DE102007000344A1 - System for detection of sensor disturbances and for signal disturbances, has sensor for recording electromagnetic radiation and for generation of image sequences - Google Patents

Abstract

The system has a sensor (25) for the recording electromagnetic radiation and for the generation of image sequences. A memory unit (18) and an analysis unit (13) are formed for the analysis of the sensor signal (1). The sensor signal is analyzed for the detection of the sensor disturbances and the signal disturbances. An independent claim is also included for a method for the detection of sensor disturbances and for signal disturbances.

Description

Technical area

The The present invention relates to a system for detecting sensor disturbances and / or of signal interference, comprising at least one sensor for detecting electromagnetic Radiation and to generate image sequences.

The The present invention further relates to a method of recognition of sensor disturbances and / or signal interference by means of at least one sensor for detecting electromagnetic Radiation and to generate image sequences.

State of the art

systems, in which sensors for detecting electromagnetic radiation and are used to generate image sequences are widely used. Starting with surveillance systems with video cameras where a situation analysis is automatic or on the part of a human observer, via automatic Robotic vision systems or fingerprint reading systems all the way to optical scanner can different types of sensor disturbances or signal interference occur, both by accidental Events, such as contamination of the sensor surface, as well can be caused by targeted sabotage actions.

From the publication EP 1 672 604 A1 From the prior art, a method and a device for detecting sabotage sabotage attempts or on a surveillance camera are known. Based on the images taken by the camera, each image is compared with a previously recorded reference image and obtained therefrom a correlation value; this correlation value is compared with an alarm threshold.

The Device for implementation contains this known method a processor for the evaluation of the captured images and one on this processor implemented program for execution the said process steps.

For example, the document deals with the problem of detecting sabotage on a sensor WO 00/57381 A1 from the prior art, which describes a method and a device for detecting a masking process when turning on a motion detector using an infrared sensor and a microwave Doppler sensor.

Of the fundamental Disadvantage of this known system and similar systems for the detection of Sabotage on a sensor is its cost intensity. This one realized fundamental Idea is to use a sensor with the help of other sensors secure. An example for is the equipment of a video camera with a variety of additional sensors, such as motion sensors, glass breakage sensors, pressure sensors or the like; this is technically possible But brings high installation and maintenance costs.

Object of the invention

outgoing from the disadvantages and shortcomings set out above as well as in appreciation of the prior art is the present invention the task is based, a system of the type mentioned above as well a method of the type mentioned in such a way that automatic detection of sensor disturbances and / or signal interference, which occur without targeted human influence, guaranteed is.

This is especially for Systems with automatic signal evaluation for the purpose of situational analysis important. Most conventional analytical systems can do without human help does not realize if the received signal is still for the Analysis is appropriate and just assume.

To the first calibration of the sensors (at initialization) maintain the determined calibration parameters, regardless of whether the situation has changed in the meantime or not. It can for example, the spatial Location or the focus settings change, as is the case for example at a video camera installed on a railway station due to the mechanical influences passing by Trains often happens.

Illustration of the present invention: Solution, advantages

These Task is achieved by a system with the specified in claim 1 Characteristics and by a method with the specified in claim 23 Characteristics solved. Advantageous embodiments and expedient developments of the present Invention are characterized in the respective subclaims.

A partial solution of the above-described technical problem already provides the document WO 01/60074 A1 in the prior art, in which a method and an apparatus for detecting errors in signal processing equipment security technology, namely transmission errors, are described.

This known method essentially consists of generating a changing visual signal in addition to the camera signal and displaying it in the display unit. This makes it possible for a user to recognize whether the currently displayed signal is the video signal recorded by the camera unit. Other types of sensor interference and / or sabotage are in the document WO 01/60074 A1 not dealt with in the prior art.

The The present invention is based on a system and method to detect sensor interference and / or signal interference in plants that have at least one sensor for detecting electromagnetic Use radiation and to generate image sequences.

The System has at least one storage unit and at least one Analysis unit for analyzing the sensor signal and is inventively designed so that the detection of sensor interference and / or signal interference and sabotage from the sensor signal itself, so that the appropriate countermeasures through the system itself or by a message to an observer.

The proposed according to the invention Detection of sabotage from the sensor signal itself means one high cost savings.

at some advantageous embodiments of present invention, the detection of a signal failure, a change the spatial Position of the sensor, a change the focusing settings of the sensor, a masking and / or contamination of the sensor surface, changes in illumination in the detection range of the sensor, a replacement of the sensor signal by another signal or similar Operations.

at The present invention is a system and a method that detects the detection of sensor interference and / or signal interference ensure the sensor signal itself. The realization consists in an installation of new modules or also in adaptation to new tasks of possibly existing modules for automatic evaluation of the sensor signal. These modules are at least one storage unit and at least one analysis unit for analyzing the sensor signal.

The inventive system as well as the method according to the invention are designed so that the sensor signal measured at short intervals and in the analysis unit is analyzed using the methods contained in the storage unit. The temporal signal changes are considered and their equivalent to those in the storage unit contained fault scenarios examined.

at the systems that benefit particularly from this invention these are automatic video-based systems for security surveillance, for traffic monitoring, for market research, driver assistance and the like.

