Smart park management system and method based on Internet of Things

Abstract

The invention relates to the field of park management, and particularly discloses a smart park management system and method based on the Internet of Things, and the method comprises the steps: firstly obtaining park air quality parameter data collected by a plurality of sensors and park noise parameter data, collected by a sound sensor, at a plurality of preset time points, and then employing a deep learning technology to carry out the recognition of the park air quality parameter data and the park noise parameter data; and feature extraction and correlation analysis are carried out on the two parameters, and finally, whether to send out park environment detection early warning is judged through a classifier, so that the air pollution problem is found in time, related personnel are helped to quickly respond, and the health of park residents is protected.

Description

Technical Field

[0001] This application relates to the field of park management, and more specifically, to an Internet of Things-based intelligent park management system and method. Background Art

[0002] A park is an area that integrates various functional facilities and services, aiming to promote industries, technological innovation, scientific research cooperation, and business activities. Parks usually provide complete infrastructure, modern management, and supportive policies to help enterprises develop better and promote industrial agglomeration and economic growth. Common parks include science and technology parks, industrial parks, and business parks, which attract enterprises to settle in through industrial advantages and form industrial chains and innovation ecosystems.

[0003] With the popularization of the Internet and intelligent hardware, the management of parks has started to introduce Internet of Things technology to improve the intelligent level of facility management. In terms of environmental safety monitoring, parks use sensors to monitor parameters such as smoke and dust in real time to ensure the safety of the park. However, problems such as sensor failures and data deviations may occur in the monitoring system, which will affect the monitoring accuracy and pose potential safety hazards. Currently, the main method to solve these problems is to regularly check the accuracy of the monitoring data of sensors. However, due to the different scales of parks and the types of monitoring parameters, the checking process often requires a large amount of manpower and is rather cumbersome.

[0004] Therefore, an Internet of Things-based intelligent park management system and method are desired. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an Internet of Things-based intelligent park management system and method, which first obtain the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by a sound sensor, then use deep learning technology to perform feature extraction and correlation analysis on the two, and finally use a classifier to determine whether to issue a park environmental detection warning, so as to timely discover air pollution problems, help relevant personnel respond quickly, and protect the health of park residents.

[0006] According to one aspect of this application, an Internet of Things-based intelligent park management system is provided, which includes:

[0007] A park environmental data acquisition module, configured to obtain the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by a sound sensor;

[0008] A park environment data extraction module, configured to extract a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by a sound sensor;

[0009] A park environment warning module, configured to determine whether to issue a park environment detection warning based on the park air quality word granularity feature map and the park noise parameter global feature map.

[0010] According to another aspect of the present application, there is provided an Internet of Things-based intelligent park management method, which includes:

[0011] Obtain park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by a sound sensor;

[0012] Extract a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by a sound sensor;

[0013] Based on the park air quality word granularity feature map and the park noise parameter global feature map, determine whether to issue a park environment detection warning.

[0014] Compared with the prior art, an Internet of Things-based intelligent park management system and method provided by the present application first obtains park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by a sound sensor, then uses deep learning technology to perform feature extraction and correlation analysis on the two, and finally uses a classifier to determine whether to issue a park environment detection warning, so as to timely discover air pollution problems, help relevant personnel respond quickly, and protect the health of park residents. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0016] Figure 1 It is a block diagram schematic diagram of an Internet of Things-based intelligent park management system according to an embodiment of the present application.

[0017] Figure 2It is a block diagram schematic of the park environmental data extraction module in the Internet of Things-based smart park management system according to an embodiment of the present application.

[0018] Figure 3 It is a block diagram schematic of the park air quality parameter feature extraction unit in the Internet of Things-based smart park management system according to an embodiment of the present application.

[0019] Figure 4 It is a block diagram schematic of the park environmental warning module in the Internet of Things-based smart park management system according to an embodiment of the present application.

