CN117251700A

CN117251700A - Artificial intelligence-based environmental monitoring sensor data analysis method and system

Info

Publication number: CN117251700A
Application number: CN202311539512.0A
Authority: CN
Inventors: 雷济民; 翁发梅
Original assignee: Tops Sensor Taicang Co ltd
Current assignee: Tops Sensor Taicang Co ltd
Priority date: 2023-11-18
Filing date: 2023-11-18
Publication date: 2023-12-19
Anticipated expiration: 2043-11-18
Also published as: CN117251700B

Abstract

According to the artificial intelligence-based environment monitoring sensor data analysis method and system, by the application of the embodiment of the application, the attribute knowledge is accurately extracted from the environment monitoring data, the association degree between different target attribute extraction branches is reduced, the intersectivity is improved, the future trend of the environment state is accurately predicted, and therefore the environment state is efficiently, accurately and reliably monitored, and powerful support is provided for environment treatment.

Description

Artificial intelligence-based environmental monitoring sensor data analysis method and system

Technical Field

The application relates to the technical field of artificial intelligence, in particular to an environment monitoring sensor data analysis method and system based on artificial intelligence.

Background

Environmental monitoring, also known as environmental quality monitoring, is the systematic, continuous and comprehensive observation and assessment of various factors in the environment (e.g., air, water, soil, noise, etc.). It involves collecting environmental data, analyzing this data, and using this information to assess environmental health for better understanding and protection of our environment.

The environmental monitoring can provide accurate information about environmental factors such as air quality, water quality, soil conditions and the like, and help us know the current situation of the environment. Through continuous observation and tracking of the environment, possible environmental problems can be pre-warned, so that measures are timely taken to avoid or reduce environmental pollution and damage. Environmental monitoring data is an important basis for formulating and adjusting environmental policies, and can also be used to evaluate the effectiveness of existing policies. The large amount of data generated by environmental monitoring has important value for environmental science research, and can promote understanding of an environmental system. By publishing the environmental monitoring result, the environmental awareness of the public can be enhanced, and the social world is promoted to participate in environmental protection together.

In practical environmental monitoring tasks, how to implement accurate and reliable environmental state trend prediction is still one aspect of the present need for improvement.

Disclosure of Invention

The application provides at least an artificial intelligence-based environmental monitoring sensor data analysis method and system.

The application provides an artificial intelligence-based environmental monitoring sensor data analysis method, which is applied to a sensor data analysis system, and comprises the following steps:

collecting data of an environment monitoring sensor to be identified;

Loading the to-be-identified environment monitoring sensor data into a target attribute extraction branch matched with a plurality of monitoring situation labels in a target environment data analysis network to perform attribute extraction operation, so as to obtain environment monitoring data attribute knowledge corresponding to each monitoring situation label; the target environment data analysis network is obtained by debugging a basic monitoring situation association component in a basic attribute extraction branch through exclusive cost data, the exclusive cost data represents a commonality index of environment association attribute data, and the environment association attribute data is an environment association attribute vector generated by the basic monitoring situation association component corresponding to any two monitoring situation labels;

and determining the trend viewpoint of the target environmental state corresponding to the environmental monitoring sensor data to be identified according to the environmental monitoring data attribute knowledge corresponding to each monitoring situation label and the target discrimination branch corresponding to each monitoring situation label.

Optionally, the target environment data analysis network is obtained through the following steps:

acquiring a first environmental monitoring sensor data case set;

loading the first environment monitoring sensor data case set to a basic attribute extraction branch matched with each of the plurality of monitoring situation labels in a set AI network to perform attribute extraction operation, so as to obtain environment association attribute vector cases corresponding to each monitoring situation label;

Determining exclusive cost data corresponding to a target context label pair based on an environment association attribute vector case corresponding to the target context label pair; the target context label pair is any two of the plurality of monitoring context labels;

and debugging the corresponding basic monitoring situation association component by the target situation label according to the corresponding exclusive cost data of the target situation label to obtain the target environment data analysis network.

Optionally, the basic attribute extraction branch corresponding to each monitoring situation label comprises a basic convolution component and a basic knowledge mining component, and the basic knowledge mining component comprises a first focusing component, a first feature mapping component, a basic monitoring situation association component and a first feature integration component;

the loading the first environmental monitoring sensor data case set to a basic attribute extraction branch matched with each of the plurality of monitoring situation labels in a set AI network to perform attribute extraction operation, so as to obtain an environmental association attribute vector case corresponding to each monitoring situation label, including:

loading the first environment monitoring sensor data case set to a basic convolution component corresponding to each monitoring situation label to carry out convolution operation, so as to obtain environment monitoring sensor data case convolution characteristics corresponding to each monitoring situation label;

Loading the environmental monitoring sensor data case convolution characteristics corresponding to each monitoring situation label to a first focusing component corresponding to each monitoring situation label to perform focus strengthening operation, so as to obtain a first focus strengthening vector case corresponding to each monitoring situation label;

loading the first focus reinforcement vector cases corresponding to the monitoring situation labels and the environmental monitoring sensor data case convolution features corresponding to the monitoring situation labels to a first feature mapping component corresponding to the monitoring situation labels to perform feature mapping operation to obtain first basic monitoring mapping vectors corresponding to the monitoring situation labels;

loading the first focus reinforcement vector cases corresponding to the monitoring situation labels and the environmental monitoring sensor data case convolution characteristics corresponding to the monitoring situation labels to a basic monitoring situation association component corresponding to the monitoring situation labels for association operation to obtain first environmental association attribute vectors corresponding to the monitoring situation labels;

and loading the first basic monitoring mapping vector corresponding to each monitoring situation label, the first environment association attribute vector corresponding to each monitoring situation label, the first focus reinforcement vector case corresponding to each monitoring situation label and the environment monitoring sensor data case convolution characteristic corresponding to each monitoring situation label to the first characteristic integration component corresponding to each monitoring situation label for vector integration operation to obtain the second basic monitoring mapping vector corresponding to each monitoring situation label.

Optionally, the method further comprises:

acquiring first environmental state trend viewpoint cases corresponding to all the environmental monitoring sensor data cases in the first environmental monitoring sensor data case set;

loading the second basic monitoring mapping vector corresponding to each monitoring situation label to a basic judging branch corresponding to each monitoring situation label in the set AI network to judge operation, so as to obtain a third environment state trend prediction viewpoint corresponding to each monitoring situation label;

loading the first environment monitoring sensor data case set into a basic confidence identification component in the set AI network for confidence identification to obtain confidence cases corresponding to all environment monitoring sensor data cases in the first environment monitoring sensor data case set;

weighting the third environmental state trend prediction views respectively matched with the plurality of monitoring situation labels according to the confidence level cases to obtain fourth environmental state trend prediction views corresponding to the environmental monitoring sensor data cases in the first environmental monitoring sensor data case set;

determining first confidence cost data according to the fourth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case;

The debugging of the basic monitoring context association component corresponding to the target context label pair according to the exclusive cost data corresponding to the target context label pair to obtain the target environment data analysis network comprises the following steps:

debugging the corresponding basic monitoring situation association component of the target situation label according to the corresponding exclusive cost data of the target situation label, and debugging the basic confidence identification component according to the first confidence cost data to obtain the target environment data analysis network.

Optionally, the basic attribute extraction branch corresponding to each monitoring situation label includes a basic convolution component and a basic knowledge mining component, the basic knowledge mining component includes a second focusing component, a second feature mapping component, a basic monitoring situation association component and a second feature integration component, and the basic monitoring situation association component includes a first basic monitoring situation linkage processing node and a second basic monitoring situation linkage processing node;

loading the environmental monitoring sensor data case convolution characteristics corresponding to each monitoring situation label to a second focusing component corresponding to each monitoring situation label to perform focus strengthening operation, so as to obtain a second focus strengthening vector case corresponding to each monitoring situation label;

loading the second focus reinforcement vector cases corresponding to the monitoring situation labels to a first basic monitoring situation linkage processing node corresponding to the monitoring situation labels to perform association operation, so as to obtain second environment association attribute vectors corresponding to the monitoring situation labels;

loading the second environment association attribute vector corresponding to each monitoring situation label and the environment monitoring sensor data case convolution feature corresponding to each monitoring situation label to a second feature mapping component corresponding to each monitoring situation label to perform feature mapping operation, so as to obtain a third basic monitoring mapping vector corresponding to each monitoring situation label;

Loading a third basic monitoring mapping vector corresponding to each monitoring situation label to a second basic monitoring situation linkage processing node corresponding to each monitoring situation label to perform association operation, so as to obtain a third environment association attribute vector corresponding to each monitoring situation label;

and loading the third environment association attribute vector corresponding to each monitoring situation label, the second environment association attribute vector corresponding to each monitoring situation label and the environment monitoring sensor data case convolution feature corresponding to each monitoring situation label to a second feature integration component corresponding to each monitoring situation label for vector integration operation to obtain a fourth basic monitoring mapping vector corresponding to each monitoring situation label.

Optionally, the method further comprises:

loading a fourth basic monitoring mapping vector corresponding to each monitoring situation label to a basic judging branch corresponding to each monitoring situation label in the set AI network to judge operation, so as to obtain a fifth environmental state trend prediction viewpoint corresponding to each monitoring situation label;

weighting the fifth environmental state trend prediction viewpoints matched by the monitoring situation labels according to the confidence coefficient cases to obtain sixth environmental state trend prediction viewpoints corresponding to the environmental monitoring sensor data cases in the first environmental monitoring sensor data case set;

determining second confidence cost data according to the sixth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case;

debugging the corresponding basic monitoring situation association component of the target situation label according to the corresponding exclusive cost data of the target situation label, and debugging the basic confidence identification component according to the second confidence cost data to obtain the target environment data analysis network.

