Atmospheric component observation inspection system and method based on artificial intelligence

Abstract

The invention provides an atmospheric component observation and inspection system and method based on artificial intelligence, and is applied to the technical field of artificial intelligence, and the system comprises augmented reality glasses which comprise a sensor module and a data processing module and are used for collecting image data of an observation station and audio data of inspection personnel in real time, preprocessing the image data and the audio data to obtain collected data; the abnormity judgment module is used for analyzing whether the running state of the observation station is abnormal or not according to the preprocessed collected data; the exception handling module is used for handling the condition that the operation state output by the exception judgment module is abnormal; and the self-adaptive scheduling module is used for determining corresponding parameter weights according to the output conditions of the three, and scheduling the execution sequence of the three according to the parameter weights. According to the invention, timely guidance and assistance are provided for maintenance and overhaul of inspection personnel, the efficiency is improved, and a good foundation is laid for further standardizing the atmospheric component observation inspection work and improving the observation quality.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based atmospheric composition observation and inspection system and method. Background Technology

[0002] The instruments used for atmospheric composition observation are complex in structure, highly sensitive and accurate, and highly automated. During the observation process, in addition to the instruments themselves requiring various parameter checks, regular or irregular inspections, maintenance, and calibrations by patrol personnel are also necessary. Due to differences in the knowledge level, professional skills, instrument operation and maintenance proficiency, and understanding of observation specifications among patrol personnel, as well as subjective and objective factors such as each patrol personnel's work attitude, sense of responsibility, meticulousness, physical condition, and mental state, individual differences can occur in the observation process, sometimes directly or indirectly affecting the quality of the observation data.

[0003] Furthermore, the high standards and stringent requirements of atmospheric composition observation operations inevitably require a significant investment of human, material, and financial resources, leading to low work efficiency and limited benefits. This also prevents inspection personnel from truly and rigorously implementing or enforcing the required standards. Even when strict standards are established, considerations regarding the capabilities, energy, and technical skills of inspection personnel necessitate relaxation or leniency of certain requirements. Therefore, to ensure the normal operation of observation instruments, standardize and unify observation procedures and the various operations of inspection personnel, and obtain accurate, reliable, comparable, and truly high-quality observation data that represents and reflects the true state of atmospheric composition, it is essential to maintain strict requirements and unify procedures, methods, requirements, and standards. Only then can all work be carried out in a structured and systematic manner.

[0004] Therefore, there is an urgent need to develop an artificial intelligence-based atmospheric composition observation and inspection system and method to solve the above problems. Summary of the Invention

[0005] This disclosure provides an artificial intelligence-based atmospheric composition observation and inspection system and method to solve the above-mentioned problems. It addresses the issue in the prior art where differences among inspection personnel lead to individual variations in the observation process, which may directly or indirectly affect the observation quality.

[0006] According to a first aspect of this disclosure, an artificial intelligence-based atmospheric composition observation and inspection system is provided. The system includes: augmented reality glasses, comprising a sensor module and a data processing module, for real-time acquisition of image data from an observation station and audio data from inspection personnel, and for preprocessing the image data and the audio data to obtain preprocessed acquisition data; An anomaly detection module, which is communicatively connected to the augmented reality glasses, is used to analyze whether there are any anomalies in the operating status of the observation station based on the preprocessed collected data; An exception handling module, which is communicatively connected to the exception judgment module, is used to handle situations where there are abnormalities in the operating status output by the exception judgment module; An adaptive scheduling module, which is communicatively connected to the augmented reality glasses, the anomaly detection module, and the anomaly handling module, is used to determine the parameter weights of the three components based on their respective outputs, and to schedule the execution order of the three components according to the parameter weights.

[0007] Furthermore, the data processing module includes a speech recognition unit and an image recognition unit; The speech recognition unit is used to convert the acquired audio data of the inspection personnel into text commands and calculate the text confidence level corresponding to the text commands. The image recognition unit is used to perform target detection and recognition on the captured observation station image data to obtain equipment status data, and to calculate the image confidence level corresponding to the equipment status data. The preprocessed acquired data includes the text commands and their corresponding text confidence scores, device status data and their corresponding image confidence scores.

[0008] Furthermore, the anomaly detection module includes an intelligent recognition algorithm library, a data analysis engine, and a data management module; wherein, The intelligent identification algorithm library is used to store and identify the conversion algorithms corresponding to different atmospheric composition observation equipment; The data analysis engine, which is connected to the intelligent recognition algorithm library, is used to transform the preprocessed collected data based on the transformation algorithm matched with the current device to obtain standardized observation data. The data management module, which is connected to the data analysis engine, includes a data receiving subunit, a threshold comparison subunit, and a result output subunit. It is used to receive the standardized observation data and compare it with a preset threshold range to analyze whether there are any abnormalities in the operating status.

[0009] Furthermore, the data management module compares the standardized observation data with a preset threshold range to analyze whether there are any anomalies in the operating status, specifically including the following steps: A real-time change curve is constructed based on the standardized observation data; The real-time change curve is compared with the historical change curve by dynamic time warping to detect three typical patterns in the real-time change curve, which include slow drift, periodic fluctuation and instantaneous jump. Based on a preset baseline, it is determined whether the real-time change curve of the detected corresponding typical mode exceeds a threshold range. If it does, it is determined that the operating status of the observation station equipment is abnormal; wherein, The threshold range includes a dynamic threshold range and a static threshold range. The static threshold range is the allowable deviation range based on a preset baseline, and the dynamic threshold range is the boundary fluctuation range based on the deviation range of the static threshold range.

[0010] Furthermore, the exception handling module is used to perform the following steps: The preprocessed data collected at the time of the anomaly is automatically saved as evidence. The preprocessed collected data corresponding to the abnormal moment is input into the equipment health status model, and the observation station outputs the predicted fault risk and potential faulty components. The system retrieves maintenance guidance information corresponding to the potentially faulty component from a pre-set maintenance database and provides it to the inspection personnel via the augmented reality glasses; wherein, The device health status model is constructed through the following steps: Acquire and preprocess the historical abnormal state dataset of the observation station; The machine learning model is trained using the preprocessed historical abnormal state dataset until the loss function in the machine learning model converges, thus obtaining the equipment health status model.

[0011] Furthermore, the adaptive scheduling module is used to perform the following steps: The current task type, the image and audio data collected by the augmented reality glasses, and the output of the augmented reality glasses, the anomaly detection module, and the anomaly handling module are obtained and input into the finite state machine. The finite state machine determines the current working state and module call range of the three components based on a preset state matching matrix. The calculation module calls the augmented reality glasses, the anomaly detection module, and the anomaly handling module, and generates a module scheduling priority list based on the parameter weights. The execution order of the three is scheduled according to the module scheduling priority list.

