EP2254104B1 - Method for automatic recognition of a change in a situation - Google Patents

Description

The invention relates to a method for automatic detection of a situation change within a spatial area to be monitored by means of sensors. The invention also relates to a device for this purpose.

Frequently, from the point of view of security, the sometimes very complex requirement is made to monitor places or areas in which many actors or objects move and interact with each other by means of corresponding sensors. Examples of such places or areas are streets, parks, railway stations, public squares, airports, highways or shopping malls. Within these locations or areas, the actors or objects attached to each other interact with each other, which may be, for example, pedestrians, cyclists, cars, ships, aircraft, athletes or customers. In this case, surveillance is particularly geared in particular to ensuring that the objects or actors located in the area move within the specified boundary conditions, so that atypical or even threatening events can be quickly detected by the respective supervisors (eg air traffic controllers, security services),

The most commonly used sensors for monitoring are cameras that record the area to be monitored. The captured image of a camera is then forwarded in most cases to a central, where it is then displayed on a screen for an operator. Especially in areas with a very large spatial extent it comes very quickly to a large number of required cameras to capture and cover all areas of the area to be monitored. For the operator, there is the problem that with an increase in the number of cameras and the number of images that need to be monitored, are increased. Ultimately, this leads to increased complexity, which in the worst case is at the expense of safety.

On the one hand, in order to reduce this monitoring complexity, there is the possibility of reducing the number of images to be monitored, or generally speaking, the number of information required for monitoring, which can ultimately lead to the loss of information. On the other hand, there is the possibility of analyzing the image signals with the aid of image recognition methods and filtering them on to suitably programmed signatures.

For example, describes the US 2003/0107650 A1 a monitoring and security system which automatically issues a warning when corresponding events that were previously programmed occur in the monitoring of an area. For example, this system can alert you to possible shoplifting if a person opens a plastic bag inside the area to be monitored (here a shop). For this purpose, the system is set up in such a way that it picks up the noise within the store and examines it for the characteristic acoustic pattern of an opening plastic bag.

The disadvantage here is in particular the fact that this system recognizes only those events to which it has been specially programmed. So this system is not easily transferable to another scenery or environment.

Surveillance also plays an important role in traffic control and traffic situation detection. Thus, the traffic density is constantly detected on German motorways by means of appropriate sensors, so as. B. traffic guidance devices, as they are installed on the A2 motorway to be able to control depending on the traffic density.

To this topic is from the EP 0 798 684 A1 a method and a system for traffic position detection by stationary data acquisition devices known, in which the data collection devices forward their data to a central, in which they are then evaluated accordingly. The result of the evaluation is then sent back to the data collection devices, so that they are enabled to pay attention only to those events that ultimately have to lead to a message to the control center. This is intended in particular to ensure that the data to be transmitted from the data acquisition device to the center is kept as low as possible.

A disadvantage of this system as well as the other known from the prior art systems, the fact that signatures of the events to be detected must be set by the operator. In addition, in the weeks following installation, often additional maintenance runs are required to re-adjust the system through changes in the observed area, weather changes, seasons, and so on. These systems are usually only able to detect exactly defined processes within their range.

So z. B. in the US 2004/0130620 Al discloses a method for video analysis, in which a plurality of image sensors record an overlapping area, the objects within the surveillance area moving objects can be tracked by means of a corresponding tracking algorithm, even across the recording limits of an image sensor.

Also the documents EP 1074958 . US 2006/0074546 and US 5091780 belong to the state of the art.

In view of the disadvantages known from the prior art, it is an object of the present invention to provide an improved method for detecting events within a spatial area to be monitored.

1. a) Erfassen von Werten mittels der Sensoren hinsichtlich mindestens einer Eigenschaft von Objekten, die sich innerhalb des zu überwachenden räumlichen Gebietes befinden, und deren Felder, innerhalb dessen der jeweilige Wert der Eigenschaft erfasst wurde,
2. b) Lernen zumindest eines Verhaltens der Objekte bezüglich mindestens einer der Objekteigenschaften anhand einer statistischen Auswertung der ermittelten Werte für mindestens ein Feld, und
3. c) Erkennen einer Situationsänderung in Abhängigkeit eines Vergleichs zwischen dem erlernten Verhalten mindestens eines Feldes und zumindest einem aktuell ermitteltem Wert des mindestens einen Feldes bezüglich mindestens einer der Objekteigenschaften.