Optionally, the method further comprises:

acquiring a second environment monitoring sensor data case set and second environment state trend viewpoint cases corresponding to the environment monitoring sensor data cases in the second environment monitoring sensor data case set;

loading the second environmental monitoring sensor data case set to a basic attribute extraction branch matched with each of the plurality of monitoring situation labels in the set AI network to perform attribute extraction operation, so as to obtain a fifth basic monitoring mapping vector corresponding to each monitoring situation label;

loading a fifth basic monitoring mapping vector corresponding to each monitoring situation label to a basic judging branch corresponding to each monitoring situation label to judge, so as to obtain a seventh environment state trend prediction viewpoint corresponding to each monitoring situation label;

determining environmental monitoring cost data corresponding to each monitoring situation label according to a seventh environmental state trend prediction viewpoint corresponding to each monitoring situation label and the second environmental state trend viewpoint case;

Debugging corresponding basic monitoring situation association components of the target situation label according to the corresponding exclusive cost data of the target situation label, and debugging corresponding initial network components of each monitoring situation label in the set AI network according to the environment monitoring cost data corresponding to each monitoring situation label to obtain the target environment data analysis network;

the initial network component corresponding to each monitoring situation label is a component except for the basic monitoring situation association component corresponding to each monitoring situation label in the basic attribute extraction branch corresponding to each monitoring situation label.

Optionally, the determining, according to the environmental monitoring data attribute knowledge corresponding to the monitoring context labels and the target discrimination branch corresponding to the monitoring context labels, the trend viewpoint of the target environmental state corresponding to the environmental monitoring sensor data to be identified includes:

loading the environmental monitoring data attribute knowledge corresponding to each monitoring situation label to a target discrimination branch corresponding to each monitoring situation label to perform discrimination operation, so as to obtain a first environmental state trend prediction viewpoint corresponding to each monitoring situation label;

And determining a target environment state trend viewpoint corresponding to the environment monitoring sensor data to be identified according to the first environment state trend prediction viewpoints matched with the monitoring situation tags.

Optionally, the method further comprises:

loading the to-be-identified environment monitoring sensor data into a target confidence identification component in the target environment data analysis network to carry out confidence identification so as to obtain target confidence;

the determining, according to the first environmental state trend prediction views respectively matched with the plurality of monitoring context labels, the target environmental state trend view corresponding to the environmental monitoring sensor data to be identified includes:

weighting the first environmental state trend prediction viewpoints matched by the monitoring situation labels according to the target confidence level to obtain a second environmental state trend prediction viewpoint corresponding to the environmental monitoring sensor data to be identified;

and determining the target environmental state trend viewpoint according to the second environmental state trend prediction viewpoint.

Optionally, the loading the to-be-identified environmental monitoring sensor data into a target attribute extraction branch matched with each of a plurality of monitoring context labels in a target environmental data analysis network performs attribute extraction operation to obtain environmental monitoring data attribute knowledge corresponding to each monitoring context label, including:

Loading the environment monitoring sensor data to be identified into a target convolution component in a target attribute extraction branch corresponding to each monitoring situation label to carry out convolution operation, so as to obtain the environment monitoring sensor data convolution characteristics corresponding to each monitoring situation label;

and loading the environmental monitoring sensor data convolution characteristics corresponding to the monitoring situation labels to a target knowledge mining component in a target attribute extraction branch corresponding to the monitoring situation labels to carry out knowledge mining to obtain the environmental monitoring data attribute knowledge corresponding to the monitoring situation labels.

Optionally, the target knowledge mining component corresponding to each monitoring context label includes a third focusing component, a third feature mapping component, a target monitoring context association component and a third feature integration component;

the loading the environmental monitoring sensor data convolution characteristic corresponding to each monitoring situation label to the target knowledge mining component in the target attribute extraction branch corresponding to each monitoring situation label to perform knowledge mining to obtain the environmental monitoring data attribute knowledge corresponding to each monitoring situation label, including:

loading the environmental monitoring sensor data convolution characteristics corresponding to each monitoring situation label to a third focusing component corresponding to each monitoring situation label to perform focus strengthening operation, so as to obtain a third focus strengthening vector case corresponding to each monitoring situation label;

Loading the environmental monitoring sensor data convolution characteristics corresponding to each monitoring situation label and the third focus enhancement vector cases corresponding to each monitoring situation label to a third feature mapping component corresponding to each monitoring situation label to perform feature mapping operation, so as to obtain a first target monitoring mapping vector corresponding to each monitoring situation label;

loading the environmental monitoring sensor data convolution characteristics corresponding to each monitoring situation label and the third focus enhancement vector cases corresponding to each monitoring situation label to a target monitoring situation association component corresponding to each monitoring situation label for association operation, so as to obtain a fourth environmental association attribute vector corresponding to each monitoring situation label;

and loading a fourth environment association attribute vector corresponding to each monitoring situation label, a first target monitoring mapping vector corresponding to each monitoring situation label, an environment monitoring sensor data convolution characteristic corresponding to each monitoring situation label and a third focus enhancement vector case corresponding to each monitoring situation label to a third characteristic integration component corresponding to each monitoring situation label for vector integration operation, so as to obtain environment monitoring data attribute knowledge corresponding to each monitoring situation label.

Optionally, the target knowledge mining component corresponding to each monitoring situation label includes a fourth focusing component, a fourth feature mapping component, a target monitoring situation association component and a fourth feature integration component, and the target monitoring situation association component corresponding to each monitoring situation label includes a first target monitoring situation linkage processing node and a second target monitoring situation linkage processing node;

loading the environmental monitoring sensor data convolution characteristics corresponding to each monitoring situation label to a fourth focusing component corresponding to each monitoring situation label to perform focus strengthening operation, so as to obtain a fourth focus strengthening vector case corresponding to each monitoring situation label;

loading fourth focus reinforcement vector cases corresponding to the monitoring situation labels to a first target monitoring situation linkage processing node corresponding to the monitoring situation labels to perform association operation, so as to obtain fifth environment association attribute vectors corresponding to the monitoring situation labels;

Loading a fifth environment association attribute vector corresponding to each monitoring situation label and an environment monitoring sensor data convolution feature corresponding to each monitoring situation label to a fourth feature mapping component corresponding to each monitoring situation label for feature mapping operation to obtain a second target monitoring mapping vector corresponding to each monitoring situation label;

loading the second target monitoring mapping vector corresponding to each monitoring situation label to a second target monitoring situation linkage processing node corresponding to each monitoring situation label to perform association operation, so as to obtain a sixth environment association attribute vector corresponding to each monitoring situation label;

and loading the sixth environment-related attribute vector corresponding to each monitoring situation label, the fifth environment-related attribute vector corresponding to each monitoring situation label and the environment monitoring sensor data convolution feature corresponding to each monitoring situation label to a fourth feature integration component corresponding to each monitoring situation label for vector integration operation to obtain the environment monitoring data attribute knowledge corresponding to each monitoring situation label.

The application also provides a sensor data analysis system, which comprises a processor and a memory; the processor is in communication with the memory, and the processor is configured to read and execute a computer program from the memory to implement the method described above.

The present application also provides a computer readable storage medium having stored thereon a computer program which, when run, implements the method described above.

The technical scheme provided by the embodiment of the application can comprise the following beneficial effects: the method comprises the steps of collecting the to-be-identified environmental monitoring sensor data, loading the to-be-identified environmental monitoring sensor data to a target attribute extraction branch matched with a plurality of monitoring situation labels in a target environmental data analysis network to perform attribute extraction operation to obtain environmental monitoring data attribute knowledge corresponding to each monitoring situation label, wherein the target environmental data analysis network is obtained by debugging a basic monitoring situation association component in a basic attribute extraction branch through exclusive cost data, the exclusive cost data represents a commonality index between environment association attribute vectors generated by the basic monitoring situation association component corresponding to any two monitoring situation labels in the plurality of monitoring situation labels, the relativity between the target attribute extraction branches corresponding to different monitoring situation labels can be weakened and the dissimilarity can be strengthened through the exclusive cost data debugging, the method can reduce the association degree of attribute knowledge generated by different target attribute extraction branches, improve the difference of attribute knowledge generated by different target attribute extraction branches, thereby improving the linkage quality among a plurality of branches, obtain a target environment data analysis network through a debugging basic monitoring situation association component, reduce the operation amount of a debugging process, combine the environment monitoring data attribute knowledge corresponding to each monitoring situation label and the target discrimination branches matched with a plurality of monitoring situation labels in the target environment data analysis network respectively, determine the target environment state trend view corresponding to the environment monitoring sensor data to be identified, improve the environment state trend discrimination precision of the environment monitoring sensor data with a plurality of monitoring situation labels by the target environment data analysis network, realize accurate and reliable environment state trend prediction, thereby providing a credible basis for subsequent environmental management.

For a description of the effects of the above-described sensor data analysis system, computer-readable storage medium, see the description of the above-described method.

