EP2064666A1

EP2064666A1 - Method for monitoring congregations

Info

Publication number: EP2064666A1
Application number: EP07803514A
Authority: EP
Inventors: Wolfram Klein; Gerta KÖSTER
Original assignee: Siemens AG
Current assignee: Siemens AG
Priority date: 2006-09-20
Filing date: 2007-09-17
Publication date: 2009-06-03
Also published as: DE102006044298B3; WO2008034780A1

Abstract

A method for monitoring congregations in a monitoring area (1) repeats the steps of an image recording apparatus recording a picture of the congregation, an evaluation apparatus detecting the position of a plurality of people on the picture, and comparison with the positions on a previous picture. A direction and speed are then respectively attributed by the evaluation apparatus. The resultant positions, directions of movement and speeds of movement of the people are compared with comparison patterns for a plurality of people and this makes an assessment for an event which has occurred or a future event.

Description

description

Procedure for monitoring the collection of persons

The invention relates to a method for monitoring the accumulation of persons according to the preamble of claim 1.

For safety reasons and for statistical and economic ^¬ Liche purposes, there is a need, the number of people at different venues, eg at a

Volksfest such as the Oktoberfest. One goal here is the identification of critical masses, i. the identification of dangers from people's agglomerations and critical flows of persons as well as a prognosis of their time course in order to be able to take appropriate measures.

Two methods are known for the detection, in particular counting, and the surveil ^¬ monitoring of persons accumulations.

First, the counting of the people who arrive at the entrances and exits as well as go away. This can be done manually or automatically, but is error prone and only allows statements about the number of people in the entire observation area, not in subareas.

On the other hand by two-dimensional images and counting the people on the images. There are automated procedures for counting. However, can take place no Erfas ^¬ measurement and evaluation of movements of persons and the surveying of pedestrian flows.

From DE 200 15 559 Ul an arrangement for detecting streams of people is known in which a sensor module detects the location coordinates of individual persons in a room to be monitored from entering to leaving. As a sensor module can serve a CCD camera. In an embodiment of the arrangement described there, means and methods or algorithms of the image processing are used to detect or identify the objects and subjects located in a room. From gray or color values and circumscribing shapes, the movement behavior of persons is derived according to position and speed.

Next 200 15 559 Ul is known from DE to divide the space into meh ^¬ eral areas and capture the persons each in these areas.

A disadvantage of this prior art is that thereby no recognition of dangerous incidents or potentially developing dangerous situations is possible.

It is therefore an object of the present invention to provide a method for, with the monitoring of persons accumulations is possible to the effect that entstande ^¬ ne or potentially dangerous conditions are detected.

This object is achieved by a method for monitoring persons with the features of claim 1. Advantageous developments of the method according to the invention are specified in the subclaims.

By a method having the features of claim 1, an evaluation and thus prediction of the development over a limited period is possible. Potential dangerous situations, such as critical masses, overcrowding, accident patterns, can thus be better anticipated and measures can be initiated more quickly. The forecasts allow a forward-looking STEU ^¬ ern of pedestrian flows. A control of the crowd is possible or more efficient. Likewise, the evaluation by the evaluation, such as a PC automatically, so that automation of security precautions is possible, such as the closing of a crowded subway station. Finally, better statistical statements for Economic Zwe ^¬ blocks such as tourism can be won. Conveniently embodiment of the monitoring area is divided into several sub-areas which are respectively associated with its own Ver ^¬ same pattern.

This allows a precise adaptation to the situation in the sub-area, e.g. a ride at a folk festival. Also, a comparison of the states in adjacent subareas allows a quick statement of expected flows of people without detailed consideration.

The monitoring area can not open to the imaging device, concealed subregions, de ^¬ ren additions are in the monitored area, and a review of the hidden sub-regions can based on comparison ^¬ patterns for the inflow and outflow through the entrances SUC ^¬ gen.

In particular, by observing gradients in sub-areas that allow access to the hidden area, e.g. a beer tent form, a statement about the number of people in the hidden sub-area is possible.

In an advantageous embodiment, the comparison patterns for the subregions are dependent on the weather, the season, the time of day, the day of the week and / or the type of event.

In a favorable embodiment, a model for the behavior of the persons in the subregions is created for the evaluation, in which the subareas are assigned parameters depending on the weather, the season, the time of day, the day of the week and / or the type of event.

As a result, a better prognosis can be made for typical situations depending on circumstances such as time or weather. For example, the assessment will be more accurate in the event that many people leave a folk festival in the late evening in bad weather. An accident can be detected by a comparison pattern in which the speed of individual persons within ei ^¬ nes flow of people is significantly lower and these move to a center and / or a circle without other people trained around this.

