Deep learning-based pathogen identification and classification method and processing device

Abstract

The invention discloses a pathogen identification and classification method based on deep learning and a processing device. The processing device comprises an air disinfection processing device and a related subsystem. The pathogen identification and classification method comprises the following steps: receiving multi-modal sensor data about a target space; wherein the multi-modal sensor data comprises at least one of environmental data, chemical data, biological data and optical data; analyzing the data of the multi-modal sensor to obtain the pathogen type and concentration; working parameters of the air disinfection treatment device and related subsystems are adjusted according to the pathogen type and concentration, and targeted disinfection is conducted on the target space. In this way, rapid recognition and response to different pathogens and chemical pollutants are achieved, and the energy use efficiency of the system is improved.

Description

Technical Field

[0001] The present application relates to the field of deep learning identification and classification, and in particular to a pathogen identification and classification method and processing device. Background Art

[0002] Traditional air treatment systems have a delayed response, are unable to accurately process the characteristics of different pathogens, and are unable to detect biological pathogens and harmful chemical pollution at the same time. Existing systems lack intelligent adjustment capabilities, have low energy efficiency, and are separated from air safety monitoring and control systems, making it difficult to form a closed loop. Summary of the invention

[0003] The deep learning-based pathogen identification and classification method and processing device provided in this application can achieve rapid identification and response to different pathogens and chemical pollutants, and improve the energy utilization efficiency of the system.

[0004] In a first aspect, the present application provides a pathogen identification and classification method based on deep learning, which pathogen identification and classification method includes: receiving multimodal sensor data about a target space; wherein the multimodal sensor data includes at least one of environmental data, chemical data, biological data, and optical data; analyzing the multimodal sensor data to obtain the type and concentration of the pathogen; adjusting the working parameters of the air disinfection treatment device and related subsystems according to the type and concentration of the pathogen to perform targeted disinfection of the target space.

[0005] Among them, the analysis of multimodal sensor data also includes: extracting environmental characteristics, pathogen characteristics and chemical pollutant characteristics from the multimodal sensor data; detecting pathogens based on pathogen characteristics, and detecting chemical pollutants based on chemical pollutant characteristics; monitoring environmental status based on environmental characteristics, and monitoring pathogen status based on pathogen detection results, and monitoring chemical pollutant status based on chemical pollutant detection results; locating pathogen sources based on environmental data and pathogen detection results, and locating chemical pollutant sources based on environmental data and chemical pollutant detection results; conducting an overall risk assessment based on environmental status monitoring information, pathogen status monitoring information, chemical pollutant status monitoring information, pathogen source location information, and chemical pollutant source location information; and formulating a disinfection strategy based on the overall risk assessment results.

[0006] Among them, analyzing the multimodal sensor data includes: analyzing the multimodal sensor data using a deep learning model; wherein the deep learning model is trained using distributed federated learning.

[0007] Among them, the deep learning model uses distributed federated learning for training process as follows: use reinforcement learning to dynamically adjust the communication frequency between edge computing, optimize the edge computing cluster, and adjust the segmentation strategy of the deep learning model; perform edge forward calculations for the adjusted segmentation strategy, generate intermediate representations, and upload the intermediate representations to the regional nodes; perform regional aggregation calculations based on the optimized edge computing cluster and intermediate representations, fuse multi-node data, and upload the fusion results to the server; perform central joint modeling training in the server; calculate the global gradient based on the adjusted communication frequency and send it to the regional node; update the regional gradient in combination with the local data of the node; send the gradient to the edge device so that the edge device can update the local model.

[0008] Among them, multimodal sensors are deployed in a distributed manner in the target space to form a distributed sensing network.

[0009] After receiving the multimodal sensor data about the target space, it includes: preprocessing the multimodal sensor data and selecting a data transmission protocol; distributing tasks based on the preprocessed multimodal sensor data, and selecting corresponding edge computing nodes to process the tasks.

[0010] Among them, a digital twin model is built for the air disinfection treatment device and related systems, and the digital twin model is used for simulation and prediction.

[0011] In a second aspect, the present application provides a processing device, which includes an air disinfection processing device and related subsystems, for implementing the deep learning-based pathogen identification and classification method provided in the first aspect.

