EP4632711A1

EP4632711A1 - Method and system for anomaly detection

Info

Publication number: EP4632711A1
Application number: EP24168980.1A
Authority: EP
Inventors: Ju Hoon Kim; Ming Yin
Original assignee: Deutsche Telekom AG
Current assignee: Deutsche Telekom AG
Priority date: 2024-04-08
Filing date: 2024-04-08
Publication date: 2025-10-15

Abstract

A method for anomaly detection is proposed, comprising the steps of:
a) determining (711) object location information of a plurality of objects (620, 630, 640, 650) within a predetermined spatial area,
b) determining (721, 722, 731, 732, 733) object movement information of the plurality of objects (620, 630, 640, 650) based on the determined object location information,
c) based on the determined object movement information, selecting (744) at least one (640, 650) of the plurality of objects,
d) identifying at least one of the selected objects as a human object (650') based on detectability of at least one vital characteristic of the selected object (640, 650),
e) determining (741, 742) vital characteristics of the identified human object (650'), wherein
- steps a) to e) are repeated in pre-defined time intervals, and
- an anomaly is detected (743), if vital characteristics are determined in step e), which differ from pre-determined vital parameters defining a normal vital condition.

Further, an anomaly detection device (100) and an anomaly detection system (30) for performing the method are proposed.

Description

The invention generally relates to monitoring of human objects and especially to detecting a safety incident, such as a medical emergency of a human worker in a factory. In particular, the invention relates to a method and a system for anomaly detection.

Background

The current trend indicates a growing inclination towards constructing industry facilities in a more autonomous fashion, such as the implementation of smart factories and autonomous logistics. Consequently, these advancements result in factories operating with significantly reduced human involvement when compared to conventional factory structures. However, in the majority of cases, human intervention is still necessary for carrying out maintenance work. Functional safety measures are employed to minimize the risk of physical injury or of damage to the health of people, comprising safety functions on technical equipment.
However, with increasing automation there is a possibility of a single individual working without the supervision of other human colleagues in a vast section of a factory. In a traditional setup of industrial facilities where all tasks are performed by humans, safety incidents or emergencies, such as a heart attack, involving a human worker could be readily identified and communicated by fellow human colleagues. Yet, in an autonomous industry facility like a smart factory, it is less probable to receive notifications from other human workers.
It is known to employ video surveillance systems with distributed cameras to monitor workers in a factory, wherein a human supervisor monitors the factory from a control room. Instead of a human supervisor, also autonomous monitoring may be employed based on video surveillance, wherein machine learning (ML) techniques may be utilized for this purpose. It is, however, difficult for such a system to detect the condition of a human object.
Further surveillance techniques comprise aerial surveillance utilizing camera drones or the utilization of autonomous robots for surveillance purposes. Such methods, however, are difficult and expensive to maintain.

Summary of the Invention

It is an object of the present invention to provide an improved and/or simplified way for detecting a safety incident or medical emergency of a human, for example of a human worker in an industrial environment.
A core aspect of the invention can be seen in providing a detection mechanism, which analyzes the movement patterns of detected objects and based thereon filters out non-human objects, so that detailed analysis is focused only on detected human objects, wherein preferably quantum sensors are utilized in order to increase sensing precision and resolution and to enable the detection of vital characteristics of a detected human, such as heart beat rate. Furthermore, quantum sensors preferably are employed to determine environmental parameters, such as temperature, moisture or the presence of certain substances.
The technical problem cited above is solved by the features of the independent claims. Preferred embodiments are set forth in the dependent claims.
Accordingly, a method for anomaly detection is provided, which comprises the steps of

a) determining object location information of a plurality of objects within a predetermined spatial area,
b) determining object movement information of the plurality of objects based on the determined object location information,
c) based on the determined object movement information, selecting at least one of the plurality of objects,
d) identifying at least one of the selected objects as a human object based on detectability of at least one vital characteristic of the selected object,
e) determining vital characteristics of the identified human object,

steps a) to e) are repeated in pre-defined time intervals, and
an anomaly is detected, if vital characteristics are determined in step e), which differ from pre-determined vital parameters defining a normal vital condition.

