CN104463121A - Crowd density information obtaining method - Google Patents

Crowd density information obtaining method Download PDF

Info

Publication number
CN104463121A
CN104463121A CN201410743783.2A CN201410743783A CN104463121A CN 104463121 A CN104463121 A CN 104463121A CN 201410743783 A CN201410743783 A CN 201410743783A CN 104463121 A CN104463121 A CN 104463121A
Authority
CN
China
Prior art keywords
crowd
density
crowd density
image
current
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.)
Pending
Application number
CN201410743783.2A
Other languages
Chinese (zh)
Inventor
靳薇
李济朝
李彬
曲寒冰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
BEIJING NEW TECHNOLOGY APPLICATION INST
Original Assignee
BEIJING NEW TECHNOLOGY APPLICATION INST
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by BEIJING NEW TECHNOLOGY APPLICATION INST filed Critical BEIJING NEW TECHNOLOGY APPLICATION INST
Priority to CN201410743783.2A priority Critical patent/CN104463121A/en
Publication of CN104463121A publication Critical patent/CN104463121A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands

Landscapes

  • Engineering & Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a crowd density information obtaining method. The crowd density information obtaining method includes the steps of extracting a current crowd sampling image from current crowd collecting images; obtaining a crowd density sampling value according to the current crowd sampling image; conducting texture-based method density analysis or pixel-based method crowd density analysis to obtain current crowd density information according to judgment results of the crowd density sampling value and a set threshold value; generating crowd density alarm information according to the current crowd density information. The problem that the existing crowd density accuracy is low is accordingly solved. The condition that the density analysis accuracy is reduced due to the fact that an analysis method is improper is avoided. In this way, the crowd density collecting accuracy is improved, the crowd pre-storage cost is reduced, and the information effectiveness is improved.