In order to make the above objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are necessary for use in the embodiments are briefly described below, which drawings are incorporated in and form a part of the present description, these drawings illustrate embodiments consistent with the present application and together with the description serve to explain the technical solutions of the present application. It is to be understood that the following drawings illustrate only certain embodiments of the present application and are therefore not to be considered limiting of its scope, for the person of ordinary skill in the art may derive other relevant drawings from the drawings without inventive effort.

FIG. 1 is a block diagram of a sensor data analysis system shown in an embodiment of the present application.

FIG. 2 is a flow chart of an artificial intelligence based environmental monitoring sensor data analysis method according to an embodiment of the present application.

Detailed Description

Reference will now be made in detail to exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, the same numbers in different drawings refer to the same or similar elements, unless otherwise indicated. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application.

Fig. 1 is a schematic diagram of a sensor data analysis system 10 according to an embodiment of the present application, including a processor 102, a memory 104, and a bus 106. The memory 104 is used for storing execution instructions, including a memory and an external memory, where the memory may also be understood as an internal memory, and is used for temporarily storing operation data in the processor 102 and data exchanged with the external memory such as a hard disk, where the processor 102 exchanges data with the external memory through the memory, and when the sensor data analysis system 10 operates, the processor 102 and the memory 104 communicate through the bus 106, so that the processor 102 executes the environmental monitoring sensor data analysis method based on artificial intelligence in the embodiment of the present application.

Referring to fig. 2, fig. 2 is a schematic flow chart of an artificial intelligence-based environmental monitoring sensor data analysis method applied to a sensor data analysis system according to an embodiment of the present application, where the method may include steps 110-130.

Step 110, collecting the data of the environmental monitoring sensor to be identified.

The environmental monitoring sensor data to be identified is pre-collected, sensor data for identifying or understanding environmental conditions. Such data may originate from one or more types of environmental monitoring sensors, such as temperature sensors, humidity sensors, illumination sensors, etc., which may collect corresponding environmental parameters.

In some possible examples, it is desirable to know and analyze the ecology status of a forest area through a set of environmental monitoring sensors. First, various environmental monitoring sensors including a temperature sensor, a humidity sensor, an illumination sensor, etc. are disposed in the forest area. These sensors will then periodically collect and record environmental data, such as current temperature, humidity, illumination intensity, etc., of the location where they are located. These recorded data are the environmental monitoring sensor data to be identified.

In the monitoring of forest ecoconditions, the environmental monitoring sensor data to be identified plays a very important role. Different types of sensors can help to obtain detailed information of various parameters of the forest.

Temperature sensor: the temperature has a significant impact on the biological population and ecosystem in the forest. For example, some species may only survive within a specific temperature range. Thus, monitoring temperature changes can help to understand forest health, predicting problems that may occur.

Humidity sensor: humidity is also a key factor affecting forest ecosystems. Changes in humidity can affect plant growth and animal activity. Too high or too low a humidity may result in the reduction or disappearance of certain species.

Illumination sensor: illumination is critical to plant growth in forests, and plants require sufficient illumination to perform photosynthesis. By monitoring the illumination intensity, the vegetation status of the forest can be analyzed, as well as predicting possible vegetation changes in the future.

These above environmental monitoring sensor data to be identified are all loaded as input data into the target environmental data analysis network for attribute extraction and generation of environmental monitoring data attribute knowledge. Finally, according to the attribute knowledge, the current state of the forest ecological system can be determined, the future trend of the forest ecological system can be predicted, and ecological problems possibly existing can be found in advance.

And 120, loading the environment monitoring sensor data to be identified into a target attribute extraction branch matched with each of a plurality of monitoring situation labels in a target environment data analysis network to perform attribute extraction operation, so as to obtain environment monitoring data attribute knowledge corresponding to each monitoring situation label.

The target environment data analysis network is obtained by debugging a basic monitoring situation association component in a basic attribute extraction branch through exclusive cost data, the exclusive cost data represents a commonality index of environment association attribute data, and the environment association attribute data is an environment association attribute vector generated by the basic monitoring situation association component corresponding to any two monitoring situation labels.

In some possible examples, the target environmental data resolution network is primarily tasked with receiving input environmental monitoring sensor data, processing and resolving it.

Further, the exclusive cost data is one indicator used to characterize the variability between different monitoring scenarios. In some machine learning or deep learning models, this type of data is used when the output of the model is desired to be exclusive (i.e., mutually exclusive) in some way.

For example, the assumed target environment data analysis network includes two monitoring context labels: drought and flood. These two tags represent two diametrically opposed environmental states. Thus, when the attribute extracted from the environmental monitoring data is more prone to drought, then it should represent less of a flood and vice versa. This is called exclusive or mutually exclusive.

The exclusive cost data is information that measures such exclusivity. If an input data is misclassified into multiple mutually exclusive monitoring scenarios, then the error is given a higher cost value when calculating the cost (or penalty). This encourages the model to classify as much as possible so that each input data belongs to only one most appropriate monitoring context, rather than simultaneously belonging to multiple mutually exclusive monitoring contexts. This is the primary role of the exclusive cost data.

In optimizing such models, a special loss function, such as multiple logistic regression (Multinomial logistic regression) or cross entropy loss (cross-entropy loss), is typically used to calculate the exclusive cost for each input data. By minimizing this cost, a model can be trained that correctly classifies the environmental conditions.

In addition, the monitoring context tags are used to mark different environmental monitoring scenarios or conditions, such as forest fires, floods, etc.

In the embodiment of the present application, the target attribute extraction branch and the basic attribute extraction branch may be understood as two key parts of the environment data analysis network, which are respectively responsible for extracting a specific attribute and a general basic attribute under a specific monitoring situation.

Target attribute extraction branch (target extraction subnetwork): the network is mainly responsible for extracting relevant attributes of corresponding monitoring situation labels from input environment monitoring data. Each monitoring context tag has its own target attribute extraction branch for identifying and extracting attributes associated with the tag. For example, in the context of forest ecosystem monitoring, if a drought monitoring context label is defined, the corresponding target attribute extraction branch will focus on extracting attributes characterizing drought conditions, such as high temperature, low humidity, etc., from the input data.

Basic attribute extraction branch (sample extraction sub-network): unlike the target attribute extraction branch, the base attribute extraction branch is not focused on a particular monitoring context, but is used to extract base attributes that may be needed in all contexts. These basic properties may include general environmental parameters such as temperature, humidity, illumination intensity, etc. The task of this part of the network is to extract these generic attributes and provide the basis data for the subsequent environmental state decisions.

By simultaneously transmitting the input environment monitoring data into the two sub-networks, the detailed information required under each specific situation can be ensured to be acquired, and the basic attribute which is possibly important for all the situations can not be ignored.

Further, through the attribute extraction operation, key information or features can be extracted from the input environmental monitoring data. Based on this, knowledge of the attributes of the environmental monitoring data can be understood as a set of information or features extracted and generated from the raw environmental monitoring data, which are represented in the form of vectors for describing and characterizing a particular environmental state.

The environment monitoring data attribute vector is composed of a series of values, each representing a particular environment attribute. For example, in monitoring a forest ecosystem, this vector may include parameters such as temperature, humidity, illumination intensity, etc. Different combinations of parameters may reflect different environmental conditions such as normal, drought, flood, etc.

In addition, this attribute vector can also be used to represent the trend in time. For example, if environmental monitoring data of a forest is collected for several consecutive days, the change situation of the forest environment can be observed by comparing attribute vectors at different time points, so as to predict possible future trends.

Thus, knowledge of the attributes of the environmental monitoring data is not just a simple generalization of the raw data, but rather is an important tool to accurately reflect and predict the environmental state.

Still further, the basic monitoring context correlation component (or sample monitoring context matching module) is a key part of the target environmental data resolution network. Its main task is to match the environmental monitoring data with a predefined monitoring context and generate a corresponding environmental correlation attribute vector.

The working principle of this component is typically based on some form of machine learning or deep learning model. First, it receives the basic attribute data extracted from the basic attribute extraction branch and then converts the basic attribute data into a higher-level attribute vector representing a particular monitoring context through a trained model. For example, in monitoring a forest ecosystem, the component may generate an attribute vector representing normal, drought, flood, etc. conditions based on input basic attribute data such as temperature, humidity, illumination intensity, etc.

This attribute vector is called an environment-associated attribute vector because it contains not only the original environment attribute data, but also the association information between these data and the particular monitoring context. Such correlation information is learned through a training process of the model, which can help more accurately understand and predict environmental conditions.

As can be seen, the basic monitoring context association component is a module that associates raw environmental attribute data with a predefined monitoring context and generates an environmental association attribute vector.

The monitoring of the forest ecosystem is taken as an example to carry out noun interpretation on the environment-related attribute data, the commonality index and the environment-related attribute vector.

Environmental correlation attribute data (matching feature): this is a feature vector generated by the underlying monitoring context correlation component that characterizes the state of the environment. Assuming that a temperature sensor, a humidity sensor and an illumination sensor are arranged in a forest, the characteristic vector reflecting the forest environment state can be obtained after the data collected by the sensors are processed and converted. For example, the feature vector may contain three elements: temperature value, humidity value and illumination intensity value, which are environmental related attribute data.

Commonality index (similarity): the commonality index is an index that measures the degree of similarity between two or more environment-associated attribute vectors. For example, if forest environment data is collected in two consecutive days and two environment-related attribute vectors are obtained, then the similarity (such as cosine similarity or euclidean distance) between the two vectors can be calculated to obtain their commonality index. If the commonality index is higher, the forest environment states in the two days are similar; if the commonality index is low, a large change in the environmental state is indicated.