Step a) may in particular comprise monitoring of the predetermined spatial area, automatically detecting objects within the monitored spatial area and determining location information of each detected object, e.g. in the form of a position in a pre-defined coordinate system. Monitoring may preferably comprise capturing at least one digital image with a camera. Preferably, digital images are captured by several different cameras in regular time intervals, wherein each camera is arranged at a different location, so that a different part of the predetermined spatial area is respectively captured. However, any suitable sensor devices may be utilized for determining object location information in step a). For determining object location information, captured sensor data, e.g. captured digital images, preferably is processed by a central anomaly detection device, to which the sensor device is connected. However, processing or pre-processing may also be performed directly by the respective sensor device.
For the normal condition preferably at least one threshold value for at least one specific vital parameter may be defined, wherein an anomaly is detected, if the determined vital characteristics comprise a value for the respective vital parameter which exceeds the threshold value. The normal condition preferably is represented by a threshold value for at least one vital parameter, wherein the at least one vital parameter may preferably comprise body temperature, heart beat rate, respiration rate, blood oxygen saturation, movement or body position. Therefore, an anomaly may for example be detected, if a significantly increased heart beat rate is determined for a detected human object exceeding a pre-defined normal value, and/or if no movement is determined for the detected human object and/or a horizontal body position is detected of the detected human object.
Preferably, the vital characteristics detected in step e) are stored and based on the stored vital characteristics of human objects the vital parameters and respectively associated threshold values are determined, which define a normal vital condition. In other words, the at least one threshold value for at least one specific vital parameter defining a normal vital condition may preferably be automatically learned based on a history of stored vital characteristics of human objects.
In a preferred embodiment, environmental parameters are determined and/or a notification or alarm is generated, if an anomaly is detected. Environmental parameters, which may for example be determined, comprise a temperature, a moisture level and/or the presence or content of certain chemicals in the air, e.g. the oxygen or carbon dioxide content. The notification or alarm preferably is automatically transmitted to a safety administration center, such as a hospital or an emergency center. In order to react to the emergency optimally, the detected vital characteristics of the detected human object that triggered the alarm and /or the determined environmental parameters are transmitted within the notification or alarm message.
For determining the vital characteristics in step e) and/or for determining the environmental parameters preferably a sensor device is used, which comprises at least one quantum sensor. Quantum sensors are devices that utilize quantum phenomena to measure physical quantities with high precision and sensitivity, wherein in particular principles of quantum mechanics, such as superposition and entanglement, are exploited to achieve performance beyond classical limits.
An example of a quantum sensor is a so-called quantum dot sensor. Quantum dots are nanoscale semiconductor particles that exhibit quantum mechanical behavior. Quantum dot sensors are devices that utilize the unique properties of quantum dots to detect and measure various physical quantities, wherein the working principle of quantum dot sensors can vary depending on their specific application.
Quantum dot sensors can with advantage be used as image sensors with a very high sensitivity. They can also be used for chemical sensing, wherein quantum dots for this purpose can be functionalized with specific molecules or ligands that selectively bind to target substances, e.g., gases or biomolecules. It is to be noted that not only quantum dot sensors may be utilized, but also any other suitable quantum sensor. Due to their high sensitivity, quantum sensors can also preferably be utilized for determining vital characteristics of a human object from a distance. Detectable characteristics comprise for example heartbeat rate, body temperature, respiratory rate and/or blood oxygen saturation.
Another example of a quantum sensor is a Rydberg atom-based sensor, which utilizes the sensitivity of Rydberg atoms to oscillating electromagnetic fields, so that the sensor is particularly useful in high-resolution imaging and the detection of electromagnetic signals.
In addition to the examples given, there are further types of quantum sensors with quantum sensing being a rapidly evolving field. In general, any suitable, also future quantum sensor may be employed for the invention.
It is to be noted, however, that a quantum sensor preferably is only utilized, if the required sensing precision and/or resolution cannot be provided by any other type of sensor, since the implementation of quantum sensors still involves a high level of effort and cost.
For determining the object location information of the plurality of objects within the predetermined spatial area in step a), camera images may be analyzed using known object recognition techniques, wherein for this step the high sensitivity of quantum sensors is not necessarily needed. The sensor device therefore may comprise different sensors, e.g. a camera. It is to be noted that the sensor device may comprise a combination of a sensor, such as a camera, and at least one quantum sensor.
In a preferred embodiment, the anomaly detection method performs a pre-filtering of static objects, i.e. objects that do not move, and of objects, which exhibit a predictable movement pattern. That way, the anomaly detection can preferably be concentrated on human objects, thereby enhancing efficiency, accuracy, and the frequency of detection.
Accordingly, step b) preferably comprises identifying static objects and storing object location information of the identified static objects, and identifying moving objects that are moving in a pre-defined movement pattern and storing object location information and/or movement information of the identified moving objects. Preferably, object location information and/or movement information is stored only for newly detected objects.
In a preferred embodiment, a digital map is generated comprising a digital representation of the plurality of objects, wherein in step c) the digital map is modified to comprise only the selected objects, wherein the selected objects are potentially human objects.
The method is utilized with advantage for detecting an anomalous condition of a human object in an industrial environment. In such an environment, the plurality of objects typically comprises objects, which are components of an autonomously operating industrial system. Typically, the majority of the objects are either static objects or objects that exhibit a static or predictable movement pattern, such as for example autonomous machines.
An anomaly detection device comprises a processor, a memory and at least one interface for connecting to a sensor device, wherein in said memory executable instructions are stored, which are adapted, when executed by said processor, to perform the steps of:

An anomaly detection system comprises the described anomaly detection device and at least one sensor device connected to the anomaly detection device.
It is to be noted that the anomaly detection device and/or the anomaly detection system preferably are configured in such a way that they are adapted to perform any aspect described above with respect to the method.

Brief Description of the Drawings

The invention is described in detail below by way of preferred embodiments in connection with the accompanying drawings, wherein

Figure 1: schematically shows a preferred, but exemplary anomaly detection system, which is adapted to perform the inventive method,
Figure 2: schematically shows an exemplary process of filtering objects within a predetermined spatial area,
Figure 3: shows the exemplary anomaly detection system of Fig. 1, schematically depicting functional modules of the anomaly detection device, and
Figure 4: schematically shows an exemplary flowchart of the communication between the devices and functional modules involved in a preferred embodiment of the method.

Detailed Description of exemplary Embodiments

Referring now to figure 1, an exemplary anomaly detection system 30 is schematically depicted.
In the shown exemplary embodiment, the system 30 comprises an anomaly detection device 100, to which sensor devices 201-207 are connected, wherein in the shown embodiment sensor devices 201-207 are connected to the anomaly detection device 100 via a controller 10, wherein the controller 10 is adapted to collect and forward data from the sensor devices 201-207. Preferably, the controller 10 may perform a pre-processing of the received sensor data and transmit the pre-processed data to the anomaly detection device 100. In the shown embodiment, all sensor devices 201-207 are connected to a common controller 10, which is connected to the anomaly detection device 100, but it might also be provided a separate controller for each sensor device, wherein the controller may also be integrated in the respective sensor device, in which case the sensor devices are connected directly to the anomaly detection device 100.
For the communication connection between one of the sensor devices 201-207 and the controller 10 and/or between the controller 10 and the anomaly detection system 100 any suitable wired or wireless communication technique may be utilized.
The anomaly detection device 100 comprises a processor 150, a program memory 140 and a data memory 130 and at least one interface 160 for connecting to a sensor device, wherein the interface 160 preferably is adapted for communicating via a wired or wireless communication network with at least one sensor device, such as one or more of sensor devices 201 to 207, directly or via a controller 10 of the sensor device. In the shown embodiment, the anomaly detection device 100 has one interface 160 for connecting to the controller 10, wherein the interface 160 may for example be a WLAN (Wireless Local Area Network) interface.
In the shown embodiment, in the program memory 140 executable instructions are stored, which are adapted, when executed by the processor 150, to perform the steps of:

Preferably, the determined object location information, the determined object movement information and the determined vital characteristics are stored, either in a memory of the anomaly detection device 100, such as data memory 130, or in an external storage, to which the anomaly detection device 100 has access, such as an external database 300. The external database 300 may for example be located within a LAN (Local Area Network), to which the anomaly detection device 100 is connected. For this purpose, the anomaly detection device 100 preferably comprises a respective interface 170.
When an anomaly is detected, preferably a notification or an alarm message is automatically transmitted to a safety administration instance 20, in particular if the detected anomaly represents a human in the condition of a medical emergency. For this purpose, the anomaly detection device 100 is connected to the safety administration instance 20, wherein in the shown embodiment the anomaly detection device 100 is connected to the safety administration instance 20 exemplarily via the Internet 400, wherein the anomaly detection device 100 comprises a respective interface 180 for connecting to the Internet 400. The anomaly detection device 100 might also be connected to the safety administration instance 20 by any other suitable communication connection.
The safety administration instance 20 may for example be a monitoring center, wherein by the notification or alarm message surveillance personnel is informed of the detected anomaly and is thus enabled to make an emergency call. With advantage, the notification or alarm message may also be transmitted directly to an emergency center or to a hospital. Accordingly, the safety administration instance 20 may advantageously also be an emergency center or a hospital.
Although not shown in Fig. 1, the external database 300 may for example also be a cloud-based database, to which the anomaly detection device 100 has access via the Internet 400.
In a preferred embodiment, a digital map is generated of the plurality of objects, which are detected and for which in step a) respective object location information is determined. The digital map preferably comprises a digital representation of the plurality of objects, wherein in step c) the digital map is modified to comprise only the selected objects, wherein the selected objects are potentially human objects.
For selecting objects, with advantage a pre-filtering is performed of static objects, i.e. objects that do not move, and of objects, which exhibit a predictable movement pattern. That way, the anomaly detection can preferably be concentrated on human objects, thereby enhancing efficiency, accuracy, and the frequency of detection.
Fig. 2 shows schematically shows an exemplary process of the pre-filtering of objects within the predetermined spatial area.
Fig. 2 a) schematically shows a pre-defined spatial area 601, such as a production hall or storage hall of a factory. In the shown example, seven sensor devices 610 are arranged at different positions of the pre-defined spatial area 601, preferably in such a way that a complete monitoring of the pre-defined spatial area 601 is enabled. As necessary, depending on the number of objects and of the size of the pre-defined spatial area 601 size any suitable other number of sensor devices might be employed.
In the shown example, in the pre-defined spatial area 601 different objects are present, comprising static objects 620, such as permanently installed inventory, objects 630, which are moving in a predictable movement pattern, such as machines or robot arms, and objects 640 and 650, which are moving in an unpredictable movement pattern. In the shown example, an autonomously guided vehicle 640 and a human object 650 are present.
In a first filtering step, the static objects 620 are filtered out, preferably removing the static objects 620 from the digital map. The result of this filtering step is shown in Fig. 2 b). If the location of a detected object does not change over consecutive detection instances, it is identified as a static object, wherein preferably the locations of identified static objects are stored, e.g. in the data memory 130 or in the database 300. That way a detected object may advantageously be identified as a static object by comparing the detected location with the stored locations of previously detected static objects.
In a second filtering step, objects 630 having a predictable movement pattern are filtered out, preferably removing these objects 630 from the digital map. The result of this second filtering step is shown in Fig. 2 c). For identifying a predictable movement pattern of an object, the object location preferably is monitored during a learning period with a pre-determined duration. A predictable movement pattern may for example be an object moving along a certain trajectory or moving within a restricted movement area. If an object is identified as an object, which moves in a predictable movement pattern, information on the movement pattern is also preferably stored, e.g. in the data memory 130 or in the database 300.
That way, the anomaly detection can preferably be concentrated on human objects, thereby enhancing efficiency, accuracy, and the frequency of detection.
Fig. 2 c) shows an example of objects that are selected in step c), wherein in the shown example the selected objects comprise objects 640 and 650, wherein object 640 is an autonomously guided vehicle (AGV) that may move in a an unpredictable manner, at least unpredictable for the anomaly detection device 100, and object 650 is a human object.