[0012] Furthermore, the state matching matrix is ​​set based on a two-dimensional decision table within the finite state machine, which establishes matching relationships between different inputs of the finite state machine and different working states.

[0013] Furthermore, it also includes: The report generation module is used to generate standardized inspection reports based on the data output by the data processing module and the anomaly detection module. The file upload module is used to upload the standardized inspection reports, image data and audio data to the server in real time or at regular intervals. The remote assistance module is used to establish an audio and / or video communication connection between the augmented reality glasses and the remote expert terminal; A Chinese-foreign language translation module is used for real-time translation of the audio data or text materials; The guided tour service module is used to provide customized guided tour services through the augmented reality glasses based on preset narration data.

[0014] According to a second aspect of this disclosure, an artificial intelligence-based method for atmospheric composition observation and inspection is provided, the method comprising the following steps: Image data from the observation station and audio data from the inspection personnel are collected using augmented reality glasses. The image data and audio data are then preprocessed to obtain preprocessed collected data. The anomaly detection module analyzes the preprocessed collected data to determine if there are any anomalies in the operating status of the observation station; the anomaly handling module processes any cases where the anomaly detection module outputs an abnormal operating status. The adaptive scheduling module determines the parameter weights of the augmented reality glasses, the anomaly detection module, and the anomaly handling module based on their outputs, and schedules the execution order of the three modules according to these parameter weights.

[0015] Furthermore, it also includes: An environmental perception module is installed inside the augmented reality glasses to collect environmental data in real time; The environmental data and the preprocessed acquired data are synchronized in time, and the spatial location of the device in the environmental data and the image data is matched to obtain a multimodal data packet; The corresponding interference patterns are identified and marked based on the multimodal data packets; The data compensation engine set in the anomaly judgment module calls the corresponding compensation model according to the marked interference mode, and outputs the corrected observation value based on the input marked multimodal data packet. Within the anomaly detection module, the environmental data is input into the dynamic threshold model to calculate the dynamic threshold range under the corresponding interference mode. It is then determined whether the corrected observation value is within the dynamic threshold range. If not, proceed to the next step. The adaptive scheduling module performs scheduling according to the preset scheduling strategy corresponding to the interference mode.

[0016] Compared to existing technologies, the beneficial effects of this disclosure are: This disclosure sets up an adaptive scheduling module that calls other modules, enabling these other modules to perform the following tasks: the augmented reality glasses acquire image and audio data during inspection through the sensor module, use the anomaly detection module to perform risk assessment and alarms, and the anomaly handling module to handle equipment anomalies; at the same time, it can also provide video playback and inspection of the entire inspection process.

[0017] This disclosure addresses the issues of standardized processes, methods, requirements, and standards in atmospheric composition observation inspections, maintenance, and repairs based on artificial intelligence and augmented reality technologies. It improves the accuracy of speech recognition and image recognition technologies for input from inspection personnel using dialects or non-standard languages ​​through machine learning, ensuring the precision of inspection records and anomaly assessments. It provides timely guidance and assistance to inspection personnel during maintenance and repairs, not only improving efficiency but also laying a solid foundation for further standardizing atmospheric composition observation inspection work and improving observation quality.

[0018] It should be understood that the description in the Summary of the Invention is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0019] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. The drawings are provided for a better understanding of the invention and are not intended to limit the scope of this disclosure. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 A structural block diagram of an artificial intelligence-based atmospheric composition observation and inspection system provided in an embodiment of this disclosure is shown. Figure 2 A structural block diagram of the anomaly detection module provided in an embodiment of this disclosure is shown; Figure 3 A flowchart of an artificial intelligence-based atmospheric composition observation and inspection method provided in an embodiment of this disclosure is shown; Figure 4 A block diagram of an electronic device provided according to an embodiment of the present disclosure is shown. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0021] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0022] This disclosure specifically provides an artificial intelligence-based atmospheric composition observation and inspection system 100, see [link to relevant documentation]. Figure 1 It includes the following modules: Augmented reality glasses 110 include a sensor module and a data processing module, used to collect image data from the observation station and audio data from the inspection personnel in real time, and to preprocess the image data and audio data to obtain preprocessed collected data.

[0023] Augmented reality (AR) is an artificial intelligence technology that cleverly integrates virtual information with the real world. It uses various technologies such as multimedia, real-time tracking and registration, intelligent interaction, and sensing to simulate and apply computer-generated text, images, audio, and video to the real world, thereby "enhancing" the real world.

[0024] Augmented reality glasses are worn on the face of inspection personnel and have automatic video recording, audio recording, and photo taking functions. They can record audio and video images of the environment around the observation station in a 360° direction as well as the inspection process, thereby capturing multi-angle information of the scene from different observation angles and providing more comprehensive data support to meet the complex data collection needs in atmospheric composition inspection work.

[0025] Specifically, the data-enhancing glasses are equipped with a sensor module and a data processing module.

[0026] The sensor module includes an image sensor and a microphone array. The image sensor is responsible for capturing images and videos of the surrounding scene of the observation station, achieving comprehensive content capture by optimizing the lens structure and the sensor's omnidirectional field of view configuration. The microphone array is used to collect voice commands or feedback from inspection personnel, and can utilize beamforming algorithms to achieve spatial audio localization, improving the targeting and accuracy of voice pickup.

[0027] In addition, the microphone array is directly connected to the microphone of the augmented reality glasses, while the image sensor is connected to the camera module in the lens of the data augmentation glasses. The two transmit audio data and image data to the data processing module respectively through a dedicated data bus.

[0028] The data processing module includes a speech recognition unit and an image recognition unit. It performs preliminary processing on the data collected by the sensor module to obtain pre-processed data, such as pre-noise reduction and filtering of audio. Specifically, the speech recognition unit converts the acquired audio data of the inspection personnel into text commands and calculates the text confidence score corresponding to the text commands for subsequent analysis. The image recognition unit performs target detection and recognition on the captured observation station image data to obtain equipment status data and calculates the image confidence score corresponding to the equipment status data. Deep learning algorithms can be used for accurate recognition. The pre-processed data includes the text commands and their corresponding text confidence scores, and the equipment status data and their corresponding image confidence scores.

[0029] In one specific embodiment, the image recognition unit performs image segmentation and target detection on the instrument's dial to extract target data. For images with poor quality or where no effective information can be extracted, a secondary acquisition can be achieved through a subsequent adaptive scheduling module, which will not be elaborated on here. Meanwhile, the speech recognition unit can convert the speech signal captured by the microphone array into a text format that is easy to parse using a speech signal decoding model. Furthermore, the image recognition unit and the speech recognition unit can be installed collaboratively in the same functional area of ​​the augmented reality glasses to maximize the data transmission efficiency of both, thereby ensuring the integrity and real-time nature of the data collected during the inspection process.