Description

Crowd density information getting method
Technical field
The present invention relates to image procossing and mode identification technology, particularly relate to crowd density information getting method.
Background technology
Along with the development of society, occur that the increasing stream of people converges place, immense pressure has been caused to the public infrastructure of large size city, as railway station, subway station, bus stop and gymnasium etc. often can welcome the peak traffic of short-term.Crowd is highly crowded easily tramples accident, and there is very large potential safety hazard, it is of common occurrence at every country that a lot of crowd tramples accident.Such as: during the fireworks display celebration that on Dec 31st, 2012, Cote d'lvoire held a celebration New Year, occur seriously to trample accident, cause at least 61 people dead; On January 14th, 2011 there is great tread event when celebrating church festival in India, at least causes 100 people dead; Within 2013, accident is trampled because of crowded by primary school of Hubei Province one, causes 4 student's death.In order to tackle these accidents, many public arenas have all been installed supervisory system and have been monitored the stream of people in recent years.But traditional supervisory system is all monitored by the artificial different scenes to closed-circuit television, this method has subjectivity, can not quantitative test, and monitor staff is easily tired, cause ignoring the emergency situations on monitor, thus cause irremediable consequence.Therefore the crowd density appraisal procedure based on intelligent video is most important for avoiding major safety problems.
Summary of the invention
For above-mentioned defect of the prior art, the object of this invention is to provide crowd density information getting method, solve the problem that existing crowd density accuracy is low.
The invention discloses crowd density information getting method, comprising:
Step S101, gathers image from current crowd, extracts current crowd's sampled images;
Step S102, obtains crowd density sampled value according to described current crowd's sampled images;
Step S103, judges whether described crowd density sampled value is greater than setting threshold values, if so, then gathers image according to texturing method to described current crowd and carry out density analysis, obtain current crowd density information; If not, then according to pixel method, image is gathered to described current crowd and carry out crowd density analysis, obtain current crowd density information;
Step S104, generates crowd density warning message according to described current crowd density information.
As one of the present invention preferred embodiment, described step S101 comprises:
Gather the geomorphologic map of image from current crowd and isolate, the first geomorphic province and the second geomorphic province; In described first geomorphic province and described second geomorphic province, carry out image acquisition respectively, obtain when ex-first lady mine massively collection image and the second crowd gather image; Mine massively collection image as ex-first lady from described, extract when ex-first lady mines massively sampled images, gather image from described current second crowd, extract when the second forefathers mine massively sampled images.
As one of the present invention preferred embodiment, described step S102 comprises:
To mine massively sampled images as ex-first lady according to described, obtain the first population density sampled value, according to described current second crowd's sampled images, obtain the second crowd density sampled value.
As one of the present invention preferred embodiment, described step S103 comprises:
Judge whether described the first population density sampled value is greater than setting threshold values, if so, then according to texturing method to described when ex-first lady mine massively collection image carry out crowd density analysis, obtain the first population density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain the first population density information;
Judge whether described second crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current second crowd and carry out crowd density analysis, obtain the second crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain the second crowd density information
The density map that current crowd gathers image is generated according to the geomorphologic map that described the first population density information, the second crowd density information and described current crowd gather image.
As one of the present invention preferred embodiment, described step S104 comprises:
Gather density map and the setting warning density thresholds of image according to described crowd, generate the warning message in each region in described density map, generate crowd density warning figure according to this warning message.
As one of the present invention preferred embodiment, also comprise in described step S101:
Gather the geomorphologic map of image from current crowd, extract key area;
The convergence region be connected with this landforms section is extracted according to described emphasis landforms section;
In described key area and described convergence region, carry out image acquisition respectively, obtain current key area crowd and gather image and current convergence region crowd gathers image; Gather image from described current key area crowd, extract current key area crowd's sampled images, gather image from described current convergence crowd, extract when convergence crowd sampled images.
As one of the present invention preferred embodiment, described step S102 comprises:
According to described current key area crowd's sampled images, obtain key area crowd density sampled value, according to described convergence crowd sampled images, obtain and converge region crowd density sampled value.
As one of the present invention preferred embodiment, described step S103 comprises:
Judge whether described key area crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current key area crowd and carry out crowd density analysis, obtain key area crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain key area crowd density information;
Judge whether described convergence region crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current convergence region crowd and carry out crowd density analysis, obtain and converge region crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain and converge region crowd density information.
As one of the present invention preferred embodiment, described step S104 comprises:
Judge whether described key area crowd density information exceedes crowd density alarming threshold value, if so, then generate crowd density warning message according to described key area crowd density information; If not, then according to people's Flow Velocity and distance, key area crowd density information, the convergence region crowd density information in described current convergence region, and crowd density alarming threshold value, obtain warning density delay time;