Environment-associated attribute vector (environment attribute matching feature): the feature vector is obtained by further processing and optimizing the environment-related attribute data. In the optimization process, some machine learning or deep learning techniques, such as Principal Component Analysis (PCA), auto encoder (auto encoder), etc., may be used to reduce the data dimension while preserving as much important information as possible in the original data. In this way, a more compact and efficient feature vector representing the forest environment status can be obtained.

In step 120, the collected environmental monitoring sensor data to be identified (such as temperature, humidity, illumination intensity, etc.) is first loaded into the target environmental data analysis network. This network consists of two parts: a target attribute extraction branch (i.e., a target extraction sub-network) and a base attribute extraction branch (i.e., a sample extraction sub-network).

In the target attribute extraction branch, each preset monitoring situation label (such as normal, drought, flood and the like) has a corresponding sub-network, and is responsible for extracting the attribute related to the label from the input data. For example, for drought scenarios, the target attribute extraction branch may be focused on higher temperature, lower humidity, etc. features. In this way, the environment monitoring data attribute vector, that is, the environment monitoring data attribute knowledge, corresponding to each monitoring context label can be obtained.

At the same time, the basic attribute extraction branch will extract some general environmental parameters that may be important for all situations, such as temperature, humidity, illumination intensity, etc. The extracted attribute information is sent to a basic monitoring context association component (i.e. a sample monitoring context matching module) for processing to generate an environment association attribute vector.

The exclusive cost data (i.e., the mutually exclusive penalty data) is then used to measure the similarity or difference between these environment-associated attribute vectors. If the commonality index (i.e., similarity) between two environment-associated attribute vectors is high, then it is stated that their corresponding environment states may be more similar; otherwise, it indicates that the environmental state has a large change.

Through the steps, useful information can be extracted from the original environment monitoring sensor data, and the information can be displayed in a more efficient and compact mode, so that the state of the forest ecological system can be better understood and predicted.

And 130, determining the trend viewpoint of the target environmental state corresponding to the environmental monitoring sensor data to be identified according to the environmental monitoring data attribute knowledge corresponding to each monitoring situation label and the target discrimination branch corresponding to each monitoring situation label.

In some possible examples, the target discrimination branch and target environmental state trend perspective are key components for final environmental state identification and prediction.

Target discrimination branch: this is an important part of the target environmental data analysis network, which is responsible for making a final environmental state judgment based on the previously obtained environmental monitoring data attribute knowledge corresponding to each monitoring context label. In other words, it is a classifier or decision module that determines the monitoring context label that best matches the current environmental state based on the input knowledge of the environmental monitoring data attributes.

Target environmental state trend perspective: this is a prediction or perspective of future trends in environmental conditions based on the results of the target discrimination branch. It can help to know how environmental conditions may change, such as whether the forest ecosystem will change from a normal state to a drought state.

For example, in step 130, it is assumed that knowledge of the attributes of the environmental monitoring data corresponding to each monitoring context tag has been obtained through step 120. Then, these knowledge are input into the target discrimination branch. The target discrimination branch can determine which monitoring situation label the current environmental state is most suitable for, such as normal, drought or flood, according to the knowledge.

Then, based on this result, a target environmental state trend perspective can be generated. For example, if during monitoring for several consecutive days, it is found that the environmental state changes gradually from normal to drought, it can be predicted that the forest may continue to be in drought for a period of time in the future. This is the perspective of the trend of the target environmental state.

The following description of steps 110-130 is presented by way of a complete example.

First, assume that the environmental status of a forest is being monitored. Various sensors are deployed in forests, including temperature, humidity, and illumination intensity sensors. These sensors periodically collect and record environmental data of their location.

Secondly, the collected environmental data is input into a pre-constructed environmental data analysis network. This network comprises two main parts: a target attribute extraction branch and a base attribute extraction branch. In the target attribute extraction branch, each preset monitoring situation label (such as normal, drought and flood) has a corresponding sub-network, and is responsible for extracting the attribute related to the label from the input data. At the same time, the basic attribute extraction branch will extract some general environmental parameters that may be important for all contexts.

The extracted attribute information is then sent to a basic monitoring context association component for processing to generate an environment associated attribute vector. The exclusive cost data is also used to measure similarity or variability between these environment-associated attribute vectors to adjust the underlying monitoring context-associated components in the underlying attribute extraction branches.

Finally, the knowledge of the attributes of the environmental monitoring data is input into the target discrimination branch. The branch determines which monitoring context label the current environmental state best meets based on the knowledge. Based on this result, a target environmental state trend perspective can also be generated to help predict how the environmental state of the forest may change. For example, if during monitoring for several consecutive days, it is found that the environmental state changes gradually from normal to drought, it can be predicted that the forest may continue to be in drought for a period of time in the future.

It can be seen that applying steps 110-130, first, by extracting the knowledge of the environmental monitoring data attribute corresponding to each monitoring context tag from the environmental monitoring sensor data to be identified, enables a better understanding and description of the environmental state. Each monitoring context tag has its own target attribute extraction branch, focusing on extracting attributes associated with the tag from the input data. In this way, different environmental states can be accurately distinguished, thereby improving the accuracy of environmental state identification.

Secondly, by using the exclusive cost data, the association degree between attribute knowledge generated by different target attribute extraction branches can be reduced, and the dissimilarity of the attribute knowledge can be enhanced. The method is characterized in that the exclusive cost data characterizes the commonality index between the environment association attribute vectors corresponding to different monitoring situation labels, and attribute knowledge generated by different monitoring situation labels can be more independent by adjusting the basic monitoring situation association component, so that the robustness of the system is improved.

In addition, based on the target environment data analysis network and the environment monitoring data attribute knowledge, the target environment state trend viewpoint corresponding to the environment monitoring sensor data to be identified can be determined. This view can help predict possible future changes in environmental conditions, providing an important basis for environmental management and decision making. For example, in forest ecosystem monitoring, if it is predicted that a forest may be transformed from a normal state to a drought state, measures such as irrigation or adjustment of forest management strategies may be taken in time to prevent or mitigate the effects of drought on the forest.

In general, the technical scheme reduces the association degree between the extraction branches of different target attributes by accurately extracting attribute knowledge from environment monitoring data, improves the mutual variability and accurately predicts the future trend of the environment state, thereby realizing efficient, accurate and reliable monitoring of the environment state and providing powerful support for environmental management.

In some alternative embodiments, the target environmental data resolution network is obtained by steps 210-240 as follows.

Step 210, a first environmental monitoring sensor data case set is acquired.

Step 220, loading the first environmental monitoring sensor data case set to a basic attribute extraction branch matched with each of the plurality of monitoring situation labels in a set AI network to perform attribute extraction operation, so as to obtain an environmental association attribute vector case corresponding to each monitoring situation label.

Step 230, determining exclusive cost data corresponding to the target context label pair based on the environment association attribute vector case corresponding to the target context label pair; the target context tag pair is any two of the number of monitoring context tags.

Step 240, according to the exclusive cost data corresponding to the target context label pair, debugging the basic monitoring context association component corresponding to the target context label pair to obtain the target environment data analysis network.

First a set of data cases for a first environmental monitoring sensor needs to be collected. For example, if climate change is being monitored, temperature, humidity, wind speed, etc. related environmental parameters may be collected.

Next, the data case set collected in step 210 is loaded into an AI network. In this AI network, several monitoring context labels are set, such as sunny days, rainy days, storms, etc. Each data case is then associated with its corresponding context label and the underlying attributes of the context are extracted. For example, for a tag on a sunny day, attributes such as high temperature, low humidity, etc. may be extracted. In this way, the environment-associated attribute vector case corresponding to each monitoring context label is obtained.

Further, exclusive cost data corresponding to the target context label is to be determined. This means that context labels that cannot exist at the same time, such as sunny days and rainy days, are to be found. These tags are mutually exclusive so their corresponding exclusive cost data is calculated.

And finally, debugging the basic monitoring situation association component according to the exclusive cost data corresponding to the target situation label. For example, the weights of the sunny and rainy labels may be adjusted so that they are not activated at the same time. Thus, a network that can parse the target environment data is obtained.

It can be seen that applying steps 210-240, all available data can be effectively used by acquiring data from the environmental monitoring sensors and loading into the AI network. This means that each data point is taken into account, improving the utilization of the data. Relevant attributes are extracted based on the specific context labels, so that the system can more accurately understand and classify different environmental contexts. This helps to predict and analyze environmental data more accurately. By calculating the exclusive cost data corresponding to the target context label, the AI network can better judge which contexts are mutually exclusive, thereby avoiding contradictions or errors when making decisions. The monitoring context association component is debugged according to the exclusive cost data, so that the flexibility of the system for processing the complex environment can be improved. For example, if a new environmental context occurs, the system can adapt to the new context by adjusting the components. After the steps, the generated target environment data analysis network can more accurately predict and analyze the environment data, so that a more accurate prediction result is provided in practical application.

Under some other design ideas, the basic attribute extraction branches corresponding to the monitoring context labels comprise a basic convolution component (a sample convolution module) and a basic knowledge mining component (a sample feature mining module), wherein the basic knowledge mining component comprises a first focusing component (an attention module), a first feature mapping component (a nonlinear module), a basic monitoring context association component and a first feature integration component (a feature fusion module). Based on this, in step 220, the first environmental monitoring sensor data case set is loaded to the basic attribute extraction branch matched with each of the plurality of monitoring context labels in the set AI network to perform attribute extraction operation, so as to obtain an environmental association attribute vector case corresponding to each monitoring context label, which includes steps 221-225.