[0030] Furthermore, the image recognition module establishes domain-specific strategy libraries for six types of instruments (pointer type, digital tube, standard screen, indicator light) to identify parameter values ​​in image data.

[0031] Among them, the pointer type uses an improved sub-pixel level key point detection network. Specifically, it uses the YOLOv8 model to segment instances to obtain the region of interest on the dial, then regresses the sub-pixel coordinates of the pointer center line and the scale center, and finally outputs the pressure value through angle and range mapping. The resolution is improved from 100 psi / division of the physical scale to 1 psi, which meets the monitoring requirement of a slight decrease of 10 psi / day under the full scale of 2500 psi. The digital tube and non-standard font formats adopt a structure combining convolutional recurrent neural networks and attention mechanisms, supporting 0-9, AZ, special symbols and manufacturer-customized fonts, with a single character recognition accuracy of ≥99.5%; The standard screen class uses the YOLOv8 model to segment instances to obtain the regions that need to be recognized, and then uses the OCR recognition function to obtain the data values ​​that need to be extracted.

[0032] The indicator light shape is quickly determined to be in three states: "normal," "alarm," and "fault" using HSV color space clustering and morphological opening / closing operations. First, the shape features of the indicator light are optimized through morphological processing. Then, the dominant color is analyzed using HSV color space clustering. Finally, the state is determined based on the mapping relationship between the color and preset rules.

[0033] In one specific embodiment, for a tilted or inverted dashboard in the preprocessed image data, the image recognition unit uses IMU data to calculate the pitch and roll angles in real time, and then calls the perspective transformation and rotation correction network to correct it to a frontal view, with a correction error of less than 0.3°.

[0034] All collected image data is written to the local NVMe cache using timestamps as indexes and asynchronously transferred to the server for storage, ensuring data integrity for subsequent tracing and model retraining.

[0035] It should be noted that the augmented reality glasses incorporate a control chip that works in conjunction with the image sensor to automate data capture tasks. The control chip coordinates the recording and photographing logic and manages the storage location and naming format of the data files. For example, multiple high-sensitivity cameras can be arranged in an array around the perimeter of the glasses frame, while integrating synchronous triggering signal technology to ensure consistency and timeliness of image frames from different angles. The cameras are connected via micro-circuits, transmitting the acquired image information to the control chip for further storage or analysis.

[0036] In addition, augmented reality glasses can also incorporate optical display components and audio players. The optical display component uses a CMOS imaging element, combined with a high-performance analog-to-digital converter, to project data on atmospheric composition collected by the observatory, equipment parameters, inspection routes, and other relevant information onto the augmented reality glasses' screen for real-time viewing by inspection personnel. The sensor component includes a gyroscope and accelerometer to track the inspection personnel's head movements, ensuring that the data display aligns with their line of sight. For example, the optical display component can project images onto a transparent screen using micro-projection technology, allowing inspection personnel to see key inspection information while observing the site environment.

[0037] The anomaly detection module 120, which is communicatively connected to the augmented reality glasses, is used to analyze whether there are any anomalies in the operating status of the observation station based on the preprocessed collected data.

[0038] The anomaly detection module can be directly connected to the output of the data processing module in the data augmentation glasses to analyze the preprocessed collected data output by the data processing module and evaluate the operating status of the device.

[0039] In atmospheric composition observation missions, different observation instrument parameters correspond to specific effective parameter ranges. Therefore, it is necessary to set different threshold standards according to mission requirements to determine the validity of inspection data.

[0040] The anomaly detection module is located in the logical computation center of the overall system and works closely with the data processing module. See details... Figure 2 The anomaly detection module includes an intelligent recognition algorithm library, a data analysis engine, and a data management module.

[0041] The intelligent identification algorithm library stores and identifies conversion algorithms corresponding to different atmospheric composition observation devices. The data analysis engine, acting as a driver for anomaly detection, is connected to the intelligent identification algorithm library and is used to convert the preprocessed collected data based on the conversion algorithm matched to the current device to obtain standardized observation data; for example, data conversion is performed using an infrared spectrum conversion algorithm corresponding to carbon dioxide concentration. The data management module includes a data receiving subunit, a threshold comparison subunit, and a result output subunit, used to receive the standardized observation data and compare it with a preset threshold range to analyze whether there are any anomalies in the operating status.

[0042] The data receiving subunit receives the standardized observation data, i.e., numerical or signal-type information; the threshold comparison subunit compares the received standardized observation data with the corresponding threshold range to obtain the comparison result; and the result output subunit feeds back a signal to the anomaly handling module based on the comparison result. This structural setup ensures a seamless data flow from external input to internal judgment and then to the next step.

[0043] Specifically, the received preprocessed collected data is processed by an intelligent recognition algorithm library, and the corresponding atmospheric components are converted into a calculable data format by a data analysis engine. The data management module then compares and analyzes the data against a preset threshold range.

[0044] For example, in the inspection task for carbon dioxide concentration observation, in the threshold comparison sub-unit, a threshold range can be preset for the inspection results. When the actual detection result of carbon dioxide concentration exceeds the threshold range, it is further determined whether the current value fluctuation is due to natural environmental changes, equipment abnormalities, etc., and abnormal status prompt information is promptly pushed to subsequent links.

[0045] In one specific embodiment, after acquiring the transformed data, the threshold comparison subunit constructs a real-time change curve based on the standardized observation data, converting the discrete real-time data into a continuous time series curve. The sampling interval can be set according to the actual situation, and the continuous time series curve is smoothed to obtain the real-time change curve.

[0046] The real-time change curve is dynamically time-normalized and compared with the historical change curve of the past few days (to eliminate the influence of periodic fluctuation characteristics), and three typical patterns in the real-time change curve are detected: slow drift, periodic fluctuation and instantaneous jump.

[0047] The characteristics of slow drift are that the real-time change curve deviates unidirectionally and at a constant speed over time, and the deviation amplitude accumulates to exceed a certain range over time; periodic fluctuations show fluctuations with a fixed period and fixed amplitude; instantaneous jumps indicate a large deviation in a short period of time and an inability to recover after the jump. Based on the aforementioned characteristics, three typical patterns are detected, such as calculating the offset, extracting the fluctuation period, and calculating the fluctuation amplitude, which will not be explained in detail here.

[0048] Further anomaly assessment is conducted based on three typical patterns, as these patterns are associated with equipment status and may be immediate manifestations of equipment aging, poor contact, or other issues. Therefore, these patterns provide insight into the equipment's condition, thus eliminating potential risks. Initial screening using historical trend curves identifies dissimilar trends, and then a final anomaly assessment is performed based on a pre-set baseline.

