AI-based smart city monitoring system and method

Abstract

The present invention relates to the technical field of urban monitoring systems, and in particular to an AI-based smart city monitoring system and method. First, the urban monitoring data stream is acquired and analyzed to identify target information elements and behavior patterns in the urban monitoring data stream; then, data optimization and standardization processing are performed on the identified target information elements and behavior patterns; distributed processing is performed on the optimized and standardized data to obtain event summaries and statistical analysis results; finally, the data is securely encrypted and privacy-protected to obtain encrypted and desensitized data, and the data is subjected to advanced data fusion analysis with the data of the associated data source to obtain comprehensive analysis results, which are then calibrated. After adaptive identification and prediction of abnormal event monitoring are performed on the calibrated data, reports and warnings of abnormal events are obtained. This solves the technical problems of the existing technology in the inaccuracy of data processing in smart city monitoring and the inaccuracy and reliability of smart city monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of urban monitoring systems, and in particular to an AI-based smart city monitoring system and method. Background Art

[0002] With the acceleration of urbanization and advancements in technology, traditional urban management models are no longer able to meet the efficient and sustainable development needs of modern cities. The concept of smart cities has emerged, aiming to improve the intelligence and automation of urban management and enhance the efficiency and quality of urban services through advanced information technology and data analysis methods. Currently, many cities have deployed extensive surveillance systems, including video surveillance, traffic monitoring, and environmental monitoring. However, these systems often suffer from information silos, inefficient data processing, and insufficient intelligence. This has led to the accumulation of large amounts of data, but this data has not been fully utilized to improve urban management. The rapid development of artificial intelligence (AI) technology offers a potential solution to these problems. AI technologies, particularly machine learning and deep learning, have demonstrated tremendous potential in areas such as image recognition, data analysis, and predictive modeling.

[0003] There are many methods for smart city monitoring. Yuan Junjiang proposed a Chinese invention patent, "A Smart City Monitoring System Based on Image Recognition", with application number "CN202110336357.7", which mainly includes: a central processing unit, an image collection module, a facial recognition module, a color recognition module, an alarm module and a touch screen display; the facial recognition module internally includes a feature point recognition unit, a displacement output unit and a distance calculation unit. The feature point recognition unit is used to detect facial images and locate key facial feature points and complete the extraction of facial feature points through a recognition algorithm. The displacement output unit is used to output the displacement and azimuth of the single-lens imaging element when it receives facial recognition information at different positions. The invention has a reasonable design and ingenious conception. It does not require constant viewing of the monitoring, greatly reducing the search time. It is highly convenient and suitable for promotion.

[0004] However, the above technology has at least the following technical problems: the data processing in smart city monitoring is not accurate enough and the monitoring of smart cities is not precise and reliable enough. Summary of the Invention

[0005] The embodiments of the present application provide an AI-based smart city monitoring system and method to solve the technical problems of the existing technology in smart city monitoring, such as inaccurate data processing and insufficient precision and reliability in smart city monitoring, thereby achieving the technical effects of high-quality data processing and high-accuracy and reliable monitoring of smart cities.

[0006] This application provides an AI-based smart city monitoring system and method, which specifically includes the following technical solutions:

[0007] An AI-based smart city monitoring system includes the following parts:

[0008] Intelligent data analysis module, data optimization and standardization module, distributed data processing module, security encryption and privacy protection module, advanced data fusion analysis module, comprehensive information calibration module, and adaptive abnormal event detection module;

[0009] The intelligent data analysis module uses advanced deep learning algorithms to intelligently analyze the content of real-time urban monitoring data streams to obtain data stream analysis results, which include identified objects and behavior patterns, and transmits the data stream analysis results to the data optimization and standardization module;

[0010] The data optimization and standardization module preprocesses the data flow analysis results, including noise reduction, formatting and optimization, and transmits the preprocessed data to the distributed data processing module;

[0011] The distributed data processing module processes and fuses the pre-processed data using distributed computing resources, performs data analysis, obtains distributed processed data, and transmits the distributed processed data to the security encryption and privacy protection module;

[0012] The security encryption and privacy protection module encrypts and protects the distributed processed data to obtain encrypted and desensitized data, and provides the encrypted and desensitized data to the advanced data fusion analysis module;

[0013] The advanced data fusion analysis module decrypts the encrypted and desensitized data and performs advanced fusion and analysis on the data from the associated data source to obtain a comprehensive analysis result, which is then transmitted to the comprehensive information calibration module;

[0014] The comprehensive information calibration module calibrates the output data of the advanced data fusion analysis module and transmits the calibrated data to the adaptive abnormal event detection module;

[0015] The adaptive abnormal event detection module uses an adaptive algorithm to identify and predict potential abnormal events based on the output data of the comprehensive information calibration module, and generates abnormal event reports and early warning notifications, providing decision support and emergency response for urban management based on the abnormal event reports and early warning notifications.