Step 221, loading the first environmental monitoring sensor data case set to the basic convolution component corresponding to each monitoring situation label to perform convolution operation, so as to obtain the environmental monitoring sensor data case convolution characteristics corresponding to each monitoring situation label.

Step 222, loading the environmental monitoring sensor data case convolution characteristic corresponding to each monitoring situation label to the first focusing component corresponding to each monitoring situation label to perform focus enhancement operation, so as to obtain a first focus enhancement vector case corresponding to each monitoring situation label.

Step 223, loading the first focus reinforcement vector case corresponding to each monitoring situation label and the environmental monitoring sensor data case convolution feature corresponding to each monitoring situation label to the first feature mapping component corresponding to each monitoring situation label to perform feature mapping operation, so as to obtain a first basic monitoring mapping vector corresponding to each monitoring situation label.

Step 224, loading the first focus reinforcement vector case corresponding to each monitoring situation label and the environmental monitoring sensor data case convolution feature corresponding to each monitoring situation label to the basic monitoring situation association component corresponding to each monitoring situation label to perform association operation, so as to obtain a first environmental association attribute vector corresponding to each monitoring situation label.

Step 225, loading the first basic monitoring mapping vector corresponding to each monitoring situation label, the first environment association attribute vector corresponding to each monitoring situation label, the first focus reinforcement vector case corresponding to each monitoring situation label and the environmental monitoring sensor data case convolution feature corresponding to each monitoring situation label to the first feature integration component corresponding to each monitoring situation label for vector integration operation, so as to obtain the second basic monitoring mapping vector corresponding to each monitoring situation label.

The process of steps 221-225 involves performing data loading, convolution operations, focus reinforcement, feature mapping, association operations, and vector integration in the AI network to obtain a second base monitoring map vector for each monitoring context label.

First, a collected environmental monitoring sensor data case set is loaded into a basic convolution component corresponding to each monitoring situation label to carry out convolution operation. For example, for a sunny day, the environmental parameters such as temperature, humidity, etc. may be converted into feature vectors through convolution operation, which is the so-called environmental monitoring sensor data case convolution feature.

And then, loading the environmental monitoring sensor data case convolution characteristics corresponding to each monitoring situation label into the first focusing component for focus strengthening operation. This operation is mainly to extract those features that are most important and to increase their weight. For example, in a sunny situation, temperature may be a more important feature, so the weight of the temperature feature may be increased after focus strengthening.

And then, loading the convolution characteristics of the first focus reinforcement vector case and the environment monitoring sensor data case corresponding to each monitoring situation label into a first characteristic mapping component for characteristic mapping operation. This operation is to represent these features in a nonlinear space in order to better capture the relationship between them. The result is a first underlying monitoring mapping vector for each monitoring context label.

Then, the first focus reinforcement vector case and the environmental monitoring sensor data case convolution characteristic corresponding to each monitoring situation label are loaded into a basic monitoring situation association component to carry out association operation. This is mainly done to find context labels that have a common property or some kind of correlation in order to relate them. For example, it may be found that the two labels are similar in some attribute on a sunny day and drying, so they can be correlated.

And finally, loading the first basic monitoring mapping vector, the first environment association attribute vector, the first focus reinforcement vector case and the environment monitoring sensor data case convolution feature corresponding to each monitoring situation label into a first feature integration component for vector integration operation. This operation is to integrate all the feature information to form a more comprehensive and representative feature vector, namely a second basic monitoring mapping vector corresponding to each monitoring situation label.

By applying steps 221-225 and performing convolution operation on the environmental monitoring sensor data through the basic convolution component, useful features can be effectively extracted from the original data, and the accuracy and effectiveness of feature extraction are improved. The first focusing component performs focus strengthening operation on the convolution characteristics, so that the network can focus on the key characteristics with larger influence on the result, and the prediction accuracy of the model is improved. The first feature mapping component maps the features in the nonlinear space, and can capture complex nonlinear feature relations, so that the richness of feature expression is enhanced, and the generalization capability of the model is also improved. The basic monitoring context association component can find out context labels with similar attributes or relativity through association operation, so that the association analysis capability of the model to various environment contexts is improved, and the accurate recognition and prediction of environment changes are facilitated. The first feature integration component integrates all feature information through vector integration operation, so that the model can comprehensively understand and utilize various features, and the prediction accuracy of the model is further improved.

In some possible design considerations, the method further includes steps 310-350.

Step 310, obtaining a first environmental state trend perspective case corresponding to each environmental monitoring sensor data case in the first environmental monitoring sensor data case set.

Step 320, loading the second basic monitoring mapping vector corresponding to each monitoring situation label to the basic discrimination branch corresponding to each monitoring situation label in the set AI network to perform discrimination operation, so as to obtain a third environmental state trend prediction viewpoint corresponding to each monitoring situation label.

And 330, loading the first environmental monitoring sensor data case set into a basic confidence identification component in the set AI network to perform confidence identification, so as to obtain confidence cases corresponding to the environmental monitoring sensor data cases in the first environmental monitoring sensor data case set.

And 340, weighting the third environmental state trend prediction views respectively matched with the plurality of monitoring situation labels according to the confidence coefficient cases to obtain fourth environmental state trend prediction views corresponding to the environmental monitoring sensor data cases in the first environmental monitoring sensor data case set.

Step 350, determining first confidence cost data according to the fourth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case.

Based on this, the step 240 of debugging the corresponding basic monitoring context association component according to the target context tag pair to obtain the target environment data analysis network includes a step 2401.

Step 2401, debugging the corresponding basic monitoring situation association component by the target situation label according to the exclusive cost data of the target situation label pair, and debugging the basic confidence identification component according to the first confidence cost data to obtain the target environment data analysis network.

Step 310-step 350 and step 2401 involve obtaining environmental state trend perspective cases, performing discrimination and confidence recognition, then performing weighting operation on the predicted perspective according to the confidence, and finally determining confidence cost data and performing network debugging.

First, a first environmental state trend perspective case corresponding to each environmental monitoring sensor data case in a first environmental monitoring sensor data case set is obtained. For example, these point of view cases may include a temperature being raised or a humidity being lowered.

And then, loading a second basic monitoring mapping vector corresponding to each monitoring situation label into a basic judging branch in the AI network to carry out judging operation. In this step, the prediction of the environmental state trend is performed according to the extracted and integrated feature information, so as to obtain a third environmental state trend prediction viewpoint corresponding to each monitoring situation label.

The first set of environmental monitoring sensor data cases is then loaded into a base confidence identification component in the AI network for confidence identification. The purpose of this step is to determine the confidence of the model in its predicted outcome.

And weighting the third environmental state trend prediction viewpoint according to the confidence level case. This means that the model will adjust the weight of each prediction view according to its confidence level for the prediction result, and get the fourth environmental state trend prediction view corresponding to each environmental monitoring sensor data case.

And finally, determining the first confidence cost data according to the fourth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case. This is an indicator of the gap between the predicted result and the actual point of view.

Further, the basic monitoring situation association component is debugged according to the exclusive cost data corresponding to the target situation label, and the basic confidence identification component is debugged according to the first confidence cost data, so that the network is optimized and adjusted, and the target environment data analysis network capable of analyzing the environment data more accurately is obtained.

By applying the embodiment, the environmental state trend can be predicted more accurately by acquiring the environmental state trend perspective case and performing the discrimination operation in the AI network. Based on the confidence recognition and weighting operation, the model can have higher confidence on the predicted result, so that the reliability of the predicted result is improved. By determining the confidence cost data and performing network debugging, the network can be optimized and adjusted, so that the environment data can be more accurately analyzed, and the performance of the network is improved. By considering the exclusive cost data and the confidence cost data, different components are debugged, so that the network can be better adapted to various complex environmental conditions, and the adaptability of the network is enhanced.

In other possible embodiments, the basic attribute extraction branch corresponding to each monitoring context label comprises a basic convolution component and a basic knowledge mining component, wherein the basic knowledge mining component comprises a second focusing component, a second feature mapping component, a basic monitoring context association component and a second feature integration component, and the basic monitoring context association component comprises a first basic monitoring context linkage processing node and a second basic monitoring context linkage processing node; the basic monitoring context linkage processing node can be understood as a sample monitoring context matching node. Based on this, in step 220, the first environmental monitoring sensor data case set is loaded to the basic attribute extraction branch matched with each of the plurality of monitoring context labels in the set AI network to perform attribute extraction operation, so as to obtain an environmental association attribute vector case corresponding to each monitoring context label, which includes steps 220 a-220 f.

Step 220a, loading the first environmental monitoring sensor data case set to a basic convolution component corresponding to each monitoring situation label to perform convolution operation, so as to obtain environmental monitoring sensor data case convolution characteristics corresponding to each monitoring situation label.

Step 220b, loading the environmental monitoring sensor data case convolution characteristics corresponding to the monitoring situation labels to the second focusing components corresponding to the monitoring situation labels to perform focus enhancement operation, so as to obtain second focus enhancement vector cases corresponding to the monitoring situation labels.