[0049] Specifically, based on a preset baseline, it is determined whether the real-time change curve of the detected corresponding typical mode exceeds a threshold range. If it does, it is determined that the operating status of the observation station equipment is abnormal; otherwise, the operating status of the observation station equipment corresponding to the converted data is not abnormal.

[0050] The system sets a preset baseline for each type of observation equipment. This preset baseline serves as a quantitative criterion and must be based on changes in the observation equipment over at least one week. As time progresses, a new preset baseline needs to be recalculated each time data is added, and the preset baseline is adjusted within a preset threshold range.

[0051] Specifically, the threshold range includes a dynamic threshold range and a static threshold range. The static threshold range is the allowable deviation range based on a preset baseline. The dynamic threshold range is the boundary fluctuation range set based on the deviation range of the static threshold range.

[0052] For example, if the average pressure of a certain observation device over a week is taken as 300 psi at 9 am every day, then the allowable deviation range of the device is 280 psi-320 psi (static threshold range). If the average daily change of the device is 10 psi, then the static threshold range can be further widened by 10 psi ± 5 psi (dynamic threshold range).

[0053] It should be noted that fluctuations under normal circumstances should not exceed the threshold range. For example, when the pressure reading on the gauge drops below 300 psi, it indicates that the pressure cylinder needs to be replaced. 300 psi is a fixed threshold, and the daily pressure consumption is approximately 10 psi. However, the daily inspection interval is not necessarily a fixed 24-hour interval, so it is also necessary to dynamically calculate the hourly consumption and eliminate numerical deviations in the gauge readings caused by tilting. If the real-time change curve decreases by 12 psi daily, although there is a difference from the historical change curve, it is still within the threshold range compared to the preset baseline, so it is considered a normal state.

[0054] In one specific embodiment, the steps after entering the anomaly detection module can be summarized as follows: 1. By combining and comparing historical change curves with real-time change curves, the parts with dissimilar trends can be identified.

[0055] 2. Compare the real-time change curve with the preset baseline to determine whether it exceeds the threshold range.

[0056] 3. If the value exceeds the threshold range, it indicates an abnormality.

[0057] In summary, the anomaly detection module outputs the comparison results, which are then sent to the anomaly handling module. The specific faulty device still needs to be further judged and processed in the anomaly handling module.

[0058] An exception handling module 130 is communicatively connected to the exception judgment module and is used to handle situations where the operation status output by the exception judgment module is abnormal.

[0059] The anomaly handling module is used to receive anomaly signals from the anomaly judgment module, generate evidence data, and provide suggested solutions or maintenance steps for potentially problematic devices.

[0060] Specifically, upon receiving a maintenance and repair query command, the anomaly handling module uses an AI-powered big data model to search the configured maintenance and repair database or the internet environment, providing corresponding solutions or maintenance and repair operation steps. The anomaly handling module has a storage unit to store detailed logs of each discovery and handling event. Its internal architecture incorporates a multi-layered logical judgment structure to assess specific risk levels and formulate operation plans with different priorities accordingly.

[0061] When an anomaly is detected, evidence collection functions such as taking photos, recording videos, and recording audio can be activated, and relevant images, audio, and video evidence can be uploaded to the cloud for storage in a timely manner. At the same time, predictive maintenance mechanisms such as artificial intelligence can be used to assess the future failure risk of the equipment.

[0062] The anomaly handling module receives the analysis conclusions from the anomaly judgment module and combines them with a built-in algorithm library to generate a device health status model. This model can be used to estimate the performance degradation curves of key components and provides a quantitative reference for future failure rates. Furthermore, to achieve more accurate risk assessment, the anomaly handling module can further integrate key information from image and audio data from sensor modules.

[0063] For example, a machine learning model is embedded in the anomaly handling module. The machine learning model is trained using the preprocessed historical anomaly state dataset until the loss function in the machine learning model converges, resulting in an equipment health status model. This equipment health status model is used as input to the preprocessed collected data to output the observation station's predicted fault risk. The input of this model can be directly connected to the anomaly judgment module to ensure that all necessary parameters are received, while the output is directly linked to the subsequent report generation module.

[0064] In summary, the complete process of the exception handling module can be summarized as follows: 1) Automatically save the preprocessed data collected at the time of the anomaly as evidence, and generate images and / or videos and / or audio recordings for evidence preservation; 2) Input the preprocessed collected data corresponding to the abnormal moment into the equipment health status model, and output the observation station's predicted fault risk and potential faulty components; use the equipment health status model to locate potential faulty components (such as pressure reducing valve diaphragm aging, pipeline micro-leakage, sensor zero-point drift, etc.), and give the probability of each potential faulty component failing. 3) Retrieve maintenance guidance materials corresponding to the potentially faulty components from the preset maintenance database and provide them to the inspection personnel through the augmented reality glasses. This can be done by directly overlaying the guidance materials onto the glasses; alternatively, a structured work order corresponding to the guidance materials can be generated and uploaded to the computerized maintenance management system. The structured work order includes maintenance suggestions, spare part numbers, and estimated man-hours.

[0065] If remote support from engineers is needed at the observation station, a remote assistance module can be deployed to activate a two-way 1080p AR video with a single click. Remote experts can annotate faulty components in real time and provide guidance on operation via voice and / or gestures, while simultaneously storing evidence.

[0066] The anomaly handling module is located within the core control area of ​​the system and connects to the anomaly detection module via an interface protocol, ensuring real-time data exchange. The anomaly handling module also reserves an interface to access third-party maintenance databases, thereby expanding the data coverage and accuracy adaptability of its predictive maintenance algorithms. This constitutes a highly integrated and flexibly expandable technical system.

[0067] An adaptive scheduling module 140 is communicatively connected to the augmented reality glasses, the anomaly judgment module, and the anomaly handling module, respectively. It is used to determine the parameter weights of the three components based on their outputs and schedule their execution order according to the parameter weights.

[0068] First, it should be noted that the adaptive scheduling module is an independent module used to decide which specific module function should be executed in the current process. For example, when identifying the reading of a pressure gauge through image acquisition, the adaptive scheduling module will first activate the sensor module to take a picture, and then call the image recognition unit to judge the quality of the picture. If the quality does not meet the requirements, the adaptive scheduling module will analyze the lighting, distance, and other factors of the picture, and decide whether to use functions such as flash or zoom to ensure that the gauge data can be read correctly.

[0069] The adaptive scheduling module addresses challenges such as the diverse types of images, variable environments, insufficient instrument reading accuracy, and difficulties in fault location during atmospheric composition observations through a unified scheduling algorithm. This module employs a decoupled architecture between the terminal and server to ensure conflict-free scheduling when multiple terminals are operating concurrently.

