DE19651172A1 - Automatic surveillance system for area or building - Google Patents

Abstract

The system has sensor (S1-S5), e.g. infra-red motion sensors, that are positioned to provide surveillance of an area. These provide inputs to a computer (RM) based system. Output channels (LAK1-LAK3) have filters (NF1-NF3) coupled to outputs (A1-A3) . These control the alarm stages (AL1,AL2) to activate video cameras (K1,K2) to record the event detected.

Description

Automatic video and audio surveillance systems are characterized in that they are in are able to automatically trigger alarm states via sensors connected to the system capture and a permanent record of the events that z. B. via cameras or Microphones can be detected.

Because environmental false alarms such as in the area of external surveillance (outdoor area) Sun, wind, movement of trees etc. cannot be ruled out To continuously improve sensors and tries to intercept false alarms already in the sensor. The happens on the one hand through ever better hardware versions (e.g. exclusion of the Daylight or the temperature on the switching effect of passive infrared detectors or their Shielding to protect against high-frequency radiation) or by software and Hardware filter in the sensor itself. DE 42 36 618 A1 describes e.g. B. an arrangement for Prevention of false alarms in motion detectors with an infrared sensor. Further Known measures for increasing false alarm security and switching security are implemented by Circuitry AND or OR operations of several sensors implemented. So are Combined infrared microwave sensors are known which have an optional AND or OR ver provide a link between the two signals at a common output. software and hardware within the sensors enables e.g. B. the summation of a certain number of Alarm events or waiting for a certain time for which the alarm is pending must be activated before the actual sensor output is activated. Such techniques are adequate known from fire alarm and fault reporting technology. Another interesting aspect is provided by hazard detection technology. This is how AS 28 17 089 describes it a hazard detection system that logically groups detectors over different Detection lines allowed. This logical grouping serves to increase clarity by summary, with the practical advantage of a clear addressing of the individual detectors and thus easier programming of the system and query of the individual detector signals, their sensitivity threshold and their belonging to discrete Groups. However, "additional information" according to the invention is not provided by this technique won.  

Perimeter monitoring, especially fence monitoring, sometimes uses the Information of several sensors of a fence field combined to form a kind of array in order to Make conclusions about false alarms or real alarms. Wind pressure z. B. usually leads an increase in the signals of all sensors in a fence field; a signal increase between two neighboring Arranged sensors of a fence field near the ground allows the conclusion to crawl under Fences too. It is remarkable that the sensors have a certain local dependence Stand to the fence field, since they are in their detection area on the area of their fence field Respectively. The evaluation and decision on alarm or false alarm takes place on site. B. by Microcontroller and is forwarded to a remote center.

Video sensors in which the image of a connected camera changes on the Image content is examined, depending on the version offer a relatively high degree False alarm security. Often they have outputs to the alarm signal for easy automatic To provide further processing. Some of these sensors leave a selection and logical linkage of the image areas to be monitored. There is a local one Dependence of the monitored image areas on the overall image of the camera.

Methods are also known in which several cameras for monitoring the same Object area. This is a dependency of the individual information given by the cameras from this object area.

From DE 195 09 277 C1 a method for robot control is known that by indirect Measurement of quantities - in this case pixel intensities or their derivatives and over functional relationships in the form of systems of equations to the actual so-called "desired" size (in this case the position of the robot in the room) closes. By skillful weighting and combination of the equations and the measured quantities can be achieved now maximizing the information gain of the "desired" sizes. It is interesting Method in that it uses a relatively high computing power Gain information in less time than conventional image processing algorithms having. This is achieved, among other things, by avoiding iterations in the manufacturing Algorithm. In contrast, there are the disadvantages such. B. pre-setting the Systems of equations and a relatively high requirement for the resolution of the sensors as well a constant intensity of the individual pixels regardless of Observation point of view. For outdoor surveillance with its strongly changing This procedure is therefore not suitable for environmental influences. Furthermore, the described approach only applicable to imaging sensors, d. H. he does not allow  Linking the signal information "alien sensors" such. B. of infrared sensors with those of microwave sensors.