Brief description of the drawings

As already discussed above, there are various possibilities for embodying and developing the teaching of the present invention in an advantageous manner. For this purpose, on the one hand to the claim 1 and the claim 23 respectively subsequent claims referenced, on the other hand further embodiments, features and advantages of the present invention will be described below with reference to 1 to 12 illustrated embodiments explained in more detail.

It shows

1 a schematic representation of a first embodiment of the system according to the invention, which operates according to the inventive method;

2 a schematic representation of a second embodiment of the system according to the invention, which operates according to the inventive method;

3 a schematic representation of an embodiment for the construction of a unit for deciding on the presence or absence and, if appropriate, the nature of a fault;

4 an exemplary sequence of the processing of the sensor signal;

5 a schematic representation of an embodiment of the construction of an operating according to the invention unit for analyzing the sensor signal;

6 a schematic representation of an embodiment of the structure of the method according to the invention operating system with advantageous features;

7 a schematic representation of an embodiment of the method with advantageous features;

8th in a schematic representation of an embodiment of the structure of an after he Working method according to the invention secondary unit for analyzing the sensor signal;

9 a schematic representation of an embodiment of the construction of an operating according to the invention feature extraction unit;

10 a schematic representation of an embodiment of the detectors for individual types of disorders;

11 in schematic representation, six embodiments of two-dimensional calibration pattern; and

12 schematic representation of two embodiments of three-dimensional calibration objects.

Same or similar embodiments, elements or features are in 1 to 12 provided with identical reference numerals

Best way to execute the present invention

When preferred embodiment becomes an automatic monitoring system described with at least one video camera.

An example structure is in 1 shown. The system has at least one area of the real world 24 observing sensor 25 for detecting electromagnetic radiation and for generating image sequences, at least one storage unit 18 and at least one analysis unit 13 for analyzing the sensor signal 1 on.

That of the sensor 25 generated signal 1 will be in the analysis unit 13 with the help of the memory unit 18 by request message 2 requested data 3 analyzed. The analysis results 4 are stored and / or forwarded.

At an in 2 illustrated development of the system and the method, the system includes a message unit 17 , which signals a fault when a fault is detected 5 in terms of telecommunications and to a remote receiver unit 28 can send. The message may indicate the presence of a fault in general and / or the nature of the fault.

In 4 an example procedure is visualized:
The sensor signal 1 at time t is in the analysis unit 13 with its time lags mT and time periods (n + m) T delayed versions 60 and 69 compared, the units 22 . 23 the corresponding time delay elements (areas in the memory unit 18 for buffering the signal 1 ).

Furthermore, it is possible, in addition to or instead of the storage of time-delayed versions of the signal 1 create virtual reference images and save them. A reference image may be generated, for example, by a weighted averaging of the incoming images or by a caching of the first image of the sequence.

Save the storage of the signal 1 can the analysis results 4 together with feature vectors 12 be cached, which underlie the analysis and are extracted from the actual signal (shown in 1 ). In this way, not the original signal itself, but are subjected to the characteristics obtained and their temporal change of the analysis.

The feature vectors 12 can be generated for individual pixels as well as for pixel groups (as for pixel blocks) or for the entire image, as well as having a combination of descriptions of the local and global properties.

at a training of the system finds a comparison of the extracted Characteristics and the result values of the feature processing are not or not only with the feature vectors obtained from the current image sequence and reference values, but also with the previously determined empirical values for individual fault scenarios instead of. This can experimentally determined thresholds, distributions, change histories over the Time or the like.

In a further embodiment of the system, after detecting a fault, a change of at least a part of the one in the memory unit 18 data. This may be expressed, for example, in a resetting or adaptation of thresholds and distributions, storage of new reference images, or the like.

In a development of the system, at least a partial change takes place in the storage unit 18 contained data depending on the type of failure detected. For example, when detecting the spatial position change of the sensor 25 a re-storage of reference images are caused.

5 shows a variant of the realization of the system according to the invention and the corresponding method:
The unit 13 to analyze the sensor signal is made up of a unit 14 for image segmentation, a unit 15 for feature extraction and an Ent decision unit 16 together. For those in the image segmentation unit 14 formed pixel groups are in the feature extraction unit 15 feature vectors 7 extracted.

These feature vectors 7 be together with the segmentation information 6 and with the record 3 in the decision-making unit 16 processed, the record 3 by means of a message 2 from the storage unit 18 is requested.

The decision-making unit 16 decides on the presence or absence of a fault and, if necessary, on the type of fault. The analysis results 4 become in positive message in the message unit 17 in terms of telecommunications and as a message signal 5 forwarded. The results 6 . 7 . 4 be in the storage unit 18 stored and used in the next work cycle.

at One development of the system allows the system between at least two types of disorders so that the generated message signal is not only different the presence of a fault, but also the type of fault according to the analysis catalog of the system.

For example, a distinction can be made between a change in the focus settings and a change in the spatial position and / or the orientation of the sensor. The corresponding decision is made in the decision-making unit 16 met.

3 visualizes an embodiment of the structure and the operation of the decision unit 16 : In the detectors developed for different types of sensor disorders 41 . 42 . 43 . 44 . 45 . 46 . 47 . 48 become the reference values 51 . 52 . 53 . 54 . 55 . 56 . 57 . 58 calculated for the presence of a fault type, which is then weighted and the decision function 29 be supplied.