[0070] The server sets up an independent scheduling session for each terminal to ensure that scheduling commands from different terminals do not conflict. The terminal and the server maintain a long connection through a gRPC streaming interface, supporting hundreds of concurrent terminals. When there is network jitter, it automatically degrades to a local lightweight model to ensure business continuity.

[0071] Specifically, for the server side, its internal core is a finite state machine, which realizes global decision-making and module scheduling, and includes the following types of inputs: ① Current task type (including inspection, maintenance, tour guiding, etc.); different tasks correspond to different module call priorities. For example, maintenance tasks by default prioritize calling the exception judgment module and the exception handling module. The task type can be set directly by the user.

[0072] ② Image data; In this disclosure, unprocessed image data acquired through augmented reality glasses is used to reflect the environmental conditions at the scene. For example, the sound of an air compressor collected through augmented reality glasses (used to analyze the operating status of the air compressor, such as: excessive noise may be due to loose parts).

[0073] ③ User voice commands in audio data; directly triggering corresponding modules based on user voice commands, such as calling the remote assistance module to communicate online with remote experts. The acquisition of user commands can be obtained through training a machine learning model, which will not be elaborated here.

[0074] ④ Output status of other modules; update the latest status based on the real-time status of other modules so that various situations can be handled in a timely manner.

[0075] The adaptive scheduling module disclosed herein integrates with the internal characteristics and exposed interfaces of other modules. The specific scheduling method includes the following steps: 1) The finite state machine obtains the output of each other module and determines the current working state and the scope of module calls based on the state matching matrix; The task type determines the core objective, image data reflects the quality of data acquisition, and user voice commands reflect user needs. Based on the real-time progress of other modules, the four work together to enable the finite state machine to accurately determine the current working state.

[0076] Image data and user voice commands need to be standardized and parsed to allow the current working state to be locked through a preset state matrix. The state matching matrix is ​​set based on a two-dimensional decision table within a finite state machine, establishing matching relationships between different inputs of the finite state machine and different working states.

[0077] For example, image data is parsed to map the environment into several labels: no interference, low interference, and high interference. User voice commands are parsed and updated with labels such as immediate operation, process intervention, and feedback confirmation. Simultaneously, the relevant modules are identified, such as invoking expert commands to associate with remote assistance modules. Finally, based on the real-time status of each module, the decision-making basis of the finite state machine is supplemented. For example, the real-time status of the augmented reality glasses is obtained to determine whether the collected data meets requirements or whether the augmented reality glasses have continuous data collection capabilities. This allows the scope of module invocation in the current working state to be determined. If multiple modules are available, further matching items need to be narrowed down.

[0078] 2) Determine the parameter weights for each module within the module call scope. The parameter weights are mainly determined by the results of the finite state machine and the user's voice commands.

[0079] Based on the module call scope determined in the previous step, if there are multiple options, it is necessary to further calculate the parameter weights of each module.

[0080] First, set the user's voice commands to the highest priority. If the user command is an immediate action, increase the parameter weight of the associated module to meet the user's real-time operation needs. Different increase amounts can be set for different command types; for example, set the increase amount for immediate actions to 0.3, for process interventions to 0.2, and for confirmation feedback to 0.1. Finally, add all the increase amounts together to obtain the parameter weight.

[0081] 3) Place the user instructions and the decision results from the previous steps into a first-in-first-out queue, and schedule other modules according to the queue order.

[0082] The terminal completes data acquisition, namely, acquiring the current task type, the image and audio data acquired by the augmented reality glasses, and the output status of the augmented reality glasses, the anomaly detection module, and the anomaly handling module, and inputs them into the finite state machine; the finite state machine determines the current working state and module call range of the three based on a preset state matching matrix, so as to initially determine the required module call range.

[0083] For the initially determined module call range, calculate the parameter weight of each module, sort the modules according to the parameter weight to generate a module scheduling priority list, determine the module to be called next, and input it into the first-in-first-out queue.

[0084] Each subsequent module is identified sequentially and transferred to a first-in-first-out queue to ensure the orderly calling of each module and avoid resource contention.

[0085] The execution order of the three modules is scheduled according to the module scheduling priority list. Modules without dependencies can be executed in parallel using multiple threads to improve efficiency, while modules with dependencies need to be triggered sequentially according to the queue order to ensure the correctness of data flow.

[0086] After each module is executed, the results can be synchronized to the cloud or displayed visually for subsequent tracking and statistics.

[0087] In a specific inspection embodiment, the server performs one of the following operations based on the preprocessed data collected from the augmented reality glasses: 1) Conduct a complete exception judgment and exception handling process.

[0088] 2) Read the preprocessed data and verify its compliance in the image recognition unit, such as determining whether the collected pointer table data value exceeds the maximum scale value of the pointer table, and decide whether to collect the data again.

[0089] 3) Read the acquisition parameters. If the adaptive scheduling module cannot obtain usable information under intelligent decision-making, such as overexposed image data or lack of device information, it will integrate the required information and send it to the user's front-end visualization interface to guide the user to make targeted acquisition improvements. After the user makes improvements, the adaptive scheduling module will reschedule the augmented reality glasses to acquire data.

[0090] Specifically, the image recognition unit of augmented reality glasses includes the following steps: a) The image recognition unit analyzes whether the preprocessed collected data meets quality requirements and whether usable information can be obtained. If recognition fails, the analysis results can be directly sent to relevant modules, such as the user visualization interface and the adaptive scheduling module. The adaptive scheduling module can then make judgments and calls based on the analysis results.

[0091] b. Transform the target for identification to facilitate identification, such as transforming it for the four types of instrument forms.

[0092] c. The YOLO model can be used to classify domain-specific identification targets, and a method for the identification target can be selected based on a domain-specific policy library.

[0093] d. The preprocessed acquisition data and confidence level are input into the adaptive scheduling module, which makes decisions based on the preprocessed acquisition data and confidence level.

[0094] This adaptive scheduling module dynamically adjusts the scheduling order according to different scenarios during the inspection process, as detailed below: First, it is necessary to obtain the output of other modules, such as the preprocessed data and confidence level of the augmented reality glasses, the judgment results of the anomaly judgment module and the detection results of three typical patterns; the output results of the device health status model and the work order generation status in the anomaly handling module, etc.

[0095] (1) Routine inspection (no abnormality trigger) 1. The augmented reality glasses collect preprocessed data and confidence scores, and upload them to the server in JSON Schema format. The adaptive scheduling module determines whether the confidence scores meet the preset conditions. If they do, proceed to the next step.

[0096] 2. The anomaly detection module analyzes the preprocessed collected data to determine whether there is an anomaly in the device. If no anomaly is found, the adaptive scheduling module schedules the augmented reality glasses to perform the next round of data collection.