The object is achieved with the method of the aforementioned type, wherein the spatial area to be monitored is subdivided into a plurality of fields, with the following steps:

1. a) detecting values by means of the sensors with regard to at least one property of objects which are located within the spatial area to be monitored and their fields within which the respective value of the property was detected,
2. b) learning at least one behavior of the objects with regard to at least one of the object properties on the basis of a statistical evaluation of the determined values for at least one field, and
3. c) detecting a situation change in dependence on a comparison between the learned behavior of at least one field and at least one currently determined value of the at least one field with respect to at least one of the object properties.

This makes it possible to monitor an area with corresponding sensors without having to program it specifically for certain events it is to recognize. For this purpose, using the sensors monitoring the area, values are acquired from the objects which are located within the spatial area to be monitored. The recorded values are values of corresponding properties of the objects which can be detected or determined or measured with the aid of the sensors. Such an object property could be z. As the speed, direction or in general terms, the activity of an object within a corresponding field. But it is also conceivable that further information about the objects can be determined as properties, such. As the fuel supply of an aircraft, depending on what kind of sensors is used.

In addition, when acquiring the values, it is determined within which field the corresponding value was acquired, so that a data pair results from a value of an object property and associated field of the spatial area. In the next step, a behavior of the objects is then determined for at least one of the fields at least one of its object properties is learned by evaluating the values of the object property stored in the field by means of statistical methods. This results in a corresponding behavior of the objects with respect to the monitored property of the objects for each field of the entire spatial area.

Thus, for example, when monitoring a road for each field, a behavior in terms of speed can be learned what z. B. expressed in terms of average speed. In order to be able to detect a change in the observed situation, which ultimately results from a spontaneous behavioral change of one or more objects within the overall scene, the learned behavior of at least one field is compared with currently determined values of the corresponding field, whereby based on the comparison on a situation change is closed.

Of course, not only a behavior regarding a property and a field can be learned, but also several behaviors per field and property. Such constellations occur e.g. in crossing areas where e.g. There are two accumulation points for speed, once for vehicles and once for pedestrians. The currently determined values are then compared with both the one and the other behavior (step c)), wherein a change of situation is assumed if the current values can not be subsumed under any of the behaviors.

So recognizes the above method z. B. a sudden change in the average speed of the vehicles on the road, so that thus the formation of a jam or the resolution of a jam is detected. The advantage of this method is that by the division of the spatial area in a plurality of fields local situation changes within a large spatial area are readily detectable and the integration of multiple sensors that record the spatial area from different perspectives, is easily possible. In addition, in this method, no special programming on corresponding events to be detected is necessary.

Advantageously, the behavior in step b) is learned by the last ascertained values of the corresponding field, which lie within a certain period of time. So it is z. B. conceivable that only those values are taken into account when learning the behavior that within the last z. For example, five minutes were recorded. This makes it possible for the system to re-learn the changed situation after a certain time and to view it as typical behavior.

However, it is also conceivable that a behavior is learned for a plurality of discrete periods, which is the basis for the comparison for recognizing a situation change in step c). It is conceivable that a behavior of a period that lies farther back in the comparison is given a lower weighting than those periods that were not so long ago, so that young periods receive a higher weighting than older ones.

However, it is also particularly advantageous for the respective behavior of the majority of the time periods to be combined to form a common behavior, in which case older periods can also be given a different weighting than younger ones. By combining different learned behaviors of the most varied time periods, a typical behavior of the objects within the to be observed scene to be observed. The differently learned behavior z. B. be merged by means of averaging.

At this point it should be mentioned that per field and time period several behaviors can be learned in accordance with the object properties. Thus it is conceivable that a behavior is learned for an object property 1 and at the same time for an object property 2, the z. B. is recorded by other sensors, also a behavior is learned, so that for each object property a correspondingly learned behavior results. These behaviors learned per object property can then be viewed individually as well as merged.

It is advantageous if the fields are masked before the comparison in step c), d. H. Only those fields are used for comparison, for which corresponding values are stored at all, whereby all those fields for which no or only a very small number of values are stored are ignored.

It is particularly advantageous if, in addition to the detection of the values, a position indication is determined which represents the corresponding spatial position of the object within the spatial area. This so-called determination position, which indicates the position at which the value was determined within the spatial area at the corresponding object, is then used to determine the field with which the determined value is linked.