[0020] Figure 5 It is a flowchart of the Internet of Things-based smart park management method according to an embodiment of the present application. Detailed implementation manners

[0021] Various exemplary embodiments, features, and aspects of the present application will be described in detail below with reference to the accompanying drawings. The same reference numerals in the drawings denote elements having the same or similar functions. Although various aspects of the embodiments are shown in the drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0022] The special term "exemplary" here means "serving as an example, embodiment, or illustration". Any embodiment described as "exemplary" here need not be construed as superior to or better than other embodiments.

[0023] In addition, for a better description of the present application, numerous specific details are given in the following detailed implementation manners. Those skilled in the art should understand that the present application can be implemented without some specific details. In some instances, methods, means, elements, and circuits well known to those skilled in the art are not described in detail so as to highlight the gist of the present application.

[0024] Furthermore, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, "a plurality" means two or more unless otherwise specifically defined.

[0025] Figure 1 It is a block diagram schematic of the Internet of Things-based smart park management system according to an embodiment of the present application. As Figure 1As shown, according to the embodiment of the present application, the smart park management system 100 based on the Internet of Things includes: a park environment data acquisition module 110, which is used to obtain park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by sound sensors; a park environment data extraction module 120, which is used to extract a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor; a park environment early warning module 130, which is used to determine whether to issue a park environment detection early warning based on the park air quality word granularity feature map and the park noise parameter global feature map.

[0026] In the above-mentioned smart park management system 100 based on the Internet of Things, the park environment data acquisition module 110 is used to obtain park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by sound sensors. It should be understood that the air quality parameters in the park, such as PM2.5, PM10, CO2 concentration, temperature and humidity, etc., can be set by setting air quality sensors, which are placed in different locations in the park to regularly collect air quality data. The sensor is connected to the data acquisition platform and can upload data in real time and store and analyze it. At the same time, the collection of park noise data is completed by sound sensors, which are arranged at predetermined time points or continuously monitored to accurately record the noise parameters in the park. Among them, in order to ensure the accuracy and timeliness of the data, a timed collection function is set to ensure that the noise level at each time point is effectively recorded. Here, all sensors are connected to the data processing center through wireless or wired networks, and the collected air quality and noise data are transmitted in real time. The platform monitors and analyzes these data and generates reports. These data help to comprehensively monitor the park environment and provide important reference for environmental management and improvement.

[0027] Specifically, in terms of environmental safety monitoring, the park uses sensors to detect environmental parameters such as smoke and dust in real time to ensure safety. However, the sensors may fail or have data deviations, resulting in reduced monitoring accuracy, which poses a potential safety hazard. At present, the main means to solve this problem is to regularly calibrate the monitoring data of the sensor, but due to the large scale of the park and the large number of monitoring projects, the calibration work often requires a lot of manpower and is tedious and time-consuming. Therefore, in the technical solution of the present application, by obtaining the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by sound sensors, and combining deep learning technology, it is determined whether it is necessary to trigger the park environment detection warning, timely identify air pollution problems, and help relevant personnel to quickly take countermeasures to protect the health of park residents.

[0028] In the above-mentioned Internet of Things-based smart park management system 100, the park environmental data extraction module 120 is used to extract the park air quality word granularity feature map and the park noise parameter global feature map from the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor. In this way, it aims to deeply analyze the collected data and extract valuable information that helps evaluate the park environmental quality. By combining the granularity features and global feature maps of air quality and noise parameters, it can provide comprehensive decision-making support for optimizing the park environmental quality and improve the scientificity and accuracy of park management.

[0029] Figure 2 It is a block diagram schematic of the park environmental data extraction module in the Internet of Things-based smart park management system according to an embodiment of the present application. As Figure 2 shown, in a specific embodiment of the present application, the park environmental data extraction module 120 includes: a park air quality parameter feature extraction unit 121, which is used to perform feature extraction on the park air quality parameter data collected by multiple sensors to obtain the park air quality word granularity feature map; a park noise parameter feature extraction unit 122, which is used to perform feature extraction on the park noise parameter data at multiple predetermined time points collected by the sound sensor to obtain the park noise parameter global feature map.