Step 220c, loading the second focus reinforcement vector cases corresponding to the monitoring situation labels to the first basic monitoring situation linkage processing nodes corresponding to the monitoring situation labels to perform association operation, so as to obtain second environment association attribute vectors corresponding to the monitoring situation labels.

Step 220d, loading the second environment association attribute vector corresponding to each monitoring context label and the environmental monitoring sensor data case convolution feature corresponding to each monitoring context label to the second feature mapping component corresponding to each monitoring context label to perform feature mapping operation, so as to obtain a third basic monitoring mapping vector corresponding to each monitoring context label.

Step 220e, loading the third basic monitoring mapping vector corresponding to each monitoring situation label to the second basic monitoring situation linkage processing node corresponding to each monitoring situation label to perform association operation, so as to obtain a third environment association attribute vector corresponding to each monitoring situation label.

Step 220f, loading the third environment association attribute vector corresponding to each monitoring situation label, the second environment association attribute vector corresponding to each monitoring situation label and the environment monitoring sensor data case convolution feature corresponding to each monitoring situation label to the second feature integration component corresponding to each monitoring situation label for vector integration operation, so as to obtain a fourth basic monitoring mapping vector corresponding to each monitoring situation label.

Steps 220 a-220 f involve further data processing flows including convolution operations, focus reinforcement, correlation operations, feature mapping, vector integration, and the like.

Firstly, loading an environment monitoring sensor data case set to a basic convolution component corresponding to each monitoring situation label to carry out convolution operation, and obtaining the environment monitoring sensor data case convolution characteristics corresponding to each monitoring situation label. Then, the convolution features are loaded to a second focusing component for focus enhancement operation to emphasize the most important features, and a second focus enhancement vector case is generated. Next, a second focus reinforcement vector case is loaded to the first base monitoring context linkage processing node for association operations. The association operation herein may include finding a relationship or interaction between different features, generating a second environment-associated attribute vector. And then, loading the second environment-associated attribute vector and the environment monitoring sensor data case convolution feature to a second feature mapping component to perform feature mapping operation to obtain a third basic monitoring mapping vector. This step may be understood as re-expressing the data in the new feature space, possibly including some non-linear transformations or dimension reduction, etc. And then, loading the third basic monitoring mapping vector to a second basic monitoring situation linkage processing node to perform association operation, so as to obtain a third environment association attribute vector. This step is to further analyze the correlation between features to better understand the environmental situation. And finally, integrating all the characteristic information, including a third environment-related attribute vector, a second environment-related attribute vector and the environment monitoring sensor data case convolution characteristics, and loading the third environment-related attribute vector, the second environment-related attribute vector and the environment monitoring sensor data case convolution characteristics into a second characteristic integration component to perform vector integration operation to obtain a fourth basic monitoring mapping vector corresponding to each monitoring situation label.

By applying the steps 220 a-220 f and performing convolution, focus reinforcement and association operations for a plurality of times, features can be extracted more deeply and accurately, and the prediction accuracy of the model is improved. Through carrying out association operation on two different linkage processing nodes, the association between the features can be mined more deeply, and the understanding capability of the model on the environment state is improved. In the last step, all the characteristic information is integrated, so that not only is the characteristic expression more comprehensive, but also the model can be predicted from various aspects, and the robustness of the model is enhanced.

In a preferred embodiment, the method further comprises steps 410-450.

Step 410, obtaining a first environmental state trend perspective case corresponding to each environmental monitoring sensor data case in the first environmental monitoring sensor data case set.

And step 420, loading the fourth basic monitoring mapping vector corresponding to each monitoring situation label to a basic discrimination branch corresponding to each monitoring situation label in the set AI network to perform discrimination operation, so as to obtain a fifth environmental state trend prediction viewpoint corresponding to each monitoring situation label.

Step 430, loading the first environmental monitoring sensor data case set to a basic confidence identification component in the set AI network to perform confidence identification, so as to obtain confidence cases corresponding to each environmental monitoring sensor data case in the first environmental monitoring sensor data case set.

And step 440, performing a weighting operation on the fifth environmental state trend prediction viewpoints matched by the plurality of monitoring context labels according to the confidence level cases, so as to obtain a sixth environmental state trend prediction viewpoint corresponding to each environmental monitoring sensor data case in the first environmental monitoring sensor data case set.

Step 450, determining second confidence cost data according to the sixth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case.

Based on this, step 240 describes debugging the corresponding basic monitoring context association component by the target context tag according to the corresponding exclusive cost data by the target context tag, to obtain the target environment data analysis network, which includes step 2402.

Step 2402, debugging the corresponding basic monitoring situation association component by the target situation label according to the exclusive cost data of the target situation label pair, and debugging the basic confidence identification component according to the second confidence cost data to obtain the target environment data analysis network.

First, a first environmental state trend perspective case corresponding to each environmental monitoring sensor data case is obtained from a first environmental monitoring sensor data case set. And then, loading a fourth basic monitoring mapping vector corresponding to each monitoring situation label into a basic judging branch in the AI network to carry out judging operation, so as to obtain a fifth environment state trend prediction viewpoint corresponding to each monitoring situation label. This step is in effect a prediction of environmental state trends based on the extracted features. And then, loading the first environment monitoring sensor data case set into a basic confidence identification component in the AI network to carry out confidence identification, so as to obtain a confidence case corresponding to each environment monitoring sensor data case. The purpose of this step is to evaluate the confidence of the model in the self-predicted outcome. And weighting the fifth environmental state trend prediction viewpoint according to the confidence coefficient cases to obtain a sixth environmental state trend prediction viewpoint corresponding to each environmental monitoring sensor data case in the first environmental monitoring sensor data case set. The purpose of this step is to adjust the weight of the prediction horizon according to the confidence of the model in the prediction result. And finally, determining second confidence cost data according to the sixth environmental state trend prediction viewpoint and the first environmental state trend viewpoint case. The purpose of this step is to measure the predictive performance of the model by calculating the gap between the predicted result and the actual point of view.

On the basis, the basic monitoring situation association component and the basic confidence identification component can be debugged by using the exclusive cost data and the second confidence cost data corresponding to the target situation label, so that the network structure is optimized, and the prediction precision of the model is improved.

By applying the embodiment, the environment state trend can be predicted more accurately through deeper feature processing and distinguishing operation. Through confidence level identification and weighting operation, the model has higher confidence on the prediction result of the model, and the reliability of the prediction result is improved. By determining the confidence cost data and performing network debugging, the network can be optimized and adjusted, so that the environment data can be more accurately analyzed, and the performance of the network is improved. By considering the exclusive cost data and the confidence cost data, different components are debugged, so that the network can be better adapted to various complex environmental conditions, and the adaptability of the network is enhanced.

In other possible embodiments, the method further comprises steps 510-540.

Step 510, obtaining a second environmental monitoring sensor data case set and a second environmental state trend perspective case corresponding to each environmental monitoring sensor data case in the second environmental monitoring sensor data case set.

Step 520, loading the second environmental monitoring sensor data case set to a basic attribute extraction branch matched with each of the plurality of monitoring context labels in the set AI network to perform attribute extraction operation, so as to obtain a fifth basic monitoring mapping vector corresponding to each monitoring context label.

And 530, loading the fifth basic monitoring mapping vector corresponding to each monitoring situation label to a basic judging branch corresponding to each monitoring situation label to judge, thereby obtaining a seventh environmental state trend prediction viewpoint corresponding to each monitoring situation label.

Step 540, determining environmental monitoring cost data corresponding to the monitoring context labels according to the seventh environmental state trend prediction perspective corresponding to the monitoring context labels and the second environmental state trend perspective case.

Based on this, step 240 describes debugging the corresponding basic monitoring context association component by the target context tag according to the corresponding exclusive cost data by the target context tag, to obtain the target environment data analysis network, which includes step 2403.

Step 2403, according to the exclusive cost data corresponding to the target context label pair, debugging the basic monitoring context association component corresponding to the target context label pair, and according to the environment monitoring cost data corresponding to each monitoring context label pair, debugging the initial network component corresponding to each monitoring context label in the set AI network to obtain the target environment data analysis network; the initial network component corresponding to each monitoring situation label is a component except for the basic monitoring situation association component corresponding to each monitoring situation label in the basic attribute extraction branch corresponding to each monitoring situation label.

First, a second environmental monitoring sensor data case set and a corresponding second environmental state trend perspective case are obtained. This step is similar to the previous step 410, except that another set of environmental monitoring data is now being processed. And then, loading the second environment monitoring sensor data case set into basic attribute extraction branches corresponding to the monitoring situation labels in the AI network to perform attribute extraction operation, so as to obtain a fifth basic monitoring mapping vector corresponding to each monitoring situation label. This step is similar to the previous step 220f, in that the basic monitoring map vector is generated by feature extraction and feature integration. And then, loading the fifth basic monitoring mapping vector to a basic judging branch to carry out judging operation, so as to obtain a seventh environmental state trend prediction viewpoint corresponding to each monitoring situation label. This step is similar to step 420 in that the environmental state trend prediction is based on the characteristics. And finally, determining the environmental monitoring cost data corresponding to each monitoring situation label according to the seventh environmental state trend prediction viewpoint and the second environmental state trend viewpoint case. This step evaluates the predictive performance of the model by comparing the predicted results with the actual views.

On the basis, the basic monitoring situation association component and the initial network component can be debugged by using the exclusive cost data and the environment monitoring cost data corresponding to the target situation label, so that the network structure is optimized, and the prediction precision of the model is improved.