[0012] Among them, the relevant subsystems include: at least one of an active air purification system, an ultrasonic air purification system, an intelligent fan control system, and a variable rate air supply system.

[0013] Among them, multiple sensors corresponding to multimodal sensor data are integrated on the microfluidic chip.

[0014] The beneficial effect of the present application is as follows: Different from the prior art, the present application provides a deep learning-based pathogen identification and classification method and processing device, the processing device includes an air disinfection processing device and related subsystems, and the pathogen identification and classification method includes: receiving multimodal sensor data about the target space; wherein the multimodal sensor data includes at least one of environmental data, chemical data, biological data and optical data; analyzing the multimodal sensor data to obtain the type and concentration of the pathogen; adjusting the working parameters of the air disinfection processing device and related subsystems according to the type and concentration of the pathogen, and performing targeted disinfection on the target space, which can achieve rapid identification and response to different pathogens and chemical pollutants through the establishment of an intelligent air safety prevention and control system, thereby improving the energy efficiency of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them: Figure 1 It is a flow chart of an embodiment of a pathogen identification and classification method and processing device based on deep learning provided by the present application; Figure 2 It is a flow chart of another embodiment of a pathogen identification and classification method and processing device based on deep learning provided by the present application; Figure 3 It is a flowchart of an embodiment of deep learning model training provided by the present application; Figure 4 It is a structural schematic diagram of an embodiment of a processing device provided in the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. It will be understood that the specific embodiments described herein are only used to explain the present application, rather than to limit the present application. It should also be noted that, for ease of description, only some but not all structures related to the present application are shown in the drawings. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0017] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0018] See also Figure 1 , Figure 1 1 is a flow chart of an embodiment of a pathogen identification and classification method and processing device based on deep learning provided by the present application. The device includes an air disinfection processing device and related subsystems. The pathogen identification and classification method includes: Step 11: Receive multimodal sensor data about the target space; wherein the multimodal sensor data includes at least one of environmental data, chemical data, biological data and optical data.

[0019] In some embodiments, the multimodal sensor data includes at least one of environmental data, chemical data, biological data, and optical data.

[0020] In some embodiments, the multimodal sensor data includes at least two of environmental data, chemical data, biological data, and optical data.

[0021] In some embodiments, the multimodal sensor data includes at least three of environmental data, chemical data, biological data, and optical data.

[0022] In some embodiments, the multimodal sensor data includes environmental data, chemical data, biological data, and optical data.

[0023] Step 12: Analyze the multimodal sensor data to obtain pathogen type and concentration.

[0024] In some embodiments, a deep learning model can be used to analyze multimodal sensor data to obtain pathogen types and concentrations.

[0025] In some embodiments, the multimodal sensor data may be analyzed using corresponding rules to obtain the pathogen type and concentration, such as using an expert system rule base, a conditional logic tree, and / or predefined threshold rules to analyze the multimodal sensor data to obtain the pathogen type and concentration.

[0026] Step 13: Adjust the operating parameters of the air disinfection treatment device and related subsystems according to the type and concentration of pathogens to carry out targeted disinfection of the target space.

[0027] In some embodiments, the type, concentration and treatment time of disinfectants are dynamically adjusted according to the test results and environmental conditions to achieve personalized disinfection. The ventilation system is automatically adjusted according to parameters such as indoor air quality and pathogen concentration to optimize the indoor air environment. The fan air volume is automatically adjusted according to the indoor air quality and pathogen concentration to achieve efficient disinfection. Differentiated control of different spaces is achieved by adjusting the air supply speed.

[0028] Differentiated control of different spaces can be achieved by adjusting the air supply speed of the variable rate air supply system.

[0029] Active air purification systems can continuously purify the air by actively generating negative ions and photocatalysis.

[0030] In other embodiments, the indicators monitored by the system may include: TEMP (temperature), HUMI ​​(humidity), NH3 (ammonia), CO (carbon monoxide), CH2O (formaldehyde), O2 (oxygen), CO2 (carbon dioxide), O3 (ozone), SO2 (sulfur dioxide), C6H6 (benzene), C7H8 (toluene), C8H 10 At least one of (xylene), RN (radon), TVOC (total volatile organic compounds), PM0.3 (particulate matter entering the lungs), PM0.5 (fine particulate matter), PM1.0 (fine particulate matter), PM2.5 (fine particulate matter), PM10 (inhalable particulate matter), Beta-ray (β-ray), Gamma-ray (γ-ray) and X-ray (X-ray).

