caveat according to the output probability of abnormal behaviour sequence and the departure degree that sets threshold value, when reaching the alarm threshold value that sets, system can notify administrative staff in time to intervene automatically and handle also and automatically elevator be stopped at nearest floor;
Further, here at first need to calculate the abnormality degree of abnormal behaviour, among the present invention the probable value of being exported departed from that minimum threshold is big more thinks that abnormality degree is high more, abnormal behaviour process in the various elevators is judged to be different result in " affirmation violence ", " severely subnormal ", " unusually ", " propensity to violence ", " should be noted that " etc. 5 respectively, and decision method and corresponding measure are as shown in table 1;
| The judgement scope | Result of determination | Measure |
| Minimum threshold-δ (＞20) % | Confirm serious act of violence | Voice serve a grave warning, and notice related management personnel, and the control elevator is in floor stop recently |
| Minimum threshold-δ (＞15～20) % | Severely subnormal | Voice reminder warning, and notice related management personnel, the control elevator is in floor stop recently |
| Minimum threshold-δ (＞10～15) % | Unusually | Voice reminder caution, and notice related management personnel |
| Minimum threshold-δ (＞5～10) % | Propensity to violence | The voice reminder caution |
| Minimum threshold-δ (0～5) % | Should be noted that | Voice reminder is noted |
Table 1 abnormality degree decision table
Annotate: the δ in the table 1 represents the probit range that many group normal sequence are exported.
Description of drawings
Fig. 1 is based on behavior testing process figure in the act of violence intelligent detection device in the lift car of computer vision;
Fig. 2 carries out digital wireless video and audio communication system topological figure for the car inner video image adopts embedded software and hardware and radio receiving transmitting module;
Fig. 3 can change the separately color model figure of assessment to change color and brightness;
Fig. 4 is a kind of typical hidden Markov model figure;
Fig. 5 sets up process flow diagram for hidden Markov model;
Fig. 6 discerns the flow chart of behavior for adopting hidden Markov model;
Fig. 7 is composition module and the process chart based on act of violence intelligent detection device in the lift car of computer vision;
Fig. 8 is the figure as a result that judges normal behaviour and abnormal behaviour based on hidden Markov model.
Embodiment
Below in conjunction with accompanying drawing the present invention is further described.
With reference to Fig. 1～Fig. 8, a kind of based on act of violence intelligent detection device in the lift car of computer vision, comprise the video sensor that is installed in elevator car roof, the embedded system that is used for transmitting video data, be used to monitor the Surveillance center's computer in the lift car, described video sensor is connected with described embedded system by USB interface, described embedded system is connected with the Surveillance center computer by communication, described Surveillance center computer comprises the car inner video image read module that is used for showing in real time video data in the lift car, described car inner video image read module connects display unit, described Surveillance center computer is the center of the WLAN (wireless local area network) in building, elevator place, described video sensor and Surveillance center's computer carry out the wireless video data communication, described supervisory control comuter comprises the microprocessor that is used for safety precaution in the lift car, described microprocessor also comprises: background modeling and human body foreground object extraction module, crowd behaviour characteristic sequence extraction module, the MBM of hidden Markov model, the identification module of act of violence and warning and phonetic warning module, as shown in Figure 7;
Described embedded system, be used to gather be used in the car gather, the embedded system of transmission of video information and cooperating of radio receiving transmitting module, select embedded Linux system, mainly reach collection sensing, communication, move and be the purpose of one, concrete selection Samsung S3C2410X is an embedded microprocessor, the combining wireless local area network technology.Comprised software and hardware technology in the embedded system, wherein built-in Linux software is core technology, and it can realize looking the function of audio server.
Described embedded microprocessor S3C2410X is a 16/32 RISC embedded microprocessor based on the ARM920T kernel, and this processor designs for handheld device and high performance-price ratio, low-power consumption microcontroller.It has adopted the new bus architecture of a kind of AMBA of being called (Advanced Microcontroller Bus Architecture).The main resource of S3C2410X inside has memory management unit MMU, system administration manager, respectively is the instruction and data buffer memory of 16KB, lcd controller (STN﹠amp; TFT), NAND FLASH Boot Loader, 3 passage UART, 4 passage DMA, 4 PWM clocks, 1 internal clocking, 8 path 10s are ADC, touch screen interface, multimedia card interface, I2C and I2S bus interface, 2 usb host interfaces, 1 USB device interface, SD main interface, 2SPI interface, pll clock generator and general purpose I/O port etc., as shown in Figure 2;
Described embedded microprocessor S3C2410X inside comprises a memory management unit that is MMU, can realize the mapping of virtual memory space to amount of physical memory.Usually the program of embedded system leaves among the ROM/FLASH, program can access preservation behind the system cut-off, but ROM/FLASH compares with SDRAM, it is slow many that speed is wanted, and usually the aborted vector table is left among the RAM in the embedded system, utilize memory-mapped mechanism can solve this needs.