In 6 and in 7 Further exemplary embodiments for the system and the method are visualized:
The system does not only detect sensor disturbances and / or signal interferences, what happens in the primary analyzer 19 for analyzing the sensor signal 1 is realized additional tasks included in the secondary analysis unit 20 for analyzing the sensor signal 1 be solved.

These tasks may include the functionality of an automatic monitoring and alarm system, a traffic monitoring system, a market research system or a driver assistance system. Depending on the design of the secondary analysis unit 20 In particular systems for burglary security, fire detection or the like can be realized.

In 8th is an advantageous embodiment of a secondary analysis unit 20 for analyzing the sensor signal 1 shown. The secondary analysis unit 20 has a classification unit 34 for classifying the segmented pixel groups, a tracking module 35 , a grouping unit 36 for grouping the segmented pixel groups into objects, a classification unit 37 for classifying the objects and a classification unit 38 to classify actions.

The described additional functionalities can be defined by the corresponding characteristics of the classification unit 38 will be realized. Thus, data, such as the number and type of objects in the image, the object speeds, the movement trajectories of the objects, the average residence time in image areas and the like, are evaluated according to different viewpoints.

From this information, for example, the presence and possibly the behavior of persons in the detection range of the sensor 25 getting closed. Information about customer flows, customer interest in specific merchandise displays or advertising presentations can be obtained from the transaction data for the purposes of market research. Forgotten objects in the detection area of the sensor 25 can be detected for the purpose of security surveillance.

In one embodiment of the system as a traffic monitoring system, in the classification unit 37 For the classification of objects for example the distinction between cars, humans and animals are carried out, so that in the classification unit 38 for the classification of actions, for example, the detection of unauthorized parking, the crossing of a road in an unauthorized place or the like can be detected.

In 9 is an embodiment of a possible structure of the feature extraction unit 15 shown:
The feature extraction unit 15 processes the segmented pixel groups 6 and has a calculation unit 30 for calculating texture information, a calculation unit 31 for calculating the relevance of texture information, a calculation unit 32 for calculating brightness distributions and a calculation unit 33 for calculating the relevance of brightness distributions.

The features determined for pixel groups become corresponding feature vectors 40 (= Texture histogram at time t), 50 (= Brightness histogram at time t) summarized.

In the calculation unit 32 For example, the conventional brightness histograms can be calculated. The number of allowed brightness levels in such histograms is typically set to 256 so that all brightness values in the image are quantized accordingly.

The frequency of a quantized brightness value in a pixel group is plotted along an axis with values from 0 to 255. In the calculation unit 33 determine the homogeneity, the average brightness and similar properties of the individual pixel groups and thus calculate a measure of the relevance of the brightness information for these pixel groups.

The determination of texture information may be in the calculation unit 30 via a determination of the number of gray level changes in the 3x3 neighborhood of each pixel. The determined values are summed up, with each edge direction entering into the sum with a different weighting, preferably a power of two.

The calculated texture values can be plotted along an axis with values from 0 to 255, similar to the brightness histogram, to produce a texture histogram. In the calculation unit 31 the homogeneity of the pixel groups is analyzed, which can be done by evaluating the number of pixels that have no gray value changes to their neighbors.

For this is the number of pixels in each pixel group that is 0 in value Texture histogram were compared with a reference value. For example, the reference value may be at half the number of pixels in a pixel group.

The basic idea in this embodiment is the comparison of texture properties and brightness properties of pixel groups in the current image and corresponding pixel groups in past images. Among other things, it exploits the fact that the texture and the brightness change differently in the breakdown scenarios enumerated, as follows:
With a moderate change in illumination, the brightness values in recorded images change greatly, whereas the texture changes only slightly.

When changing the focus settings of the sensor 25 the textures in the captured image are heavily smeared, whereas the brightness values of pixel groups change only slightly.

A simultaneous change of textures and brightness distributions in the image provides indications of a change in the spatial position and / or the orientation of the sensor 25 , on a change of scenery in the detection range of the sensor 25 or on occlusion or contamination of the surface of the sensor 25 ,

These cases can be distinguished, for example, by being in the analysis unit 13 an automatic tracking of movements in the detection range of the sensor 25 carried out and determined changed areas in the image and their characteristic properties.

An example of an application of the described approaches is in 10 visualized:
With 40 and 50 designated feature vectors calculated for pixel groups in the current image are tagged with the corresponding feature vectors 49 and 59 compared to a past image. Depending on the result of the individual evaluations, messages are generated 51 . 52 . 53 . 54 . 55 . 56 . 57 . 58 with the corresponding reference values to the decision function 29 (see. 3 ).

In the overall evaluation in the decision function 29 In this exemplary embodiment, reference values flow for the following types of disturbance: change in illumination (= message 51 ); Cover the sensor 25 (= Message 52 ); Change in the spatial position and / or orientation of the sensor 25 (= Message 53 ); Contamination of the surface of the sensor 25 (= Message 54 ); Change the focus settings of the sensor 25 (= Message 55 ); Moisture on the surface of the sensor 25 (= Message 56 ); Object movement in the detection area of the sensor 25 (= Message 57 ); Signal failure (= message 58 ).