In order for several sensors to be able to cover overlapping areas of the area to be monitored, it is particularly advantageous if the above-described determination position is converted into a uniform position coordinate system or uniform world coordinates so that the position information of the determined values can be compared with one another. In this way, the values can be entered in the corresponding fields regardless of the sensor position.

In a specific embodiment, it is particularly advantageous if the detection of a situation change in step c) takes place as a function of the comparison between a learned behavior from step b) and a behavior based on the statistical evaluation of currently determined values. Here, a current behavior for one or more fields is continuously determined and compared with a previously learned behavior for the corresponding fields. This makes it possible to detect changes in the situation that affect the behavior of the objects within one or more fields, with respect to one or more object properties.

In another embodiment, it is particularly advantageous if a situation change is detected as a function of the comparison between the learned behavior from step b) and current values of a trajectory of a specific object. In this case, values relating to one or more object properties of a specific object are detected, while the object moves on a specific path (trajectory) through the area to be monitored. These values recorded in this way are then assigned to a specific trajectory of the object and compared with the learned behavior of the other objects. If the behavior with respect to one or more object properties of the specific object deviates significantly compared with the learned behavior of the other objects, it is possible to conclude that the situation has changed. Thus, individual objects can be identified, which behave atypically compared to the typical behavior of the entire scene, which also represents a situation change.

Advantageously, mean values, sums, products and / or standard deviations are formed for the statistical evaluation, which can then be compared with one another. In principle, all arithmetic operations are possible which the expert would use for statistical purposes. Particularly suitable sensors for monitoring the spatial area are imaging sensors which are advantageously equipped with image recognition methods, radar sensors or RFID transponders. As objects to be monitored come in particular people, animals, ie living beings in general, road, rail, air or water vehicles into consideration as well as particles and other objects, It is particularly advantageous if the sensors mentioned for determining properties such as activity of Objects that are set to speed the objects or their direction.

mit S_{i, j} als Schwellenwert in dem Feld (i, j), mit V_{obj (i, j)} als das erlernte Verhalten bezüglich einer Eigenschaft der Objekte im Feld (i, j), mit k_empf als eine Empfindlichkeitsfaktor, der für die gesamte Szene bzw. Karte gilt und mit σ_{i, j} als Standardabweichung in dem Feld (i, j) bezüglich der erlernten Objekteigenschaft. Durch k kann in der gesamten Karte eine Empfindlichkeit eingestellt werden, die dann aber in jedem Feld zu einem anderen Ergebnis in Abhängigkeit von σ führt. So kann in örtlich nahe beieinander liegenden Feldern in einem Feld eine Situationsänderung erkannt werden, in benachbarten Feldern jedoch nicht, weil hier beispielsweise die Standardabweichung σ sehr hoch ist (z.B. in Kreuzungsgebieten). Somit kann der Schwellwert für jede Zelle einzeln und dynamisch geregelt werden, was das Verfahren noch genauer macht.Moreover, it is particularly advantageous if the comparison in step c) by means of a threshold comparison takes place so that not even the smallest changes lead to the recognition of a situation change. Thus, the threshold value can be designed in such a way that sudden changes in the behavior of the objects are recognized as a situation change. Thus, for each field or each cell, a threshold value can be determined, which is then stored for each field, whereby the threshold value can for example result from the product of the standard deviation of the respective field with a sensitivity factor as follows: $${\mathrm{S}}_{\mathrm{i}.\mathrm{j}}={\mathrm{V}}_{\mathrm{obj}\left(\mathrm{i}\mathrm{j}\right)}\pm {\mathrm{k}}_{\mathrm{Rec}}*{\mathrm{\sigma }}_{\mathrm{i}.\mathrm{j}}$$

with S _{i, j} as a threshold in the field (i, j), with V _{obj (i, j)} as the learned behavior with respect to a property of the objects in the field (i, j), with k as a sensitivity _factor for the entire scene or map holds and with σ _{i, j} as the standard deviation in the field (i, j) with respect to the learned object property. Through k, a sensitivity can be set in the entire map, which then leads to a different result in each field as a function of σ. Thus, a situation change can be detected in fields close to one another in a field, but not in adjacent fields, because here, for example, the standard deviation σ is very high (eg in intersections). Thus, the threshold for each cell can be controlled individually and dynamically, making the process even more accurate.

In this case, several threshold values can also be stored for a field if several typical behaviors arise with regard to an object property.

Furthermore, the invention relates to a computer program product for carrying out the above method and a device for this purpose.