Described ROM/FLASH adopts the K9S1208VOM of the 64MB of Samsung.It can carry out 100,000 times program/erase, and data are preserved and reached 10 years, are used to loading operation system image and large-capacity data.
Described SDRAM is the K4S561632C that adopts Samsung, is used for needed data in operation system and the stored program running, and it is the synchronous dram of 4M*16bit*4bank, and capacity is 32MB.Realize the position expansion with two K4S561632C, making data-bus width is 32bit.
Described embedded software system mainly comprises writing of the installation of transplanting, driver of operating system, ICP/IP protocol and user application etc.
Adopted Linux as embedded OS among the present invention, Linux develops from UNIX, inherited the most advantage of UNIX, the disclosed kernel source code of Linux makes it become present most popular operating system, and Linux can be from its hardware-software of application cutting, this is concerning towards very necessary based on this special requirement of the elevator inner violence-proof apparatus of image recognition technology, here we are referred to as the customization operations system, and customization step is as follows: (1) writes plate base support package BSP; (2) each parts of cutting and configuration operation system, and revise corresponding configuration file; (3) compiling Kernel, assembly and BSP, generating run system image file; (4) image file is downloaded on the Target Board, debug.
Further, looking Voice Surveillance information in the elevator is to transmit in the mode of packing data, transmission through WLAN (wireless local area network) by ICP/IP protocol, therefore under the operating system support, realize ICP/IP protocol, just need carry out task division, the realization of TCP/IP can be divided into 4 tasks realizes: 1. IP task, the reorganization that is mainly used to solve IP fragmentation; 2. the TCP incoming task is mainly used to handle the TCP message segment that receives; 3. TCP output task is mainly used to packing data, the transmission that will export; 4. TCP task of timer, being mainly used to provides clock for various time delay incidents (as the repeating transmission incident).
Further, need two USB interface in the elevator inner violence-proof apparatus based on image, speech recognition technology, one of them USB interface is that monitoring camera is connected with S3C2410X, another USB interface is that wireless network card is connected with S3C2410X, because S3C2410X self-carried USB principal and subordinate interface, do not need special USB chip support, as long as can carry out USB transmission data to its install driver.On S3C2410X, dispose speech interface, microphone directly is connected the acquisition function that just can finish voice messaging with speech interface.
Described USB driver comprises following several sections: (1) establishment equipment, create two parameter calls of equipment function band, and a parameter is to point to the pointer of driver object, another parameter is to point to the pointer of physical device object; (2) closing device; (3) fetch equipment data, when client applications has requiring of fetch equipment data, system requires this to pass to function driver with the IRP form of IRP_MJ_READ, D12Meter_Read program by equipment is carried out, and then specifies the direct and equipment realization information interaction of usb bus driver by D12Meter_Read; (4) equipment is write data, when client applications has requiring of write device data, system requires this to pass to function driver with the IRP form of IRP_MJ_WRITE, and carry out by D12Meter_Write, and then by D12Meter_Write specify the usb bus driver directly with equipment realization information interaction.The USB driver is discerned USB device by PID (product identification number) in the installation file (.inf file) and VID (vendor identification number).
After the embedded OS loading is finished, driver and other corresponding application of wireless network card just can be installed.The driver of wireless network card is bundled in the operating system as a module, can avoids the WLAN Device Driver of at every turn all will resetting after system's power down.
Described car inner video image read module is used for the video information in the lift car is collected and send to Surveillance center's computer, and Surveillance center's computer real-time reads the video data that sends;
Described background modeling and human body foreground object extraction module are used for the video data that is read is extracted the human body foreground object;
Adopt the Codebook algorithm to extract the prospect human object among the present invention, this algorithm is by observing video sequence for a long time, utilizing quantification and clustering technique to make up background model.It is set up one to each pixel and comprises one or more code books, and brightness and the intensity of variation of color in the continuous sampling process according to each pixel generates code word earlier, and the code word of all represent pixels is referred to as code book.At last, judge by the code word of record whether pixel belongs to prospect or background, but the realization list of references [1] of this algorithm;
The building process of code book is: suppose that at first χ is the training sequence to a specific pixel, it forms χ={ x by N RGB vector
1, x
2.., x
N.In addition, make C={c
1, c
2..., c
LRepresent the code book of this pixel relatively, comprise L code word.In the code book of each pixel, the quantity of code word is different, depends on the fluctuation of sample.Each code word c
i, i=1...L has comprised a RGB vector v
i=(R
i, G
i, B
i) and one 6 tuple
In order to solve the problem of illumination variation,, also used a special color model that colouring information and monochrome information are separated such as shade and high light.The color model of accompanying drawing 3 expressions can change separately assessment to change color and brightness;
Can exist noise spot by the two-value video image that obtains after the above-mentioned processing, still need further to carry out the morphology denoising, modal solution is to use a kind of method based on criterion distance, and promptly expansion-erosion operator is connected as a single entity mutual approaching foreground blocks or target; Among the present invention, filter isolated point by the corrosion operation, the cavity is then removed by expansive working, but realization list of references [2], [3] of this algorithm;
Be to detect at crowd behaviour in the present invention, less than two man-hours in detecting car, system are automatically brought to described car inner video image read module, do not carry out follow-up calculating;
Described crowd behaviour characteristic sequence extraction module, be used in human body foreground object that sequence of video images extracted, and the related data that changes of the length and width that calculate the variation of area change, foreground area mass centre of the connected region of this foreground object and boundary rectangle, set up the three-dimensional feature sequence vector;
In detecting car during the above situation of two people or two people; in described background modeling and the calculating of human body foreground object extraction module, extracted the sequential binary map of human body prospect; these data have obtained characteristic vector sequence after treatment, the behavior detection that can be used in the modeling of setting up hidden Markov model or pass through hidden Markov model.The characteristic vector here comprises by the length and width variation of the variation of crowd's foreground area, crowd's prospect boundary rectangle, the center of gravity set of variations of crowd's prospect becomes the three-dimensional feature vector data;
1) the variation AC of crowd's foreground area
The computational methods of the variation AC of crowd's foreground area are provided by formula (1), and the size of this value is represented the size of the variable quantity of foreground pixel quantity, can reflect the degree of human motion fierceness;
In the formula: Ap is the area of the prospect human object that obtained of former frame image, and An is the area of the prospect human object that current frame image obtained.