[0097] (2) The quality of the collected data is substandard (low confidence level / data does not meet requirements) During routine inspections, if the adaptive scheduling module determines that the confidence level does not meet the preset conditions, such as a confidence level of 0.8 which is less than the preset condition of 0.92, the specific cause is analyzed through a finite state machine. If it is due to insufficient lighting, it is determined that the poor acquisition quality is caused by environmental interference. If the shooting angle is tilted, it is determined that the acquisition pose is improper.

[0098] The adaptive scheduling module schedules the augmented reality glasses to re-acquire data and issues corresponding adjustment commands, such as pose adjustment commands and supplementary lighting commands. After acquisition, the adaptive scheduling module verifies the data again. If it still cannot be corrected to meet the recognition requirements, an AR guide arrow can be used in the user's field of vision to indicate the best shooting pose to ensure image quality; this continues until the confidence level is higher than the set conditions or the maximum number of retries is reached, such as 5 times.

[0099] (3) Abnormal inspection status During routine inspections, if the anomaly detection module determines that the equipment is abnormal based on the preprocessed collected data, the adaptive scheduling module will switch the system status from inspection analysis to anomaly and increase the parameter weight of the anomaly handling module, such as increasing the weight from 0.3 to 0.9.

[0100] If remote support from engineers is needed at the observation station, a remote assistance module can be deployed to activate a two-way 1080p AR video with a single click. Remote experts can annotate faulty components in real time and provide guidance on operation via voice and / or gestures, while simultaneously storing evidence.

[0101] In summary, the output of the finite state machine and the parameter weights enter the first-in-first-out queue together. The adaptive scheduling module drives the image recognition, anomaly analysis and predictive maintenance modules concurrently or serially according to the queue order, thereby completing the closed loop of "acquisition-result-decision-feedback" in milliseconds.

[0102] In addition, this disclosure includes the following: The report generation module generates standardized inspection reports based on the data output from the data processing and anomaly detection modules. It supports standard or custom template formats to meet diverse needs. The module connects to the data processing module to obtain inspection record information and connects to the anomaly detection module to determine the validity of inspection parameters. It then processes this data to create report content that conforms to specific rules, satisfying the diverse needs of different users.

[0103] Specifically, the report generation module includes a data organization unit, which combines user settings and predefined report templates to specify the required data fields and their presentation formats through parameterized configuration. For example, when a user needs to generate a report suitable for environmental protection departments, they can select specific report parameter indicator fields through the report generation module to generate a report file with user-specified content.

[0104] The file upload module is used to upload standardized inspection reports, image data, and audio data to the server or a designated FTP server in real time or on a scheduled basis. The file upload module connects to the report generation module to obtain standardized inspection reports and to the sensor module to obtain image and audio data. The file upload module interacts with the server to ensure that generated reports, audio and video files are automatically uploaded and backed up within a preset timeframe.

[0105] Specifically, a connection to an FTP server is established by combining IP address, path, username, and access password, thereby enabling real-time or scheduled transmission of standardized inspection reports, audio and video files, etc.

[0106] The remote assistance module is used to establish an audio and / or video communication connection between the augmented reality glasses and a remote expert terminal, and to invite several experts to assist in real time, so as to realize audio and video interaction and provide remote guidance and assistance for on-site inspection, maintenance and repair.

[0107] The remote assistance module communicates with the augmented reality glasses to achieve bidirectional signal transmission. This module allows external technical personnel to participate in the problem-solving process via the public internet or a private enterprise network. It incorporates a video stream push encoding engine, an audio interaction line controller, and other functional modules to meet high-quality audio-visual synchronization communication standards.

[0108] The Chinese-foreign language translation module is used to translate the audio data or text materials in real time.

[0109] The Chinese-foreign language translation module in the atmospheric composition observation and inspection system provides language processing support other than Chinese. This module communicates with augmented reality glasses to transmit collected text information to a server or local terminal. After understanding and processing the multilingual data, it returns the translation results and simultaneously provides voice broadcasts. This helps staff better understand foreign language equipment manuals, technical documents, or other reference materials, ensuring operational accuracy during on-site inspections.

[0110] A Chinese-foreign language translation module can be obtained by embedding a natural language processing engine into a remote assistance module. Specifically, this module integrates a neural machine translation model and a speech synthesis unit, which are connected and controlled through a software framework. When receiving language text transmitted from augmented reality glasses, the neural machine translation model first preprocesses and performs language recognition on the language text content, then generates the target language translation, and finally the speech synthesis unit converts the translation into speech and returns it to the inspection personnel for real-time broadcast.

[0111] The guided tour service module is used to provide customized guided tour services through the augmented reality glasses based on preset narration data.

[0112] This module can intelligently retrieve information from a pre-entered knowledge base and adjust its content and format according to the needs of different scenarios. For example, during visits to atmospheric composition observation equipment, the guide service module can retrieve corresponding explanatory information based on the type of monitoring equipment, historical data, and specific environmental factors in the current area, and present it to the user in real time through augmented reality glasses. Furthermore, the guide service module also supports user interaction, allowing inspection personnel to ask questions and receive immediate feedback.

[0113] The augmented reality glasses integrate a processor unit for running the navigation service module. This processor unit can access expert opinions via a network interface and dynamically match them with currently collected image and audio data. The navigation service module also relies on a cloud computing platform for big data support and continuous learning to improve the accuracy and usability of the narration. The entire system requires no additional modifications to the glasses' appearance; its functional expansion primarily relies on software algorithms and cloud collaboration, resulting in a more compact and efficiently integrated component composition.

[0114] In addition, the augmented reality glasses feature zoom functionality. Users can view key areas of the observed object in high definition, thus meeting the needs for detailed analysis in atmospheric composition observation scenarios. This function is achieved through an optical module and image processing unit integrated within the augmented reality glasses.

[0115] Specifically, the optical module is responsible for receiving real-time image data and transmitting it to the image recognition unit. The image recognition unit then performs digital scaling and image quality optimization before re-displaying the data to the inspection personnel. To ensure stability, the augmented reality glasses' mounting structure adopts a layered frame design, with the optical module located within the lenses of the glasses and closely fitted to the augmented reality display area.

[0116] In one specific embodiment, the optical module of the augmented reality glasses includes a lens group, an image sensor, and a processor assembly. These components are connected via a high-speed communication interface to ensure low latency in information transmission. Based on the aforementioned design, the augmented reality glasses can quickly recognize the user's zoom commands and display the corresponding magnification to the user, enabling the user to efficiently complete detailed observation tasks of complex environments during inspections.

