Monitoring device and method for image-based monitoring of monitored area, computer program and storage medium

By setting the selection analysis component and the main analysis component in the monitoring device, using the deep neural network to perform rough analysis and success probability assessment of the input signal, the problem of limited resources of the monitoring device is solved, and efficient coordination of analysis application and resource utilization is achieved, improving the improvement of The efficiency and reliability of the analytical processing are obtained.

CN118946908APending Publication Date: 2024-11-12ROBERT BOSCH GMBH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380030889.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-01-26
Filing Date
2023-01-25
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The limited computing resources of the monitoring device result in that in the presence of multiple analytical applications, not all analytical applications can be implemented simultaneously, and the possibility of expanding, changing or acquiring analytical applications exists, resulting in low resource utilization efficiency.

Method used

By setting the selection analysis component and the main analysis component, using a deep neural network to perform rough analysis of the input signal, estimating the probability of success, and selecting suitable analysis applications for implementation based on this, optimizing the calculation workload and time, and realizing the coordination of multiple analysis applications .

Benefits of technology

Effectively implement efficient analytical application implementation under limited resources, improve the resource utilization rate of monitoring equipment and the efficiency of analysis processing, and ensure the reliability and success rate of analysis applications.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118946908A_ABST
    Figure CN118946908A_ABST
Patent Text Reader

Abstract

The invention relates to a monitoring device (1) for image-based monitoring of a monitoring area (5), comprising an implementation module (6), a selection module (4) and a plurality of evaluation applications (7), in which the monitoring device (1) is supplied with an input signal and / or the monitoring device (1) has a sensor device for supplying the input signal, the input signal comprises an image of the monitoring region (5), the evaluation applications (7) each comprise a selection evaluation means (9) and a main evaluation means (10), the selection module (4) is designed to determine and / or estimate a probability of success of the application of the main evaluation means (10) to the input signal for the evaluation application (7) by applying the selection evaluation means (9) to the input signal, and the selection module (4) is designed to determine and / or estimate a probability of success of the application of the main evaluation means (10) to the input signal by applying the selection evaluation means (9) to the input signal. The selection module (4) is designed to select at least one of the analysis applications (7) as an implementation application (11) on the basis of the probability of success, and wherein the implementation module (6) is designed to implement a main analysis component (10) of the implementation application (11).
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a monitoring device for monitoring a monitoring area in an image-based manner, wherein an input signal is provided to the monitoring device and / or the monitoring device has a sensor device for providing the input signal, wherein the input signal comprises an image of the monitoring area. Background Art

[0002] Document DE 20 2018 104 373 U1, which may constitute the closest prior art, describes a device for operating a machine learning system, in particular for controlling the calculations of a machine learning system. Summary of the invention

[0003] A monitoring device for image-based monitoring of a monitoring area is proposed, having the features of claim 1. Furthermore, a method, a computer program and a machine-readable storage medium with a computer program are proposed for image-based monitoring of a monitoring area. Further advantages, effects and embodiments are derived from the dependent claims, the description and the figures.

[0004] A monitoring device for monitoring a monitoring area in an image-based manner is proposed. The monitoring device can form, for example, a hardware module, which is preferably configured as a chip system or a module component, in particular for integration into a camera device. Specifically, the monitoring device is configured to analyze and process input signals. Specifically, the monitoring device itself forms a camera device. Alternatively and / or in addition, the monitoring device is connected to at least one camera device in terms of data technology. Image-based monitoring is particularly understood as video-based monitoring. Specifically, image-based monitoring is based on color images, thermal images and / or additional data (Zusatzdaten). Additional data may include, for example, temperature, distance measurement and / or grating information. Specifically, monitoring may include access monitoring, personnel identification and / or object identification or tracking, hazard monitoring, fire monitoring and / or smoke monitoring of an area. Specifically, the monitoring area includes a plurality of sub-areas, wherein the sub-areas are arranged coherently and / or split. The monitoring area may be an outdoor area, such as a park or a road, or an interior space, such as an airport or a railway station.

[0005] The monitoring device includes an implementation module (specifically, multiple implementation modules), one or more selection modules and multiple analysis applications. Specifically, the implementation module and the selection module are summarized in a common module, such as a chip. The implementation module and / or the selection module particularly have processor resources. The implementation module and the selection module particularly include software parts. The analysis application preferably includes and / or forms a software application. The monitoring device particularly includes at least ten analysis applications. Specifically, the analysis application is provided by a third party, can be acquired (beziehbar), can be expanded, can be changed and / or can be removed. The analysis application is particularly configured to implement and / or support image-based monitoring. For example, the analysis application is configured to be used for image recognition, object recognition, personnel recognition, hazard recognition, fire recognition, smoke recognition and / or access control. For example, a user of the monitoring device can save and / or store an analysis application on the monitoring device, and the analysis application can implement the desired analysis, analysis processing and / or monitoring based on the input signal and / or image provided.