2) length and width of crowd's prospect boundary rectangle change WHC
The length and width of crowd's prospect boundary rectangle change WHC and are provided by formula (2), and the size of this value is represented the size of the length and width variable quantity of crowd's prospect boundary rectangle, can reflect the degree that human body attitude changed when act of violence took place;
In the formula: Hp, Wp are respectively the length of boundary rectangle of previous image frame prospect human object and wide, and Hn, Wn are the length of boundary rectangle of prospect human object of current image frame and wide.
3) center of gravity of crowd's prospect changes CC
The center of gravity of crowd's prospect changes CC and is provided by formula (3), and the size of this value is represented the size of the center of gravity variable quantity of crowd's prospect, can reflect that human body moved the degree of variation when act of violence took place;
CC＝||C
p-C
n||(3)
In the formula: Cp be previous image frame prospect human object position of centre of gravity, Cn is the position of centre of gravity of the prospect human object of current image frame.
The MBM of described hidden Markov model, be used for each three-dimensional feature sequence vector is converted to a concrete symbol, it is observed value, by obtaining the characteristic vector data of the normal behaviour in a large amount of elevators, by the K-Means algorithm full feature vector data collection is carried out cluster then, obtained the code book set.At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set.These observed values are formed observation sequence, by the parameter of the best HMM model of Baum-Welch algorithm acquisition, set up the HMM model of normal behaviour;
Because the hidden Markov model of subsequent treatment needs a concrete symbolic feature to come the representation feature vector, therefore each characteristic vector need be converted to a concrete symbol that is to say observed value.At this moment can realize by the method for using vector quantization or cluster; In fact way is by certain methods the characteristic vector training data to be divided into K classification, makes the similarity of this K classification minimum, and the central point of this K classification is formed the code book set.In case training is finished, and has obtained the code book set, just can will be arbitrarily characteristic vector transfer index value with its nearest code book in code book is gathered to, this index value conduct is an observed value.Algorithm commonly used at present has LBG vector quantization algorithm, with reference to list of references [4] with reference to list of references [5]; What adopt among the present invention is the K-Means clustering algorithm;
The K-Means clustering algorithm is accepted input variable k; Then n data object is divided into k cluster so that make the cluster that is obtained satisfy, the object similarity in the same cluster is higher, and the object similarity in the different cluster is less.The cluster similarity is to utilize the average of object in each cluster to obtain one " center object ", and promptly center of attraction is calculated;
The roughly flow process of this algorithm is as follows:
(1) select k object as initial cluster center arbitrarily from n data object;
(2) circulation (3) to (4) till each cluster no longer changes;
(3), calculate the distance of each object and these center object, and again corresponding object is divided according to minimum range according to the average (center object) of each cluster object;
(4) recomputate the average (center object) of each (changing) cluster.
Hidden Markov model below is represented with HMM, it be a kind of with parametric representation, be used to describe the probabilistic model of statistics of random processes characteristic.A typical hidden Markov model as shown in Figure 4;
HMM can be designated as λ=(N, M, π, A, B),
The HMM model specifically can be described by following parameters:
(1) N: the state number in the model.Remember N state S
1..., S
N
(2) M: the possible observed value number of each state correspondence.Remember that M observed value is O
1..., O
M
(3) π: initial condition probability vector, π=(π
1..., π
N) wherein, π
i=P (q
1=S
i), 1≤i≤N, and satisfy π
i〉=0,
It is used to describe observation sequence when initial belongs to different state when t=1 probability.