[0030] It should be understood that the generation of the park air quality word granularity feature map is achieved by subdividing and analyzing the data of different air quality parameters (such as PM2.5, PM10, CO2 concentration, temperature and humidity, etc.) in time or space. The feature map can show the changes of air quality over time, location or conditions, revealing the severity, change trend and even potential pollution sources of air pollution during a specific period. In this way, it can more intuitively understand the detailed features of air quality, thus providing data support for formulating more accurate environmental monitoring and management strategies.

[0031] Furthermore, performing feature extraction on the park noise parameter data at multiple predetermined time points collected by the sound sensor to obtain the park noise parameter global feature map is mainly to deeply understand the overall distribution and change trend of the park noise, so as to reveal the noise hot spots and time periods in the park, help discover and address noise problems in a timely manner, and ensure the comfort and safety of the park environment.

[0032] Figure 3 It is a block diagram schematic of the park air quality parameter feature extraction unit in the Internet of Things-based smart park management system according to an embodiment of the present application. As Figure 3As shown, in a specific embodiment of the present application, the park air quality parameter feature extraction unit 121 includes: a park air quality parameter preprocessing subunit 1211, configured to perform data preprocessing on the park air quality parameter data collected by multiple sensors to obtain a park air quality word embedding matrix; a park air quality word embedding semantic encoding subunit 1212, configured to pass the park air quality word embedding matrix through a park air quality word convolutional neural network based on a convolutional neural network to obtain the park air quality word granularity feature map.

[0033] It should be understood that the primary task of the data preprocessing process is to clean and standardize the collected air quality parameters (such as PM2.5, PM10, CO2 concentration, etc.) to ensure the quality and consistency of the data. In a specific embodiment of the present application, first, the data is cleaned by means of denoising, missing value filling, etc., to ensure that the data collected by each sensor can accurately reflect the actual situation in terms of time and space. Then, the data is normalized or standardized to eliminate the dimensionality differences between different parameters, enabling various air quality indicators to be compared on the same scale. Thereafter, in order to construct the word embedding matrix, these cleaned air quality data need to be converted into a vector form suitable for processing by machine learning models. Usually, the data can be segmented into time windows or spatial units, and the air quality indicators within each time period are regarded as a "word", and then word embedding techniques (such as Word2Vec, GloVe, etc.) are used to convert these "words" into low-dimensional vector representations, which can capture the similarities and correlations between air quality indicators. Finally, these embedding vectors will form a matrix, called the air quality word embedding matrix, which can effectively represent the dynamic changes of the park air quality and provide richer and more meaningful features for subsequent machine learning models. Through this process, park managers can better understand the spatio-temporal patterns of air quality and provide a scientific basis for environmental monitoring and governance.

[0034] Furthermore, the park air quality word embedding matrix is passed through a park air quality word convolutional neural network (AQ-CNN) based on a convolutional neural network (CNN) to obtain a park air quality word granularity feature map, in order to extract deeper spatial and temporal features from the embedding matrix and reveal the potential fine-grained change patterns in the air quality data. Through the convolutional neural network, local features can be automatically learned from these embedding vectors and transformed into feature maps with higher semantic information. Specifically, the convolutional neural network filters the word embedding matrix through the convolutional layer to capture local features in the time or space dimension. The convolutional layer extracts the features of the local area through the sliding window mechanism and reduces the dimension through the pooling layer to further highlight the most important feature information. After multiple convolutional and pooling operations, the obtained park air quality word granularity feature map will present the fine-grained changes in air quality on the time and space scales, so as to perform more accurate air quality monitoring and control. Specifically, each layer of the park air quality word convolutional neural network based on the convolutional neural network performs convolutional processing, average pooling processing based on the local feature matrix, and non-linear activation processing on the input data respectively during the forward pass of the layer, and the park air quality word granularity feature map is output by the last layer of the park air quality word convolutional neural network based on the convolutional neural network, where the input of the park air quality word convolutional neural network based on the convolutional neural network is the park air quality word embedding matrix.