The texture and brightness histograms obtained by the feature extraction unit 15 , as described above, for each pixel group, are multidimensional matrices. The dimensionality of the matrices is M × N, where M is the number of pixel groups and N is the number of bins.

The in the calculation units 31 . 33 calculated vectors provide information about the relevance of the individual pixel groups for the comparisons: the textur rich regions are given a greater weighting than the texturized regions; the overexposed areas with maxima Brightness values get smaller weights for further calculations than the non-homogeneous areas.

The individual calculations run on the condition functions 70 . 71 . 72 . 73 . 74 . 75 . 76 . 77 . 78 . 79 . 80 each checking the fulfillment of certain conditions. These functions return values of 0 or 1 or messages 51 . 52 . 53 . 54 . 55 . 56 . 57 . 58 which contain reference values corresponding to the probabilities of the existence of a fault type.

Further below are the condition functions 70 . 71 . 72 . 73 . 74 . 75 . 76 . 77 . 78 . 79 . 80 briefly described:
The condition 70 is satisfied when the sum over the M pixel groups of the relevance weighted results of the comparison function Fkt 1 is greater than a threshold1. The fulfillment of the condition 70 means that the texture properties of the pixel groups have changed dramatically. The comparison function Fkt1 can be realized with arbitrary comparison functions for histograms, such as with the known chi-squared intersection.

The condition 71 is satisfied when the sum over the M pixel groups of the relevance weighted results of the comparison function Fct2 is greater than a threshold2. The fulfillment of the condition 71 means that the brightness characteristics of the pixel groups have changed greatly. The comparison function Fct2 can be realized with arbitrary comparison functions for histograms, such as with the known Chi-Squared Intersection.

The condition 72 is met if both the condition 70 as well as the condition 71 is fulfilled and at the same time the total area of the changed regions is greater than a threshold3.

The condition 73 is met if the condition 72 is not fulfilled and at the same time the size and location of the changed regions in the picture have not or only slightly changed over a longer period, in particular more than a fixed number of pictures. Failure to meet the condition 73 becomes the reference value for an object movement within the detection range of the sensor 25 set to the appropriate probability and with the message 57 sent; otherwise the reference value will be in the message 57 the value 0.

The condition 74 is satisfied if the changed range is homogeneous and the condition 72 is satisfied.

Failure to meet the condition 75 and if the calculated motion vectors in the image are similar (greater than 0 and the mean deviation is less than a threshold 4), the reference value for the change in the spatial position and / or the orientation of the sensor 25 according to its probability set and with the message 53 sent; otherwise the reference value will be in the message 53 the value 0.

In the condition 76 becomes a distinction between a signal failure and a masking of the sensor 25 carried out. If the condition 74 is satisfied and the mean value of the changed area is close to 0 (condition 76 ), a high probability of a signal failure is detected, what with the message 57 is signaled; otherwise the reference value will be in the message 57 set to 0. If the mean of the changed area is greater than the sum of 0 and a threshold5, the message is 52 a masking of the lens signals; otherwise the reference value will be in the message 52 set to 0.

At the condition 77 is searched for a certain type of possible change of textures in the image, especially after smearing the textures. This search can be performed by simply comparing bins of texture histograms for each pixel group. If the textures have changed so much that the number of individual bins for the values 1 to 255 has decreased in contrast to the increase in the value associated with 0, and the sum of the reference values across all pixel groups is greater than a threshold 6 the condition 77 Fulfills.

The condition 78 is fulfilled if at the same time the condition 77 is satisfied, the condition 71 is not satisfied and a uniformity of the texture change is detected. This uniformity can be checked by calculating the mean of the change across all pixel groups and the variances of the individual reference values. If the sum of the variances is greater than a predetermined threshold 7, moisture is likely to be on the surface of the sensor 25 or on other transparent surfaces in front of the sensor 25 (Glass) detected and with the message 56 reported; otherwise the reference value will be in the message 56 set to 0. If the change is judged to be even, it will change the focus settings of the sensor 25 recognized and with the message 55 signals; otherwise the reference value will be in the message 55 set to 0.

The condition 79 is fulfilled if at the same time the condition 71 is met and the condition 70 is not fulfilled. The condition 79 shows that up the texture properties have changed little in comparison to a large change in gray scale properties. If this occurs, it is very likely that a change in illumination will be detected and the message 51 signals; otherwise the reference value will be in the message 51 set to 0.

A contamination of the lens will be when fulfilling the condition 73 and failing to meet the condition 75 detected (condition 80 ), whereupon the message 54 is sent.

The Decision over the presence of a specific type of fault can also be directly via the sending corresponding messages are taken, which then only the values 0 or 1 include. But cases are not excluded which several types of errors occur simultaneously or be detected incorrectly. thereupon it makes sense to always send the corresponding messages and the final one Make a decision centrally.

The reference values in the messages are set according to the probabilities calculated in the condition functions for the existence of individual types of disturbances. These reference values are in the value range from 0 to 1 and are included with the weighting values 61 . 62 . 63 . 64 . 65 . 66 . 67 . 68 weighted, making the final decision in a common cost function 29 is taken. The procedure is in 3 visualized.