[0031] In this embodiment, the sensor data collected by various sensors such as spectral, electrochemical, mass, biological, and gas sensors can be combined to construct a multimodal fusion model to achieve accurate identification of different types of pathogens and chemical pollutants while realizing the pollution source location and pollution status monitoring functions.

[0032] In this embodiment, the processing device uses multiple sensors to monitor indoor air quality, pathogen concentration and other parameters in real time. Through multi-sensor fusion pattern recognition, different types of pathogens and chemical pollutants can be accurately identified while realizing pollution source positioning and pollution status monitoring functions. According to the analysis results, the system automatically adjusts the working parameters of the ventilation system and disinfection device to achieve targeted disinfection of different types of pathogens. Through the establishment of an intelligent air safety prevention and control system, it can realize rapid identification and response to different pathogens and chemical pollutants, and improve the energy efficiency of the system.

[0033] See also Figure 2 , Figure 2 It is a flow chart of another embodiment of the pathogen identification and classification method and processing device based on deep learning provided in the present application.

[0034] Step 21: Receive multimodal sensor data about the target space; wherein the multimodal sensor data includes at least one of environmental data, chemical data, biological data and optical data.

[0035] In some embodiments, multimodal fusion may be performed on the received multimodal sensor data to obtain multimodal fusion data.

[0036] Step 22: Extract environmental features, pathogen features, and chemical pollutant features from the multimodal sensor data.

[0037] In some embodiments, environmental features, pathogen features, and chemical pollutant features may be extracted from the multimodal fusion data.

[0038] Step 23: Perform pathogen detection based on pathogen characteristics, and perform chemical pollutant detection based on chemical pollutant characteristics.

[0039] In some embodiments, biological data can be collected by biosensors. A biosensor is a device that combines a biological recognition element with sensor technology, which converts target molecules (such as pathogenic microorganisms) into measurable models through a specific biological recognition process. Its working principle can be divided into immune sensing, cell sensing and enzyme sensing.

[0040] In some embodiments, chemical data can be collected by chemical sensors. The principle is to select appropriate chemical substances in advance so that when pollutants come into contact with chemical materials, they can react, thereby changing the electrical properties of the sensor, and then realizing the detection of pollutants. For example, the concentration of pollutants such as PM2.5, PM10, VOCs, etc. in the air can be monitored.

[0041] Step 24: Conduct environmental status monitoring based on environmental characteristics, conduct pathogen status monitoring based on pathogen detection results, and conduct chemical pollutant status monitoring based on chemical pollutant detection results.

[0042] In some embodiments, environmental parameters are analyzed to implement environmental status monitoring and obtain environmental status information.

[0043] In some embodiments, pathogen status information is obtained by using pathogen dynamic monitoring based on time series analysis for environmental data and pathogen detection results.

[0044] In some embodiments, chemical pollutant status information is obtained by using dynamic monitoring of chemical pollutants based on time series analysis for environmental data and chemical pollutant detection results.

[0045] Step 25: Locate the source of pathogens based on environmental data and pathogen detection results, and locate the source of chemical pollutants based on environmental data and chemical pollutant detection results.

[0046] In some embodiments, a pathogen source positioning method based on triangulation positioning and diffusion model is used for environmental data and pathogen detection results to obtain the pathogen release location and realize pathogen source positioning.

[0047] In some embodiments, a chemical pollutant source location method based on a diffusion model is used for environmental data and chemical pollutant detection results to obtain the chemical pollutant release location and achieve chemical pollutant source location.

[0048] Step 26: Conduct an overall risk assessment based on the environmental status monitoring information, pathogen status monitoring information, chemical pollutant status monitoring information, pathogen source location information, and chemical pollutant source location information.

[0049] Step 27: Develop a disinfection strategy based on the overall risk assessment results.

[0050] In some embodiments, a deep learning model can be used to formulate a disinfection strategy based on the overall risk assessment results. Furthermore, during the disinfection process, the latest multimodal sensor data can be collected in real time, and the disinfection strategy can be updated according to the latest multimodal sensor data and the control logic.

[0051] In some embodiments, a deep learning model may be used to analyze multimodal sensor data; wherein the deep learning model is trained using distributed federated learning.