In step d) the object 650 is identified as a human object based on detectability of at least one vital characteristic of the selected object 650, wherein the object that has been identified as being a human object is shown with reference numeral 650'.
A vital characteristic may for example be a detectable temperature corresponding to the body temperature of a human object, an object form corresponding to the form of a human object, or a detectable heart beat rate, respiration rate or blood oxygen saturation. Preferably, the at least one detectable vital characteristic comprises at least one detectable vital parameter.
Sensor data for detecting vital characteristics preferably are requested by the anomaly detection device 100 only for the selected objects. For this purpose, a request may be transmitted to at least one dedicated sensor device, which may with advantage be a quantum sensor device, to capture sensor data for the selected objects, wherein the request may comprise the information on the current location of the respective selected object.
Preferably, the vital characteristics detected in step e) are stored, e.g. in the data memory 130 or in the database 300, and based on the stored vital characteristics of human objects the vital parameters are determined, which define a normal vital condition.
The normal condition preferably is represented by a threshold value for at least one vital parameter, wherein the at least one vital parameter may preferably comprise body temperature, heart beat rate, respiration rate, blood oxygen saturation, movement or body position. An anomaly preferably is detected, if the determined vital characteristics comprise a value for the respective vital parameter, which exceeds the threshold value. Accordingly, an anomaly may for example be detected, if a significantly increased heart beat rate is determined for a detected human object exceeding a pre-defined normal value, and/or if a termination of movement is determined for the detected human object and/or a horizontal body position is detected of the detected human object.
In the shown embodiment, in the program memory 140 executable instructions are stored, which are adapted, when executed by the processor 150, to perform the above-described steps.
Fig. 3 shows the exemplary anomaly detection system 30 of Fig. 1, schematically depicting functional modules 510, 520, 530, 540 and 550 of the anomaly detection device 100. In particular, the functional modules 510, 520, 530, 540 and 550 are software modules, which are executed when the executable instructions stored in the program memory 140 are executed by the processor 150. In Fig. 3, storage modules 310 and 320 are depicted, each of which may for example be the data memory 130 or the database 300.
In Fig. 4, schematically an exemplary flowchart of the communication between the devices and functional modules involved in a preferred, but exemplary embodiment of the method is shown. In the shown embodiment, the controller 10, the functional modules 510, 520, 530, 540 and 550 of the anomaly detection device 100, the storage modules 310 and 320 of the anomaly detection device 100, and the emergency center 20 are involved in the communication.
In step 701, at regular time intervals general sensing information is transmitted from the controller 10 to the anomaly detection device 100. The transmitted information in particular comprises sensor data generated by one or more of the sensor devices 201-207 and/or data derived from such sensor data. The information received by the anomaly detection device 100 is passed to the functional module 510, which processes the data. The functional module 510 in step 711 in particular performs object recognition, i.e. it identifies and localizes objects based on the data, wherein the data preferably comprises image data. With advantage, a digital map is generated comprising a digital representation of the plurality of objects.
In step 712, the functional module 510 transmits, preferably for each detected object individually, a look-up request to the storage module 310 to look up, whether the respective detected object is stored in storage module 310 as a static object. In step 713, a respective response is transmitted back from the storage module 310, indicating whether the respective object is stored as a static object. Each of the detected objects, for which a response is received indicating that the respective object is a static object, is excluded from further processing, wherein it is in particular deleted from the digital map.
In step 714, data of the remaining detected objects is forwarded from the functional module 510 to the functional module 520, wherein in particular the modified digital map is forwarded, in which all objects have been deleted that are stored in storage module 310 as static objects.
Each detected object that is not stored in the storage module 310 as a static object, preferably is put under inspection on whether the respective object is a static object. For this purpose, preferably location information is stored in the storage module 310 together with a time information that indicates at which point in time the respective object was at the respective location. That way, a static object can be newly identified, for example if the location of the respective object does not change for a pre-defined duration.