The mentioned Thresholds and weights are estimated or determined experimentally, including a workout with the help of artificial methods Intelligence, in particular with the help of neural networks can.

In a further embodiment, additionally replacing the sensor signal with another signal and obscuring the sensor 25 with a non-homogeneous object, such as an image of the background.

The The recognition principle is based on the fact that in these cases the movement patterns in the taken picture in comparison to past pictures in unusual Change way would. This can express itself, for example, that the calculated Abruptly cancel movement trajectories of detected objects. To this, too will determine a continuous tracking of movements in the Image performed.

At the Recognition of several movement trajectories, their properties stronger than the specified change thresholds to distinguish from properties of normal motion trajectories, a replacement of the signal is detected by another signal.

For that too Detecting abnormal behaviors may be such criteria as the ends of the object trajectories in the computations, not for leaving the detection area are defined. These areas can manually (for example, a door) or by the sensor structure (for example, image edges).

If additionally the condition is fulfilled is that the new image is a saved reference image for the background strongly resembles can with great probability the corresponding sabotage type can be detected.

A training of the system, which also in 2 is illustrated, includes the local storage in the storage unit 18 and / or sending to a remote receiver unit 28 at least one image that is close in time to the fault detection.

As in 2 shown, the system can also be a controllable device 27 , such as a sprinkler system, a light source or a speaker, via a control unit 39 and the corresponding messages 5 . 9 Taxes. In the case of systems for robot vision, for example, when certain types of failure are detected, the engine can be switched off.

The System components can both spatially close to each other as well as spatially separated from each other and connected to message lines. As an embodiment can lead a video camera here wirelessly connected to a personal computer.

As sensors 25 not only a video camera but also a laser scanner, an infrared camera, a terahertz wave detector or a radar device can be used to detect the electromagnetic radiation and to generate image sequences; the basic structure of the system does not change.

Depending on the sensor used 25 For example, the number and type of types of disturbances to be detected may change to suit the shape and design of individual detectors. This adaptation takes place via a definition of corresponding calculation rules in the analysis unit 13 and corresponding comparison data in the storage unit 18 instead of.

When processing the image sequences of For example, an infrared camera can also cause measurement errors if the thermal calibration of the camera no longer corresponds to the situation within its detection range. This can be done for example by a continuous comparison of the respective current image with a past image and by an analysis of the course of the temperature change. Upon detection of the described fault, recalibration of the camera can be initiated.

In 2 Further advantageous features of the system are visualized. In the analysis unit 13 For example, signals may be processed by other sensors, not all of which are necessarily sensors for detecting electromagnetic radiation and for generating image sequences.

So, as in 2 shown in the analysis of the image sequence 1 in the analysis unit 13 also the messages 11 of the sensor 26 evaluated. As a sensor, for example, a smoke detector or a motion detector can occur.

A particularly advantageous expression The system is the use of at least two sensors for Detection of electromagnetic radiation and generation of image sequences, their coverage areas overlap at least partially.

In a previous, often referred to as calibration determination of the internal sensor parameters and the external sensor parameters can be obtained from the image sequences using stereobased techniques three-dimensional information for the scene in the common detection range. The scene in the common detection range of the sensors becomes three-dimensional:
The actual sizes of objects can be calculated so that, for example, an insect near the sensor surface can not be confused with a faraway car. The motion vectors of objects in the scene can be determined three-dimensionally:
The approach of a person to the sensor is detected, so that a sabotage can be detected before the sabotage reaction. Overall, the number of wrong decisions is greatly reduced.

The Obtaining the three-dimensional information from the image sequence of a single sensor is by other known methods from the field of computer vision possible, is more difficult than the calculation with stereobased Techniques.

The absolute depth information or the relative depth information can with the methods of shape extraction from motion (shape from motion), from the texture (shape from texture), from the focused and defocused images (shape from focus or shape from defocus) or estimated from the shadow (shape from shadow).

The However, algorithms used are generally more complex than the stereobased algorithms; the accuracy of the estimates is strongly dependent on the accuracy of the parameterization, so that at a small to medium calculation cost only rough estimates possible are. But this is for the purposes described here are sufficient.

A problem with many image-based applications is the homogenous areas in the sensing area of the sensor 25 For example, a white-colored wall in the detection range of a video camera or the temperature-homogeneous ground in the detection range of an infrared camera.

The Mapping a homogeneous area into a picture sequence is approximately homogeneous, so that no distinction is made between the values of the associated pixels can be. If these homogeneous areas are very large, There is also a problem with some of the methods described to detect sensor interference and / or signal interference.

For example, a change in the spatial position and / or orientation of the video camera is not recognized, as far as the video camera is aligned with another part of the same white wall, because the analysis unit 13 sees no difference between the corresponding pixel values.

Analogously, a change in the focus settings of a video camera can only be detected if sufficiently relevant structure-characterizing information, in particular edges and textures in the detection range of the sensor 25 , is present.