[0052] In some embodiments, features used for training can be obtained from multimodal sensor data. For example, multimodal fusion can be performed on the received multimodal sensor data to obtain multimodal fusion data. High-dimensional feature representation is performed on the multimodal fusion data, and then feature dimension reduction, feature enhancement and semantic association modeling are performed to obtain a high-dimensional feature representation. The data is classified and labeled, and the classified and labeled features are used as training features. Among them, classification based on rules and classification based on machine learning can be performed.

[0053] See also Figure 3 ,The deep learning model uses distributed federated learning for training process is as follows: Step 31: Use reinforcement learning to dynamically adjust the communication frequency with edge computing, optimize the edge computing cluster, and adjust the segmentation strategy of the deep learning model.

[0054] Step 32: Perform edge forward calculation for the adjusted segmentation strategy, generate an intermediate representation, and upload the intermediate representation to the regional node.

[0055] Step 33: Perform regional aggregation calculations based on the optimized edge computing cluster and intermediate representation, fuse multi-node data, and upload the fusion results to the server.

[0056] Step 34: Perform central joint modeling training in the server.

[0057] Step 35: Based on the adjusted communication frequency, calculate the global gradient and send it to the regional nodes.

[0058] Step 36: Update the regional gradient based on the local data of the node.

[0059] Step 37: Send the gradient to the edge device so that the edge device updates the local model.

[0060] After completing a round of training, determine whether the edge model (local model) reasoning and fusion requirements are met. If so, perform multimodal model fusion to improve generalization capabilities.

[0061] In some implementations, multimodal sensors are deployed in a distributed manner within the target space to form a distributed sensing network that can achieve real-time monitoring and dynamic response to the environment within the range of a single device and between multiple devices.

[0062] After forming a distributed sensor network, after receiving multimodal sensor data about the target space, the multimodal sensor data is preprocessed and the data transmission protocol is selected; tasks are distributed based on the preprocessed multimodal sensor data, and the corresponding edge computing nodes are selected to process the tasks.

[0063] For example, for distributed sensor networks, data collection involves multimodal sensor access, data collection frequency and synchronization settings, modal data alignment and timestamp synchronization, raw data quality assessment and calibration, and then data preprocessing and compression at the edge node, as well as the selection of data transmission protocols and QOS calculation.

[0064] Among them, edge data preprocessing mainly involves noise filtering and smoothing operations on the data, and then data standardization / normalization operations, and further data fusion, such as early fusion / late fusion. The preprocessed data is stored at the edge, such as storage according to the corresponding cache strategy, and local data backup is performed to generate corresponding logs.

[0065] After data preprocessing, cluster management and monitoring are performed, and then task distribution and load balancing are performed based on the preprocessed data, and the corresponding tasks are assigned to the corresponding edge computing nodes. Such as edge computing node 1, edge computing node 2 and edge computing node 3. And the edge model can be updated and iterated using the preprocessed data. And the edge model can be used to perceive the environment of the preprocessed data and obtain corresponding intelligent decisions. And the environmental perception system also needs to set up corresponding API interfaces, such as between the edge and the cloud server, between the edge and the device, and the API interface needs to be securely authenticated.

[0066] Furthermore, the cloud server can aggregate and compress the corresponding intelligent decision and preprocessed data, store / analyze the cloud data, and evaluate the corresponding analysis results using the first evaluation network. Furthermore, the edge model can be updated and iterated using the analysis results.

[0067] Among them, the relevant subsystems include: at least one of an active air purification system, an ultrasonic air purification system, an intelligent fan control system, and a variable rate air supply system.

[0068] Among them, a digital twin model is built for the air disinfection treatment device and related systems, and the digital twin model is used for simulation and prediction.

[0069] See also Figure 4 , Figure 4 1 is a schematic diagram of the structure of an embodiment of a processing device provided by the present application. The processing device 100 includes an air disinfection processing device 10 and a related subsystem 20, which is used to implement the control of the pathogen identification and classification method based on deep learning as in any of the above embodiments. The related subsystem 20 includes: at least one of an active air purification system, an ultrasonic air purification system, an intelligent fan control system, and a variable rate air supply system.