In an advantageous embodiment, a panic situation is detected by a comparison pattern in which flows of people with high speed of movement, high density of people occur at the same time forming areas without or with a few people.

Advantageously, the gradient of the person density is evaluated.

In a favorable embodiment, the first and second Ablei ^¬ tion of the person density and / or the speed of movement of the people is evaluated.

The number of people in an area or subarea is a function of time and place. By comparing the positions of the persons at different times, the derivative of this function is determined by time or approximated by difference quotients. The derivation according to the location (gradient) provides information about the local changes in the number of persons or persons (number of people / area). Even higher derivatives can be approximated by differential quotients. By extrapolation, for example using the time derivative and possibly higher derivatives of the value of the density of people and / or the speed of movement, better evaluation of the situ ation ^¬ can be delivered in the future. The number of persons and the flows of persons can be predicted.

An evaluation is possible even if not all persons can be recorded. It is also possible to additionally include selected persons in order to obtain information about the distribution of certain groups of persons in the mass of persons. hold. This can be, for example, children or wheelchair users, who need a higher level of protection than other people. Such persons may be recognized by special markings such as a cap or cap, or physical characteristics such as height.

The present invention will be explained in more detail below with the aid of ^exemplary embodiments with reference to the attached drawings. It shows

1 is a schematic plan view of a monitoring area with subregions on the basis of the example of a ^Volksfest ,

FIG. 2 schematically shows an assignment of parameters to subregions, FIG.

3A schematically shows a comparison pattern for an accident,

Fig. 3B schematically shows a second comparison pattern for an accident and

FIGS. 4A-4C schematically show the sequence of a comparison pattern for a panic situation.

1 shows schematically in ^plan view a monitoring range 1 with subregions on the basis of the example of a popular ^event . There are different subareas with different characteristics, such as the behavior of the subcategory, subunit function, commercial interest or safety requirements. So there are street areas 2, a ride 3, here eg a roller coaster, a beer tent 4, an entrance beer tent 5, an entrance area outside 6, an entrance area inside 7 and an entrance area ride 8. The persons are indicated by dots. In the case of a flow of persons in the direction of the beer tent 4, the gradient of the streams of people outside in the sub-area entrance beer tent 5 can be used as a guideline for determining the number of persons in the subarea beer tent 4. The monitoring of the movement pattern allows an estimation of a development in a subsequent period.

FIG. 2 schematically shows an assignment of parameters to subregions. In addition to a road area, the driving business 3, the beer tent 4, a roast chicken stand 9 and a Rie ^¬ senrad 10 are arranged. For each sub-area typical properties are taken into account in the comparative samples. Under section z, as in areas that have normally gleichmä ^¬ FLOWING persons distribution as the road section 2, only sparse with people under areas such as the roller coaster as

Ride 3 and the Ferris wheel, as well as urban areas such as the beer tent 4, which can be at the same time in an aerial view a ver ^¬ covered sub-area.

By time derivative and the gradients, in particular at the subrange margins, it can be determined whether the number of persons in the subrange increases or decreases and the speed of this change.

Experience has shown that the ratio of Perso ^¬ nen on the road range 2 to people in the beer tent 3 is different from the sunshine, at night and at meal times, unlike in the afternoon, for example, when it rains.

A concrete assessment and prognosis is described ^¬ model to allow for different scenarios, a Prog ^¬ nose of the distribution of people in sub-regions, starting from the current distribution people. For each sub-area, different parameters are defined, eg P (weather) and P (time), which are dependent on the respective sub-area and correspond to an "attractiveness" for persons under the parameter condition Scenarios and for the sub-area concerned. The value range of the parameters is specified as [0,1]. For example, the attractiveness parameter P (Weather is rain) = P _{R can be set} high for rainfall in the subarea of beer tent 4 (ie many people crowd into beer tent 3 in rainy weather), but in sunshine it is lower P (weather is sun) = 0.5 , The attractiveness parameter for the subarea beer tent 3 at later time high is ^¬ is also (P (time in the evening) = P _A = 1 at 21 o'clock), while in the period from 23 pm (festival ended) and 10 am the following day on , 0 ^x and set rather low in the late morning hours (P (clock = Hours) = 0.2). The same applies to the setting of the parameters of the other sub-areas: The Ferris wheel 10 is not attractive in the rain: P _R = 0.2, but relatively attractive in the sun or at night. The roast chicken stand 9 is attractive at lunchtime and in the evening. FIG. 2 shows by way of example a parameterization with values P _A for a-bends and at the same time P _R for rain.