[0006] For example, at least one input signal is provided to the monitoring device by a camera, a sensor device and / or a sensor. Alternatively and / or in addition, the monitoring device includes a sensor device, specifically at least one camera, for providing an input signal. For example, the monitoring device is connected to one or more cameras and / or sensors, wherein the cameras and sensors provide input signals. To this end, the monitoring device has, for example, an interface for coupling with the camera and / or sensor. The input signal can be formed and / or include a digital signal or an analog signal. The input signal includes at least one image, an image sequence and / or an image stream of the monitored area. Based on the at least one image or input signal, the monitored area is monitored based on an image. Specifically, the input signal can include additional data in addition to the image, such as metadata, temperature data and / or sensor data. Preferably, the input signal is provided at a providing frequency, preferably based on a scanning frequency and / or a recording frequency.

[0007] The analysis application has a selection analysis component and a main analysis component respectively. The selection analysis component and the main analysis component form, for example, two subroutines of the analysis application. In particular, it is provided that the selection analysis component and the main analysis component of the analysis application differ in their implementation scope and / or implementation duration, for example, the implementation duration and / or complexity of the main analysis component is at least 10 times, specifically at least 100 times, that of the selection analysis component. Specifically, it is provided that the selection analysis component has similarities with the main analysis component. For example, the selection analysis component is constructed to be used to perform only the features and / or analysis of the main analysis component.

[0008] The selection module is configured to apply the selection analysis component of the analysis application to the input signal. For example, the input signal is provided to all the selection analysis components and / or processed by the selection analysis components. The selection analysis component is configured to provide, calculate and / or estimate the probability of success. When the selection analysis component is applied to the input signal, what is determined as the probability of success is, for example, how successfully, meaningfully and / or reliably the input signal can be analyzed and / or analyzed with the help of the assigned main analysis component. For example, the selection module is configured to find the following features in the input signal by applying the selection analysis component: the features are suitable, relevant and / or promising for the application of the main analysis component. Therefore, the probability of success, for example, describes how promising, meaningful and / or reliable the application of the main analysis component to the input signal is. Preferably, the probability of success is a percentage probability.

[0009] The selection module is configured to select at least one analysis application from the analysis applications as an implementation application based on the determined success probability. Specifically, it is provided that the selection module is configured to select a plurality of analysis applications as implementation applications based on the success probability. Preferably, the implementation module is configured to select an analysis application with a maximum success probability and / or a plurality of maximum success probabilities as implementation applications.

[0010] The execution module is designed to execute, apply and / or utilize a main evaluation component of an execution application and / or a plurality of execution applications.

[0011] The present invention is based on the following consideration: the resources of the monitoring device, the computer and / or the chip system, such as the computing performance, are limited, so that when there are multiple analysis applications, not all analysis applications can be implemented at the same time. If there is a possibility for the monitoring device and / or the camera device to expand, change and / or obtain the analysis application from a third party, then by setting the selection analysis component and the main analysis component for all analysis applications in the proposed manner, it is possible to estimate how meaningful it is to apply the more work-intensive main analysis component to the current input signal with low computing workload and in the shortest time by applying efficient selection analysis components instead of complete analysis applications. Then, only the time-intensive and computing-intensive part of the analysis application (here is the main analysis component) is implemented for the following analysis application: for which there is a sufficient probability of success. In this way, the coordination (Orchestrierung) of the monitoring device and multiple analysis applications can be effectively achieved.

[0012] In particular, the analysis application, the selection analysis component and / or the main analysis component are respectively based on and / or form a neural network, which is specifically a deep neural network. For example, the analysis application, the selection analysis component and / or the main analysis component include and / or form a neural network trained for its application. In particular, the analysis application is trained for different applications. In particular, the selection analysis component of the analysis application is based on the parts, applications, weights and / or features of the assigned main analysis component. The analysis application, in particular the main analysis component, has an input layer, wherein an input signal is provided to the input layer for processing. The neural network of the main analysis component also has an input layer, wherein an input signal or a signal obtained at the output of the network of the failure analysis component is provided to the input layer as an input (Eingang).