Similar restrictions are known from DE 30 44 688 C2. The patent describes generally applies a combined monitoring circuit based on digital and analog a 2 out of 3 system. The task is to monitor three similar channels using three sensors. The digital part of the system works in conventional technology, d. H. provides an error signal when the limit values of two channels are exceeded. About the analog In the event of a fault, the circuit part is determined by mutual channel signal subtraction Deviation of one channel from the other two, even if none of the 3 limit switches has addressed. The subsequent interconnection of the digital outputs with those of the analogue, allows an increase in responsiveness while being insensitive to individual tolerances of the individual channels. A disadvantage for a general one Use in surveillance systems affects the fact of exclusive Possibility of combining similar channels (similar sensor information) due to "hard difference formation", as well as the "hard comparisons" made with it. The one with it Additional information obtained therefore remains local to that monitored in the example there Machine part limited.

This procedure and intelligent decentralized signal filtering in the sensors have the following disadvantages:

• 1. Sensors are often installed separately from each other (e.g. the different ones) Sensor arrays of several fence panels), however, they can have a certain logical dependency be assigned. For example, consider a building with one camera and one each your assigned sensor on two opposite sides of the building (e.g. north and south walls), from the point of view of the cameras to one another and the sensors to one another, one cannot see one speak local dependency, since the north and south side are not in the same detection area both sensors (and therefore not in the same detection range of both cameras) lie. However, this is a certain logical connection between the signals from the sensors exists, you can see immediately when z. B. a simultaneous attempt at penetration and thus one Simultaneous alarm signaling of both sensors on the doors possibly monitored by the sensors of the building. Conventional circuit variants solve this problem (also for more than two sensors) through multi-alarm treatment. However, this means that the relatively high  Alarm trigger threshold for each of the (in the case described two) sensors must be exceeded. If the signals detected by the sensors are only slightly below theirs alarm triggers, there is no signal transmission to the control center. In order to an additional information gain becomes impossible. According to the current state of the art can be said that the sum of the information from the actual sensor signals, the Locations of the sensors at the time of the events and the functional principles of the In principle, sensors do not characterize the "overall status" of the system (from which then corresponding actuating actions for output channels can be derived), but themselves primarily on the evaluation of the sensor signals (with an assignment to individual Output channels or groups of output channels in addition to corresponding actuating actions) limited.
• 2. The prior art tends to sensors with ever higher intelligence equip to detect and suppress false alarms. That means on the one hand, on the other hand Partly considerable computing power as well as hardware and software for the sensors and associated with very high sensor costs (This is particularly clear from the state of the art with video sensors.). On the other hand, any signal filtering within the sensors (with the goal of false alarm suppression there) inevitably leads to a loss of information for the Overall system. High alarm trigger thresholds result in good false alarm suppression, however, they also carry the risk of not recognizing (especially weak) "real alarms".

The object of the invention is to develop a monitoring system with offline Event tracking for automatic control of video and / or Based on audio monitoring devices, which avoids the disadvantages described the loss of information due to signal filtering within the sensors with the aim of false alarm suppression there, based on technically simple and therefore cheaper sensors and Hardware, on a computing machine with low computing power, good false alarm suppression offers at the same time high real alarm detection.

This object is achieved in connection with the preamble of patent claim 1 solved by the characterizing features of claim 1.  

According to the invention, sensors are not divided into locally connected groups, but instead in logically connected groups resulting from the task. With that you can additional information (e.g. by evaluating the location of the sensor at the time of its Signals as well as their functional principle) to the already existing sensor signals. Since there is deliberately no strong signal filtering in the sensor itself (deliberate avoidance of intelligent sensors and their attempt to suppress false alarms) and the optional Prefilter only for filtering signal transmission interference or deliberate data reduction of the signal (lower bandwidth for signal transmission, higher signal-to-noise ratios) are used this additional information with a high influencing factor in the system. Further increases The gain in additional information can be achieved if the sensors are given local degrees of freedom to align them with the help of sensor positioning actuators at locations that supply alarm signals or to position. This procedure leads to an increase in the amount of information of the overall system, which in various forms such. B. for range or Increased sensitivity of sensors or direction evaluation of the alarming object can be implemented. Since the information from a sensor is in principle an influencing factor on the If the state of the overall system is considered, it can be seen that one and the same sensor in principle can be assigned to several (or all) logical groups.