Fig. 1a, 1b

skizzenhafte Darstellung einer aufgenommenen Szene;

Fig. 2

Grundschema der Datenstruktur;

Fig. 3

Grundschema von abgeleiteten Informationen;

Fig. 4

schematische Darstellung der Situationserkennung über mehrere Zeitbereiche;

Fig. 5

skizzenhafte Darstellung eines Trajektorienvergleiches.

The invention will be explained by way of example with reference to the accompanying drawings. Show it:

Fig. 1a, 1b

sketchy representation of a recorded scene;

Fig. 2

Basic scheme of the data structure;

Fig. 3

Basic scheme of derived information;

Fig. 4

schematic representation of the situation detection over several time ranges;

Fig. 5

sketchy representation of a trajectory comparison.

The Fig. 1a and 1b sketchy show the inclusion of a road junction, which is to be used as an embodiment of the further embodiments. The intersection is monitored by two video cameras, where Fig. 1a the image of the first video camera can be seen from a first angle while Fig. 1b See the image of a second video camera from a different angle.

By means of appropriate software, which is not the subject of the present inventive method, values of specific object properties are identified from the recorded image information. In these embodiments, the objects are vehicles passing through the intersection. The speed and direction are recognized by the image recognition software for each vehicle and stored accordingly. Thus, per vehicle, which is in the field of view of the cameras, continuously determined both the speed and the direction.

As in Fig. 1a and 1b By way of example, in the upper area of the recorded images, the recorded image area is subdivided into a plurality of fields 1. In Fig. 1a and 1b However, this is only sketchily indicated and usually extends over the entire scene. When determining the speed and direction values for each object, the position at which the corresponding values were determined is also determined. So that these values can also be utilized independently of the location of the corresponding camera, this position information is converted into a world coordinate system and stored accordingly. Thus arise for a vehicle, which by the field of view of the two cameras in Fig. 1a and 1b a plurality of speed and direction values at different positions or a plurality of speed and direction values at the same positions, since the values can be detected by a plurality of different sensors.

Each speed and direction value is now assigned the corresponding field 1 within which the determined position lies, based on its position. Thus, exactly one such field can be assigned to each measured speed and direction value. This is schematically in Fig. 2 shown. In other words, there will be at least one velocity and direction value for that object in all those fields which are affected by the object while moving through the spatial area. It is not important to identify the object itself. In Fig. 2 the hatched fields 2 are the fields in which object values of different objects have been stored. The white fields, however, have no stored values.

This arrangement results in a card-like data structure from which further card-like data structures can then be determined with the aid of statistical evaluations, from which the behavior of the objects within the scene can then be derived.

Such a derived card-like structure is in Fig. 3 to recognize. In the Fig. 3 This is a data structure in which the mean value of the values stored in this field was formed in each field. In this embodiment, this was done with the object property "Speed". But it can also derive other card-like structures, such. For example, those in which the standard deviation is determined as a statistical evaluation for each field.

First of all, the values are determined for each field whose property is to be derived statistically. In this embodiment, these would be the speed values. Then the statistical evaluation is carried out for each field. Since the statistical evaluation is carried out separately for each field, this results in different fields for each field Fig. 3 represented by a different hatching in the individual fields.

By such a statistical evaluation, as in the Fig. 3 is shown, the behavior of the objects can be determined with respect to certain object properties within the scene, whereby the learned behavior is understood as a typical behavior of the objects. Now changes the scene by z. If, for example, the speed becomes slower or faster, this ultimately also results in the statistical evaluation so that a situation change can then be concluded on the basis of a comparison between the learned behavior and the current values. Thus, such a situation change is ultimately always a deviation of the behavior of the objects from the typical behavior of the objects within this scene with respect to at least one object property.

For example, suppose the average speed is between 50 and 60 km / h, depending on the field. If there is a traffic jam, naturally the speed of the vehicles decreases drastically, which is detected by the sensors. Compared with the learned behavior, which indicates a mean speed between 50 and 60 km / h as typical behavior, is a speed of 10 or 20 km / h of an object within this scene atypical, so that a situation change can be inferred. In addition, in this case, the statistical evaluation with respect to the standard deviation would change abruptly. Thus, an atypical behavior of the objects can be concluded very quickly without the system first having to know what to look for. The congestion can be detected solely from the change in the standard deviation, since Stop & Go generates a much larger standard deviation than flowing traffic, and this is already noticeable in relatively small or weak changes.