Similar problems can occur with other types of homogeneity, such as temperature homogeneity in the sensing range of the sensor 25 and an infrared camera as a sensor or a homogeneous spatial depth and a laser scanner or radar device as a sensor 25 ,

In view of the problem described above, the present invention can be simplified according to a particularly advantageous development in that a favorable for the detection of sensor disturbances and / or signal interference recording situation is ensured, namely there in that at least one suitable two- or three-dimensional calibration object or calibration pattern enters the detection range of the sensor 25 is brought.

The Detection of sensor disturbances and / or of signal interference is calculated by comparing the images of the calibration object or calibration sample realized in the image sequence at different times.

at Systems with one or more video cameras as sensors preferably as calibration objects or as a calibration pattern signs or stickers used, their introduction to the scene not only the described To facilitate the detection of sensor disturbances and / or of Signal interference, but also serves other purposes.

So can for example, advertising stickers and advertising signs can be affixed that these are not only from the passing people, but also from the video camera are observable.

In analogous way the signs attached for data protection reasons or Stickers like "area is video-monitored "," video surveillance "or the like, the video surveillance signage or signs such as "exit", "emergency exit" or the like, suitable suspended become.

The suitable calibration objects and calibration samples can also be used for the purpose the simplification of the present invention and the increase of Robustness of the system can be created.

It For example, they may be objects or patterns based on their special design possess characteristic properties and so or in a similar Form does not occur in a typical recording situation, such as Pattern with characteristic edge distributions and / or patterns with characteristic color distributions.

Some examples of such patterns and three-dimensional objects are in 11 (= Embodiments 81 . 82 . 83 . 84 . 85 . 86 for two-dimensional calibration samples) and in 12 (= Embodiments 87 . 88 for three-dimensional calibration objects).

at Systems with one or more infrared cameras may be present the calibration objects are objects with a characteristic temperature distribution, for example, a heating element a preset constant temperature or around a heater body with a timer so that the temperature changes a specific correspond to characteristic distribution.

at Systems with one or more sensors, the spatial Depth or the three-dimensional shape of objects in the detection area of the sensor, as for example with laser scanners or at radars the case is, as a calibration object is an object with a characteristic spatial Form used.

The technical design of the method is realized in the first variant, that the Illustrations of the calibration objects or calibration pattern in the image sequence itself or the corresponding characteristic feature vectors in one Database are stored and consulted for calculation. In the second variant, the images of the calibration objects or Calibration pattern manually selected in an image of the image sequence.

The calibration objects or calibration samples described are preferably so attached, that in a typical shooting situation at least not Completely obscured by other objects.

1

sensor signal

2

Request message to the storage unit 18

3

Record from the storage unit 18

4

analysis results

5

Message about the analysis results 4

6

Results the segmentation

7

Results the feature extraction

8th

Analysis results of the secondary unit 20 for analyzing the sensor signal 1

9

signal for controlling the controllable device

10

signal another sensor

11

signal another sensor, not necessarily image based

12

feature vectors

13

Unit for analyzing the sensor signal 1

14

unit for image segmentation

15

unit for feature extraction

16

unit to decide about Presence and nature of the disorder

17

message unit

18

storage unit

19

primary unit for analyzing the sensor signal 1

20

secondary unit for analyzing the sensor signal 1

22

delay for delay around Time interval mT

23

delay for delay around Time period nT

24

Detection range of the sensor 25 in the real world

25

sensor for the detection of electromagnetic radiation and for generation of image sequences

26

Sensor, not necessarily image-based

27

controllable device

28

distant receiver device

29

decision function

30

unit for extraction of texture information

31

unit to calculate the relevance of texture information

32

unit for the extraction of brightness distributions

33

unit to calculate the relevance of brightness distributions

34

unit for the classification of segmented pixel groups

35

tracking module for individual pixel groups

36

unit for grouping segmented pixel groups into objects

37

unit for classifying objects

38

unit for the classification of actions

39

control unit

40

texture histogram at time t

41 to 48

detectors for different Types of sensor failure and sabotage, namely

41

first detector

42

second detector

43

third detector

44

fourth detector

45

fifth detector

46

sixth detector

47

seventh detector

48

eight detector

49

texture histogram at time t - mT

50

brightness histogram at time t

51 to 58

reference values for different Types of sensor failure or sabotage, namely

51

first Reference value, in particular change of illumination

52

second reference value, in particular occlusion of the sensor 25

53

third reference value, in particular change of the spatial position and / or the orientation of the sensor 25

54

fourth reference value, in particular contamination of the surface of the sensor 25

55

fifth reference value, in particular focusing settings of the sensor 25

56

sixth reference value, in particular moisture on the surface of the sensor 25

57

seventh reference value, in particular object movement in the detection range of the sensor 25

58

eight Reference value, in particular signal failure

59

brightness histogram at time t - mT

60

sensor signal at time t - mT

 61 to 68

Weightings of the reference values 51 . 52 . 53 . 54 . 55 . 56 . 57 . 58 for individual detectors 41 . 42 . 43 . 44 . 45 . 46 . 47 . 48 , namely

61

Weighting of the reference value 51 for first detector 41

62

Weighting of the reference value 52 for second detector 42

63

Weighting of the reference value 53 for third detector 43

64

Weighting of the reference value 54 for fourth detector 44

65

Weighting of the reference value 55 for fifth detector 45

66

Weighting of the reference value 56 for sixth detector 46

67

Weighting of the reference value 57 for seventh detector 47

68

Weighting of the reference value 58 for eighth detector 48

69

sensor signal at time t - (m + n) T

70 to 80

Conditional functions, namely

70

first Condition (sfunktion)