(4) A is a state transition probability matrix, A=(a
Ij)
N * NWherein A is a state transition probability matrix, A=(a
Ij)
N * NWherein, a
Ij=p (q
T+1=S
j| q
t=S
i), 1≤i, j≤N, and,
It is used to describe the probability distribution that is transformed to the state j of time t+1 from the state i at time t.
(5) B is the observed value probability matrix, B=(b
Ji)
N * MWherein, b
j(i)=p (o
t=O
i| q
t=S
j)=p (O
i| S
j) 1≤j≤N, 1≤i≤M, and
It is used to be described in time t state S
jSituation under to produce observed value be O
iProbability.
Realize that by HMM the act of violence detection in the elevator must solve three basic problems.
(a) computational problem.For given observed value sequence o=O
1O
2... O
NAnd specific HMM model parameter λ=(N, M, π, how A B), calculates the probability that this specific HMM model produces characteristic sequence o.This problem is very important, can pass through Forward-backward[6] the algorithm solution, list of references [6]:.
(b) identification problem.For given observed value sequence o=O
1O
2... O
N, how to select best status switch S so that this sequence of best explanation.This can pass through Viberti-Decoding[6] the algorithm solution.
(c) training problem, how the adjustment model parameter lambda makes P (o| λ) probability maximum.This can lead to Baum-Welch algorithm [6] and solve.
But realization list of references [6] about Forward-backward algorithm, Viberti-Decoding algorithm and Baum-Welch algorithm; After having solved above-mentioned three problems, just can come act of violence is detected by the HMM modeling;
Further, it is simple relatively, easy for normal behavioral data is obtained, and abnormal behaviours such as violence have the consideration of the infinite property enumerated, the present invention comes act of violence is detected by the normal behaviour in the elevator is carried out the HMM modeling, what select for use is that from left to right two condition shifts HMM, as shown in Figure 4;
The use of HMM model needed through training and two stages of detection.Promptly set up the HMM model stage in the training stage.At first, obtain the characteristic vector data of the normal behaviour in a large amount of elevators, by the K-Means algorithm full feature vector data collection is carried out cluster then, obtained the code book set.At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set.These observed values are formed observation sequence, obtain the parameter of best HMM model by the Baum-Welch algorithm, thereby set up the HMM model of normal behaviour.Concrete modeling process as shown in Figure 5;
The identification module of described act of violence, be used to distinguish normal behaviour or abnormal behaviour, go out the probability of observed value sequence by the Forward-backward algorithm computation, if its output probability of observed value sequence is very high, think then that greater than pre-set threshold it is a normal behaviour sequence.Otherwise,, then think an abnormal behaviour sequence if the output probability of an observation sequence is lower than setting threshold; Concrete detection identifying as shown in Figure 6;
As shown in Figure 6, the probability of observation sequence Model Calculation thus obtains, and the process of identification behavior is exactly by whether judging the output probability value less than a minimum threshold, if just this behavior sequence is classified as abnormal behaviour less than a minimum threshold;
Described minimum threshold generally can obtain by the training study of normal sequence, usually by observing the probable value of exporting in the training stage, the establishing method of minimum threshold is to deduct and organize 5% of probit range δ that normal sequence exports more one fixing on minimum value in the probable value that many group normal sequence export;
Accompanying drawing 8 has shown training and testing result, and the recognition methods of as can be seen from the figure adopting among the present invention to be proposed is to distinguish normal and abnormal behaviour by threshold value is set easily.Filling * point expression abnormal behaviour among the figure, and filling point is represented normal behaviour, as can be seen from the figure can select to select the zone as threshold value in-49～-50 scopes.
Described warning and phonetic warning module, be used for the abnormal behaviour sequence is carried out phonetic warning, system calculates abnormality degree according to the output probability of abnormal behaviour sequence with the departure degree that sets threshold value, send in various degree caveat according to abnormality degree then, when reaching the alarm threshold value that sets, system can notify administrative staff in time to intervene automatically and handle also and automatically elevator be stopped at nearest floor;
Further, here at first need to calculate the abnormality degree of abnormal behaviour, among the present invention the probable value of being exported departed from that minimum threshold is big more thinks that abnormality degree is high more, the abnormal behaviour process in the various elevators is judged to be different result in " affirmation violence ", " severely subnormal ", " unusually ", " propensity to violence ", " should be noted that " etc. 5 respectively; Decision method and corresponding measure are as shown in table 1;
| The judgement scope | Result of determination | Measure |
| Minimum threshold-δ (＞20) % | Confirm serious act of violence | Voice serve a grave warning, and notice related management personnel, and the control elevator is in floor stop recently |
| Minimum threshold-δ (＞15～20) % | Severely subnormal | Voice reminder warning, and notice related management personnel, the control elevator exists |
| Floor is stopped recently | ||
| Minimum threshold-δ (＞10～15) % | Unusually | Voice reminder caution, and notice related management personnel |
| Minimum threshold-δ (＞5～10) % | Propensity to violence | The voice reminder caution |
| Minimum threshold-δ (0～5) % | Should be noted that | Voice reminder is noted |
Table 1 abnormality degree decision table
δ in the table 1 represents the probit range that many group normal sequence are exported.