In step 721, the functional module 520 transmits, preferably individually for each detected object in the modified digital map, a look-up request to the storage module 310 to look up, whether the respective object is still under inspection on whether it is a static object, and if so, how much time has passed since the inspection has begun and how the location of the object has changed during inspection. A response is transmitted in step 722, wherein the response may preferably comprise a list of points in time with corresponding locations, e.g. in the form of coordinates. An inspection time, i.e. an information on how much time has passed since inspection has begun, is determined based on the first entry of the list and the current time. The inspection time may also be stored and transmitted back within the response, wherein the inspection time is regularly updated.
Based on the response it is checked whether the object under inspection is a static object. If movement is recognized, the object is identified as a moving object. If no movement is recognized and the inspection time exceeds a pre-defined threshold amount of time, the object is identified as a static object. If no movement is recognized and the inspection time is below the threshold amount of time, the object's location information is updated, i.e. a further entry is inserted into the list, and the inspection time preferably is updated. The update information and the information on newly detected static objects is transmitted from the functional module 520 to the storage module 310 in step 723 and is stored in the storage module 310.
In step 724, the functional module 520 transmits, preferably for each detected object individually, a look-up request to the storage module 320 to look up, whether the respective detected object is stored in storage module 320 as an object moving in a predictable moving pattern. In step 725, a respective response is transmitted back from the storage module 320, indicating whether the respective object is stored as an object moving in a predictable moving pattern. Each of the detected objects, for which a response is received indicating that the respective object is an object moving in a predictable moving pattern, is excluded from further processing, wherein it is in particular deleted from the digital map.
In step 726, data of the remaining detected objects is forwarded from the functional module 520 to the functional module 530, wherein in particular the further modified digital map is forwarded, in which also all objects have been deleted that are stored in storage module 320 as objects moving in a predictable moving pattern, wherein preferably also newly detected static objects are deleted.
Each detected object, for which data is forwarded to the functional module 530, and which is not stored in the storage module 320 as an object moving in a predictable moving pattern, preferably is put under inspection on whether the respective object is an object moving in a predictable moving pattern. For this purpose, preferably location information is stored in the storage module 320 together with a time information that indicates at which point in time the respective object was at the respective location. That way, an object moving in a predictable moving pattern can be newly identified by analyzing the movement for a pre-defined duration.
In step 731, the functional module 530 transmits, preferably individually for each detected object in the further modified digital map, a look-up request to the storage module 320 to look up, whether the respective object is still under inspection on whether it is an object moving in a predictable moving pattern, and if so, how much time has passed since the inspection has begun and how the location of the object has changed during inspection. A response is transmitted in step 732, wherein the response may preferably comprise a list of points in time with corresponding locations, e.g. in the form of coordinates. An inspection time, i.e. an information on how much time has passed since inspection has begun, is determined based on the first entry of the list and the current time. The inspection time may also be stored and transmitted back within the response, wherein the inspection time is regularly updated.
Based on the response the movement of the object under inspection is analyzed in step 733 in order to recognize whether the objects moves in a predictable movement pattern. If a predictable movement is recognized, the object is identified as an object moving in a predictable moving pattern. Such an object may for example show a repetitive or continuous movement pattern. Since the movement of such an object indicates that it is a non-autonomously moving object, it is not likely a human object. If no predictable movement pattern is recognized and the inspection time exceeds a pre-defined threshold amount of time, the object is identified as a potentially human object, since the movement indicates the object to be an autonomous object. If no predictable movement pattern is recognized and the inspection time is below the threshold amount of time, the object's location information is updated, i.e. a further entry is inserted into the list, and the inspection time preferably is updated. The update information and the information on newly detected objects moving in a predictable moving pattern is transmitted from the functional module 530 to the storage module 320 in step 734 and is stored in the storage module 320.
In step 735, data of the remaining detected objects is forwarded from the functional module 530 to the functional module 540, wherein in particular the further modified digital map is forwarded, in which also all objects have been deleted that are newly identified as objects moving in a predictable moving pattern.