[0117] This disclosure provides an artificial intelligence-based method for atmospheric composition observation and inspection, see [link to relevant documentation]. Figure 3 This includes the following steps: Image data from the observation station and audio data from the inspection personnel are collected using augmented reality glasses. The image data and audio data are then preprocessed to obtain preprocessed collected data. The anomaly detection module analyzes the preprocessed collected data to determine if there are any anomalies in the operating status of the observation station; the anomaly handling module processes any cases where the anomaly detection module outputs an abnormal operating status. The adaptive scheduling module determines the parameter weights of the augmented reality glasses, the anomaly detection module, and the anomaly handling module based on their outputs, and schedules the execution order of the three modules according to these parameter weights.

[0118] Other details can be found in the previous system section and will not be repeated here.

[0119] It is important to note that the weather environment at atmospheric composition observation stations has a dual nature: it is both the object of observation (requiring the collection of data such as temperature, precipitation, wind speed, and PM2.5 for atmospheric analysis) and a source of interference during inspections (strong light, heavy rain, strong winds, and low temperatures can affect the accuracy of augmented reality glasses, recognition effects, and the safety of personnel operation). Therefore, a balance needs to be struck between observation requirements and interference response.

[0120] Specifically, it also includes the following steps: Step 1: Set up an environmental perception module inside the augmented reality glasses to collect environmental data in real time.

[0121] For example, the environmental sensing module includes temperature and humidity sensors, wind speed sensors, light sensors, etc., and the corresponding environmental data includes temperature, humidity, wind speed, light, etc.

[0122] Step 2: Synchronize the environmental data and the preprocessed acquisition data in time, and match the spatial location of the devices in the environmental data and the image data to obtain a multimodal data packet.

[0123] Time synchronization ensures that all data streams have a unified timestamp, eliminating misjudgments caused by asynchrony; and spatial coordinate transformation is used to match and associate spatial locations to establish accurate environmental correlations for subsequent analysis, effectively distinguishing between equipment malfunctions and environmental interference.

[0124] Step 3: Identify and mark the corresponding interference modes based on the multimodal data packets.

[0125] Specifically, a rule base for interference patterns can be pre-set to perform corresponding logical judgments based on multi-mode data, thereby qualitatively identifying interference patterns and achieving the structuring and labeling of complex weather conditions, providing a clear decision-making basis for subsequent scheduling. Furthermore, the interference pattern rule base also includes preset scheduling strategies for corresponding interference patterns to ensure timely response.

[0126] For example, the real-time temperature collected is -15℃, while the temperature corresponding to low-temperature interference must be less than -10℃. At this time, the interference mode is the low-temperature interference interference mode in the low-temperature interference interference mode rule library, and the real-time temperature is marked accordingly.

[0127] Step 4: The data compensation engine set in the anomaly judgment module calls the corresponding compensation model according to the marked interference mode, and outputs the corrected observation value based on the input marked multimodal data packet.

[0128] The data compensation engine is a data-driven process based on a large amount of calibration data, and is constructed through the following steps: Step 41: Determine the observation parameters and corresponding interference modes; A combination of statistical analysis algorithms and domain experts can be used to screen out key disturbance patterns that have a decisive impact on the accuracy of observation parameters from a large amount of environmental data.

[0129] Step 42: Input the observation parameters and corresponding interference modes into the corresponding observation instruments, adjust individual observation parameters, record the readings of the observation instruments, and obtain the calibration dataset.

[0130] By controlling environmental conditions using the controlled variable method, data on the relationship between weather and disturbances can be obtained, i.e., a calibration dataset.

[0131] Step 43: Select a suitable model to describe the calibration dataset. This model is the compensation model, which is used to output corrected observations.

[0132] The distribution characteristics of the scatter plot can be used to select an appropriate model to describe the relationship between observed parameters and disturbance patterns. Commonly used models include linear models, polynomial models, and exponential models.

[0133] It suppresses or eliminates quantifiable systemic weather interference at the data level, fundamentally reducing the risk of false alarms and warnings caused by environmental interference.

[0134] Step 5: In the anomaly detection module, the environmental data is input into the dynamic threshold model to calculate the dynamic threshold range under the corresponding interference mode. It is then determined whether the corrected observation value is within the dynamic threshold range. If not, proceed to the next step; if so, it indicates that there is no anomaly, and proceed to Step 1 to continue collecting environmental data in a loop.

[0135] For example, the dynamic threshold model can be trained on a neural network using historical environmental data and observation parameters collected by synchronous observation instruments. The dynamic threshold model outputs a dynamic threshold range, which includes an upper limit and a lower limit.

[0136] Step 6: The adaptive scheduling module schedules the execution order of the augmented reality glasses, the anomaly detection module, and the anomaly handling module according to the interference mode. Other modules are scheduled according to the interference mode to adapt to or counteract the current weather conditions.

[0137] The system can match preset scheduling strategies corresponding to interference patterns in the interference pattern rule base. For example, under high temperature conditions, the adaptive scheduling module can reduce the equipment risks caused by high temperatures by initiating cooling and reducing load.

[0138] Based on the above technical solutions, this disclosure sets up an adaptive scheduling module to call other modules, enabling the other modules to complete the following tasks: the augmented reality glasses acquire image and audio data during inspection through the sensor module, use the anomaly judgment module to perform risk judgment and alarm, and the anomaly handling module to handle equipment anomalies; at the same time, it can also provide video playback and inspection of the entire inspection process.

[0139] This disclosure addresses the issues of standardized processes, methods, requirements, and standards in atmospheric composition observation inspections, maintenance, and repairs based on artificial intelligence and augmented reality technologies. It improves the accuracy of speech recognition and image recognition technologies for input from inspection personnel using dialects or non-standard languages ​​through machine learning, ensuring the precision of inspection records and anomaly assessments. It provides timely guidance and assistance to inspection personnel during maintenance and repairs, not only improving efficiency but also laying a solid foundation for further standardizing atmospheric composition observation inspection work and improving observation quality.

[0140] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this disclosure is not limited to the described order of actions, because according to this disclosure, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this disclosure.

[0141] The acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0142] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0143] Figure 4A schematic block diagram of an electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0144] Electronic device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in ROM 402 or a computer program loaded into RAM 403 from storage unit 408. RAM 403 can also store various programs and data required for the operation of electronic device 400. The computing unit 401, ROM 402, and RAM 403 are interconnected via bus 404. I / O interface 405 is also connected to bus 404.

[0145] Multiple components in electronic device 400 are connected to I / O interface 405, including: input unit 406, such as keyboard, mouse, etc.; output unit 407, such as various types of displays, speakers, etc.; storage unit 408, such as disk, optical disk, etc.; and communication unit 409, such as network card, modem, wireless transceiver, etc. Communication unit 409 allows electronic device 400 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0146] The computing unit 401 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 401 performs the various methods and processes described above, such as the AI-based atmospheric composition observation and inspection method. For example, in some embodiments, the AI-based atmospheric composition observation and inspection method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the electronic device 400 via ROM 402 and / or communication unit 409. When the computer program is loaded into RAM 403 and executed by the computing unit 401, one or more steps of the AI-based atmospheric composition observation and inspection method described above can be performed. Alternatively, in other embodiments, computing unit 401 may be configured to perform the AI-based atmospheric composition observation and inspection method by any other suitable means (e.g., by means of firmware).