[0016] An AI-based smart city monitoring method includes the following steps:

[0017] S1. Obtain the city monitoring data stream, analyze the content of the city monitoring data stream, and identify the target information elements and behavior patterns in the city monitoring data stream;

[0018] S2. Optimize and standardize the data of the identified target information elements and behavior patterns;

[0019] S3. Perform distributed processing on the optimized and standardized data to obtain event summaries and statistical analysis results;

[0020] S4. Securely encrypt and protect the data to obtain encrypted and desensitized data;

[0021] S5. Perform advanced data fusion analysis on the encrypted and desensitized data with the data from the associated data source to obtain comprehensive analysis results, calibrate the comprehensive analysis results, and provide data for abnormal event monitoring. After adaptively identifying and predicting abnormal event monitoring on the calibrated data, obtain reports and warnings of abnormal events.

[0022] Preferably, the S1 specifically includes:

[0023] Data is collected on the urban monitoring data stream, and the monitoring network of the urban monitoring system is connected to obtain real-time urban monitoring data streams; the collected monitoring data streams are preprocessed and feature extracted from the preprocessed monitoring data streams. Target recognition and classification are performed based on the extracted features combined with at least two object detection deep learning models. The major categories, including vehicles and people, are first identified, and then the minor categories, including car types and pedestrian characteristics, are subdivided. Advanced trajectory analysis technology is used to analyze the recognition results, and the analysis results are integrated using multi-dimensional data fusion technology to obtain the final recognition and analysis results, which include target information elements and behavior patterns in the monitoring data stream.

[0024] Preferably, the S2 specifically includes:

[0025] The received target information elements and behavior pattern data are cleaned; the cleaned data are subjected to noise reduction and optimization processing; the noise reduction and optimization data are formatted, during which text information, including license plate numbers, are unified to obtain a standard format; and standardized data are further obtained.

[0026] Preferably, the S3 specifically includes:

[0027] A machine learning classification model is applied to the optimized and standardized data to automatically identify data categories. When the data flows into the distributed data processing module, it is preliminarily classified using the pre-trained machine learning classification model. The resource scheduler is used to dynamically allocate computing resources, and tasks are intelligently assigned according to data processing requirements and server load conditions to obtain data after distributed resource configuration. The data is further fused, and then a time series analysis is performed on the fused comprehensive data set. The autoregressive model and the sliding average model are used to identify trends and patterns in the time series data, and future trends are predicted based on the results of the time series analysis. Finally, a text generation algorithm is applied based on the results of the time series analysis and trend prediction to automatically generate an event summary. At the same time, data mining technology is used to perform statistical analysis to obtain a statistical analysis report.

[0028] Preferably, in said S3, it further includes:

[0029] In the data fusion stage, a fusion enhancement algorithm and an efficient collaborative algorithm are introduced; the fusion enhancement algorithm adjusts and unifies the data formats and scales of different data sources through an adaptive adjustment mechanism; the efficient collaborative algorithm adopts advanced data structures and parallel computing technology.

[0030] Preferably, the S4 specifically includes:

[0031] Set classification and evaluation criteria to subdivide input data into sensitive data and non-sensitive data; select appropriate encryption algorithms based on the nature of the data and protection requirements, perform encryption operations on sensitive data, and use security protocols to protect data during the data transmission stage; perform data desensitization processing on the data, which includes using masks, disguises, hiding and replacing sensitive data, so that even if the data is leaked, personal information will not be exposed; implement access control and permission management.

[0032] Preferably, the S5 specifically includes:

[0033] When conducting advanced data fusion analysis, the input encrypted monitoring data is first decrypted and then integrated with the associated data source data to form a data set. Data cleaning tools are used to clean the data to remove erroneous, duplicate or irrelevant information data, and data standardization and normalization are performed. The standardized and normalized data are fused using database connection and data fusion algorithms to create a comprehensive data set containing information from at least two data sources. Machine learning algorithms and time series analysis are used to conduct in-depth analysis of the comprehensive data set to obtain comprehensive analysis results. The comprehensive analysis results are then verified and calibrated using data verification rules to obtain calibrated data. Finally, the calibrated data is used to perform adaptive identification and prediction monitoring algorithms to adaptively identify and predict abnormal event monitoring, and obtain reports and warnings of abnormal events.

[0034] Preferably, in said S5, it further includes:

[0035] In the monitoring of abnormal events, the specific implementation process of the adaptive recognition and prediction monitoring algorithm is as follows:

[0036] An intelligent abnormal event recognition algorithm is introduced; the algorithm distinguishes between normal behavioral pattern changes and abnormal events through an algorithm based on time series data analysis, combined with probability models and machine learning. Furthermore, an environmental adaptive adjustment algorithm is introduced to perform abnormality judgment based on the adjusted abnormality score: when the adjusted abnormality score exceeds the preset adjusted threshold, it is judged as abnormal; based on the abnormality judgment result, a report and warning of the abnormal event are obtained to provide decision support and emergency response for urban management departments.

[0037] Beneficial effects:

[0038] The multiple technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0039] 1. This application applies machine learning classification models to automatically and accurately identify and process large amounts of heterogeneous data, ensuring effective management of data flows and improving the accuracy of subsequent processing. It uses an adaptive adjustment mechanism to adjust and unify the data formats and scales of different data sources, greatly improving the consistency and comparability of data. It utilizes an efficient collaborative algorithm using advanced data structures and parallel computing technologies to significantly improve the efficiency of data fusion, ensuring that the system can respond quickly and update information in real time when processing large-scale data.