[0013] One configuration of the present invention provides that the selection analysis component is constructed to determine the probability of success by a rough analysis of the input signal. For example, the rough analysis is understood as a simplified analysis of the input signal, especially in a similar and / or rougher manner than the main analysis component. For example, the rough analysis is a partial, simplified, interrupted application of the main analysis component to the input signal. For example, the selection analysis component is constructed to analyze only a subset of the features studied by the assigned main analysis component. For example, less than 1 / 100 or 1 / 1000 of the features of the main analysis component. In particular, the selection analysis component is based on a neural network with a first number of nodes, wherein the assigned main analysis component is based on a neural network with a second number of nodes, wherein the ratio of the second number of nodes to the first number of nodes is at least 100, specifically at least 1000. In particular, it can be provided that the selection analysis component is constructed to apply the analysis based on the main analysis component to an input signal of reduced quality, for example, to a coarsened image, a smaller dimension of the image and / or an input signal density.

[0014] It is particularly preferred that the selection module is configured to determine a normalized success probability based on the success probability. For example, the normalization is performed based on the sum of all success probabilities. Specifically, the normalized success probability Determined as Among them, p i is the success probability of each analysis component, and n is the number of analysis components.

[0015] The selection module is designed to perform a selection of an implementation application based on a normalized probability of success. This configuration is based on the consideration that each implementation application statistically occurs with a frequency of a normalized probability. If there are two applications, one application (A) with a normalized probability of 0.9 and one application (B) with a normalized probability of 0.1, A is selected with a probability of 90% and B is selected with a probability of 10%, for example by drawing a random number between 0 and 1, selecting A in the case of 0-0.9 and selecting B in the case of 0.9-1.0.

[0016] Specifically, it is provided that the selection module is configured to select an implementation application based on a success probability and an additional parameter. The additional parameter includes, for example, device resources, such as available memory, processor. In addition, the additional parameter may include device context (such as communication speed and / or application profile), analysis application context (such as related security applications) and / or metadata of the analysis application. This configuration is based on the following consideration: in order to coordinate the analysis application, in addition to considering the pure success probability, other data should be considered to make full use of existing resources so that a safe, complete and fail-proof implementation of the analysis application on the monitoring device can be achieved.

[0017] For example, it is provided that the metadata of the analysis application includes the minimum execution frequency, execution duration, accepted pause time and / or priority level of the analysis application. For example, it is provided that the analysis application is executed at regular intervals, for example, once every 10 seconds or every 10 image frames to check for smoke or fire, so that the analysis application must be executed even in the case of an inherently low probability of success. In addition, analysis applications that accept pause times can be considered, for example, for background adaptation, the analysis application accepts a pause time of more than one hour.

[0018] For example, it is provided that the selection module is configured to select an implementation application from the analysis applications based on the implementation history. For example, there is an analysis application with the lowest implementation frequency or the longest non-implementation duration. Thus, based on the implementation history, for example, the analysis application with the longest non-implementation, the following analysis application is set as the implementation application: the analysis application has not been implemented for a very long time and should be implemented again. This configuration is based on the following consideration: although the analysis application has a low probability of success, it should still be implemented regularly.

[0019] Specifically, the input signal is provided, scanned and / or updated again for each provision time interval. For example, the provision time interval is based on a scanning frequency of the camera device and / or the sensor device. The selection module is designed to select, for each n-th provision interval, for example, for each second or each tenth provision time interval, the analysis application with the longest non-application duration, the lowest execution frequency and / or the highest priority as the implementation application.

[0020] In one embodiment of the present invention, a selection module is configured to select a plurality of implementation applications from the analysis application based on the probability of success. In particular, the selection module is configured to determine an implementation list based on the plurality of implementation applications, wherein the implementation list includes the selected implementation applications, and in particular determines the order of the selected implementation applications. For example, the implementation list includes the implementation application with the greatest probability of success, wherein the implementation application with the greatest probability of success forms the implementation application to be prioritized and / or the implementation application at the first position in the implementation list. The implementation module is configured to implement, for example, based on the order, sequentially or in parallel, the main analysis component of the implementation application of the implementation list. In particular, the selection module is configured to determine the length of the list, for example, the number of implementation applications and / or the estimated duration of the implementation of the implementation applications, based on resource data, such as processor performance or memory.