Fig. 4 shows by way of example the comparison between the learned behavior and a current behavior with respect to the two object properties speed (S) and direction (D) on the basis of the statistical evaluation "averaging" (Sμ, Dμ).

Over three past discrete time periods t _-3 , t _-2 and t _-1 , map-type data structures have been learned as described above. For this purpose, data structures with respect to the speed and direction of the objects within the observed scene were learned. From the determined values for the speed S and the direction D, the average value was then derived for each field, which is referred to as card-like data structure 41 and 42 in FIG Fig. 4 can be seen.

From these three derived card-like data structures in the periods t _-3 to t _-1 then a common behavior was derived, which is also stored in the form of a card-like data structure 43 and 44 respectively. In the formation of a common learned behavior can be done with the help of different weights z. B. younger learned behavior to be considered more than older.

In addition, a current behavior has been learned from currently ascertained values with regard to the speed and the direction, which is reflected in the derived card-like data structures 45 and 46 in FIG Fig. 4 reflected. In all of them With the aid of averaging, the behavior of the objects within the scene was mapped as a statistical evaluation. On the basis of the currently determined behavior 45 or 46, the comparison with the jointly learned behavior 43 or 44 is now carried out. In this exemplary embodiment, the difference is simply formed for this purpose, ie the mean value of a field in the learned behavior is subtracted from the mean value of the corresponding same field in the current behavior, so that a difference map results. However, other possibilities of comparison, in particular of a mathematical nature, are also conceivable. This comparison is carried out both with the object property "Speed" and with the object property "Direction". After subtraction, the individual values of all fields are added up.

As a result of this comparison, a concrete scalar emerges in this embodiment representing the differences between the current and learned behaviors. It is easy to see that the larger the difference between the learned and current behavior, the larger the scalar is. Furthermore, it can be seen from this arrangement that small changes do not significantly affect the overall result, so that it is not immediately on a situation change is concluded, as soon as the values vary locally even minimally.

Furthermore, it can be seen in this arrangement that, if the situation changes, the system is self-learning and it adapts to the new situation. Based on the above example of congestion, this means that although the congestion is initially recognized as a change in situation (eg by means of a large standard deviation), after a certain time the slow speed of the congestion is learned as a typical behavior (the standard deviation becomes smaller again). The system automatically adapts to the given situation without having to readjust it. If the traffic jam clears up again after a while, the average speed increases abruptly, which is again recognized as a change in the situation, as is the behavior of the objects from the typical behavior learned differs. Thus, the system can recognize both the formation of a congestion and its resolution afterwards without any problems and manual intervention and adapts by learning the standard deviation also to expected levels of change.

It is also possible with this method to respond to certain external influences such as weather conditions or failure of certain sensors accordingly, without the need for manual intervention. Falls z. B. a sensor or the sensor is partially covered by a foreign body, this is indeed recognized by the method first as a change in situation, but perceived in the further course as typical. For the later course of the failure or the partial failure of a sensor has no meaning.

It should be noted at this point that both the statistical evaluation "averaging" and the mentioned object properties are not limited to these and should not be understood as limiting.

It is also possible here to work with threshold values which are determined for each behavior and the relative standard deviation. Thus, the average value (mean) in a large scene often varies only slightly, although e.g. has jammed on a roadway, e.g. because the traffic jam initially forms only on one lane and then "spreads" to the other lanes or the jam forms only in one direction, while the opposite lane is free. This can lead to compensatory changes that cancel each other out, so that no change in the situation is detected in the sum.

So it is particularly advantageous if the standard deviation is stored for each field, which is then used for a threshold value comparison, so as to be able to determine selectively for individual fields situation changes. By factoring the standard deviation, a sensitivity can be set so that in some fields situation changes result, since here the standard deviations are relatively low, while in other fields this is recognized as normal behavior due to relatively large standard deviations.

Thus, the method of the invention adjusts the standard deviation in the fields where a large change occurs over time (eg, when a congestion occurs), while the other fields remain sensitive to changes (eg, the oncoming lane where no congestion has occurred ).

Fig. 5 schematically shows another embodiment in which a particular or specific object was tracked by means of a corresponding tracking algorithm. For this purpose, a trajectory 51 is recorded by the object to be observed, which has a plurality of data points. At these data points were then z. B. corresponding values of object properties to be measured and stored. These data points are then compared to the learned behavior of the overall scene to determine if the object behaves in a typical or atypical manner relative to the overall behavior of all objects. This is done at the same time for all objects in the scene individually.