[0040] 2. This application uses an intelligent abnormal event recognition algorithm, combined with time series data analysis, probability models and machine learning technology, to accurately distinguish between changes in normal behavior patterns and true abnormal events, significantly reducing false alarms. The environmental adaptive adjustment algorithm considers the impact of external environmental factors on the judgment of abnormal events, thereby improving the accuracy and reliability of predictions.

[0041] 3. The technical solution of the present application can effectively solve the technical problems of inaccurate data processing and inaccurate and unreliable monitoring of smart cities in smart city monitoring. In addition, the above-mentioned system or method has undergone a series of effect surveys. By applying a machine learning classification model, it can automatically and accurately identify and process a large amount of heterogeneous data, ensuring the effective management of data flow and improving the accuracy of subsequent processing; adjusting and unifying the data format and scale of different data sources through an adaptive adjustment mechanism, greatly improving the consistency and comparability of data, and significantly improving the efficiency of data fusion by utilizing an efficient collaborative algorithm of advanced data structures and parallel computing technology. When processing large-scale data, it ensures that the system can respond quickly and update information in real time. The intelligent identification algorithm for abnormal events combines time series data analysis, probability models and machine learning technology to accurately distinguish between changes in normal behavior patterns and true abnormal events, significantly reducing false alarms. The environmental adaptive adjustment algorithm considers the impact of external environmental factors on the judgment of abnormal events, thereby improving the accuracy and reliability of predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a module diagram of the AI-based smart city monitoring system described in this application;

[0043] Figure 2 This is a flowchart of the AI-based smart city monitoring method described in this application. DETAILED DESCRIPTION

[0044] The embodiments of the present application solve the technical problems of inaccurate data processing and inaccurate and unreliable monitoring of smart cities in the prior art by providing an AI-based smart city monitoring system and method. The overall idea is as follows: First, data collection is performed on the city monitoring data stream to obtain real-time monitoring data stream and analyze the content of the monitoring data stream. Next, feature extraction is performed on the pre-processed data stream, and background modeling technology is used to dynamically separate the foreground and background, and key features are extracted from the video frames. Based on the extracted features, a variety of object detection deep learning models are used to perform target recognition and classification. According to the recognition results, advanced trajectory analysis technology is used to analyze the target behavior pattern, and the long short-term memory network model is used to continuously update the behavior recognition model to include new behavior patterns and trends. The results of the above models are integrated using multi-dimensional data fusion technology to finally obtain comprehensive recognition and analysis results. Then, the identified target signals are analyzed. The method optimizes and standardizes the data based on information elements and behavior patterns; further, a machine learning classification model is applied to the optimized and standardized data, and the machine learning classification model, such as a support vector machine, automatically identifies the data category. When the data flows into the distributed data processing module, a preliminary classification is performed through a pre-trained model; then, an efficient resource scheduler is used to dynamically allocate computing resources to obtain data after distributed resource configuration; further, an efficient collaborative algorithm is introduced, and advanced data mining technology is used for final fusion to obtain a fully integrated comprehensive data set; time series analysis is performed on the integrated data set after fusion, and autoregressive models and sliding average models are used to identify trends and patterns in time series data; based on the results of the time series analysis, a prediction model is used to predict future trends, and based on the results of the time series analysis and trend prediction, a text generation algorithm is applied to automatically generate event summaries; at the same time, data mining technology is used to reveal the deep-seated connections behind the data. Furthermore, for the security and privacy of the data, the data is securely encrypted and privacy-protected to obtain encrypted and desensitized data; finally, the securely encrypted and desensitized data is subjected to advanced data fusion analysis with the data from the associated data source to obtain a comprehensive analysis result, and the comprehensive analysis result is calibrated. After adaptively identifying and predicting abnormal event monitoring on the calibrated data, reports and warnings of abnormal events are obtained, which are provided to urban management departments for decision-making support and emergency response, thereby realizing monitoring of smart cities.

[0045] In order to better understand the above technical solution, the above technical solution will be described in detail below with reference to the accompanying drawings and specific implementation methods.

[0046] Refer to the attached Figure 1 The AI-based smart city monitoring system described in this application includes the following parts:

[0047] Intelligent data analysis module, data optimization and standardization module, distributed data processing module, security encryption and privacy protection module, advanced data fusion analysis module, comprehensive information calibration module, and adaptive abnormal event detection module;

[0048] The intelligent data analysis module uses advanced deep learning algorithms to intelligently analyze the content of real-time urban monitoring data streams to obtain data stream analysis results, which include identified objects and behavior patterns, and transmits the data stream analysis results to the data optimization and standardization module for further processing;

[0049] The data optimization and standardization module pre-processes the data flow analysis results, including noise reduction, formatting and optimization, to improve the efficiency of subsequent processing, obtains standardized and optimized data, and transmits the standardized and optimized data to the distributed data processing module;

[0050] The distributed data processing module uses distributed computing resources to efficiently process and integrate large amounts of data after standardization and optimization, performs complex data analysis, obtains distributed processed data, and transmits the distributed processed data to the security encryption and privacy protection module;