[0021] Another embodiment of the invention provides that the execution module is designed to migrate the execution application externally, for example to the cloud, for execution. For example, the selection module has determined an execution list with a plurality of execution applications, wherein the execution module has and / or determines limited resources, wherein the execution module is designed to, in the event of a determined resource shortage, provide the execution application externally and / or to the cloud and / or execute it there.

[0022] It is particularly preferred that the monitoring device has a basic network and a plurality of special networks (also referred to as main networks). The basic network and the main networks respectively form a neural network. The basic network has an input layer and a boundary layer, wherein the input layer is preferably provided with an input signal. The boundary layer is particularly understood as the last layer of the basic network, for example, as an output layer. A plurality of layers (also referred to as hidden layers) are arranged between the input layer and the boundary layer. The basic network is constructed for and / or trained to implement, execute the selection analysis component. In particular, the basic network can be understood as a part of the selection module. In particular, the boundary layer forms an output of the probability of success.

[0023] The special network is trained and / or constructed for implementing the main analysis component. For example, for an analysis application with a number of N, the analysis processing device has N special networks. Preferably, the special networks are independent of each other and / or separated in terms of data technology. The special networks respectively have a special network input layer and a special network output layer. In particular, a plurality of layers are arranged between the special network input layer and the special network output layer. In particular, the special network is part of the implementation module. In particular, it is provided that the output layer of a basic network and / or a plurality of basic networks is connected to the input layer of the special network in terms of data technology. Specifically, it is provided that the analysis application is based on two neural networks and / or is implemented on two neural networks, namely the basic network and the special network, wherein the basic network is constructed and / or trained for implementing the analysis selection component, wherein the main analysis component is implemented by the special network. In particular, for the corresponding analysis component, the result of the basic network is transferred as input data to the input layer of the special network of the main analysis component. Therefore, the basic network can be understood as a network backbone (Netzwerkrückgrat) for implementing the selection analysis component, and its result is passed to the special network or the main analysis component.

[0024] A configuration of the present invention is provided with at least two basic networks and a plurality of special networks. The first basic network is configured to implement a selection analysis component, wherein a success probability for selecting an implementation application is obtained as a result on its output layer, for example, the selection module and / or the monitoring device includes another module, such as a CPU, for this purpose, thereby selecting the implementation application. For example, the configuration is configured that the second basic network, during the selection of the implementation application by another module and / or by the monitoring device, provides the result located on the output layer of the first basic network as an input signal to the second basic network for further processing, wherein the second basic network is configured to further process the data and provide it to a special network and / or a plurality of special networks on the output layer of the second basic network for implementing the main analysis component. The configuration is based on the following consideration: during the selection of the implementation application from the analysis application, time is utilized by further processing the data in the second basic network and providing it to the special network for executing the main analysis component.

[0025] A method for monitoring a monitoring area forms another subject of the present invention. For monitoring, a plurality of analysis applications should be provided for implementation, wherein the analysis applications respectively include a selection analysis component and a main analysis component. The method is provided to apply the selection analysis component of the analysis application to the input signal and determine the success probability of applying the corresponding main analysis component to the input signal. Based on the determined success probability, at least one or more implementation applications are selected from the analysis applications, for example, an implementation application with a maximum success probability and / or an implementation application with multiple maximum success probabilities. The main analysis component of the implementation application is implemented.

[0026] A computer program for implementation on a computer, a monitoring device or a camera device forms a further subject matter of the invention. The computer program is designed and / or configured to implement and / or control the described method.

[0027] A machine-readable storage medium on which the computer program is stored forms a further subject matter of the invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Further advantages, effects and embodiments emerge from the drawings and the description thereof. Here, it is shown:

[0029] Further advantages, effects and embodiments emerge from the drawings and the description thereof. Here, it is shown:

[0030] Figure 1 A first embodiment of a monitoring device for monitoring in an image-based manner is shown;

[0031] Figure 2 A further embodiment of a monitoring device having a neural network backbone is shown;

[0032] Figure 3 An embodiment of a monitoring device having two underlying networks is shown;

[0033] Figure 4 A schematic diagram showing the network layers. DETAILED DESCRIPTION

[0034] Figure 1 The exemplary embodiment of a monitoring device 1 is shown. The monitoring device 1 is connected to a sensor device in terms of data technology, which includes a camera device 2. The images of a monitoring area 5 recorded by the camera device 2 are provided as input signals to the monitoring device 1, in particular to a selection module 4, via a data technology connection 3.