Crowd density warning message is generated according to described warning density delay time.
As one of the present invention preferred embodiment, described step S104 comprises:
The density map that current crowd gathers image is generated according to the geomorphologic map that described key area crowd density information, convergence region crowd density information and described current crowd gather image.
Beneficial effect of the present invention is, by the judgement to current crowd density sampled value, by diverse ways, current crowd density is analyzed, thus obtain crowd density information more accurately, avoid situation that is improper due to analytical approach and the density analysis precise decreasing caused.Thus improve the accuracy that crowd density gathers, the cost that the crowd of reducing prestores also improves the validity of information.
Accompanying drawing explanation
In order to be illustrated more clearly in the embodiment of the present invention or technical scheme of the prior art, be briefly described to the accompanying drawing used required in embodiment or description of the prior art below, apparently, accompanying drawing in the following describes is only some embodiments of the present invention, for those of ordinary skill in the art, under the prerequisite not paying creative work, other accompanying drawing can also be obtained according to these accompanying drawings.
Fig. 1 is in one embodiment of the present invention, the treatment scheme schematic diagram of crowd density information getting method;
Fig. 2 is in one embodiment of the present invention, the change schematic diagram of the different angles of the contrast that four different directions are corresponding;
Fig. 3 is in one embodiment of the present invention, the change schematic diagram of the different angles of the correlativity that four different directions are corresponding;
Fig. 4 is in one embodiment of the present invention, the change schematic diagram of the different angles of the energy that four different directions are corresponding;
Fig. 5 is in one embodiment of the present invention, the change schematic diagram of the different angles of the unfavourable balance square that four different directions are corresponding;
Fig. 6 is in one embodiment of the present invention, the treatment scheme schematic diagram of crowd density information getting method.
Embodiment
Below in conjunction with accompanying drawing of the present invention, be clearly and completely described technical scheme of the present invention, obviously, described embodiment is only the present invention's part embodiment, instead of whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art, not making the every other embodiment obtained under creative work prerequisite, belong to the scope of protection of the invention.
At a kind of crowd density information getting method of the present invention, as shown in Figure 1, comprising:
Step S101, gathers image from current crowd, extracts current crowd's sampled images.
In this step, by the crowd is dense, shooting, photographic goods that place installs gather current crowd's image.Simultaneously, as the region such as park or mountain region, the personnel of different landforms converge situation difference, and as the gate place in park, personnel converge can be comparatively concentrated, also can according to different geomorphic types, existing landforms are divided into, high density geomorphic province, i.e. the first geomorphic province, and low-density geomorphic province, i.e. the second geomorphic province.Specifically can gather the geomorphologic map of image from current crowd and isolate, the first geomorphic province and the second geomorphic province.In the first geomorphic province and the second geomorphic province, carry out image acquisition respectively respectively, obtain when ex-first lady mine massively collection image and the second crowd gather image.From mining massively as ex-first lady collection image, extract when ex-first lady mines massively sampled images, i.e. crowd's image of high density region; Gather image from current second crowd, extract when the second forefathers mine massively sampled images, i.e. crowd's image of low-density region.Above-mentioned Gather and input image graph as be from video sequence extract single width still image, also may be video sequence.Meanwhile, in this step, also can carry out pre-service to collection image, pre-service generally comprises illumination pretreatment, shields for the template background of single width still image, for the foreground extraction etc. of video sequence.The impact of illumination variation can be reduced by histogram equalization; Template background shielding template masks most of complex background, only retains area-of-interest; Background subtracts the region in order to obtain moving target object by reducing background.
This patent will adopt mixed Gaussian (Mixture of Gaussians, MoG) background extracting is carried out, mixed Gaussian background modeling method is proposed in traffic monitoring scene by Friedman etc. at first, and Stauffer etc. proposed a kind of more general mixed Gaussian background modeling method afterwards [35].Mixed Gauss model can describe the multi-mode states of pixel, can tackle the change of background that relative complex, illumination slowly change or there is repeating motion by a small margin, and can modeling comparatively accurately, is thus subject to applying comparatively widely.
Mixed Gauss model is that K the Gauss model of each pixel in background image is represented the state that this pixel is different within certain period.Suppose X tfor certain pixel is at the gray-scale value of moment t, then its probability density function is as formula 3-1:
P ( X t ) = Σ i = 1 K ω i , t * η ( X , μ i , t , σ i , t 2 ) - - - ( 3 - 1 )
Wherein, K represents the quantity of Gauss model, and general value is the integer between 3 to 5; ω i, tfor the weights of t i-th Gauss model; for the probability density function of t i-th Gaussian distribution, wherein μ i, tfor average, for covariance.
Mixture Gaussian background model is utilized to extract the process of background:
(1) parameter upgrades
Mixture Gaussian background model belongs to typical parametrization background model, therefore must have the process that parameter upgrades.Mixed Gauss model is mainly determined by average and variance two parameters, needs to take the study that different mechanism realizes average and variance.In practical application, when carrying out modeling to follow-up frame of video, each pixel needs carry out mating with K Gauss model and determine whether upgrade, as formula 3-2:
ω i , t = ( 1 - α ) ω i , t - 1 + α ( M i , t ) μ i , t = ( 1 - ρ ) μ i , t - 1 + ρ X t σ i , t 2 = ( 1 - ρ ) σ i , t - 1 2 + ρ ( X t - μ i , t ) T ( X t - μ i , t ) ρ = αη ( X t , μ i , t , σ i , t 2 ) - - - ( 3 - 2 )
Wherein, ρ is parameter turnover rate, and α is right value update rate.