[0070] In some embodiments, the related subsystem 20 includes at least one of an active air purification system, an ultrasonic air purification system, an intelligent fan control system, and a variable rate air supply system.

[0071] Among them, multiple sensors corresponding to multimodal sensor data are integrated on the microfluidic chip. Integrating multiple sensors into the microfluidic chip can achieve miniaturized, portable, and high-throughput detection.

[0072] In one application scenario, the air disinfection treatment device 10 can be modularly designed. For example, the scene can be customized to modularize the scenes such as hospitals, offices, and homes. And the corresponding universal interfaces can be standardized to achieve quick plug-in and pull-out.

[0073] The processing device 100 has a corresponding environmental perception system, which can perform multi-physics field simulations. For example, thermodynamic simulation can optimize the evaporation and diffusion of disinfectants. For example, CFD optimization can analyze air flow and pollutant distribution, as well as experimental verification and simulation feedback to dynamically adjust design parameters.

[0074] The environmental perception system can make intelligent decisions based on sensor data, evaluate the intelligent decisions, and optimize based on the evaluation, such as optimizing the process of data fusion and analysis, such as optimizing the process of edge computing and cloud computing. And optimize the corresponding disinfection strategy based on the evaluation. Then the strategy is issued according to the disinfection strategy. For example, it is issued to the active air purification system, ultrasonic air purification system, intelligent fan control system, and variable rate air supply system, so that the active air purification system, ultrasonic air purification system, intelligent fan control system, and variable rate air supply system can perform corresponding work. Furthermore, a second evaluation can be conducted in combination with the disinfection strategy, so as to use the corresponding disinfection strategy loss function to adjust the parameters in the disinfection strategy.

[0075] Furthermore, the processing device 100 also has an intelligent integration platform, which can interact with the active air purification system, the ultrasonic air purification system, the intelligent fan control system, the variable rate air supply system, the sensor module, and the air disinfection processing device to achieve remote visual control.

[0076] Furthermore, the processing device 100 can be divided into an intelligent sensor fusion system layer, an intelligent control layer and an execution system layer.

[0077] The intelligent sensor fusion system layer can be divided into sampling unit and detection unit.

[0078] The sampling unit may include a multi-channel injection, a pre-concentration device, and an automatic sampler.

[0079] The detection unit may include a spectrum detection module, an electrochemical sensing module, and a mass sensor.

[0080] The intelligent control layer can involve signal processing and AI analysis.

[0081] Signal processing can include multi-source data fusion, feature extraction, and pattern recognition.

[0082] AI analysis can utilize deep learning model analysis and cross-validation to provide real-time warnings.

[0083] The execution system layer may include modular disinfection processing units, intelligent ventilation control systems, variable frequency ultrasonic collaborative air treatment systems, etc.

[0084] Specifically, the intelligent ventilation control system based on environmental perception can automatically adjust the ventilation system and optimize the indoor air environment according to parameters such as indoor air quality and pathogen concentration.

[0085] The variable frequency ultrasonic collaborative air purification system can improve air purification efficiency and reduce energy consumption.

[0086] The modular design can be easily customized according to different scenarios and needs.

[0087] The variable frequency ultrasonic system can produce cavitation effect through ultrasound to destroy the cell structure of pathogens.

[0088] The intelligent fan control system can automatically adjust the fan air volume according to the indoor air quality and pathogen concentration to achieve efficient disinfection.

[0089] The variable rate air supply system can achieve differentiated control of different spaces by adjusting the air supply speed.

[0090] The adaptive disinfection strategy can dynamically adjust the type, concentration and treatment time of disinfectants according to test results and environmental conditions to achieve personalized disinfection.

[0091] Collaborative multi-physics simulation can combine CFD simulation, thermodynamic simulation, etc. to optimize air flow and disinfectant distribution and improve disinfection efficiency.

[0092] Active air purification can continuously purify the air by actively generating negative ions, photocatalysis, etc.

[0093] In summary, the processing device 100 uses multiple sensors to monitor indoor air quality, pathogen concentration and other parameters in real time. Through multi-sensor fusion pattern recognition, different types of pathogens and chemical pollutants can be accurately identified while achieving pollution source positioning and pollution status monitoring functions.

[0094] The processing device 100 uses a deep learning model to analyze the sensor data to accurately identify the type and concentration of pathogens.