The IEEE802.11b radio communication that the present invention adopts is as based on the communication technology between the embedded system of act of violence intelligent detection device in the lift car of computer vision and the Surveillance center's computer.Described embedded system is an embedded Linux system, and described Surveillance center computer is employed to be PC or server, and the user program module among the present invention is realized by C and Java language.
Embodiment 2
With reference to Fig. 1-Fig. 8, all the other are identical with embodiment 1, different is wireless video transmission aspect, adopted the video capture processor of video information in this embodiment, be used to carry out the video compression chip of small echo video compression, be used for the dsp chip that calculates the quantization parameter of each field picture in real time and finish some important algorithm, be used for high complexity programmable logic device (CPLD) that video capture processor and dsp chip are carried out logic control, the communication serial port that is used for radio communication, the digital video of gathering is after the compression of video compression chip, by DSP to packing, sent the video data of compression then by radio receiving transmitting module, radio receiving transmitting module meets the communication standard of IEEE802.11b.
The invention effect that the above embodiments 1 and 2 are produced is to have made full use of technology such as increasingly mature wireless video communication calculating, embedded system, computer vision and behavior identification, by the various actions in the car are calculated, various abnormal behaviours can be judged to be 5 kinds of different results such as " affirmation violence ", " severely subnormal ", " unusually ", " propensity to violence ", " should be noted that " respectively.Realized the anti-violence monitoring of remote intelligent of elevator, taken place to steal wealth, stealing and to notify the relevant personnel to take the rescue measure in the very first time when elevator incident of violence such as look takes place; Reduced the crime rate in the elevator, improved elevator user's the sense of security, the crime in the prevention elevator has been had positive effect; Some behaviors person agaainst the law, attempt robber are had beyond the fright effect, some weak persons are also had bigger psychological placebo effect, can play a safeguard protection psychological application especially.
List of references is in this specification:
[1]: Kyungnam Kim, Thanarat H.Chalidabhongse, David Harwood, LarryDavis, Real-time foreground-background segmentation using codebook model[J], Real-time Imaging, 2005,11 (3): 167-256.Kyungnam Kim, Thanarat H.Chalidabhongse, David harwood, Larry Davis; That is, based on the real-time prospect and the background segment [J] of code book model, real time imagery, 2005,11 (3): 167-256;
[2]: A.Senior.Tracking people with probabilistic appearance models[C]. //Proceedings of International Workshop on Performance Evaluation of Tracking andSurveillance.2002.A.Senior.; That is, based on the human body tracking [C] of probability display model. about following the tracks of and the collection of thesis .2002. of international symposium of monitoring performance evaluation;
[3]: C.Stauffer and E.Grimson.Learning patterns of activity using real-timetracking[J] .IEEE Transactions on Pattern Analysis and Machine Intelligence.2000,22 (8): 747-757.C.Stauffer, E.Grimson; That is, based on Activity Type study [J] the .IEEE pattern analysis and the machine intelligence journal .2000 of real-time tracking, 22 (8): 747-757;
[4]: Y.Linde, A.Buzo, R.M.Gray, An Algorithm for Vector Quantizer Design[J] .IEEE Transactions on Communications.1980,28:702-710.Y.Linde, A.Buzo, R.M.Gray; That is, vector quantizer algorithm for design [J] .IEEE communication journal .1980,28:84-95; And K-Means clustering algorithm;
[5]: Nair, V., Clark, J.J.Automated Visual Surveillance Using Hidden MarkovModels[C] Proceeding of the 15th Vision Interface Conference.Calgary, Canada, 2002:88-92.Nair, V., Clark, J.J; That is, monitor [C] based on the automatic vision of HMM. the 15th visual interface meeting.The Calgary, Canada, 2002:88-92.
[6]: L.R.Rabiner, " A Tutorial on Hidden Markov Models and SelectedApplications in Speech Recognition ", Proceedings of the IEEE, Vol.77, No.2, pp.257-286,1989.[6] L.R.Rabiner; That is, " HMM guide and the application in speech recognition " IEEE meeting, Vol.77, No.2, pp.257-286,1989.