Functional module 540 accordingly only receives information on selected detected objects, e.g. objects 640 and 650 as shown in Fig. 2 b), with information on most of the detected objects being already filtered out. That way, the further processing may be focused on potentially human objects.
In step 741, a request is transmitted from the functional module 540 to the controller 10, preferably individually for each remaining selected detected object in the digital map, to detect vital signs of the respective object. Upon receiving the request, the controller 10 causes at least one of the sensor devices 201-207, which is adapted to detect at least one vital parameter, to capture sensor data of the respective object. For this purpose, the request may comprise information on the location of the respective object, wherein based thereon at least one of the sensor devices 201-207 is selected for capturing sensor data.
In step 742, the requested sensor data is transmitted form the controller 10 to the functional module 540. Based thereon, the functional module 540 identifies at least one of the selected objects as a human object based on detectability of at least one vital characteristic of the selected object, e.g. object 650' as shown in Fig. 2 c), wherein the at least one vital characteristic comprises at least one vital parameter, for which a value may be determined based on the sensor data.
The above-described steps are repeated in pre-defined time intervals, thus enabling an efficient monitoring of at least one vital characteristic of objects, which are identified as human objects. A continuous monitoring is performed by the functional module 540 in step 743, wherein continuously values are determined for at least one vital parameter for the respective object based on respective sensor data. The monitoring of the at least one vital parameter in particular has the purpose to detect anomalies, in which the condition of the monitored human object differs from a normal condition, which may indicate an emergency.
For the normal condition preferably at least one threshold value for at least one specific vital parameter is defined, wherein the at least one vital parameter may preferably comprise body temperature, heart beat rate, respiration rate, blood oxygen saturation, movement or body position of the human object. An anomaly is detected by the functional module 540, if the determined vital characteristics comprise a value for the respective vital parameter, which exceeds the threshold value. Accordingly, an anomaly may for example be detected, if a significantly increased heart beat rate is determined for a detected human object exceeding a pre-defined normal value, and/or if no movement is determined for the detected human object and/or a horizontal body position is detected of the detected human object.
Preferably, the detected vital characteristics are stored, e.g. in data memory 130 or in database 300. Based on the stored vital characteristics of human objects the vital parameters and respectively associated threshold values may preferably be determined, which define a normal vital condition. In other words, the at least one threshold value for at least one specific vital parameter defining a normal vital condition may preferably be automatically learned based on a history of stored vital characteristics of human objects.
If an anomaly is detected, information on the detected anomaly is transmitted from the functional module 540 to functional module 550 in step 744.
In step 751, a request is transmitted from the functional module 550 to the controller 10, to detect at least one environmental parameter, e.g., temperature, moisture, oxygen level, CO₂ level, level of toxic gases, or the like. Upon receiving the request, the controller 10 causes at least one of the sensor devices 201-207, which is adapted to detect the at least one environmental parameter, to capture respective environmental sensor data. For this purpose, the request may comprise information on the location of the respective object, wherein based thereon at least one of the sensor devices 201-207 is selected for capturing sensor data. The request may also comprise information on which type of environmental parameter is to be detected, wherein based thereon at least one of the sensor devices 201-207 is selected for capturing sensor data.
In step 752, the requested sensor data is transmitted from the controller 10 to the functional module 550. Based thereon and based on the information received from the functional module 540, the functional module 550 in step 753 analyzes the anomaly in the context of the environmental condition and preferably identifies a cause of the anomaly, e.g. a fire, a flood or lack of oxygen. Preferably, a level for the urgency and/or severity of the detected incident may be determined.
The transmission in steps 701, 741, 742, 751 and 752, which is respectively indicated in Fig. 3 with reference numerals 101, 102, 103, 104 and 105, is for example performed via the interface 160 and controlled by processor 150 as shown in Fig. 1.
In step 754, a notification or alarm is automatically transmitted to safety administration instance 20. In order to optimally react to the emergency, the identified cause of the incident and/or the determined urgency and/or severity level are preferably transmitted within the notification or alarm message. Also the detected vital characteristics of the detected human object that triggered the alarm and /or the determined environmental parameters may be transmitted within the notification or alarm message.