[0035] The monitoring device 1 includes a selection module 4 and an implementation module 6. In particular, the selection module 4 and the implementation module 6 are configured as a common module, for example, as a chip. In addition, the monitoring device 1 includes a plurality of analysis applications 7. The analysis application 7 is stored and / or can be stored in a memory module 8, for example. The analysis application 7 is configured to analyze and / or analyze, especially user-specifically analyze and / or analyze an input signal, which is specifically an image of the camera device 2. For example, the analysis application is configured for image recognition, hazard recognition, access monitoring and / or is configured as a fire alarm device. In particular, a plurality of various types of analysis applications 7 can be stored and / or can be stored on the memory module Sp. The analysis application 7 can be expanded, changed and / or provided by a third party, for example from an application store. The following configuration is particularly preferred: in this configuration, the monitoring device 1 is integrated into the camera device 2, for example as a hardware module or a software module.

[0036] The analysis application 7 has a selection analysis component 9 and a main analysis component 10. For example, the selection analysis component 9 and the main analysis component 10 are designed as separate program structure groups and / or neural networks. In particular, the resource requirements of the main analysis component 10 of the analysis application 7, such as CPU requirements, memory and / or execution time, are greater than the resource requirements of the corresponding failure analysis component 9. In particular, it is provided that the selection analysis component 9 is implemented and / or applied by the selection module 4 and the main analysis component 10 is implemented and / or applied by the execution module 6.

[0037] The selection module 4 is designed to apply a selection analysis component 9 of all analysis applications 7 to the provided input signal. By applying the selection analysis component 9 to the input signal, a success probability p for the corresponding (i-th) analysis application 7 is determined. i , which describes the probability of successfully and / or meaningfully applying the main analysis component 10 to the current input signal. For example, a surveillance image of a door is provided as an image, wherein the analysis application 7 is designed for fire detection and another analysis application is designed for access control based on faces. The selection analysis component applied by the selection module 4 studies the image very roughly, for example, with regard to relevant features, such as whether there is a face or a luminous object in the image, which would be necessary in order to be able to show a flame or a fire. The analysis application 7 that is intended to detect a fire, for example, provides a low success probability p1 for an image that only shows people, because no luminous areas are detected. The analysis application 7 that is designed for access control will not detect a face and will therefore provide a low success probability p2 by applying the selection analysis component 9 to an image without people but with a campfire (Lagerfeuer).

[0038] The selection module 4 is designed to, based on the success probability p i, one or more analysis applications 7 are selected as (one or more) implementation applications, in particular with a maximum and / or multiple maximum success probabilities p i Implementation module 6 is configured as a main analysis component 10 for implementing only the analysis application 7 that has been selected as implementation application 11. Main analysis component 10 is configured to analyze the input signal (image here) more accurately and / or more specifically according to its task, for example, whether a person is a person allowed to access, or whether a bright and high-contrast element is a fire. The result from the application of main analysis component 10 in implementation module 6 is output as result 12, wherein result 12 is supplemented with the input signal itself, for example.

[0039] Figure 2 An embodiment of a monitoring device 1 is shown, which is essentially based on Figure 1 Embodiment on the above. The monitoring device 1 has a network backbone 13, which is also called a backbone network and / or a basic network. An input signal is provided to the network backbone 13 by a camera device 2, wherein the basic network processes the input signal and provides an intermediate result 14. A selection module 6 is connected to the output layer of the network backbone 12 in terms of data technology, wherein the intermediate result 13 is provided to the selection module. In particular, the selection module 6 can form an extended selection module 6′, so that the network backbone 12 is part of the selection module 6′. In the selection module 6 or in the selection module 6′, based on the intermediate result, a selection analysis component 9 of the analysis application 7 is applied to the intermediate result, in particular directly to the input signal, for determining the success probability p i The selection module 6 or 6 ′ is designed to, based on the success probability p i For example, the selection module 6 comprises a selector 14 for this purpose, which selects the implementation based on the success probability p i For this purpose, additional information can be provided, such as metadata 15 of the execution history and / or the CPU load rate. The execution module 10 is designed as a main analysis component for executing the selected application and outputting it as a result 12.

[0040] Figure 3 A further exemplary embodiment is shown, in which the monitoring device 1 has two network backbones 12 , 16 , which are also referred to as first basic network 12 and second basic network 16 . The selection module 6 and the execution module 10 are part of a CPU 17 .

[0041] The first basic network 12 is designed to process the input signals provided by means of a neural network, which takes time and should last until time T1. The result of this analysis is provided to the selection module 6, which determines the success probability p on this basis. i . With the help of the obtained success probability pi , at least one implementation application is selected by a selector 14 , in particular with the aid of additional information 15 , wherein the implementation module 10 is designed to implement the implementation application.