[0035] In a specific embodiment of the present application, the park air quality parameter preprocessing subunit 1211 includes: performing word segmentation processing on the park air quality parameter data collected by multiple sensors to obtain multiple park air quality parameter data items; arranging the multiple park air quality parameter data items into the park air quality word embedding matrix.

[0036] It should be understood that air quality data is usually collected by multiple sensors at different time points and contains multi-dimensional parameters. Although these data can reflect the air quality situation in the park, due to the complexity of time and space, it is often difficult to directly analyze these raw data. Therefore, word segmentation plays a role in converting complex data structures into units that are easy to process. Specifically, word segmentation can be regarded as an operation of "segmenting" air quality parameter data. First, according to differences in time or space, the data can be segmented into different time windows or regional segments. Each time period or region represents an independent "word" or data item. Spatially, the park can be divided into several regions, and the sensor data within each region can also be regarded as an independent data item. In this way, the original continuous air quality data is segmented into multiple smaller units that can be analyzed and processed individually. These data items can be obtained through simple segmentation operations (such as dividing by timestamp or geographical location), or through more complex natural language processing (NLP) techniques for feature extraction. Each data item contains the air quality characteristics within that time period or region, in a form similar to a "word", and each data item can be connected through similarity or correlation. In this way, the processed data is not only convenient for storage and query, but also can reveal the laws and trends of air quality changes through further analysis.

[0037] Furthermore, considering that air quality parameter data items are collected at different times, in different spaces or regions. Arranging these data items into a word embedding matrix can simplify complex multi-dimensional data into a form that is convenient for machine learning algorithms to process, and further reveal the potential laws and associations in the data. Specifically, the air quality parameter data items in the park can be regarded as "words", and the word embedding matrix is to arrange these "words" into a matrix according to certain rules and structures. Among them, each row in the matrix represents a data item (that is, the set of air quality parameters for a certain time period or region), and the columns of the matrix represent different air quality dimensions (such as PM2.5, CO2 concentration, humidity, etc.). Through this arrangement method, each item of data on the air quality in the park can find its corresponding position in the matrix, thus facilitating further analysis and processing. By constructing an air quality word embedding matrix, the relationship between air quality parameters in time and space can be effectively captured.

[0038] In a specific embodiment of the present application, the park noise parameter feature extraction unit 122 includes: obtaining a plurality of park noise parameter feature values by passing the park noise parameter data at a plurality of predetermined time points collected by the sound sensor through a park noise parameter encoder; constructing the plurality of park noise parameter feature values into a park noise parameter feature matrix; and obtaining the park noise parameter global feature map by passing the park noise parameter feature vector through a park noise parameter convolutional neural network based on a multi-scale neighborhood feature extraction module.

[0039] It should be understood that the noise parameter data usually includes multi-dimensional information such as sound intensity, frequency, and noise waveform. These data are often presented in the form of a time series and exhibit different change patterns at different time points and in different regions. Directly processing these raw data may lead to information loss or overly high computational complexity. Therefore, processing the data through a noise parameter encoder helps to extract more representative features. Among them, the role of the park noise parameter encoder is to extract features and perform mapping on the collected noise data, converting the noise parameters at each time point or time window into a set of high-dimensional feature values. In the specific operation process, first, the raw noise data at each time point or time window is preprocessed to remove noise and outliers to ensure the accuracy and consistency of the data. Then, the encoder performs mapping on these data to convert different types of noise parameters (such as sound intensity, frequency, waveform, etc.) into features of a fixed length. These feature values can reflect the different patterns of noise in different time periods and regions, thus providing more detailed noise pollution monitoring data for park managers. Specifically, the park noise parameter data at multiple predetermined time points collected by the sound sensor is passed through the park noise parameter encoder to obtain multiple park noise parameter feature values, including: using the fully connected layer of the park noise parameter encoder to perform fully connected encoding on the park noise parameter data at multiple predetermined time points collected by the sound sensor to extract the high-dimensional hidden features of the feature values at each position in the park noise parameter data at multiple predetermined time points collected by the sound sensor; and, using the one-dimensional convolutional layer of the park noise parameter encoder to perform one-dimensional encoding on the park noise parameter data at multiple predetermined time points collected by the sound sensor to extract the high-dimensional hidden correlation features of the correlation between the feature values at each position in the park noise parameter data at multiple predetermined time points collected by the sound sensor.