71

second Condition (sfunktion)

72

third Condition (sfunktion)

73

fourth Condition (sfunktion)

74

fifth condition (sfunktion)

75

sixth Condition (sfunktion)

76

seventh Condition (sfunktion)

77

eighth Condition (sfunktion)

78

ninth Condition (sfunktion)

79

tenth Condition (sfunktion)

80

eleventh Condition (sfunktion)

 81 to 86

Examples for two-dimensional Calibration pattern, namely

81

first example for two-dimensional calibration pattern

82

second example for two-dimensional calibration pattern

83

third example for two-dimensional calibration pattern

84

fourth example for two-dimensional calibration pattern

85

fifth example for two-dimensional calibration sample

86

sixth example for two-dimensional calibration pattern

87 to 88

Examples for three-dimensional Calibration objects, namely

87

first example for three-dimensional calibration object

88

second example for three-dimensional calibration object

Claims (50)

System for detecting sensor disturbances and / or signal interference, comprising at least one sensor ( 25 ) for detecting electromagnetic radiation and for generating image sequences, characterized by at least one memory unit ( 18 ) and at least one for analyzing the sensor signal ( 1 ) provided analysis unit ( 13 ), wherein the sensor signal ( 1 ) is itself analyzed for the detection of sensor disturbances and / or signal disturbances. System according to claim 1, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that when a fault is detected, a signal is generated. System according to claim 1 or 2, characterized by at least one message unit ( 17 ), which is designed so that when a fault is detected at least one signal ( 5 ) and to at least one remote receiver unit ( 28 ) is sent. System according to at least one of claims 1 to 3, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that at least two types of interference can be distinguished. System according to at least one of claims 1 to 4, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are executed so that, depending on the detected type of disturbance, at least one signal indicating the type of disturbance is generated. System according to at least one of claims 1 to 5, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are executed so that when a fault is detected in general and / or when a specific type of fault is detected, a change of at least a part of the one in the memory unit ( 18 ) is initiated. System according to at least one of claims 1 to 6, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that when detecting a fault generally and / or when detecting a particular type of fault at least one controllable device is caused to at least one physical action. System according to at least one of claims 1 to 7, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that at least one change of the spatial position and / or the orientation of the sensor ( 25 ) is / will be recognized. System according to at least one of claims 1 to 8, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that a failure of the sensor signal ( 1 ) is recognized. System according to at least one of claims 1 to 9, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that a complete or partial occlusion of the sensor ( 25 ) is recognized. System according to at least one of claims 1 to 10, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are carried out such that a replacement of the sensor signal ( 1 ) is detected by at least one other signal. System according to at least one of claims 1 to 11, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that a contamination of the surface of the sensor ( 25 ) is recognized. System according to at least one of claims 1 to 12, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are carried out so that a change of the focus settings of the sensor ( 25 ) is recognized. System according to at least one of claims 1 to 13, characterized in that the memory unit ( 18 ) and the analysis unit ( 13 ) are designed so that illumination changes in the detection range of the sensor ( 25 ) be recognized. System according to at least one of Claims 1 to 14, characterized in that the sensor ( 25 ) for detecting electromagnetic radiation and for generating images at least one video camera. System according to at least one of Claims 1 to 15, characterized in that the sensor ( 25 ) for detecting electromagnetic radiation and for generating image sequences by at least one laser scanner. System according to at least one of claims 1 to 16, characterized in that it is at the sensor ( 25 ) for detecting electromagnetic radiation and for generating image sequences by at least one infrared camera. System according to at least one of claims 1 to 17, characterized in that it is at the sensor ( 25 ) for detecting electromagnetic radiation and generating image sequences by at least one terahertz wave detector. System according to at least one of claims 1 to 18, characterized in that it is at the sensor ( 25 ) for detecting electromagnetic radiation and for generating image sequences around at least one radar device. System according to at least one of the claims 1 to 19, characterized in that at least one component the system is not in spatial Close to located at least one other part of the system and that at least one component of the system with this other component the system by means of a cable via at least one message line or wirelessly connected. System according to at least one of claims 1 to 20, characterized by at least two sensors ( 25 ) for detecting electromagnetic radiation and for generating image sequences with at least partially overlapping detection areas, wherein the memory unit ( 18 ) and the analysis unit ( 13 ) are carried out so that an extraction and an evaluation of at least coarse three-dimensional information in the common detection range of the sensors ( 25 ) possible are. System according to at least one of claims 1 to 21, characterized in that at least one two-dimensional calibration pattern ( 81 . 82 . 83 . 84 . 85 . 86 ) and / or at least one three-dimensional calibration object ( 87 . 