As pixel value X twhen mating with one of them Gaussian distribution, M i, tvalue is 1, otherwise value is 0.If pixel value X in mixed Gauss model tdo not mate with any Gaussian distribution, so by the Gaussian distribution that a new Gaussian distribution replacement priority is minimum, then use larger variance and less weight initialization.
(2) background estimating
After having upgraded parameter, need to determine the best Gaussian distribution describing background of energy in current video frame, the standard at this moment as evaluation and test is exactly the relative value ω calculated i, t/ σ i, t, this value sorts by K Gauss model from big to small, and the Gaussian distribution that most probable describes context process is located in the top of sequence, and the distribution produced by the disturbance of instability will come the bottom of sequence, finally be replaced by newly-generated Gaussian distribution.Front B the model meeting formula 3-3 is just required background.
B = arg min b { Σ k = 1 b ω k > T } - - - ( 3 - 3 )
Wherein T represents weight threshold.
(3) foreground extraction
During foreground extraction, according to the background model that front B Gaussian distribution is set up, according to priority size order by pixel value X trespectively with B Gaussian distribution comparison one by one.If there is no the Gaussian distribution characterizing background model matches with it, then judge that this point is as foreground pixel point, otherwise be background pixel point.
Still there is a lot of noise in background extracting rear video frame, use medium filtering process to make prospect more complete clear here.After all pre-service complete, prospect bianry image comparatively clearly can be obtained, calculate the ratio that crowd's foreground pixel area accounts for image total pixel number, be considered as high-density scene when exceeding default threshold value, otherwise be then considered as low-density scene.
Step S102, obtains crowd density sampled value according to current crowd's sampled images.
In this step, also can according to sampled images (i.e. crowd's image of high density region) of mining massively as ex-first lady, obtain the first population density sampled value, according to current second crowd's sampled images (i.e. crowd's image of low-density region), obtain the second crowd density sampled value.Above-mentioned crowd density collection value and crowd characteristic value, these features contain the key message of crowd density, and the validity of feature is directly connected to the accuracy of density Estimation result.
Step S103, obtains current crowd density information.
In this step, judge whether crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to current crowd and carry out density analysis, be i.e. crowd density estimation under high density, obtain current crowd density information; If not, then according to pixel method, image is gathered to current crowd and carry out crowd density analysis, obtain current crowd density information, i.e. crowd density estimation under low-density.
Also can comprise in this step:
Judge whether the first population density sampled value is greater than setting threshold values, if so, then according to texturing method, crowd density analysis is carried out to collection image of mining massively as ex-first lady, obtain the first population density information; If not, then according to pixel method, crowd density analysis is carried out to current crowd's image, obtain the first population density information;
Judge whether the second crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to current second crowd and carry out crowd density analysis, obtain the second crowd density information; If not, then according to pixel method, crowd density analysis is carried out to current crowd's image, obtain the second crowd density information
The geomorphologic map gathering image according to the first population density information, the second crowd density information and current crowd generates the density map that current crowd gathers image.
Under above-specified high density, the concrete grammar of crowd density estimation is:
When crowd is in high density, there is more mutually blocking, the method for estimation of Corpus--based Method pixel count not too adapts to.Because crowd's image texture is mostly stochastic pattern texture, obey Statistical Distribution, therefore the crowd density estimation method based on texture analysis has better applicability, texture analysis is the significant surfaces grayscale distribution information utilizing image processing techniques to extract target, and to the technology that these features are analyzed.Many visual signatures of texture have vital role in texture description, and comprise consistance, density, the slightly property made, systematicness, directivity, frequency and phase place etc., therefore textural characteristics has multi-dimensional nature.Gray level co-occurrence matrixes is the one of texture analysis, and these features have locus independence, therefore becomes the key character of texture analysis.Final system is classified utilizing Training Support Vector Machines to Dense crowd.
Gray level co-occurrence matrixes is more traditional texture analysis method, is based on estimating that the probability density function f (i, j/d, θ) of two-dimentional combination condition realizes.Function representation is d apart from initial point distance, and angle is that the gray-scale value at θ place is respectively i, and the gray scale of j is to the probability occurred.Can represent all estimated values with a two-dimensional histogram or matrix, this matrix is exactly gray level co-occurrence matrixes.General a pair (d, θ) corresponding matrix, can by (d, θ) restriction within the specific limits.Usual θ chooses in (0 °, 45 °, 90 °, 135 °) four values.Can derive several statistical parameter in order to describe textural characteristics from this probability matrix, the calculation expression as contrast (Contrast), unfavourable balance square (Homogeneity), energy (Energy) and correlativity (Correlation) is as follows:
Contrast: S c ( d , θ ) = Σ i = 0 L - 1 Σ j = 0 L - 1 ( i - j ) 2 f ( i , j | d , θ )
Unfavourable balance square: S h ( d , θ ) = Σ i = 0 L - 1 Σ j = 0 L - 1 f ( i , j | d , θ ) 1 + ( i - j ) 2
Energy: S g ( d , θ ) = Σ i = 0 L - 1 Σ j = 0 L - 1 f ( i , j | d , θ ) 2
Correlativity: S p ( d , θ ) = Σ i = 0 L - 1 Σ j = 0 L - 1 ijf ( i , j | d , θ ) - μ x μ y σ x 2 σ y 2
Wherein L is the progression of gray-scale value.For number of greyscale levels, need to find a rational numerical value, significantly can reduce calculated amount, the textural characteristics that loss of energy is not too much again, herein in order to computing velocity and the demand of satisfied reality, the gray level of selection is 8.