Claims (5)
1. one kind based on act of violence intelligent detection device in the lift car of computer vision, comprise the video sensor that is installed in elevator car roof, the embedded system that is used for transmitting video data, be used to monitor the Surveillance center's computer in the lift car, described video sensor is connected with described embedded system, described embedded system is connected with the Surveillance center computer by communication, described Surveillance center computer comprises the car inner video image read module that is used for showing in real time video data in the lift car, described car inner video image read module connects display unit, it is characterized in that:
Described Surveillance center computer comprises the microprocessor that is used for safety precaution in the lift car, and described microprocessor comprises:
Car inner video image read module is used for the video information in the lift car is collected and send to Surveillance center's computer, and Surveillance center's computer real-time reads the video data that sends;
Background modeling and human body foreground object extraction module are used for the video data that is read is extracted the human body foreground object;
Crowd behaviour characteristic sequence extraction module, be used for the human body foreground object that video data extracted to being read, calculate the related data that the length and width of the variation of area change, foreground area center of gravity of the connected region of this human body foreground object and boundary rectangle change, set up the three-dimensional feature sequence vector;
The MBM of hidden Markov model, be used for each three-dimensional feature sequence vector is converted to a concrete symbol, be observed value:, by the K-Means algorithm full feature vector data collection is carried out cluster then and obtain the code book set by obtaining the characteristic vector data of the normal behaviour in a large amount of elevators; At last, transfer all characteristic vectors to can in the HMM model, use observed value by the code book set; Described observed value is formed observation sequence, by the parameter of the best HMM model of Baum-Welch algorithm acquisition, sets up the HMM model of normal behaviour;
The identification module of act of violence, be used to distinguish normal behaviour and abnormal behaviour, go out the probability of observed value sequence by the Forward-backward algorithm computation, for an observed value sequence, if its output probability is very high, think then that greater than pre-set threshold it is a normal behaviour sequence; Otherwise,, then think an abnormal behaviour sequence if the output probability of an observation sequence is lower than setting threshold;
Alarm module is used for sending alarm command to alarm device when being identified as the abnormal behaviour sequence, and notifies administrative staff in time to intervene and handle.
2. as claimed in claim 1 based on act of violence intelligent detection device in the lift car of computer vision, it is characterized in that: in described crowd behaviour characteristic sequence extraction module, in detecting car during the above situation of two people or two people, in described background modeling and human body foreground object extraction module, extracted the sequential binary map of human body foreground object, this sequential binary map has obtained the three-dimensional feature sequence vector after treatment, and the behavior that is used to set up the modeling of hidden Markov model or pass through hidden Markov model detects; Described three-dimensional feature sequence vector comprises by the length and width variation of the variation of crowd's foreground area, crowd's prospect boundary rectangle, the center of gravity set of variations of crowd's prospect becomes the three-dimensional feature vector data, specifically has:
1) the variation AC of crowd's foreground area:
The computational methods of the variation AC of crowd's foreground area are provided by formula (1), and the size of this value is represented the size of the variable quantity of foreground pixel quantity, can reflect the degree of human motion fierceness;
In the formula: Ap is the area of the human body foreground object obtained of former frame image, and An is the area of the human body foreground object that current frame image obtained;
2) length and width of crowd's prospect boundary rectangle change WHC:
The length and width of crowd's prospect boundary rectangle change WHC and are provided by formula (2), and the size of this value is represented the size of the length and width variable quantity of crowd's prospect boundary rectangle, can reflect the degree that human body attitude changed when act of violence took place;
In the formula: Hp, Wp are respectively the length of boundary rectangle of previous image frame human body foreground object and wide, and Hn, Wn are the length of boundary rectangle of human body foreground object of current image frame and wide;
3) center of gravity of crowd's prospect changes CC:
The center of gravity of crowd's prospect changes CC and is provided by formula (3), and the size of this value is represented the size of the center of gravity variable quantity of crowd's prospect, can reflect that human body moved the degree of variation when act of violence took place;
CC＝‖C
p-C
n‖ (3)
In the formula: Cp be previous image frame human body foreground object position of centre of gravity, Cn is the position of centre of gravity of the human body foreground object of current image frame.
3. as claimed in claim 1 based on act of violence intelligent detection device in the lift car of computer vision, it is characterized in that: in the MBM of described hidden Markov model, with probabilistic model parametric representation, that be used to describe the statistics of random processes characteristic:
HMM be designated as λ=(N, M, π, A, B),
The HMM model is specifically described by following parameters:
1) N: the state number in the model, remember N state S
1..., S
N
2) M: the possible observed value number of each state correspondence, remember that M observed value is O
1..., O
M
3) π: initial condition probability vector, π=(π
1..., π
N) wherein, π
i=P (q
1=S
i), 1≤i≤N, and satisfy π
i〉=0,
It is used to describe observation sequence when initial belongs to different state when t=1 probability;
4) A is state transition probability matrix A=(a
Ij)
N * N, wherein, a
Ij=p (q
T+1=S
j| q
t=S
i), 1≤i, j≤N, and,
It is used to describe the probability distribution that is transformed to the state j of time t+1 from the state i at time t;
5) B is the observed value probability matrix, B=(b
Ji)
N * MWherein, b
j(i)=p (o
t=O
i| q
t=S
j)=p (O
i| S
j) 1≤j≤N, 1≤i≤M, and
It is used to be described in time t state S
jSituation under to produce observed value be O
iProbability;
Realize that by HMM the abnormal behaviour detection in the elevator needs to determine following three basic problems;
(a) computational problem: for given observed value sequence o=O
1O
2... O
NAnd specific HMM model parameter λ=(N, M, π, A B), calculates the probability that this specific HMM model produces characteristic sequence, solves by the Forward-backward algorithm;
(b) identification problem: for given observed value sequence o=O
1O
2... O
N, how to select best status switch so that this sequence of best explanation solves by the Viberti-Decoding algorithm;
(c) training problem: make P (o| λ) probability greatest problem for the adjustment model parameter lambda, logical Baum-Welch algorithm solves.