Since there is generally no precise functional connection to the combination of the information the signals from the sensors can be given, but very often empirical values for Are available, the use of a "soft" combination and evaluation by means of Fuzzy logic or the use of neural networks if the later location of the system is still uncertain or you want to forego previous parameterization (are advantageous also combinations of the two techniques mentioned!). In the latter case you can reach through special training of the network an optimal combination of the individual signals of the sensors among themselves. The results of the continuous evaluation of the sensor signals and the Additional information about your logical group can be found in the logical group assigned to it Output channel again and provide the basic information after the post-filtering Activation of the devices connected to the physical output channels. Im In contrast to the prior art, the sensor signals are not filtered according to the invention in the sensor, with the aim of suppressing false alarms there, but filtering the Information of a logical output channel (post-filtering) with subsequent output via the corresponding physical output channel with the result of a better alarm False alarm differentiation with the same or higher sensitivity of the entire system. One serves  recording devices with pre-alarm recording capability continue to be so Control (directly or indirectly via additional devices) after filter longer Use filter runtime, which in addition to the advantage of better quality filter results by integrating the information of the logical output channels over time, the advantage of a lower computing load on the computing machine. The through the filter runtimes Conditional late announcement at the start of the recording can be made by using his Pre-alarm recording ability can be compensated.

Advantageous for the purpose of suppressing signal transmission interference or one Data reduction of the sensor signals prove to be directly in front of the input channels of the Calculator or pre-filters connected directly after the signal outputs of the sensors after a corresponding signal conversion performed by the input channels and thus in a format that can be processed by the computing machine has changed to an essential one Reduce the computing machine's workload.

If sensors have several different (disjoint) signal outputs, they can be used Gain additional information, each assigned to an input channel in the sense of the invention will. For example, infrared sensors (IR sensors) often have additional ones Brightness sensors, the signal information z. B. for the purpose of parameterization (e.g. Day / night operation) of the post-filter according to the invention can be used. The same applies for combined IR microwave sensors. Here can be due to their different Operating principles with simultaneous signal evaluation of both elements pull the direction of movement of the signal-triggering object. The Doppler effect of the Microwave sensors allow very precise conclusions to be drawn about object movements towards the sensor or away from the sensor. IR sensors are particularly sensitive to heat emitting Objects with direction of movement parallel to the sensor.

Digital sensor signals offer the advantage of a relatively interference-free transmission, however allow analog sensor signals to convey a high density of information. Choosing the type of The sensor signal used is expediently decided on a case-by-case basis.

For a later evaluation of the sensor information on the input channels, with the goal of exact tracing of events without loss of information, due to the filtering effect of the system according to the invention, is a logging of the same special advantage. Since a relatively large amount of data from sensors is generated in this way, it turns out  a storage in an automatically processable by another computing machine Format as particularly useful.

Calculating machines prove to be economically and technically efficient Hardware basis of microcontrollers, computers, transputers and fuzzy controllers with Appropriate software or firmware based on neural networks or fuzzy logic. This algorithmic approaches still allow an amazingly simple realization of the Linking the information of the individual input channels.

According to the invention, simple, solid are sufficient for locations with low interference parameterized post-filter. With changing environmental conditions (e.g. day-night change) it is This is particularly advantageous if the system is set to the optimally adjust to new circumstances. The parameterization can be event-driven or time-controlled. Become fast post-filters with low runtime delays hardware implementations or combined hardware soft are recommended realizations. Are particularly inexpensive and easy to parameterize software post-filters that can be advantageously implemented in the computing machine itself. In addition to logging the signals of the input channels, there is an audiovisual recording Using video recorders and tapes for offline event tracking is an advantage. Especially recording devices with pre-alarm recording capability such as B. digital Video recorder, as a result the use according to the invention is very slow, but very much so false alarm-proof, low computation-intensive post-filter is made possible.

Additional actuators expand the area of application (lamps and infrared heaters enable e.g. Night operation; Modems allow automatic alarming of remote centers; . . .).

In addition, the sensors and actuators of the system are granted local degrees of freedom (for example, by installing on a pan-tilt arm of a camera), so improves this means the overall information of the system on the one hand, and on the other hand by the actuators targeted actions at the locations that supply the alarm signal, such as B. switching on lighting be performed.  

The advantages of evaluating additional information and post-filtering according to the invention are listed in the following examples.

Further details of the invention can be found in the following drawings.

Here shows

Fig. 1 shows an embodiment of the system

Fig. 2 shows a schematic representation of the use of two sensors to the sensor range increase for the same quality of the false alarm detection alarm by evaluation of additional information.

example 1

In Fig. 1 a an embodiment of the system is shown.