[0040] Furthermore, noise data is usually multi-dimensional, including parameters such as sound intensity, frequency, waveform, etc., and has different distributions at different time points and spatial regions. A single eigenvalue cannot effectively show the correlation between data. By constructing a feature matrix, these data points can be organically combined and represented, thus better revealing the spatial distribution and temporal variation trend of noise. Specifically, the eigenvalue of the park noise parameter is a high-dimensional feature extracted by a noise parameter encoder, representing the noise characteristics at each sampling time point or time window. The first step in constructing these eigenvalues into a feature matrix is to determine the dimension of the matrix. Usually, the number of rows in the matrix represents different time windows or spatial regions, while the number of columns represents different noise feature dimensions. For example, each row can represent the noise characteristics of different regions within a certain time period, and each column corresponds to a specific noise parameter, such as noise intensity, frequency, spectral characteristics, etc. In this way, each element in the matrix corresponds to the noise characteristics of a specific time point or region. Through this method, the feature matrix not only integrates the noise feature data of each time point or region but also shows the internal connection between different time and spatial dimensions.

[0041] Specifically, constructing the multiple park noise parameter eigenvalues into a park noise parameter feature matrix includes: creating a project through Spring Initializr and selecting dependencies; defining a model NoiseParameter representing the park noise parameter eigenvalues; creating a service layer NoiseService for processing multiple noise parameter eigenvalues and constructing the park noise parameter feature matrix; creating a controller layer NoiseController for exposing REST interfaces, receiving requests, and returning results; creating a main startup class Application for scanning and starting all necessary components; building and running the project using Maven.

[0042] Among them, first define the noise parameter data model NoiseParameter.

[0043]

[0044] Then, create a service layer NoiseService for constructing the noise parameter eigenvalues into a matrix.

[0045]

[0046]

[0047] Next, create a controller layer NoiseController for exposing REST API interfaces, receiving user requests, and calling the methods of the service layer to process data.

[0048]

[0049]

[0050] Next, the main startup class Application is used to automatically scan and start all necessary components.

[0051]

[0052] In this way, a park noise parameter processing system based on the Spring Boot framework is implemented. By providing RESTful API interfaces, multiple park noise parameter feature values are constructed into a matrix. Specifically, the system includes a data model layer, a service layer, and a controller layer. The data model NoiseParameter is used to represent each noise data point, including a timestamp and a noise level. The service layer NoiseService is responsible for processing the received multiple noise feature values and constructing a two-dimensional matrix (noise feature matrix). The controller layer NoiseController receives the noise parameter data submitted by the user through the HTTP POST interface and calls the service layer method to return the constructed feature matrix. The entire system is implemented based on Spring Boot. After deployment, it can receive noise parameter data in JSON format through the interface and return the constructed noise parameter matrix. In this way, the processed noise data matrix can be obtained through a simple HTTP request, which is convenient for further data analysis and decision support.