88 ), in particular at least one object with a characteristic spatial shape and / or at least one object with a characteristic temperature distribution, in the detection range of the sensor ( 25 ) is provided. Method for detecting sensor disturbances and / or signal interference by means of at least one sensor ( 25 ) for detecting electromagnetic radiation and for generating image sequences, characterized in that the sensor signal ( 1 ) is itself analyzed for the detection of sensor disturbances and / or signal disturbances. A method according to claim 23, characterized in that for the analysis of the sensor signal ( 1 ) at least one reference value locally or globally calculated for the current image with at least one in at least one storage unit ( 18 ) is compared. Method according to claim 23 or 24, characterized in that the temporal change at least one each for the current image is calculated locally or globally calculated reference value and on their similarity with at least one for one malfunction typical change process is examined. Process according to at least one of the claims 23 to 25, characterized in that a decision on the The presence or absence of at least one fault and optionally via the nature of the present disorder by comparing the for at least two types of malfunction calculated reference values or probability values. A method according to any one of claims 23 to 26, characterized in that a decision on the presence or absence of at least one disorder and optionally on the nature of the present disorder by comparing the for at least one type of disorder and for at least one type of not by a disorder caused change in scenery within the detection range of the sensor ( 25 ) calculated reference values or probability values. Process according to at least one of the claims 23 to 27, characterized in that it is at the reference values brightness values and / or color values and / or from these values extracted features. Process according to at least one of the claims 23 to 28, characterized in that it is the reference values are values representing the texture properties of individual pixels and / or Characterize pixel groups in the picture. Method according to at least one of claims 23 to 29, characterized in that the reference values are values representing the three-dimensional information in the detection range of the sensor ( 25 ) characterize. Method according to at least one of claims 23 to 30, characterized in that in the analysis of the sensor signal ( 1 ) Gray value flows are evaluated in the image. Process according to at least one of the claims 23 to 31, characterized in that when detecting a fault generally and / or upon detection of a particular type of failure, at least one of the fault type dependent and / or independent Signal is generated. Process according to at least one of the claims 23 to 32, characterized in that when detecting a fault at least a temporally close to the time of detection image stored and / or telecommunications becomes. Method according to at least one of Claims 23 to 33, characterized in that, when a fault is detected in general and / or when a specific type of fault is detected, at least one change of at least one part of the one in the memory unit ( 18 ) is initiated. Process according to at least one of the claims 23 to 34, characterized in that when detecting a fault generally and / or upon detection of a particular type of fault at least one controllable Device too a physical action is initiated. Method according to at least one of claims 23 to 35, characterized in that the sensor signal ( 1 ) to a change in the spatial position and / or the orientation of the sensor ( 25 ) is examined. Method according to at least one of claims 23 to 36, characterized in that the sensor signal ( 1 ) to a failure of the sensor signal ( 1 ) is examined. Method according to at least one of claims 23 to 37, characterized in that the sensor signal ( 1 ) to a complete or partial occlusion of the sensor ( 25 ) is examined. Method according to at least one of claims 23 to 38, characterized in that the sensor signal ( 1 ) to a replacement of the sensor signal ( 1 ) is examined by another signal. Method according to at least one of claims 23 to 39, characterized in that the sensor signal ( 1 ) on contamination of the surface of the sensor ( 25 ) is examined. Method according to at least one of claims 23 to 40, characterized in that the sensor signal ( 1 ) to a change in the focus settings of the sensor ( 25 ) is examined. Method according to at least one of claims 23 to 41, characterized in that the sensor signal ( 1 ) to at least one illumination change in the detection range of the sensor ( 25 ) is examined. Method according to at least one of claims 23 to 42, characterized in that the sensor signal ( 1 ) is additionally analyzed for the purpose of fulfilling the functionalities of an automatic monitoring and alarm system. Method according to at least one of claims 23 to 43, characterized in that the sensor signal ( 1 ) is additionally analyzed for the purpose of fulfilling the functionalities of an automatic market research system. Method according to at least one of claims 23 to 44, characterized in that the sensor signal ( 1 ) is additionally analyzed for the purpose of fulfilling the functionalities of a traffic monitoring system. Method according to at least one of claims 23 to 45, characterized in that the sensor signal ( 1 ) is additionally analyzed for the purpose of fulfilling the functionalities of a fire alarm. Method according to at least one of claims 23 to 46, characterized in that the sensor signal ( 1 ) is additionally analyzed for the purpose of fulfilling the functionalities of a burglar alarm system. Method according to at least one of claims 23 to 47, characterized in that at least one two-dimensional calibration pattern ( 81 . 82 . 83 . 84 . 85 . 86 ) and / or at least one three-dimensional calibration object ( 87 . 88 ), in particular at least one object with a characteristic spatial shape and / or at least one object with a characteristic temperature distribution, in the detection range of the sensor ( 25 ) is introduced. Method according to claim 48, characterized in that the sensor disturbances and / or the signal disturbances are determined by comparing the images of the calibration sample ( 81 . 82 . 83 . 84 . 85 . 86 ) and / or the calibration object ( 87 . 88 ) are recognized in the image sequence at different times. A method according to claim 48 or 49, characterized in that the images of the Calibration pattern ( 81 . 82 . 83 . 84 . 85 . 86 ) and / or the calibration object ( 87 . 88 ) are stored in the image sequence or the corresponding characteristic feature vectors in a database and used for the calculation or that the images of the calibration sample ( 81 . 82 . 83 . 84 . 85 . 86 ) and / or the calibration object ( 87 . 88 ) are manually selected in an image of the image sequence.