[0051] The security encryption and privacy protection module encrypts and protects the distributed processed data to obtain encrypted and desensitized data, ensuring the security and privacy of the information, and provides the encrypted and desensitized data for use by the advanced data fusion analysis module;

[0052] The advanced data fusion analysis module decrypts the encrypted and desensitized data and performs advanced fusion and analysis on the data from the associated data source to obtain a comprehensive analysis result, which is then transmitted to the comprehensive information calibration module;

[0053] The comprehensive information calibration module calibrates the output data of the advanced data fusion analysis module to ensure the consistency and accuracy of the data, obtains the calibrated data, and transmits the calibrated data to the adaptive abnormal event detection module;

[0054] The adaptive abnormal event detection module uses an adaptive algorithm to identify and predict potential abnormal events based on the output data of the comprehensive information calibration module, and generates abnormal event reports and early warning notifications. Based on the abnormal event reports and early warning notifications, it provides decision support and emergency response for urban management, thereby realizing the monitoring of smart cities.

[0055] Refer to the attached Figure 2 , the AI-based smart city monitoring method described in this application includes the following steps:

[0056] S1. Obtain the city monitoring data stream, analyze the content of the city monitoring data stream, and identify the target information elements and behavior patterns in the city monitoring data stream;

[0057] First, data collection is performed on the urban monitoring data stream, which is a monitoring video stream. A high-speed network connection is used to access the camera network of the urban monitoring system and other monitoring networks to obtain real-time monitoring data streams. Taking monitoring videos as an example, efficient video encoding and decoding technology is used to reduce bandwidth usage and maintain high video quality. Video streams are obtained from monitoring cameras in real time, and their stability and high definition are ensured. Subsequently, the collected data streams are preprocessed. Taking the monitoring video stream as an example, a digital filter is used to remove noise from the video stream. The brightness and contrast of the video after noise removal are automatically adjusted using an illumination adaptation algorithm to adapt to different lighting conditions. Histogram equalization is then used to enhance the video quality. At the same time, an inter-frame difference analysis algorithm is used to analyze the changes between consecutive frames to identify important events, such as sudden light changes or the appearance of moving objects, thereby identifying possible important events and preparing for subsequent feature extraction.

[0058] Next, feature extraction is performed on the preprocessed data stream, using background modeling techniques to dynamically separate foreground and background, enhancing target recognition accuracy. Key features, such as shape, color, and texture, are extracted from video frames, laying the foundation for object recognition. Feature caching is performed on frequently occurring scene elements to speed up subsequent recognition processing. Objects are defined as the elements in the video stream being identified and analyzed.

[0059] Based on the extracted features, multiple deep learning models for object detection are combined to perform target recognition and classification to improve coverage and accuracy. The system first identifies broad categories (such as vehicles and people) and then further subcategories (such as car types and pedestrian characteristics). At the same time, newly identified target features are updated to the database in real time to continuously optimize and expand recognition capabilities.

[0060] Analyzing the target's behavior patterns based on the recognition results using advanced trajectory analysis techniques, such as Kalman filtering. Specifically, the Kalman filtering advanced trajectory analysis technique is used to analyze target behavior patterns in complex environments, such as congested traffic or crowded places. Long-short-term memory network models are used to continuously update the behavior recognition model to include new behavior patterns and trends.

[0061] The results of the above model are integrated using multi-dimensional data fusion technology. The multi-dimensional data fusion technology uses data formatting tools to format multi-dimensional data such as time, location, and target type, and ultimately obtain comprehensive recognition and analysis results; the recognition and analysis results include target information elements and behavior patterns in the data stream.

[0062] S2. Optimize and standardize the data of the identified target information elements and behavior patterns;

[0063] First, a stable and efficient data receiving interface is established using the HTTP protocol to ensure the integrity and accuracy of data received from the intelligent data analysis module. The SHA-1 algorithm is used to calculate the hash value of the received data to ensure that the received data has not been tampered with or damaged. The received data includes real-time data stream analysis results, such as object recognition data and behavior pattern data.

[0064] Next, the received data is cleansed to remove invalid, erroneous, or incomplete data. Data validation rules are applied during this process to check the format and scope of the data, identifying and removing data records that do not meet predefined standards. This ensures data cleanliness and consistency, providing clean data input for the subsequent noise reduction process.

[0065] After data cleaning, the cleaned data is subjected to noise reduction optimization processing to improve data quality and reduce errors. During the noise reduction process, the cleaned data is subjected to a digital filter, such as a Wiener filter, to reduce noise. Then, signal processing techniques, such as Fourier transform, are used to analyze and reduce frequency noise. This obtains clear and accurate data, providing a basis for subsequent data formatting. Furthermore, the noise reduction optimized data is formatted according to data formatting standards.

[0066] The purpose of data formatting is to convert data into a unified format to ensure consistency and facilitate subsequent processing. During the formatting process, all data is ensured to meet specific resolution and encoding standards, and text information (such as license plate numbers) is unified to obtain a standard format. Further, a standardized data set is obtained in preparation for transmission to the distributed data processing module.