4. as claimed in claim 1 based on act of violence intelligent detection device in the lift car of computer vision, it is characterized in that: described video sensor connects video processor, and described video processor comprises:
The image logging modle is used for the video information recording in the lift car is got off;
Image processing module, the video data that is used for noting carry out compressed encoding, multiplexing and be modulated into compressed video data;
First radio receiving transmitting module is used for according to communication standard, sends the video data of compression;
Described Surveillance center computer comprises:
Second radio receiving transmitting module is used for according to communication standard, receives the video data of compression;
The image decompression processing module, the data that are used for receiving decompress, demultiplexing and demodulation, revert to video data;
The output of described image decompression processing module connects car inner video image read module.
5. as claimed in claim 4 based on act of violence intelligent detection device in the lift car of computer vision, it is characterized in that: described first radio receiving transmitting module is the wireless network card that meets communication standard, described Surveillance center computer comprises ICP/IP protocol and the radio network interface that cooperates with described wireless network card, and described microprocessor is the built-in Linux microprocessor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100989198A CN101557506B (en) | 2009-05-19 | 2009-05-19 | Intelligent detecting device for violent behavior in elevator car based on computer vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2009100989198A CN101557506B (en) | 2009-05-19 | 2009-05-19 | Intelligent detecting device for violent behavior in elevator car based on computer vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101557506A CN101557506A (en) | 2009-10-14 |
| CN101557506B true CN101557506B (en) | 2011-02-09 |
Family
ID=41175421
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2009100989198A Expired - Fee Related CN101557506B (en) | 2009-05-19 | 2009-05-19 | Intelligent detecting device for violent behavior in elevator car based on computer vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101557506B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111453576A (en) * | 2020-04-21 | 2020-07-28 | 通力股份公司 | Monitoring system and elevator |
Families Citing this family (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10255374B2 (en) * | 2011-03-11 | 2019-04-09 | Gregory A. Piccionelli | Aggregation of live performances on an aggregate site on a network |
| CN101834982B (en) * | 2010-05-28 | 2012-04-25 | 上海交通大学 | Hierarchical screening method of violent videos based on multiplex mode |
| CN102457705A (en) * | 2010-10-19 | 2012-05-16 | 由田新技股份有限公司 | Method and system for detecting and monitoring fight behavior |
| CN102164270A (en) * | 2011-01-24 | 2011-08-24 | 浙江工业大学 | Intelligent video monitoring method and system capable of exploring abnormal events |
| CN102756960A (en) * | 2011-04-28 | 2012-10-31 | 肖磊 | Photoinduction alarming system for elevator |
| CN103024292A (en) * | 2011-09-20 | 2013-04-03 | 佳都新太科技股份有限公司 | Pre-background separation algorithm based on dynamic interaction |
| EP2826020A4 (en) | 2012-03-15 | 2016-06-15 | Behavioral Recognition Sys Inc | Alert volume normalization in a video surveillance system |
| CN102740121B (en) * | 2012-07-11 | 2015-12-09 | 赛特斯信息科技股份有限公司 | Be applied to video quality diagnostic control system and the method for video surveillance network |
| CN103049746B (en) * | 2012-12-30 | 2015-07-29 | 信帧电子技术（北京）有限公司 | Detection based on face recognition is fought the method for behavior |
| CN103136899B (en) * | 2013-01-23 | 2016-01-20 | 宁凯 | Based on the intelligent alarm method for supervising of Kinect somatosensory device |
| CN106006266B (en) * | 2016-06-28 | 2019-01-25 | 西安特种设备检验检测院 | A kind of machine vision method for building up applied to elevator safety monitoring |
| CN107665326B (en) * | 2016-07-29 | 2024-02-09 | 奥的斯电梯公司 | Monitoring system for passenger conveyor, passenger conveyor and monitoring method thereof |
| CN107886103B (en) * | 2016-09-29 | 2023-12-08 | 日本电气株式会社 | Method, device and system for identifying behavior patterns |
| CN108319960A (en) * | 2018-01-09 | 2018-07-24 | 和芯星通（上海）科技有限公司 | Activity recognition system and method, equipment and storage medium based on probability graph model |
| CN108462708B (en) * | 2018-03-16 | 2020-12-08 | 西安电子科技大学 | HDP-HMM-based behavior sequence detection method |
| CN109019213A (en) * | 2018-09-10 | 2018-12-18 | 宾爵电梯（中国）有限公司 | A kind of personal safety early warning elevator implementation method and its system |
| CN109052093A (en) * | 2018-09-29 | 2018-12-21 | 浙江云安达物联信息技术有限公司 | A kind of method and system of the improper boarding behaviorist risk of early warning |