[0147] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0148] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0149] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0150] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including voice input, speech input, or tactile input).

[0151] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0152] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0153] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0154] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. An artificial intelligence-based atmospheric composition observation and inspection system, characterized in that, include: Augmented reality glasses, including a sensor module and a data processing module, are used to collect image data from the observation station and audio data from the inspection personnel in real time, and to preprocess the image data and audio data to obtain preprocessed collected data; An anomaly detection module, which is communicatively connected to the augmented reality glasses, is used to analyze whether there are any anomalies in the operating status of the observation station based on the preprocessed collected data; An exception handling module, which is communicatively connected to the exception judgment module, is used to handle situations where there are abnormalities in the operating status output by the exception judgment module; An adaptive scheduling module, which is communicatively connected to the augmented reality glasses, the anomaly detection module, and the anomaly handling module, is used to determine the parameter weights of the three components based on their respective outputs, and to schedule the execution order of the three components according to the parameter weights.

2. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 1, characterized in that, The data processing module includes a speech recognition unit and an image recognition unit; The speech recognition unit is used to convert the acquired audio data of the inspection personnel into text commands and calculate the text confidence level corresponding to the text commands. The image recognition unit is used to perform target detection and recognition on the captured observation station image data to obtain equipment status data, and to calculate the image confidence level corresponding to the equipment status data. The preprocessed acquired data includes the text commands and their corresponding text confidence scores, device status data and their corresponding image confidence scores.

3. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 2, characterized in that, The anomaly detection module includes an intelligent recognition algorithm library, a data analysis engine, and a data management module; wherein... The intelligent identification algorithm library is used to store and identify the conversion algorithms corresponding to different atmospheric composition observation equipment; The data analysis engine, which is connected to the intelligent recognition algorithm library, is used to transform the preprocessed collected data based on the transformation algorithm matched with the current device to obtain standardized observation data. The data management module, which is connected to the data analysis engine, includes a data receiving subunit, a threshold comparison subunit, and a result output subunit. It is used to receive the standardized observation data and compare it with a preset threshold range to analyze whether there are any abnormalities in the operating status.

4. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 3, characterized in that, The data management module compares the standardized observation data with a preset threshold range to analyze whether there are any abnormalities in the operating status, specifically including the following steps: A real-time change curve is constructed based on the standardized observation data; The real-time change curve is compared with the historical change curve by dynamic time warping to detect three typical patterns in the real-time change curve, which include slow drift, periodic fluctuation and instantaneous jump. Based on a preset baseline, it is determined whether the real-time change curve of the detected corresponding typical mode exceeds a threshold range. If it does, it is determined that the operating status of the observation station equipment is abnormal; wherein, The threshold range includes a dynamic threshold range and a static threshold range. The static threshold range is the allowable deviation range based on a preset baseline, and the dynamic threshold range is the boundary fluctuation range based on the deviation range of the static threshold range.

5. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 1, characterized in that, The exception handling module is used to perform the following steps: The preprocessed data collected at the time of the anomaly is automatically saved as evidence. The preprocessed collected data corresponding to the abnormal moment is input into the equipment health status model, and the observation station outputs the predicted fault risk and potential faulty components. The system retrieves maintenance guidance information corresponding to the potentially faulty component from a pre-set maintenance database and provides it to the inspection personnel via the augmented reality glasses; wherein, The device health status model is constructed through the following steps: Acquire and preprocess the historical abnormal state dataset of the observation station; The machine learning model is trained using the preprocessed historical abnormal state dataset until the loss function in the machine learning model converges, thus obtaining the equipment health status model.

6. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 5, characterized in that, The adaptive scheduling module is used to perform the following steps: The current task type, the image and audio data collected by the augmented reality glasses, and the output of the augmented reality glasses, the anomaly detection module, and the anomaly handling module are obtained and input into the finite state machine. The finite state machine determines the current working state and module call range of the three components based on a preset state matching matrix. The calculation module calls the augmented reality glasses, the anomaly detection module, and the anomaly handling module, and generates a module scheduling priority list based on the parameter weights. The execution order of the three is scheduled according to the module scheduling priority list.

7. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 6, characterized in that, The state matching matrix is ​​set based on the two-dimensional decision table in the finite state machine, and the different inputs of the finite state machine are matched with different working states.

8. The artificial intelligence-based atmospheric composition observation and inspection system according to claim 1, characterized in that, Also includes: The report generation module is used to generate standardized inspection reports based on the data output by the data processing module and the anomaly detection module. The file upload module is used to upload the standardized inspection reports, image data and audio data to the server in real time or at regular intervals. The remote assistance module is used to establish an audio and / or video communication connection between the augmented reality glasses and the remote expert terminal; A Chinese-foreign language translation module is used for real-time translation of the audio data or text materials; The guided tour service module is used to provide customized guided tour services through the augmented reality glasses based on preset narration data.

9. An artificial intelligence-based method for atmospheric composition observation and inspection, applied to the system described in any one of claims 1 to 8, characterized in that, The method includes the following steps: Image data from the observation station and audio data from the inspection personnel are collected using augmented reality glasses. The image data and audio data are then preprocessed to obtain preprocessed collected data. The anomaly detection module analyzes the preprocessed collected data to determine if there are any anomalies in the operating status of the observation station; the anomaly handling module processes any cases where the anomaly detection module outputs an abnormal operating status. The adaptive scheduling module determines the parameter weights of the augmented reality glasses, the anomaly detection module, and the anomaly handling module based on their outputs, and schedules the execution order of the three modules according to these parameter weights.

10. The atmospheric composition observation and inspection method based on artificial intelligence according to claim 9, characterized in that, Also includes: An environmental perception module is installed inside the augmented reality glasses to collect environmental data in real time; The environmental data and the preprocessed acquired data are synchronized in time, and the spatial location of the device in the environmental data and the image data is matched to obtain a multimodal data packet; The corresponding interference patterns are identified and marked based on the multimodal data packets; The data compensation engine set in the anomaly judgment module calls the corresponding compensation model according to the marked interference mode, and outputs the corrected observation value based on the input marked multimodal data packet. Within the anomaly detection module, the environmental data is input into the dynamic threshold model to calculate the dynamic threshold range under the corresponding interference mode. It is then determined whether the corrected observation value is within the dynamic threshold range. If not, proceed to the next step. The adaptive scheduling module performs scheduling according to the preset scheduling strategy corresponding to the interference mode.

Images