Gray level co-occurrence matrixes generally has 0 °, 45 °, 90 °, and 135 ° of four directions, different directions can generate different gray level co-occurrence matrixes.Need to find out suitable angle, the object extracting complete textural characteristics can be reached, can calculated amount be reduced again.Herein by experiment on Dense crowd sample, with number of greyscale levels for 8, distance d is the variable of 1 to 50, calculates the eigenwert of the gray level co-occurrence matrixes of four direction, and Fig. 2 ~ 5 are the change schematic diagram of each eigenwert in different angles respectively.By comparative result, finding that direction is 45 °, 90 °, the result basic simlarity of 135 °, therefore in order to reduce computation complexity, determining herein only to calculate the textural characteristics on 45 ° of more representative directions.Therefore selected distance d is 15 herein.
Under above-mentioned low-density, the concrete grammar of crowd density estimation is:
In low-density crowd's environment, crowd's number is less, there is the part of coverage between men hardly, adopts the method based on pixels statistics can be fairly simple, efficient.Choose the number of pixels of statistics foreground moving object area herein, utilize the relation that crowd's number is directly proportional to area number of pixels, adopt the number of mode to crowd of curve to estimate, judge the density rating of crowd.
In a lot of field, generally have interpolation and curve two kinds of methods to carry out data of description, curve is with simple approximation of function complicated function (or unknown function), can fitting data best, and need not through all data points, but the trend of representation of data as far as possible.The method of curve is adopted to describe the relation of low-density crowd density and prospect crowd number of edges herein.
The definition of the least square method of curve [36]for: y=f (x) is for being defined in the function on interval [a, b], for m+1 on interval mutually different point, require at function class in find a function y=S *x (), makes error sum of squares:
| | δ | | 2 = Σ i = 0 m δ i 2 = Σ i = 0 m [ S * ( x i ) - y i ] 2 = min Σ i = 0 m [ S ( x i ) - y i ] 2 - - - ( 3 - 4 )
Wherein:
If fitting function is: y=ax+b, a represent slope, b represents intercept.For the N group data (x of equal precision measurement i, y i), i=1,2 ..., N, y irepresent the number observed, it is considered to accurate; x irepresent the crowd's foreground pixel number calculating gained, all errors are only with x irelevant.The computing method of coefficient a and b in the straight line of derivation least square fitting below.
From the definition of the least square method of curve, observed reading y be made isum of square of deviations be minimum, for the fitting a straight line of equal observation value, make the value of following formula minimum exactly:
| | δ | | 2 = Σ i = 1 N [ y i - ( a + bx i ) ] 2 - - - ( 3 - 5 )
Formula 3-6 asks local derviation to a, b respectively and is arranged and can obtain system of equations:
aΣ x i + bΣ x i 2 = Σ x i y i aN + bΣ x i = Σ y i - - - ( 3 - 6 )
Solve the best estimate that system of equations listed by above formula just can try to achieve straight line parameter a and b with
a ^ = ( Σ x i 2 ) ( Σ y i ) - ( Σ x i ) ( Σ x i y i ) N ( Σ x i 2 ) - ( Σ x i ) 2 b ^ = N ( Σ x i y i ) - ( Σ x i ) ( Σ y i ) N ( Σ x i 2 ) - ( Σ x i ) 2 - - - ( 3 - 7 )
Obtain straight-line equation according to estimation parameters obtained, for the frame of video of reality, calculate prospect profile number of pixels, if it is less than the threshold value judging high/low density crowd, then this number of pixels is substituted into straight-line equation and directly calculate crowd's number.
Step S104, generates crowd density warning message according to current crowd density information.
Also comprise in this step:
Gather density map and the setting warning density thresholds of image according to crowd, generate the warning message in each region in density map, generate crowd density warning figure according to this warning message.
As shown in Figure 6, for ease of to key area namely, early warning is carried out in the region easily had problems, and in another embodiment of the invention, also comprises in step S101:
Step S1011, extracts when convergence crowd sampled images.
In this step, gather the geomorphologic map of image from current crowd, extract key area; The convergence region be connected with this landforms section is extracted according to emphasis landforms section.Key area and converge in region and carry out image acquisition respectively, obtains current key area crowd and gathers image and current convergence region crowd gathers image; Gather image from current key area crowd, extract current key area crowd's sampled images, gather image from current convergence crowd, extract when convergence crowd sampled images.
Comprise in step s 102:
Step S1021, according to current key area crowd's sampled images, obtains key area crowd density sampled value, according to convergence crowd sampled images, obtains and converges region crowd density sampled value.
Comprise in step s 103:
Step S1031, obtains key area crowd density information and obtains and converge region crowd density information.
This step comprises: judge whether key area crowd density sampled value is greater than setting threshold values, if so, then gather image according to texturing method to current key area crowd and carry out crowd density analysis, obtains key area crowd density information; If not, then according to pixel method, crowd density analysis is carried out to current crowd's image, obtain key area crowd density information;
Judge that whether converge region crowd density sampled value is greater than setting threshold values, if so, then gathers image according to texturing method to current convergence region crowd and carries out crowd density analysis, obtain and converge region crowd density information; If not, then according to pixel method, crowd density analysis is carried out to current crowd's image, obtain and converge region crowd density information.
Comprise in step S104:
Step S1041, generates crowd density warning message.
Judge whether key area crowd density information exceedes crowd density alarming threshold value, if so, then generate crowd density warning message according to key area crowd density information; If not, then according to people's Flow Velocity and distance, key area crowd density information, the convergence region crowd density information in current convergence region, and crowd density alarming threshold value, obtain warning density delay time; Crowd density warning message is generated according to warning density delay time.
Also comprise in this step simultaneously:
Step S1042, generates the density map that current crowd gathers image.
The geomorphologic map gathering image according to key area crowd density information, convergence region crowd density information and current crowd generates the density map that current crowd gathers image.
The above; be only the specific embodiment of the present invention, but protection scope of the present invention is not limited thereto, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; change can be expected easily or replace, all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should described be as the criterion with the protection domain of claim.