[0067] S3. Perform distributed processing on the optimized and standardized data to obtain event summaries and statistical analysis results;

[0068] First, a machine learning classification model is applied to the optimized and standardized data. The machine learning classification model, such as a support vector machine, automatically identifies data categories. When the data flows into the distributed data processing module, it is initially classified using a pre-trained model to ensure that each data type is correctly identified and assigned to the corresponding processing flow.

[0069] Subsequently, an efficient resource scheduler is used to dynamically allocate computing resources. Tasks are intelligently assigned based on data processing requirements and server load, ensuring the efficiency and balance of the processing process, and obtaining data after distributed resource configuration.

[0070] In the data fusion stage, to enhance data consistency, the data after distributed resource configuration is integrated. The data after distributed resource configuration includes video analysis results and traffic flow data. A fusion enhancement algorithm is introduced. The fusion enhancement algorithm aims to adjust and unify the data formats and scales of different data sources through an adaptive adjustment mechanism. The specific implementation can be expressed as follows:

[0071] V adjusted =θ(V source )·V original +β(V source , V target )

[0072] Among them, V adjusted is the preliminary fusion dataset with improved data consistency; V original It is the original data value, which comes from video surveillance, traffic sensors, and environmental monitoring equipment, reflecting the real-time status of the city; V source Is the type of data source, indicating what kind of monitoring device the data comes from; V target is the target data format, which represents the unified format and standard expected after data fusion. α(·) is a rescaling function that adjusts the scale of the data according to the different data sources to achieve consistency between data from different sources. For example, traffic flow data and environmental monitoring data have different scales and need to be adjusted for unified processing.

[0073] α(V source )=a source ·V source +b source

[0074] Among them, a source and b source Is based on the data source type V source Adjustment coefficient; V source is the data source type, such as video surveillance, traffic monitoring; a source is the scale adjustment coefficient corresponding to the data source type, which is used to adjust the scale of the data to make it compatible with other data sources; b source It is the offset corresponding to the data source type, which is used to further fine-tune the data to achieve consistency;

[0075] β(V source , V target ) is an offset adjustment function that adjusts the data format deviation according to the data source type and target data format. For example, it unifies the timestamp format in the video surveillance analysis results with the timestamp format of the traffic monitoring system.

[0076] β(V source , V target)=c source，target ·V source +d source，target

[0077] Among them, c source，target and d source，target It is an adjustment coefficient based on the data source type and target data format, used to adjust the format deviation of the data according to the data source type and target data format;

[0078] Furthermore, in order to avoid low data fusion efficiency when processing large-scale data, this application introduces an efficient collaborative algorithm, which uses advanced data structures and parallel computing technology to improve the efficiency of data fusion; the specific implementation process is expressed by the following formula:

[0079]

[0080] Among them, R is the result after fusion, that is, the highly efficient fused data set; P i is a set of parameters used to adjust and optimize the processing of the i-th data source; f i (V adjusted，i , P i ) is an optimized conversion function that depends not only on the i-th data source V adjusted，i , also depends on a specific set of parameters P i , which can include data weights, priorities, or specific processing rules; is a parallel fusion operator that contains intelligent decision logic to optimize the order and manner of data fusion; n is the total number of data sources involved in the data fusion process;

[0081]

[0082] Among them, w ij is the weight coefficient; g ij (V adjusted，ij ) is the conversion function of the jth feature of the i-th data source; m represents the i-th data source V adjusted，i The number of features considered in ;

[0083] The processed data sets are finally integrated using advanced data mining techniques to obtain a fully integrated comprehensive data set; the advanced data mining techniques, such as cluster analysis and principal component analysis, are used to extract key information from the data;

[0084] Next, we conduct time series analysis on the integrated dataset, using specialized time series databases and analytical tools for in-depth analysis. We apply autoregressive and sliding average models to identify trends and patterns in the time series data. Based on the results of this time series analysis, we use a forecasting model to predict future trends. We train the model with extensive historical data to improve the accuracy and reliability of the forecasts. Based on historical data and current analysis results, we predict future trends, such as increases and decreases in traffic volume and changes in hotspots of urban activity.

[0085] Finally, based on the results of time series analysis and trend forecasting, a text generation algorithm is applied to automatically generate event summaries. At the same time, data mining techniques, such as association rule learning and frequent item set mining, are used to perform statistical analysis and generate comprehensive statistical analysis reports, revealing the deep connections behind the data.

[0086] By applying machine learning classification models, this application can automatically and accurately identify and process large amounts of heterogeneous data, ensuring effective management of data flows and improving the accuracy of subsequent processing; it uses an adaptive adjustment mechanism to adjust and unify the data formats and scales of different data sources, greatly improving the consistency and comparability of data; and it significantly improves the efficiency of data fusion by utilizing efficient collaborative algorithms using advanced data structures and parallel computing technologies, ensuring that the system can respond quickly and update information in real time when processing large-scale data.

[0087] S4. To ensure data security and privacy, securely encrypt and protect the data to obtain encrypted and desensitized data.