By applying the embodiment, the environment state trend can be predicted more accurately through deeper feature processing and distinguishing operation. By determining the environmental monitoring cost data and performing network debugging, the network can be optimized and adjusted, so that the environment data can be more accurately analyzed, and the performance of the network is improved. By considering the exclusive cost data and the environment monitoring cost data, different components are debugged, so that the network can be better adapted to various complex environment conditions, and the adaptability of the network is enhanced. By processing the second set of environmental monitoring data, the generalization ability of the model can be checked and improved, ensuring that the model performs well on new and unseen data.

In some preferred embodiments, determining, in step 130, a trend viewpoint of the target environmental state corresponding to the environmental monitoring sensor data to be identified according to the knowledge of the attribute of the environmental monitoring data corresponding to the respective monitoring context tag and the target discrimination branch corresponding to the respective monitoring context tag includes steps 131-132.

And 131, loading the environmental monitoring data attribute knowledge corresponding to each monitoring situation label to a target discrimination branch corresponding to each monitoring situation label to perform discrimination operation, so as to obtain a first environmental state trend prediction viewpoint corresponding to each monitoring situation label.

And 132, determining a target environmental state trend viewpoint corresponding to the environmental monitoring sensor data to be identified according to the first environmental state trend prediction viewpoints respectively matched with the monitoring situation tags.

Steps 131-132 involve predicting environmental state trends using knowledge of environmental monitoring data attributes and determining a target environmental state trend perspective. Firstly, the environmental monitoring data attribute knowledge corresponding to each monitoring situation label is loaded to a corresponding target discrimination branch to carry out discrimination operation. The knowledge of the attribute of the environmental monitoring data is understood as an understanding or knowledge of the environmental change obtained from the historical monitoring data, such as the change rule of a certain meteorological parameter, the distribution characteristic of a certain pollutant, etc. The result of the discriminating operation is a first environmental state trend prediction perspective corresponding to each monitoring context label. And then, determining the target environmental state trend view corresponding to the environmental monitoring sensor data to be identified according to the first environmental state trend prediction view corresponding to all the monitoring situation labels. This step is actually to comprehensively consider the prediction results of all monitoring situations, so as to obtain the final environment state trend prediction.

By applying steps 131-132 and utilizing knowledge of the attribute of the environmental monitoring data, the environmental state trend can be predicted more accurately, and the prediction accuracy is improved. By determining the trend perspective of the target environmental state, powerful support is provided for subsequent decisions. For example, if it is predicted that the air quality of a certain area will deteriorate in the future, measures such as limiting factory emissions, reminding residents of reducing outdoor activities, etc. may be taken in advance. The method can predict the future environmental change according to the historical data, avoids complex calculation and judgment by manpower, and greatly improves the efficiency.

In some possible embodiments, the method further comprises: and loading the to-be-identified environment monitoring sensor data into a target confidence identification component in the target environment data analysis network to carry out confidence identification so as to obtain target confidence. Based on this, the determining, in step 132, the target environmental state trend view corresponding to the environmental monitoring sensor data to be identified according to the first environmental state trend prediction views respectively matched by the plurality of monitoring context labels includes steps 1321-1322.

And 1321, performing a weighting operation on the first environmental state trend prediction viewpoints matched by the monitoring context labels according to the target confidence level, so as to obtain a second environmental state trend prediction viewpoint corresponding to the environmental monitoring sensor data to be identified.

Step 1322, determining the target environmental state trend viewpoint according to the second environmental state trend prediction viewpoint.

Steps 1321-1322 involve weighting the environmental state trend prediction perspective with the target confidence and determining the target environmental state trend perspective accordingly. In the previous step, the target confidence recognition component that loads the environmental monitoring sensor data to be recognized into the target environmental data analysis network performs confidence recognition, and the target confidence is obtained. And then, according to the target confidence, weighting the first environmental state trend prediction viewpoints matched by the monitoring situation labels. This means that the weight of the predicted result is adjusted according to the confidence level of the model on the predicted result, so as to obtain a second environmental state trend prediction viewpoint. And finally, determining the target environmental state trend viewpoint according to the second environmental state trend prediction viewpoint. This means that the prediction viewpoint with the highest confidence (i.e. the highest weight) is selected as the final target environmental state trend viewpoint.

By introducing confidence and weighting, the environmental state trend can be predicted more accurately, thereby improving the prediction accuracy, by applying steps 1321-1322. The determined trend view of the target environmental state has higher confidence, which means higher confidence in the predicted result, thereby enhancing the reliability of the decision. By considering the prediction results of different context labels and weighting, the influence of individual prediction errors can be offset to a certain extent, so that the robustness of the model is improved.

In some exemplary embodiments, in step 120, the loading the environmental monitoring sensor data to be identified into a target attribute extraction branch matched with each of a plurality of monitoring context labels in a target environmental data analysis network performs attribute extraction operation to obtain environmental monitoring data attribute knowledge corresponding to each monitoring context label, including steps 121-122.

And step 121, loading the to-be-identified environmental monitoring sensor data to a target convolution component in a target attribute extraction branch corresponding to each monitoring situation label to carry out convolution operation, so as to obtain the environmental monitoring sensor data convolution characteristics corresponding to each monitoring situation label.

Step 122, loading the environmental monitoring sensor data convolution characteristics corresponding to the monitoring situation labels to a target knowledge mining component in a target attribute extraction branch corresponding to the monitoring situation labels to perform knowledge mining, so as to obtain the environmental monitoring data attribute knowledge corresponding to the monitoring situation labels.

Steps 121-122 involve convolving the environmental monitoring sensor data to be identified to extract features, and then performing knowledge mining to obtain knowledge of the attributes of the environmental monitoring data. Firstly, loading the environment monitoring sensor data to be identified into a target convolution component in a target attribute extraction branch corresponding to each monitoring situation label to carry out convolution operation. Convolution operation is a commonly used feature extraction method, typically used to process images or other structured data. Through the operation, the data convolution characteristics of the environment monitoring sensor corresponding to each monitoring situation label can be obtained. And then, loading the obtained environmental monitoring sensor data convolution characteristics into a target knowledge mining component in a target attribute extraction branch corresponding to each monitoring situation label to carry out knowledge mining. Knowledge mining typically involves techniques of clustering, classification, association rules, etc., for finding valuable patterns and rules from data. Through the operation, the environment monitoring data attribute knowledge corresponding to each monitoring situation label can be obtained.

By applying steps 121-122, valid features can be extracted from the original environmental monitoring data by convolution operations, which facilitate subsequent predictions and analyses. Through knowledge mining, valuable patterns and rules can be found from the data convolution characteristics of the environment monitoring sensor, and the patterns and rules form environment monitoring data attribute knowledge and can enrich an environment knowledge base. The prediction is performed based on deeper and finer features, so that the prediction accuracy can be improved. The obtained knowledge of the attributes of the environmental monitoring data can be used for supporting environmental management decisions, for example, future environmental change trends can be predicted according to a history mode, and references can be provided for decision makers.

In some examples, the target knowledge mining component corresponding to the respective monitoring context tag includes a third focusing component, a third feature mapping component, a target monitoring context association component, and a third feature integration component. Based on this, in step 122, the environmental monitoring sensor data convolution feature corresponding to each monitoring context label is loaded to the target knowledge mining component in the target attribute extraction branch corresponding to each monitoring context label to perform knowledge mining, so as to obtain the environmental monitoring data attribute knowledge corresponding to each monitoring context label, which includes steps 1221-1224.

Step 1221, loading the environmental monitoring sensor data convolution characteristic corresponding to each monitoring situation label to a third focusing component corresponding to each monitoring situation label to perform focus enhancement operation, so as to obtain a third focus enhancement vector case corresponding to each monitoring situation label.

Step 1222, loading the environmental monitoring sensor data convolution feature corresponding to each monitoring context label and the third focus enhancement vector case corresponding to each monitoring context label to the third feature mapping component corresponding to each monitoring context label to perform feature mapping operation, so as to obtain the first target monitoring mapping vector corresponding to each monitoring context label.

Step 1223, loading the environmental monitoring sensor data convolution characteristic corresponding to each monitoring context label and the third focus reinforcement vector case corresponding to each monitoring context label to the target monitoring context association component corresponding to each monitoring context label for association operation, so as to obtain a fourth environmental association attribute vector corresponding to each monitoring context label.

Step 1224, loading the fourth environmental correlation attribute vector corresponding to each monitoring context label, the first target monitoring mapping vector corresponding to each monitoring context label, the environmental monitoring sensor data convolution feature corresponding to each monitoring context label and the third focus enhancement vector case corresponding to each monitoring context label to the third feature integration component corresponding to each monitoring context label for vector integration operation, so as to obtain the environmental monitoring data attribute knowledge corresponding to each monitoring context label.