Claims (10)

1. crowd density information getting method, is characterized in that, comprising:
Step S101, gathers image from current crowd, extracts current crowd's sampled images;
Step S102, obtains crowd density sampled value according to described current crowd's sampled images;
Step S103, judges whether described crowd density sampled value is greater than setting threshold values, if so, then gathers image according to texturing method to described current crowd and carry out density analysis, obtain current crowd density information; If not, then according to pixel method, image is gathered to described current crowd and carry out crowd density analysis, obtain current crowd density information;
Step S104, generates crowd density warning message according to described current crowd density information.
2. crowd density information getting method according to claim 1, is characterized in that, described step S101 comprises:
Gather the geomorphologic map of image from current crowd and isolate, the first geomorphic province and the second geomorphic province; In described first geomorphic province and described second geomorphic province, carry out image acquisition respectively, obtain when ex-first lady mine massively collection image and the second crowd gather image; Mine massively collection image as ex-first lady from described, extract when ex-first lady mines massively sampled images, gather image from described current second crowd, extract when the second forefathers mine massively sampled images.
3. crowd density information getting method according to claim 2, is characterized in that, described step S102 comprises:
To mine massively sampled images as ex-first lady according to described, obtain the first population density sampled value, according to described current second crowd's sampled images, obtain the second crowd density sampled value.
4. crowd density information getting method according to claim 3, is characterized in that, described step S103 comprises:
Judge whether described the first population density sampled value is greater than setting threshold values, if so, then according to texturing method to described when ex-first lady mine massively collection image carry out crowd density analysis, obtain the first population density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain the first population density information;
Judge whether described second crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current second crowd and carry out crowd density analysis, obtain the second crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain the second crowd density information
The density map that current crowd gathers image is generated according to the geomorphologic map that described the first population density information, the second crowd density information and described current crowd gather image.
5. crowd density information getting method according to claim 4, is characterized in that, described step S104 comprises:
Gather density map and the setting warning density thresholds of image according to described crowd, generate the warning message in each region in described density map, generate crowd density warning figure according to this warning message.
6. crowd density information getting method according to claim 1, is characterized in that, also comprises in described step S101:
Gather the geomorphologic map of image from current crowd, extract key area;
The convergence region be connected with this landforms section is extracted according to described emphasis landforms section;
In described key area and described convergence region, carry out image acquisition respectively, obtain current key area crowd and gather image and current convergence region crowd gathers image; Gather image from described current key area crowd, extract current key area crowd's sampled images, gather image from described current convergence crowd, extract when convergence crowd sampled images.
7. crowd density information getting method according to claim 6, is characterized in that, described step S102 comprises:
According to described current key area crowd's sampled images, obtain key area crowd density sampled value, according to described convergence crowd sampled images, obtain and converge region crowd density sampled value.
8. crowd density information getting method according to claim 7, is characterized in that, described step S103 comprises:
Judge whether described key area crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current key area crowd and carry out crowd density analysis, obtain key area crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain key area crowd density information;
Judge whether described convergence region crowd density sampled value is greater than setting threshold values, if so, then according to texturing method, image is gathered to described current convergence region crowd and carry out crowd density analysis, obtain and converge region crowd density information; If not, then according to pixel method, crowd density analysis is carried out to described current crowd's image, obtain and converge region crowd density information.
9. crowd density information getting method according to claim 8, is characterized in that, described step S104 comprises:
Judge whether described key area crowd density information exceedes crowd density alarming threshold value, if so, then generate crowd density warning message according to described key area crowd density information; If not, then according to people's Flow Velocity and distance, key area crowd density information, the convergence region crowd density information in described current convergence region, and crowd density alarming threshold value, obtain warning density delay time;
Crowd density warning message is generated according to described warning density delay time.
10. crowd density information getting method according to claim 8, is characterized in that, described step S104 comprises:
The density map that current crowd gathers image is generated according to the geomorphologic map that described key area crowd density information, convergence region crowd density information and described current crowd gather image.
CN201410743783.2A 2014-12-08 2014-12-08 Crowd density information obtaining method Pending CN104463121A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201410743783.2A CN104463121A (en) 2014-12-08 2014-12-08 Crowd density information obtaining method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201410743783.2A CN104463121A (en) 2014-12-08 2014-12-08 Crowd density information obtaining method