S1 to S5 represent the sensors of the system, with S4 being a sensor with 2 possibly disjoint outputs a and b. Output a indicates z. B. the Output of an internal infrared sensor element that is particularly sensitive to Changes in movement parallel to the sensor react, output b provides the signal Microwave sensor element available, which is particularly sensitive to approaches to or distances from the sensor responds. The signals come from the sensors via the optional pre-filter F1 to F5 in the input channels E1 to E6 of the computing machine RM to get there to be converted into a format that can be processed. S5 delivers its signals directly, i.e. H. without pre-filtering, to the input channel E6 of the RM calculator. In the RM the Signals processed accordingly to the via the logical output channels LAK1 to LAK3 Post-filter NF1 to NF3. Furthermore, a logging of E1 to E6 takes place pending signals by the RM with storage in the protocol PR, indicated by the Data flow D. The post-filters NF1 to NF3 decide whether and if so how the physical ones Output channels A1 to A3 their manipulated variables for controlling the devices connected to them provide. Indirect control of the control signals generated in A1 is via the Signal flow from A1 via a relay RE to the sensor positioning actuator SPA shown. This deals with the positioning of sensor S1 in the system - indicated by the dashed line Line P. A2 and A3 directly control the alarm inputs AL1 and AL2 of a digital one Video recorder G, the two video inputs VE1 and VE2 with the cameras K1 and K2 are connected.  

Example 2

FIG. 2 shows an example of increasing the sensor range while maintaining the quality of the alarm false alarm detection by using two sensors and evaluating their additional information. S1 and S2 identify the sensors with their associated physical detection areas A1, A2 and A12 for a given signal triggering threshold SA1. The range of the physical detection areas of S1 and S2 is designated by L1 taking SA1 into account. If the signal trigger threshold is reduced from S1 and S2 to SA2 <SA1, the probability of false alarms increases with conventional systems with an increased physical range (indicated by L2) of the sensor signal acquisition. In the present case, the signals of both sensors are evaluated simultaneously in the computing machine at SA2 with subsequent post-filtering, which again results in a theoretical detection range A1 and A12 for S1 and A2 and A12 for S2, ie the theoretical range L1, for each individual sensor. For simultaneous signals in the common physical detection range A12 and AZ12, the theoretical range is increased by path L2 or the theoretical detection range increases by AZ12 with a false alarm probability unchanged compared to SA1. Using the system described in FIG. 1 (without S4 and S5), range increases of 10-30% were achieved with the same parameterization of the postfilter.

Equivalent to the statement of a sensor range increase with the same quality of the alarm False alarm detection is the wording of improving the quality of the alarm False alarm detection with constant sensor range. The following, based on fuzzy logic based calculation example serves for a more detailed explanation. Vectors e are between 0 and 1 normalized sensor measured values of sensors S1 and S2. a denotes the calculated one Output vector (contains the information of the logical output channels) for further Processing the subsequent manipulated variables for actuators etc. T sets the rules in the form a table with the following freely selected values. The threshold whether a signal is an alarm or is classified as a false alarm, is freely set to 0.4 for the individual values of vector a Are defined.  

Policy table T

For signal values of e = (0.7, 0), a vector a = results according to the average-minimum method (0.35, 0.25). Since both values are below the switching threshold of 0.4, a classification is made as a false alarm. Canonical results for e = (0, 0.7) a to a = (0.25, 0.35) with an analogous classification. With simultaneous signals of the valence 0.7, ie with e = (0.7, 0.7), the same results Method a = (0.6, 0.6) d. H. it is classified as an alarm because of 0.6 <0.4.

Example 3

Using the system described in Fig. 1 (without S4 and S5), the parameterization of the post-filter with filter runtimes of only 2s resulted in improved false alarm suppression (depending on the weather conditions of the same monitored outside area) by factors between 1 and 10 compared to conventional filtering in the sensor with almost unchanged ability to detect real alarms. The average factor of a long-term test over several weeks was around 2. Longer filter run times resulted in an even better false alarm detection.

Reference list

Fig.