[0088] In the smart city monitoring system, the implementation of the security encryption and privacy protection module begins with the reception of distributed processed data. The distributed processed data contains sensitive information. To ensure its security and privacy, the following implementation is performed:

[0089] First, data classification and evaluation criteria are established using empirical methods to classify and evaluate the data, subdividing the input data into sensitive data and non-sensitive data. Sensitive data, such as personal information and location data, undergoes a risk assessment to determine the required level of protection. Next, an appropriate encryption algorithm is selected based on the nature of the data and protection requirements. Encryption algorithms such as AES, RSA, and ECC are used. For particularly sensitive data, more advanced asymmetric encryption technologies are used.

[0090] Subsequently, encryption is performed on sensitive data to ensure data security during transmission and storage. During the data transmission stage, security protocols are used to protect data and prevent it from being intercepted or tampered with during transmission. Data desensitization must be performed. The data desensitization process uses masks, disguises, or other technical means to hide or replace sensitive data. Even if the data is leaked, personal information will not be exposed.

[0091] Furthermore, strict access control and permission management must be implemented to ensure that only authorized users can access encrypted data. Appropriate data access rights must be assigned based on user roles and needs by setting different access permissions. All activities accessing encrypted data must be closely monitored to ensure there is no unauthorized data access or leakage. Regular security audits must be conducted on the above processes to assess the effectiveness of data protection measures and adjust them as needed.

[0092] After the above processing, encrypted and desensitized data is finally obtained. The data maintains the information required for the purpose while ensuring security and privacy, providing protected data for subsequent analysis modules;

[0093] S5. Perform advanced data fusion analysis on the encrypted and desensitized data with the data from the associated data source to obtain comprehensive analysis results, calibrate the comprehensive analysis results, and provide data for abnormal event monitoring. After adaptively identifying and predicting abnormal event monitoring on the calibrated data, abnormal event reports and warnings are obtained;

[0094] When performing advanced data fusion analysis, the input encrypted monitoring data and related data source data, such as real-time traffic flow data, weather information, and urban infrastructure status data, are first processed. The encrypted and desensitized data is then decrypted using symmetric or asymmetric encryption algorithms to convert it back to its original, usable form. The decryption process uses a pre-shared key (symmetric encryption) or a private key (asymmetric encryption) to ensure the security of the decryption process and prevent data leakage.

[0095] The decrypted data is then collected and integrated with data from related data sources using ETL tools to create a unified data format and structure, forming a unified, standardized data set for in-depth analysis. Next, data cleaning tools are used to clean the data, remove errors, duplications, or irrelevant information, and perform data standardization and normalization to ensure data quality and improve analysis accuracy.

[0096] The standardized and normalized data is fused using database connections and data fusion algorithms. During the data fusion process, the temporal and spatial correlations of the data are considered to ensure the logical consistency of the data, thereby creating a comprehensive data set containing information from multiple data sources.

[0097] Conduct in-depth analysis of comprehensive data sets using machine learning algorithms and time series analysis. These algorithms assist in pattern recognition and trend analysis, and apply time series analysis models to identify temporal correlation patterns in data, thereby identifying key data patterns and insights.

[0098] The comprehensive analysis results are checked and calibrated using data verification rules set by expert experience, such as range check and format check, and compared with historical data for verification and calibration to obtain calibrated data;

[0099] Finally, the calibrated data is subjected to adaptive recognition and prediction monitoring algorithms to perform adaptive recognition and prediction of abnormal events, thereby obtaining reports and early warnings of abnormal events. The specific implementation process of the adaptive recognition and prediction monitoring algorithm is as follows:

[0100] First, during the abnormal event detection process, in order to effectively distinguish normal daily behavioral changes from true abnormal events and avoid frequent false alarms, an abnormal event intelligent identification algorithm is introduced. This abnormal event intelligent identification algorithm uses an algorithm based on time series data analysis, combined with probability models and machine learning, to accurately distinguish normal behavioral pattern changes from abnormal events. The specific implementation process is as follows:

[0101] Let X t is the data vector at time point t (such as the time dynamic pattern analysis model quantity of pedestrian flow and vehicle flow), Y t is a label indicating whether an anomaly occurs at the time point (1 for anomaly, 0 for normal). A temporal dynamic pattern analysis model is introduced. The temporal dynamic pattern analysis model uses an advanced sequence pattern recognition technology to analyze the temporal dynamics of urban monitoring data to capture and predict changes in behavioral patterns. The specific implementation formula is as follows:

[0102] X t =γ1X t-1 +γ2X t-2 +…+γ p X t- p+τ1ε t-1 +τ2ε t-2 +…+τ p ε t-p +ε t

[0103] Among them, X t is the data vector at time point t, which can include pedestrian flow, vehicle flow, and environmental noise; γ1,…γ p and τ1,…τ p is the parameter of the model, which is obtained through historical data training and is used to describe the dynamics of time series data; p is the number of model parameters; ε t is the error term at time point t, representing random fluctuations or noise that cannot be explained by the model;

[0104] Furthermore, a sequence behavior classification engine is introduced to calculate the anomaly score. The sequence behavior classification engine uses an advanced technology based on pattern recognition and machine learning to classify the results of temporal dynamic pattern analysis into normal or abnormal patterns. The specific formula is expressed as:

[0105]