For this purpose, each data point of the trajectory is compared with the underlying card-like data structure, in particular with the fields in which the corresponding data point falls. On the example of Fig. 5 this is shown in a scene with the data point 52. In the process, it is first of all determined whether values for this data point of the object's trajectory are actually stored within the map. Then the corresponding object property, here the speed, is compared with the deposited speeds. So can be z. B. derived from the learned behavior, the average speed and compare with the speed of the concrete data point 52 accordingly. It can be prevented with the help of stability information and weightings that smallest differences are recognized as atypical behavior.

For example, can be with this embodiment in Fig. 5 recognize if z. B, Cyclists move very fast through a scene or pedestrians who run aimlessly or suddenly start to run or walk over otherwise barely used areas like meadows. Also can be seen with this system a ghost driver. But it can also monitor other objects such as aircraft so that when monitoring an airport z. For example, situations may be identified where service vehicles are moving in unusual areas or at atypical speeds. It is also possible to detect launches of rarely used runways as well as changes in the starting direction and the use of a test vehicle with this method.

Claims (18)

1. Method for automatic recognition of a change in situation within a spatial area to be monitored by means of sensors, the monitored spatial area being subdivided into a plurality of fields, comprising the steps of:

   a) detecting values by means of the sensors with respect to at least one property of objects which are located within the spatial area to be monitored, and their fields, within which the respective value of the property has been detected,

   b) learning at least one behaviour of the objects with reference to at least one of the object properties with the aid of a statistical evaluation of the determined values for at least one field, and

   c) detecting a change in situation as a function of a comparison between the learned behaviour of at least one field, and at least one currently determined value of the at least one field with reference to at least one of the object properties.
2. Method according to Claim 1, characterized by learning the behaviour in step b) with the aid of the statistical evaluation of the values last determined within a period.
3. Method according to Claim 1 or 2, characterized by learning the behaviour for more than one period and recognizing a change in situation by comparison between the learned behaviour of the respective periods and the current values.
4. Method according to Claim 3, characterized by determining a common learned behaviour as a function of the behaviour learned in the respective periods, and recognizing a change in situation by comparison between the common learned behaviour and the current values.
5. Method according to Claim 4, characterized by determining the common learned behaviour by averaging.
6. Method according to one of the preceding claims, characterized by comparing in step c) exclusively those fields for which at least one value is stored with reference to one of the object properties.
7. Method according to one of the preceding claims, characterized by detecting a determination position which indicates the position of the object within the spatial area during the determination of a value of a property, and determining the corresponding field as a function of the determination position.
8. Method according to Claim 7, characterized by converting the determination position into a unified position coordinate system and determining the corresponding field as a function of the determination position within the unified position coordinate system.
9. Method according to one of the preceding claims, characterized by detecting a change in situation in step c) as a function of the comparison between the learned behaviour from step b) and a behaviour which is current with the aid of statistical evaluation of currently determined values, as current values.
10. Method according to one of the preceding claims, characterized by recognizing a change in situation in step c) as a function of the comparison between the learned behaviour from step b) and current values of at least one of the object properties of a trajectory of a particular object while taking account of the fields which are being traversed by the trajectory of the object.
11. Method according to one of the preceding claims, characterized by images of mean values, sums, products and/or standard deviations as statistical evaluation.
12. Method according to one of the preceding claims, characterized by imaging sensors, in particular image recognition methods, radar sensors and/or RFID transponders as sensors.
13. Method according to one of the preceding claims, characterized by humans, animals, road, rail, air and/or watercraft and particles as objects.
14. Method according to one of the preceding claims, characterized by activity, speed and/or direction as a determinable property of the objects.
15. Method according to one of the preceding claims, characterized by comparison of step c) by means of a threshold comparison.
16. Method according to Claim 15, characterized by determining a threshold for each field as a function of at least one behaviour with reference to at least one object property and the standard deviation with reference to the object property and the respective field.
17. Computer program product with program code means for carrying out the method according to one of the preceding claims when the computer program product is run on a computing machine.
18. Device for recognition of a change in situation having at least one sensor which is set up to determine the values of at least one property of objects which are situated within a spatial area to be monitored, and having a data processing system which is set up to carry out the method according to one of Claims 1 to 16.

Images