If known typical values of the number of persons for the patterned ^¬ de prevailing external conditions and appropriate parameters to the observed actual values are compared, striking values can be automatically detected.

Fig. 3A schematically shows a comparative pattern for an accident. The parties are shown as a point with an arrow on its movement direction and a movement speed ent ^¬ speaking arrow length.

In the present comparative example, persons flow around a non-moving person. Other people move relatively quickly to this person and then surround them very closely. This comparison pattern can be interpreted as falling of a person. Even a steep gradient in the middle of a road area could indicate, for example, a traffic jam in front of a fallen person. Fig. 3B schematically shows a second comparison pattern for a possible accident. Around a few people, a circle without further people forms and the speed of the people in this circle is very low or zero.

FIGS. 4A-4C schematically show the timing of a comparison pattern for a panic situation. Persons flee from a specific location with a high gradient value, ie densely crowded persons and unusual movement directions. In addition, a high value of the temporal derivative, ie high movement ^speeds and clear differences between two temporally successive scenarios relative to the previous scenarios in FIG the timing of Figs. 4A to 4C.

Advantageously, with the method according to the invention, critical masses, traps and accident patterns can either be recognized directly or predicted via extrapolation, eg a linear extrapolation of the movements. This can be used to respond appropriately, such as automatically triggering alerts, informing and enabling security services. Next, the monitoring can automatically focuses on the danger zone, where refined ^¬ to.

It is also advantageous thresholds as for the people ^¬ tight to specify the average person speed and other data at the undershooting or overshooting warnings are given. The threshold values are specified depending on which parameters (weather, time of day, etc.) are available. The threshold values can be determined depending on a sub-range. For example, in a subway station at rush hour, higher person densities are tolerable than at 3 o'clock in the morning, when an accumulation of people may indicate a danger situation, such as a developing brawl. The method according to the invention represents a control loop, ie the information obtained can be used directly for better detection in the next steps. When a crisis area detected automatically, for example by a threshold value, so in this region, the detection can be verfei ^¬ nert. Options include zooming in or special alignment of additional cameras. Next, additional data can be determined as to the density of the BEWE ^¬ transfer speed of persons as well as the uniformity and distribution of the speed of movement.

Based on this information, the reference patterns, e.g. based on experience in the past, and the parameters are adjusted. Possible measures may be to block inputs and to only use them as an output, as well as further monitoring based on comparison patterns and parameters for the case of evacuation.

Claims

claims

1. A method for monitoring crowds in a surveillance area (1) with the repetitive steps

Capturing an image of the person collection by an image capture device,

Detecting the position of a plurality of persons on the image by an evaluation device,

Compare with the positions on a previous map and respectively assign a direction and speed by the evaluation device

characterized in that the resulting positions, directions of movement and movement speeds of the persons are compared with comparison patterns for a plurality of persons and thereby an assessment for a occurred or future event ^¬ nis takes place.

2. The method according to claim 2, characterized in that the monitoring area is divided into a plurality of subregions (2, 3, 4, 6, 7, 8, 9, 10), to each of which own comparison patterns are assigned.

3. The method according to claim 1 or 2, characterized in that the monitoring area (1) not accessible for the image pickup device, hidden sub-areas (4) whose accesses (5) are in the surveillance area (1), and that a rating for the hidden Subareas (4) based on comparison patterns for the inflow and outflow over the entrances (5).

4. The method according to claim 2 or 3, characterized in that the comparison patterns for the subregions (2,3,4,5,6,7,8,9,10) are dependent on boundary conditions, such as the weather, the season, the Time of day, the day of the week and / or the type of event.

5. Method according to one of claims 2 to 4, characterized in that a model for the behavior of the persons in the subregions (2, 3, 4, 9, 10) is created for the evaluation, in which the subareas parameters (P _A , P _R ) depending on the weather, the season, the time of day, the day of the week and / or the type of event.

6. The method according to any one of the preceding claims, characterized in that an accident is detected by a comparison pattern in which the speed of individual persons within a flow of people is much lower and they move to a center and / or a circle without other people to train them.

7. The method according to any one of the preceding claims, characterized in that a panic situation is detected by a comparison pattern, occur in the streams of people with high Bewegungsgeschwindig ^¬ speed, high density person at the same time forming areas with little or no persons or with people who are stationary are.

8. The method according to any one of the preceding claims, characterized in that the gradient of the person density is evaluated.

9. Method according to one of the preceding claims, characterized in that that the second derivative of the density of people and / or Be ^¬ motion speed of the person is evaluated.

10. The method according to any one of the preceding claims, characterized in that not all but only selected persons are recorded on the Ab ^¬ education and thus an evaluation based on a random sample takes place.