[0106] Among them, S t is the anomaly score at time point t, reflecting the possibility of an abnormal event occurring at time t; w i is the weight coefficient, which is used to weight the data at different time points to emphasize the impact of recent data on abnormal judgment; i is the data vector TDMA(X i ) represents the results obtained by the time dynamic mode analysis model; SBCE(TDMA(X i )) represents the result obtained by the machine learning model classification based on the pattern recognition algorithm; k represents the time span of the historical data considered when calculating the anomaly score at the current time point t;

[0107] Furthermore, in the abnormal event detection process, in order to adapt to the dynamic changes in the urban environment, such as the impact of weather, holidays, etc. on abnormal event detection, an environment adaptive adjustment algorithm is introduced. The specific implementation process of the environment adaptive adjustment algorithm is as follows:

[0108] Let Z t is the external environment vector at time point t, which may include weather conditions and special day marks; and introduces principal component analysis to Z t Perform dimensionality reduction processing to extract key features; introduce an adaptive environmental response algorithm to adaptively adjust the anomaly score. The adaptive environmental response algorithm combines the output of the time dynamic pattern analysis and sequence behavior classification engine with the key environmental features extracted by principal component analysis to adaptively adjust the recognition criteria of abnormal events. The specific implementation formula is as follows:

[0109]

[0110] Among them, S′ t It is an abnormality score adjusted after considering external environmental factors, providing a more accurate prediction of abnormal events; μ Z and σ Z are the mean and standard deviation of the external environmental characteristics, respectively, used to quantify the deviation between the current environmental conditions and the normal state; PCA(Z t ) is the result obtained through principal component analysis;

[0111] Adjusted abnormality judgment: If S′ tIf the threshold value exceeds the preset adjusted threshold value based on expert experience, it is considered abnormal;

[0112] Based on the abnormality determination results, we can finally obtain reports and warnings of abnormal events, which can be used by urban management departments for decision support and emergency response, thus realizing the monitoring of smart cities.

[0113] This application uses an intelligent abnormal event recognition algorithm, combined with time series data analysis, probability models and machine learning technology, to accurately distinguish between changes in normal behavior patterns and true abnormal events, significantly reducing false alarms. The environmental adaptive adjustment algorithm takes into account the impact of external environmental factors on the judgment of abnormal events, thereby improving the accuracy and reliability of predictions.

[0114] To sum up, the AI-based smart city monitoring system and method described in this application are completed.

[0115] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages:

[0116] 1. This application applies machine learning classification models to automatically and accurately identify and process large amounts of heterogeneous data, ensuring effective management of data flows and improving the accuracy of subsequent processing. It uses an adaptive adjustment mechanism to adjust and unify the data formats and scales of different data sources, greatly improving the consistency and comparability of data. It utilizes an efficient collaborative algorithm using advanced data structures and parallel computing technologies to significantly improve the efficiency of data fusion, ensuring that the system can respond quickly and update information in real time when processing large-scale data.

[0117] 2. This application uses an intelligent abnormal event recognition algorithm, combined with time series data analysis, probability models and machine learning technology, to accurately distinguish between changes in normal behavior patterns and true abnormal events, significantly reducing false alarms. The environmental adaptive adjustment algorithm considers the impact of external environmental factors on the judgment of abnormal events, thereby improving the accuracy and reliability of predictions.

[0118] Effect research:

[0119] The technical solution of the present application can effectively solve the technical problems of inaccurate data processing and inaccurate and unreliable monitoring of smart cities in smart city monitoring. In addition, the above-mentioned system or method has undergone a series of effect surveys. By applying a machine learning classification model, it can automatically and accurately identify and process a large amount of heterogeneous data, ensure the effective management of data flow, and improve the accuracy of subsequent processing; adjust and unify the data format and scale of different data sources through an adaptive adjustment mechanism, greatly improve the consistency and comparability of data, and significantly improve the efficiency of data fusion by using an efficient collaborative algorithm of advanced data structures and parallel computing technology. When processing large-scale data, it ensures that the system can respond quickly and update information in real time. Through the intelligent recognition algorithm of abnormal events, combined with time series data analysis, probability models and machine learning technology, it accurately distinguishes between changes in normal behavior patterns and real abnormal events, significantly reducing false alarms. The environmental adaptive adjustment algorithm takes into account the impact of external environmental factors on the judgment of abnormal events, thereby improving the accuracy and reliability of predictions.

[0120] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0122] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0123] Obviously, those skilled in the art may make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if such changes and modifications fall within the scope of the claims and their equivalents, the present invention is intended to include such changes and modifications.