[0042] The second basic network 16 is designed to determine the success probability p during the time from T1 to T2, that is, during the time from T1 to T2. i During the selection and execution of the application, the results and / or data provided by first infrastructure network 12 are further processed by means of second infrastructure network 16 .

[0043] Figure 4 The layers of the basic network 12 and the input layer of the special network 18 are shown by way of example. The basic network 12 has an input layer 19 and an output layer 20, wherein a plurality of layers and nodes are arranged between the input layer and the output layer. The result of the output layer 20 is provided to the selection module 6 for determining the success probability p i ,In addition, the data of the output layer 20 is provided to the input layer 21 of the special network 18 for further processing.

Claims

1. A monitoring device (1) for monitoring a monitoring area (5) in an image-based manner, The monitoring device comprises an implementation module (6), a selection module (4) and a plurality of analysis applications (7). in, An input signal is provided to the monitoring device (1), and / or the monitoring device (1) has a sensor device for providing the input signal, wherein the input signal comprises an image of the monitoring area (5), The analysis application (7) comprises a selection analysis component (9) and a main analysis component (10), The selection module (4) is designed to determine and / or estimate the probability of success of applying the main analysis component (10) to the input signal for the analysis application (7) by applying the selection analysis component (9) to the input signal. The selection module (4) is configured to select at least one of the analysis applications (7) as an implementation application (11) based on the success probability. The execution module (6) is designed as a main analysis component (10) for executing the execution application (11).

2. The monitoring device (1) according to claim 1, It is characterized in that The analysis application (7), the selection analysis component (9) and / or the main analysis component (10) are based on and / or include a neural network.

3. The monitoring device (1) according to claim 1 or 2, It is characterized in that The selective evaluation component (9) is designed to determine the success probability, to perform a rough analysis of the input signal, to perform a partial analysis of the input signal based on the main evaluation component (10), and / or to perform an analysis of a quality-reduced input signal based on the main evaluation component (10).

4. The monitoring device (1) according to any one of the preceding claims, It is characterized in that The selection module (4) is designed to determine a normalized probability of success based on the probability of success, wherein the selection of the implementation application (11) is based on the normalized probability of success.

5. The monitoring device (1) according to any one of the preceding claims, It is characterized in that The selection module (4) is designed to select the implementation application (11) based on device resources of the monitoring device (1), a device context, an analysis application context and / or metadata (15) of the analysis application.

6. The monitoring device (1) according to claim 5, It is characterized in that The metadata (15) includes the minimum execution frequency, execution duration, accepted suspension time and / or priority level of the analytical application.

7. Monitoring device (1) according to any one of the preceding claims, It is characterized in that The selection module (4) is designed to select the implementation application (11) based on an implementation history.

8. The monitoring device (1) according to any one of the preceding claims, It is characterized in that The input signal is provided again for each provision time interval, wherein the selection module (4) is designed to select the analysis application having the longest non-application duration as the execution application (11) for each n-th provision interval.

9. Monitoring device (1) according to any one of the preceding claims, It is characterized in that The selection module (4) is designed to select a plurality of implementation applications (11) based on the success probability and to determine an implementation list having the implementation applications (11), wherein the implementation module (6) is designed as a main evaluation component (10) for implementing the implementation list.

10. The monitoring device (1) according to any one of the preceding claims, It is characterized in that The implementation module (6) is configured to provide an implementation application (11) to the cloud for the implementation in the event of resource shortage.

11. Monitoring device (1) according to any one of the preceding claims, It is characterized in that The selection module (4) forms a network backbone for the implementation module (6), and the implementation module (6) forms a main network, wherein the output of the network backbone (12) is provided to the main network as input data.

12. The monitoring device (1) according to any one of the preceding claims, Features Two base networks (16) and one main network.

13. A method for monitoring a monitoring area (5) by means of a plurality of analysis applications (7), in, The analysis applications (7) each comprise a selection analysis component (9) and a main analysis component (10), wherein the selection analysis component (9) of the analysis application (7) is applied to the input signal and a success probability of applying the corresponding main analysis component (10) to the input signal is determined, wherein at least one or more implementation applications (11) are selected from the analysis application (7) based on the determined success probability.

14. Computer programs, in, The computer program is designed and / or set up to carry out and / or use the method according to claim 13 when the computer program is executed.

15. Machine-readable storage medium, in, The computer program according to claim 14 is stored on the storage medium.