1
S1, S2, S3, S5 sensors
S4 sensor with disjoint signal outputs a and b
a, b disjoint signal outputs of sensor S4
F1-F5 prefilter
E1-E6 RM input channels
RM calculator
LAK1-LAK3 logical output channels of the RM
NF1-NF3 post filter of the RM
A1-A3 physical output channels of the RM
RE relay
SPA sensor positioning actuator
P mechanical control of the sensor

1

(schematic)
D data flow
PR protocol
G digital video recorder
AL1-AL2 alarm inputs of the G
VE1-VE2 video inputs of the G
K1-K2 cameras

Fig. 2
S1-S2 sensors
A1, A2, A12, AZ12 detection areas
L1-L2 ranges

Claims (21)

1. Circuit arrangement and method for monitoring systems with offline event tracking of locations with a high probability of false alarms with two or more input channels, a computing machine and one or more filtered physical output channels for controlling one or more video and / or audio monitoring devices and / or sensor positioning actuators, characterized in that not only the optionally pre-filtered signals of one or all of the sensors connected to the input channels, optionally automatically positionable, not necessarily of the same type and / or type and / or function, but also their logical relationships within the overall system for the respective automatic calculation of the individual output variables of one or more physical output channels, each with a downstream post-filter for each physical output channel, for direct or indirect control of video and / or audio monitoring devices and / or sensor positioning actuators can be used. 2. Monitoring system according to claim 1, characterized in that the signals from the sensors connected to the input channels are both pre-filtered and are not pre-filtered and through the input channel into a processable for the computing machine Format to be changed. 3. Monitoring system according to one or more of claims 1 to 2 thereby characterized in that each sensor has several different (disjoint) signal information can deliver, which are then each assigned to a separate input channel. 4. Monitoring system according to one or more of claims 1 to 3 thereby characterized in that the signals of the sensors connected to the input channels are analog or are digital. 5. Monitoring system according to one or more of claims 1 to 4 thereby characterized in that the signals of one to all sensors connected to the input channels with information and / or time of arrival and / or duration, constantly logged and permanent to be recorded.   6. Monitoring system according to one or more of claims 1 to 5 thereby characterized in that the automatic calculation of the logical output channel sizes in one Calculator is done. 7. Monitoring system according to one or more of claims 1, 5 and 6 thereby characterized in that the signal protocol of the same and / or another Calculator is automatically captured online and / or offline. 8. Monitoring system according to one or more of claims 6 to 7 thereby characterized in that the computing machine on a microcontroller and / or fuzzy controller and / or computer and / or transputer with appropriate software and / or firmware based. 9. Monitoring system according to one or more of claims 6 to 8 thereby characterized in that the software and / or firmware based on fuzzy logic and / or based on neural networks. 10. Monitoring system according to one or more of claims 1 to 9 thereby characterized in that the post-filters are permanently parameterized or freely parameterizable. 11. Monitoring system according to claim 10, characterized in that the post-filter for the event of their free parameterization by the computing machine event-dependent and / or can be parameterized depending on the time. 12. Monitoring system according to one or more of claims 10 to 11 thereby characterized in that the post-filters are implemented in software and / or hardware. 13. Monitoring system according to one or more of claims 10 to 12 thereby characterized in that the post-filtering is carried out by the computing machine. 14. Monitoring system according to one or more of claims 1 to 13 thereby characterized in that the output signal of the physical output channels for direct and / or indirect control of video and / or audio monitoring devices and / or Sensor positioning actuators is used.   15. Monitoring system according to claim 14, characterized in that the video and / or Audio monitors have pre-alarm recording capability. 16. Monitoring system according to claim 15, characterized in that the video and / or audio monitoring device is a digital video recorder. 17. Monitoring system according to one or more of claims 1 to 16 thereby characterized in that the output signal of the physical output channels for direct and / or indirect control of additional devices of the video and / or audio surveillance construction is used. 18. Monitoring system according to one or more of claims 1 to 17 thereby characterized in that the sensors and / or actuators of the system by sensor positioning actuators the system can be freely positioned. 19. Monitoring system according to one or more of claims 1 to 18 thereby characterized in that the output signal of the physical output channels for direct and / or indirect control of signal generators and / or actuators of alarm systems becomes. 20. Monitoring system according to one or more of claims 1 to 19 thereby characterized in that the output signal of the physical output channels for direct and / or indirect control of any signal generators and / or actuators is used. 21. Monitoring system according to one or more of claims 1 to 20 thereby characterized, the algorithm underlying the post-filtering is a non-linear Filter algorithm is.