| CN109712361A (en) * | 2019-01-14 | 2019-05-03 | 余海军 | Real-time anti-violence opens platform |
| CN109785214A (en) * | 2019-03-01 | 2019-05-21 | 宝能汽车有限公司 | Safety alarming method and device based on car networking |
| CN111126176A (en) * | 2019-12-05 | 2020-05-08 | 山东浪潮人工智能研究院有限公司 | Monitoring and analyzing system and method for specific environment |
| CN111339945B (en) * | 2020-02-26 | 2023-03-31 | 贵州安防工程技术研究中心有限公司 | Video-based people group and scatter inspection method and system |
| US11544931B2 (en) * | 2020-05-26 | 2023-01-03 | Otis Elevator Company | Machine learning based human activity detection and classification in first and third person videos |
| CN113011290A (en) * | 2021-03-03 | 2021-06-22 | 上海商汤智能科技有限公司 | Event detection method and device, electronic equipment and storage medium |
| CN114524339B (en) * | 2022-01-06 | 2024-02-09 | 广东博智林机器人有限公司 | Method, device, equipment and storage medium for detecting safe operation of elevator car |
| CN116629250B (en) * | 2023-04-28 | 2024-01-12 | 华南师范大学 | Violent vocabulary analysis method, system, device and medium |
| CN116778573A (en) * | 2023-05-24 | 2023-09-19 | 深圳市旗扬特种装备技术工程有限公司 | Violence behavior detection method and device, electronic equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1889681A (en) * | 2005-07-29 | 2007-01-03 | 上海新时达电气有限公司 | Elevator long-distance monitorng device based on TCP/IP technology |
| CN1921618A (en) * | 2006-09-20 | 2007-02-28 | 浙江工业大学 | Elevator video monitoring device based on full digital wireless communication |
| CN1919711A (en) * | 2006-09-20 | 2007-02-28 | 浙江工业大学 | Elevator inner violence-proof apparatus based on image and speech recognition technique |
-
2009
- 2009-05-19 CN CN2009100989198A patent/CN101557506B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1889681A (en) * | 2005-07-29 | 2007-01-03 | 上海新时达电气有限公司 | Elevator long-distance monitorng device based on TCP/IP technology |
| CN1921618A (en) * | 2006-09-20 | 2007-02-28 | 浙江工业大学 | Elevator video monitoring device based on full digital wireless communication |
| CN1919711A (en) * | 2006-09-20 | 2007-02-28 | 浙江工业大学 | Elevator inner violence-proof apparatus based on image and speech recognition technique |
Non-Patent Citations (1)
| Title |
|---|
| JP特开2006-327746A 2006.12.07 |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111453576A (en) * | 2020-04-21 | 2020-07-28 | 通力股份公司 | Monitoring system and elevator |
| CN111453576B (en) * | 2020-04-21 | 2022-01-04 | 通力股份公司 | Monitoring system and elevator |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101557506A (en) | 2009-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101557506B (en) | Intelligent detecting device for violent behavior in elevator car based on computer vision | |
| JP7241776B2 (en) | audio/video device with viewer | |
| CN103839373B (en) | A kind of unexpected abnormality event Intelligent Recognition alarm device and warning system | |
| US10593174B1 (en) | Automatic setup mode after disconnect from a network | |
| CN202257856U (en) | Driver fatigue-driving monitoring device | |
| US11115629B1 (en) | Confirming package delivery using audio/video recording and communication devices | |
| CN105700363A (en) | Method and system for waking up smart home equipment voice control device | |
| CN106241534B (en) | More people's boarding abnormal movement intelligent control methods | |
| US11341825B1 (en) | Implementing deterrent protocols in response to detected security events | |
| CN202130312U (en) | Driver fatigue driving monitoring device | |
| CN106006266A (en) | Machine vision establishment method applied to elevator safety monitoring | |
| US11935297B2 (en) | Item monitoring for doorbell cameras | |
| US11164435B1 (en) | Audio/video recording and communication doorbell devices with supercapacitors | |
| CN103632479A (en) | Home-based care for the aged security and protection system based on human motion trajectory analysis | |
| US10733857B1 (en) | Automatic alteration of the storage duration of a video | |
| US10939120B1 (en) | Video upload in limited bandwidth | |
| US10943442B1 (en) | Customized notifications based on device characteristics | |
| US11032128B2 (en) | Using a local hub device as a substitute for an unavailable backend device | |
| CN111666780A (en) | Intelligent door control security method based on emotion recognition technology | |
| US20210374967A1 (en) | Precipitation removal from video | |
| CN202217340U (en) | Intelligent monitor device used for POS machine | |
| US10863146B1 (en) | Setup and configuration of audio/video recording and communication devices | |
| CN113420631A (en) | Safety alarm method and device based on image recognition | |
| CN103530605A (en) | Method for detecting abnormal behavior of robustness | |
| US11163097B1 (en) | Detection and correction of optical filter position in a camera device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20110209 Termination date: 20130519 |