Publications (1)

Publication Number Publication Date
CN104463121A true CN104463121A (en) 2015-03-25

Family

ID=52909137

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201410743783.2A Pending CN104463121A (en) 2014-12-08 2014-12-08 Crowd density information obtaining method

Country Status (1)

Country Link
CN (1) CN104463121A (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106096521A (en) * 2016-06-02 2016-11-09 苏州大学 A kind of swarm and jostlement method for early warning based on stress and strain model and device
CN106845344A (en) * 2016-12-15 2017-06-13 重庆凯泽科技股份有限公司 Demographics' method and device
CN106910229A (en) * 2015-12-23 2017-06-30 韩华泰科株式会社 Image processing equipment and method
CN107016696A (en) * 2017-03-31 2017-08-04 广州地理研究所 A kind of passenger flow density detection method and device
CN107025450A (en) * 2017-04-25 2017-08-08 广东兆邦智能科技有限公司 Thermal map generation method
CN107203760A (en) * 2017-06-09 2017-09-26 中国联合网络通信集团有限公司 Crowd density monitoring method and device
CN110287929A (en) * 2019-07-01 2019-09-27 腾讯科技(深圳)有限公司 The quantity of target determines method, apparatus, equipment and storage medium in group region
CN111540162A (en) * 2020-04-17 2020-08-14 佛山科学技术学院 Pedestrian flow early warning system based on raspberry group