Steps 1221-1224 further refine the knowledge mining process, including focus augmentation operations, feature mapping operations, association operations, and vector integration operations. Firstly, loading the data convolution characteristics of the environment monitoring sensor corresponding to each monitoring situation label to a third focusing assembly for focus enhancement operation to obtain a third focus enhancement vector case. The goal of this step is to highlight the most important features, thereby increasing the focus of the model attention. And then, loading the environmental monitoring sensor data convolution characteristic and the third focus reinforcement vector case to a third characteristic mapping component for characteristic mapping operation to obtain a first target monitoring mapping vector. The goal of this step is to transform or map the original features to a new feature space to better extract the information. And then, loading the environmental monitoring sensor data convolution characteristic and the third focus reinforcement vector case to a target monitoring situation association component for association operation to obtain a fourth environmental association attribute vector. The goal of this step is to discover the association between the features and provide more information for subsequent predictions. And finally, loading a fourth environment-associated attribute vector, a first target monitoring mapping vector, an environment monitoring sensor data convolution characteristic and a third focus strengthening vector case to a third characteristic integration component for vector integration operation to obtain environment monitoring data attribute knowledge. The goal of this step is to integrate all existing information to form a comprehensive, comprehensive knowledge of the attributes of the environmental monitoring data.

Applying steps 1221-1224, deeper, more abundant features can be extracted from the original environmental monitoring data through focus augmentation, feature mapping, correlation operations, and vector integration. Through the steps, valuable modes and rules can be mined from the environment monitoring sensor data, and the modes and rules form environment monitoring data attribute knowledge and can enrich an environment knowledge base. The prediction is performed based on deeper and finer features, so that the prediction accuracy can be improved.

In other possible embodiments, the target knowledge mining component corresponding to each monitoring context label includes a fourth focusing component, a fourth feature mapping component, a target monitoring context association component, and a fourth feature integration component, and the target monitoring context association component corresponding to each monitoring context label includes a first target monitoring context linkage processing node and a second target monitoring context linkage processing node. Based on this, in step 122, the environmental monitoring sensor data convolution feature corresponding to each monitoring context label is loaded to the target knowledge mining component in the target attribute extraction branch corresponding to each monitoring context label to perform knowledge mining, so as to obtain the environmental monitoring data attribute knowledge corresponding to each monitoring context label, which includes steps 122 a-122 e.

Step 122a, loading the environmental monitoring sensor data convolution characteristic corresponding to each monitoring situation label to a fourth focusing component corresponding to each monitoring situation label to perform focus enhancement operation, so as to obtain a fourth focus enhancement vector case corresponding to each monitoring situation label.

Step 122b, loading the fourth focus reinforcement vector case corresponding to each monitoring situation label to the first target monitoring situation linkage processing node corresponding to each monitoring situation label to perform association operation, so as to obtain a fifth environment association attribute vector corresponding to each monitoring situation label.

Step 122c, loading the fifth environment association attribute vector corresponding to each monitoring context label and the environment monitoring sensor data convolution feature corresponding to each monitoring context label to the fourth feature mapping component corresponding to each monitoring context label to perform feature mapping operation, so as to obtain a second target monitoring mapping vector corresponding to each monitoring context label.

Step 122d, loading the second target monitoring mapping vector corresponding to each monitoring situation label to the second target monitoring situation linkage processing node corresponding to each monitoring situation label to perform association operation, so as to obtain a sixth environment association attribute vector corresponding to each monitoring situation label.

Step 122e, loading the sixth environment-related attribute vector corresponding to each monitoring context label, the fifth environment-related attribute vector corresponding to each monitoring context label, and the environment monitoring sensor data convolution feature corresponding to each monitoring context label to the fourth feature integration component corresponding to each monitoring context label for vector integration operation, so as to obtain the environment monitoring data attribute knowledge corresponding to each monitoring context label.

Steps 122 a-122 e further refine the knowledge mining process, including two focus reinforcement operations, two association operations, feature mapping, and vector integration operations. Firstly, loading the data convolution characteristics of the environment monitoring sensor corresponding to each monitoring situation label to a fourth focusing component for focus strengthening operation to obtain a fourth focus strengthening vector case. This is a feature selection process aimed at improving the performance of the model by reducing extraneous features. And then, loading the fourth focus reinforcement vector case to the first target monitoring situation linkage processing node for association operation to obtain a fifth environment association attribute vector. This step aims to find the inherent relevance between the various features. And then, loading the fifth environment-related attribute vector and the environment monitoring sensor data convolution characteristic into a fourth characteristic mapping component to perform characteristic mapping operation to obtain a second target monitoring mapping vector. This step is mainly to convert or map the original feature space to a new feature space. And then, loading the second target monitoring mapping vector to a second target monitoring situation linkage processing node to perform association operation, so as to obtain a sixth environment association attribute vector. And performing feature relevance analysis again to further extract deep relevant features. And finally, loading the sixth environment-related attribute vector, the fifth environment-related attribute vector and the environment monitoring sensor data convolution feature into a fourth feature integration component to perform vector integration operation, so as to obtain the environment monitoring data attribute knowledge. The method is characterized in that all existing information is integrated to form comprehensive and comprehensive knowledge of the attribute of the environmental monitoring data.

By applying steps 122 a-122 e and through multi-stage focus reinforcement, association operation and feature mapping, the steps extract deeper and richer features from the original environmental monitoring data, so that the model can acquire more information and judge the environmental state more accurately. Generating environment monitoring data attribute knowledge by integrating a plurality of attribute vectors corresponding to each monitoring situation label, so as to provide a comprehensive environment monitoring knowledge base. This is of great value for subsequent decision making. By using techniques such as focus enhancement, association, and feature mapping, more representative and differentiated features can be generated, which helps to improve the accuracy of environmental state predictions. Through multiple processing and feature selection, irrelevant or noise features can be removed, and the adaptability and the robustness of the model to unknown data are improved. The finally formed environment monitoring data attribute knowledge can provide more visual and comprehensive information for a decision maker through multi-level feature processing and integration, so that the decision making efficiency is improved.

Further, there is also provided a readable storage medium having stored thereon a program which when executed by a processor implements the above-described method.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not repeated herein. In the several embodiments provided in the present application, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.

Claims

1. An artificial intelligence based environmental monitoring sensor data analysis method, characterized in that it is applied to a sensor data analysis system, the method comprising:

Collecting data of an environment monitoring sensor to be identified;

2. The method of claim 1, wherein the target environmental data resolution network is obtained by:

acquiring a first environmental monitoring sensor data case set;

3. The method of claim 2, wherein the base attribute extraction branch for each monitoring context label comprises a base convolution component and a base knowledge mining component, the base knowledge mining component comprising a first focusing component, a first feature mapping component, a base monitoring context association component, and a first feature integration component;

4. A method as claimed in claim 3, wherein the method further comprises:

5. The method of claim 2, wherein the base attribute extraction branch corresponding to each monitoring context label comprises a base convolution component and a base knowledge mining component, the base knowledge mining component comprising a second focusing component, a second feature mapping component, a base monitoring context association component and a second feature integration component, the base monitoring context association component comprising a first base monitoring context linkage processing node and a second base monitoring context linkage processing node;

loading the third environment association attribute vector corresponding to each monitoring situation label, the second environment association attribute vector corresponding to each monitoring situation label and the environment monitoring sensor data case convolution feature corresponding to each monitoring situation label to a second feature integration component corresponding to each monitoring situation label for vector integration operation to obtain a fourth basic monitoring mapping vector corresponding to each monitoring situation label;

wherein the method further comprises:

Debugging the corresponding basic monitoring situation association component according to the target situation label pair corresponding exclusive cost data to obtain the target environment data analysis network, wherein the method comprises the following steps:

6. The method of claim 2, wherein the method further comprises:

7. The method of claim 1, wherein the determining the target environmental state trend perspective corresponding to the environmental monitoring sensor data to be identified based on the environmental monitoring data attribute knowledge corresponding to the respective monitoring context tag and the target discrimination branch corresponding to the respective monitoring context tag comprises:

determining a target environment state trend viewpoint corresponding to the environment monitoring sensor data to be identified according to the first environment state trend prediction viewpoints matched by the monitoring situation tags;

wherein the method further comprises:

8. The method of claim 1, wherein loading the environmental monitoring sensor data to be identified into a target attribute extraction branch to which a plurality of monitoring context labels in a target environmental data analysis network are respectively matched performs attribute extraction operation to obtain environmental monitoring data attribute knowledge corresponding to each monitoring context label, and the method comprises:

9. The method of claim 8, wherein,

the target knowledge mining component corresponding to each monitoring situation label comprises a third focusing component, a third feature mapping component, a target monitoring situation association component and a third feature integration component;

loading the environmental monitoring sensor data convolution characteristics corresponding to the monitoring situation labels to a target knowledge mining component in a target attribute extraction branch corresponding to the monitoring situation labels to perform knowledge mining to obtain environmental monitoring data attribute knowledge corresponding to the monitoring situation labels, wherein the method comprises the following steps:

loading a fourth environment association attribute vector corresponding to each monitoring situation label, a first target monitoring mapping vector corresponding to each monitoring situation label, an environment monitoring sensor data convolution characteristic corresponding to each monitoring situation label and a third focus enhancement vector case corresponding to each monitoring situation label to a third characteristic integration component corresponding to each monitoring situation label for vector integration operation, so as to obtain environment monitoring data attribute knowledge corresponding to each monitoring situation label;

or,

the target knowledge mining component corresponding to each monitoring situation label comprises a fourth focusing component, a fourth feature mapping component, a target monitoring situation association component and a fourth feature integration component, and the target monitoring situation association component corresponding to each monitoring situation label comprises a first target monitoring situation linkage processing node and a second target monitoring situation linkage processing node;

10. A sensor data analysis system comprising a processor and a memory; the processor being communicatively connected to the memory, the processor being adapted to read a computer program from the memory and execute it to carry out the method of any of the preceding claims 1-9.