Claims

1. A smart city monitoring method based on AI, characterized in that: The following steps are involved: S1. Obtain urban monitoring data streams, analyze the content of urban monitoring data streams, and identify target information elements and behavior patterns in urban monitoring data streams; S2. Optimize and standardize the data of the identified target information elements and behavior patterns; S3. Apply a machine learning classification model to the optimized and standardized data to automatically identify data categories. Once the data flows into the distributed data processing module, it undergoes preliminary classification using a pre-trained machine learning classification model. A resource scheduler dynamically allocates computing resources, intelligently assigning tasks based on data processing requirements and server load, to generate data after distributed resource allocation. The data is further fused and time series analysis is performed on the fused dataset. Autoregressive and sliding average models are used to identify trends and patterns in the time series data, and future trends are predicted based on the results of the time series analysis. Finally, a text generation algorithm is used to automatically generate event summaries based on the results of the time series analysis and trend prediction. Statistical analysis is also performed using data mining techniques to generate statistical results. S4. Securely encrypt and protect the data to obtain encrypted and desensitized data. S5. Decrypt the encrypted and desensitized data and integrate it with data from related data sources to form a data set. This data is then cleaned using data cleaning tools and standardized. The data from these related data sources includes real-time traffic flow data, weather information, and urban infrastructure status data. Perform data fusion on the standardized and normalized data using database connection and data fusion algorithms to create a comprehensive data set containing information from at least two data sources; Performing in-depth analysis on the comprehensive data set using machine learning algorithms and time series analysis to obtain comprehensive analysis results, and performing data verification and calibration on the comprehensive analysis results using data verification rules to obtain calibrated data; The calibrated data is then used to perform adaptive recognition and prediction monitoring using an adaptive recognition and prediction monitoring algorithm. The specific implementation process is as follows: An intelligent abnormal event recognition algorithm is introduced. Based on the algorithm of time series data analysis, combined with probability model and machine learning, it distinguishes between normal behavioral pattern changes and abnormal events. Furthermore, an environmental adaptive adjustment algorithm is introduced to make abnormality judgments based on the adjusted abnormality score: when the adjusted abnormality score exceeds the preset adjusted threshold, it is judged as abnormal. Based on the abnormality judgment results, reports and warnings of abnormal events are obtained.

2. The AI-based smart city monitoring method according to claim 1, characterized in that: Said S1 specifically includes: Data is collected on the urban monitoring data stream, and the monitoring network of the urban monitoring system is connected to obtain real-time urban monitoring data streams; the collected monitoring data streams are preprocessed and feature extracted from the preprocessed monitoring data streams. Target recognition and classification are performed based on the extracted features combined with at least two object detection deep learning models. The major categories, including vehicles and people, are first identified, and then the minor categories, including car types and pedestrian characteristics, are subdivided. Advanced trajectory analysis technology is used to analyze the recognition results, and the analysis results are integrated using multi-dimensional data fusion technology to obtain the final recognition and analysis results, which include target information elements and behavior patterns in the monitoring data stream.

3. The AI-based smart city monitoring method according to claim 2, characterized in that: Said S2 specifically includes: The received target information elements and behavior pattern data are cleaned; the cleaned data are subjected to noise reduction and optimization processing; the noise reduction and optimization data are formatted, during which text information, including license plate numbers, are unified to obtain a standard format; and standardized data are further obtained.

4. The AI-based smart city monitoring method according to claim 1, characterized in that: Said S3 specifically includes: In the data fusion stage, a fusion enhancement algorithm and an efficient collaborative algorithm are introduced; the fusion enhancement algorithm adjusts and unifies the data formats and scales of different data sources through an adaptive adjustment mechanism; the efficient collaborative algorithm adopts advanced data structures and parallel computing technology.

5. The AI-based smart city monitoring method according to claim 1, characterized in that: Said S4 specifically includes: Set classification and evaluation criteria to subdivide input data into sensitive data and non-sensitive data; select appropriate encryption algorithms based on the nature of the data and protection requirements, perform encryption operations on sensitive data, and use security protocols to protect data during the data transmission stage; perform data desensitization processing on the data, which includes using masks, disguises, hiding and replacing sensitive data, so that even if the data is leaked, personal information will not be exposed; implement access control and permission management.

6. An AI-based smart city monitoring system, applied to the AI-based smart city monitoring method according to claim 1, characterized in that: Includes the following sections: Intelligent data analysis module, data optimization and standardization module, distributed data processing module, security encryption and privacy protection module, advanced data fusion analysis module, comprehensive information calibration module, and adaptive abnormal event detection module; The intelligent data analysis module uses advanced deep learning algorithms to intelligently analyze the content of real-time urban monitoring data streams to obtain data stream analysis results, which include identified objects and behavior patterns, and transmits the data stream analysis results to the data optimization and standardization module; The data optimization and standardization module preprocesses the data flow analysis results, including noise reduction, formatting and optimization, and transmits the preprocessed data to the distributed data processing module; The distributed data processing module processes and fuses the pre-processed data using distributed computing resources, performs data analysis, obtains distributed processed data, and transmits the distributed processed data to the security encryption and privacy protection module; The security encryption and privacy protection module encrypts and protects the distributed processed data to obtain encrypted and desensitized data, and provides the encrypted and desensitized data to the advanced data fusion analysis module; The advanced data fusion analysis module decrypts the encrypted and desensitized data and performs advanced fusion and analysis on the data from the associated data source to obtain a comprehensive analysis result, which is then transmitted to the comprehensive information calibration module; The comprehensive information calibration module calibrates the output data of the advanced data fusion analysis module and transmits the calibrated data to the adaptive abnormal event detection module; The adaptive abnormal event detection module uses an adaptive algorithm to identify and predict potential abnormal events based on the output data of the comprehensive information calibration module, and generates abnormal event reports and early warning notifications, providing decision support and emergency response for urban management based on the abnormal event reports and early warning notifications.

Images