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103164711A (en) * 2013-02-25 2013-06-19 昆山南邮智能科技有限公司 Regional people stream density estimation method based on pixels and support vector machine (SVM)

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103164711A (en) * 2013-02-25 2013-06-19 昆山南邮智能科技有限公司 Regional people stream density estimation method based on pixels and support vector machine (SVM)

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ZHE WANG等: ""Crowd density estimation based on local binary pattern co-occurrence matrix"", 《MULTIMEDIA AND EXPO WORKSHOPS (ICMEW),2012 IEEE INTERNATIONAL CONFERENCE ON》 *
谢鹏程: ""复杂场景下实时监控中人群密度估计的研究与实现"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 *

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106910229A (en) * 2015-12-23 2017-06-30 韩华泰科株式会社 Image processing equipment and method
CN106096521A (en) * 2016-06-02 2016-11-09 苏州大学 A kind of swarm and jostlement method for early warning based on stress and strain model and device
CN106845344A (en) * 2016-12-15 2017-06-13 重庆凯泽科技股份有限公司 Demographics' method and device
CN106845344B (en) * 2016-12-15 2019-10-25 重庆凯泽科技股份有限公司 Demographics' method and device
CN107016696A (en) * 2017-03-31 2017-08-04 广州地理研究所 A kind of passenger flow density detection method and device
CN107025450A (en) * 2017-04-25 2017-08-08 广东兆邦智能科技有限公司 Thermal map generation method
CN107025450B (en) * 2017-04-25 2020-01-07 广东兆邦智能科技有限公司 Heat map generation method
CN107203760A (en) * 2017-06-09 2017-09-26 中国联合网络通信集团有限公司 Crowd density monitoring method and device
CN110287929A (en) * 2019-07-01 2019-09-27 腾讯科技(深圳)有限公司 The quantity of target determines method, apparatus, equipment and storage medium in group region
CN110287929B (en) * 2019-07-01 2023-09-05 腾讯科技(深圳)有限公司 Method, device, equipment and storage medium for determining number of targets in group area
CN111540162A (en) * 2020-04-17 2020-08-14 佛山科学技术学院 Pedestrian flow early warning system based on raspberry group

Similar Documents

Publication Publication Date Title
CN104463121A (en) Crowd density information obtaining method
CN103632158B (en) Forest fire prevention monitor method and forest fire prevention monitor system
CN103839065B (en) Extraction method for dynamic crowd gathering characteristics
CN100545867C (en) Aerial shooting traffic video frequency vehicle rapid checking method
CN102542289B (en) Pedestrian volume statistical method based on plurality of Gaussian counting models
CN103530893B (en) Based on the foreground detection method of background subtraction and movable information under camera shake scene
CN105279772B (en) A kind of trackability method of discrimination of infrared sequence image
CN105608691A (en) High-resolution SAR image individual building extraction method
CN104091348A (en) Multi-target tracking method integrating obvious characteristics and block division templates
CN104268851A (en) ATM self-service business hall behavior analysis method based on depth information
CN105138982A (en) Crowd abnormity detection and evaluation method based on multi-characteristic cluster and classification
CN103164711A (en) Regional people stream density estimation method based on pixels and support vector machine (SVM)
CN103729854A (en) Tensor-model-based infrared dim target detecting method
CN108921076B (en) Pavement crack disease self-adaptive constant false alarm detection method based on image
CN105046683A (en) Object detection method based on adaptive-parameter-adjustment Gaussian mixture model
CN104183142A (en) Traffic flow statistics method based on image visual processing technology
CN105868734A (en) Power transmission line large-scale construction vehicle recognition method based on BOW image representation model
CN103810703A (en) Picture processing based tunnel video moving object detection method
CN110110675A (en) A kind of wavelet field of combination of edge information divides shape infrared cirrus detection method
CN104933738A (en) Visual saliency map generation method based on local structure detection and contrast
CN110443830B (en) Precipitation cloud cluster detection method based on Gaussian mixture background model
CN103871062A (en) Lunar surface rock detection method based on super-pixel description
CN107808524A (en) A kind of intersection vehicle checking method based on unmanned plane
CN105184317A (en) License plate character segmentation method based on SVM classification
CN104835142A (en) Vehicle queuing length detection method based on texture features

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20150325

RJ01 Rejection of invention patent application after publication