Claims

A method for anomaly detection, comprising the steps of:
a) determining (711) object location information of a plurality of objects (620, 630, 640, 650) within a predetermined spatial area,

b) determining (721, 722, 731, 732, 733) object movement information of the plurality of objects (620, 630, 640, 650) based on the determined object location information,

c) based on the determined object movement information, selecting (744) at least one (640, 650) of the plurality of objects,

d) identifying at least one of the selected objects as a human object (650') based on detectability of at least one vital characteristic of the selected object (640, 650),

e) determining (741, 742) vital characteristics of the identified human object (650'),
wherein
- steps a) to e) are repeated in pre-defined time intervals, and

- an anomaly is detected (743), if vital characteristics are determined in step e), which differ from pre-determined vital parameters defining a normal vital condition.
The method of claim 1, wherein in case an anomaly is detected, environmental parameters are determined (751, 752) and/or a notification or alarm message is generated (754), wherein the notification or alarm message in particular is transmitted to an emergency center (20).
The method of any one of claims 1 or 2, wherein at least one sensor device (201-207) comprising at least one quantum sensor is used to determine the vital characteristics in step e) and/or to determine the environmental parameters.
The method of any one of the preceding claims, wherein the vital characteristics detected in step e) are stored and based on the stored vital characteristics of human objects the vital parameters are determined, which define a vital normal condition.
The method of any one of the preceding claims, wherein step b) comprises
- identifying static objects and storing object location information of the identified static objects, and

- identifying moving objects that are moving in a predictable movement pattern and storing object location information and/or movement information of the identified moving objects.
The method of any one of the preceding claims, wherein a digital map (601-603) is generated comprising a digital representation of the plurality of objects (620, 630, 640, 650), and wherein in step c) the digital map is modified to comprise only the selected objects.
The method of any one of the preceding claims, wherein the plurality of objects comprises objects, which are components of an autonomously operating industrial system.
An anomaly detection device, (100), comprising
- a processor (150),

- a memory (130, 140)

- at least one interface for connecting to a sensor device (201-207),
wherein in said memory (140) executable instructions are stored, which are adapted, when executed by said processor, to perform the steps of:
a) determining (711) object location information of a plurality of objects (620, 630, 640, 650) within a predetermined spatial area,

b) determining (721, 722, 731, 732, 733) object movement information of the plurality of objects (620, 630, 640, 650) based on the determined object location information,

c) based on the determined object movement information, selecting (744) at least one (640, 650) of the plurality of objects,

d) identifying at least one of the selected objects as a human object (650') based on detectability of at least one vital characteristic of the selected object (640, 650),

e) determining (741, 742) vital characteristics of the identified human object (650'),
wherein
- steps a) to e) are repeated in pre-defined time intervals, and

- an anomaly is detected (743), if vital characteristics are determined in step e), which differ from pre-determined vital parameters defining a normal vital condition.
The anomaly detection device of claim 8, adapted to cause, in response to an anomaly being detected, at least one environmental parameter to be determined using at least one sensor device (201-207), wherein the at least one sensor device comprises at least one quantum sensor.
The anomaly detection device of any one of claims 8 or 9, adapted to generate, in response to an anomaly being detected, a notification or alarm message and transmit the notification or alarm message to an emergency center (20).
The anomaly detection device of any one of claims 8 to 10, adapted to store the vital characteristics detected in step e) and to determine, based on the stored vital characteristics of human objects, the vital parameters, which define a vital normal condition.
The anomaly detection device of any one of claims 8 to 11, wherein step b) comprises
- identifying static objects and storing object location information of the identified static objects, and

- identifying moving objects that are moving in a predictable movement pattern and storing object location information and/or movement information of the identified moving objects.
The anomaly detection device of any one of claims 8 to 12, adapted to generate a digital map (601-603) comprising a digital representation of the plurality of objects (620, 630, 640, 650), wherein in step c) the digital map is modified to comprise only the selected objects.
The anomaly detection device of any one of claims 8 to 13, wherein the plurality of objects comprises objects, which are components of an autonomously operating industrial system.
An anomaly detection system (30), comprising
- an anomaly detection device (100) according to any one of claims 8 to 14, and

- at least one sensor device (201-207) connected to the anomaly detection device (100).