[0095] The processing device 100 automatically adjusts the working parameters of the ventilation system and the disinfection device according to the analysis results to achieve targeted disinfection of different types of pathogens. On the premise of protecting data privacy, the federated learning technology is used to jointly model data from different places to improve the generalization ability of the model.

[0096] The processing device 100 continuously adjusts the control strategy according to the disinfection effect to achieve dynamic optimization of indoor air quality.

[0097] The processing device 100 deploys sensor nodes in different areas to form a distributed sensor network, thereby realizing partitioned monitoring and dynamic response to the environment within a single device and between multiple devices.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed pathogen identification and classification method and processing device based on deep learning can be implemented in other ways. For example, the device implementation described above is only schematic, for example, the division of the modules or units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0099] If the integrated units in the above other embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the pathogen identification and classification method based on deep learning described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0100] The above description is only an implementation method of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A pathogen identification and classification method based on deep learning, the pathogen identification and classification method comprising: Receiving multimodal sensor data about the target space; wherein the multimodal sensor data includes at least one of environmental data, chemical data, biological data, and optical data; Analyzing the multimodal sensor data to obtain pathogen type and concentration; The operating parameters of the air disinfection treatment device and related subsystems are adjusted according to the type and concentration of the pathogens to carry out targeted disinfection of the target space.

2. The pathogen identification and classification method according to claim 1, characterized in that: The analyzing the multimodal sensor data further includes: extracting environmental features, pathogen features, and chemical pollutant features from the multimodal sensor data; Performing pathogen detection based on the pathogen characteristics, and performing chemical pollutant detection based on the chemical pollutant characteristics; Performing environmental status monitoring based on the environmental characteristics, performing pathogen status monitoring based on pathogen detection results, and performing chemical pollutant status monitoring based on chemical pollutant detection results; and locating the source of pathogens based on the environmental data and the pathogen detection results, and locating the source of chemical pollutants based on the environmental data and the chemical pollutant detection results; Conduct overall risk assessment based on environmental status monitoring information, pathogen status monitoring information, chemical pollutant status monitoring information, pathogen source location information, and chemical pollutant source location information; Develop a disinfection strategy based on the overall risk assessment results.

3. The pathogen identification and classification method according to claim 1, characterized in that: The analyzing the multimodal sensor data includes: Analyzing the multimodal sensor data using a deep learning model; Wherein, the deep learning model is trained using distributed federated learning.

4. The pathogen identification and classification method according to claim 3, characterized in that: The deep learning model uses distributed federated learning for training as follows: Use reinforcement learning to dynamically adjust the communication frequency with edge computing, optimize edge computing clusters, and adjust the segmentation strategy of deep learning models; Perform edge forward calculation for the adjusted segmentation strategy, generate an intermediate representation, and upload the intermediate representation to the regional node; Perform regional aggregation calculations based on the optimized edge computing cluster and the intermediate representation, fuse multi-node data, and upload the fusion results to the server; Central joint modeling training is performed in the server; According to the adjusted communication frequency, the global gradient is calculated and sent to the regional nodes; Combine the local data of the node to update the regional gradient; Send gradients to edge devices so that they can update their local models.

5. The pathogen identification and classification method according to claim 1, characterized in that: Multimodal sensors are distributedly deployed in the target space to form a distributed sensor network.

6. The pathogen identification and classification method according to claim 5, characterized in that: After receiving the multimodal sensor data about the target space, the method further comprises: Performing data preprocessing on the multimodal sensor data and selecting a data transmission protocol; Tasks are distributed based on the preprocessed multimodal sensor data, and the corresponding edge computing nodes are selected to process the tasks.

7. The pathogen identification and classification method according to claim 6, characterized in that: A digital twin model is constructed for the air disinfection treatment device and related systems, and the digital twin model is used for simulation and prediction.

8. A processing device, characterized in that: The processing device includes an air disinfection processing device and related subsystems, which are used to implement the pathogen identification and classification method as described in any one of claims 1-7.

9. The processing device according to claim 8, characterized in that The related subsystems include: at least one of an active air purification system, an ultrasonic air purification system, an intelligent fan control system, and a variable rate air supply system.

10. The processing device according to claim 9, characterized in that The multiple sensors corresponding to the multimodal sensor data are integrated on a microfluidic chip.

Images