[0053] Furthermore, the noise parameter feature vector usually reflects information such as noise intensity and frequency at different time points or spatial regions. However, these local eigenvalue cannot comprehensively reflect the noise situation of the entire park. Through a convolutional neural network, especially a convolutional neural network based on a multi-scale neighborhood feature extraction module, noise features at different scales can be extracted, and then a comprehensive global feature map that can describe the noise state of the entire park can be obtained. Specifically, when the noise parameter feature vector passes through the convolutional neural network, the convolutional layer will scan the input feature vector through a series of convolutional kernels, and extract different levels of features in different scales and local neighborhoods. The role of the multi-scale neighborhood feature extraction module is to capture the local change patterns of noise data from different scales by using convolutional kernels or sliding windows of different sizes. For example, small-scale convolutional kernels can capture the changes in noise in a short period of time or a local area, while large-scale convolutional kernels help to capture the overall noise distribution law within the park. This multi-scale processing method enables the network to simultaneously focus on noise features in different ranges, and then extract multi-level and multi-dimensional comprehensive features. Through this multi-scale neighborhood feature extraction, the convolutional neural network can generate a global feature map representing the noise patterns in different regions and time periods within the park. Each element in the feature map contains the comprehensive information of the noise features at that position, which can better show the spatial distribution, temporal evolution, and potential correlations of the noise. Compared with traditional feature extraction methods, this method not only focuses on local features, but also can capture the overall trend of noise changes from a global perspective, thus improving the accurate description and prediction of the park noise state. Specifically, passing the park noise parameter feature vector through the park noise parameter convolutional neural network based on the multi-scale neighborhood feature extraction module to obtain the park noise parameter global feature map includes: using each layer of the park noise parameter convolutional neural network based on the multi-scale neighborhood feature extraction module to respectively perform the following operations on the input data during the forward pass of the layer: performing convolutional processing on the input data based on the first convolutional kernel to obtain a first convolutional feature map; performing convolutional processing on the input data based on the second convolutional kernel to obtain a second convolutional feature map; performing convolutional processing on the input data based on the third convolutional kernel to obtain a third convolutional feature map; performing convolutional processing on the input data based on the fourth convolutional kernel to obtain a fourth convolutional feature map, where the first convolutional kernel, the second convolutional kernel, the third convolutional kernel, and the fourth convolutional kernel have different sizes; concatenating the first convolutional feature map, the second convolutional feature map, the third convolutional feature map, and the fourth convolutional feature map to obtain a multi-scale convolutional feature map; performing average pooling processing on the multi-scale convolutional feature map along the channel dimension to obtain a pooled feature map; and performing non-linear activation processing on the pooled feature map to obtain an activated feature map; where the output of the last layer of the park noise parameter convolutional neural network based on the multi-scale neighborhood feature extraction module is the park noise parameter global feature map.

[0054] In the above-mentioned Internet of Things-based intelligent park management system 100, the park environment early warning module 130 is used to determine whether to issue a park environment detection early warning based on the park air quality word granularity feature map and the park noise parameter global feature map. It should be understood that the quality of the park environment is often jointly determined by air quality and noise. The two are interrelated and can reflect the overall environmental conditions of the park to a certain extent. By combining these two types of feature maps, it is possible to more comprehensively and accurately determine whether an environmental detection early warning needs to be issued.

[0055] Figure 4 It is a block diagram schematic of the park environment early warning module in the Internet of Things-based intelligent park management system according to an embodiment of the present application. As Figure 4 shown, in a specific embodiment of the present application, the park environment early warning module 130 includes: a park environment parameter feature fusion unit 131, which is used to fuse the park air quality word granularity feature map and the park noise parameter global feature map to obtain a park environment detection early warning judgment classification feature map; a park environment parameter feature optimization unit 132, which is used to perform spatial constraint adaptation adjustment based on principal component space projection on the park environment detection early warning judgment classification feature map to obtain an optimized park environment detection early warning judgment classification feature map; a park environment detection early warning classification judgment unit 133, which is used to pass the optimized park environment detection early warning judgment classification feature map through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a park environment detection early warning.

[0056] It should be understood that relying solely on the air quality or noise feature map for early warning judgment may ignore the environmental risks brought by the interaction between the two. By fusing the feature maps of the two, potential environmental problems can be better revealed, thereby improving the accuracy and reliability of the early warning system. Specifically, the park air quality word granularity feature map provides fine-grained air quality data, showing the concentration distribution of different pollutants (such as PM2.5, carbon dioxide, nitrogen oxides, etc.) in each area and time point within the park; while the park noise parameter global feature map reflects the noise intensity and its change trend in different time periods or spatial regions. When fusing the two, the key information of the two feature maps can be combined through weighted summation, splicing, or feature fusion methods in deep learning. The fused feature map contains richer environmental information and can simultaneously reflect the interaction between air pollution and noise pollution in space and time.

[0057] In particular, considering the classification feature map for environmental detection and early warning judgment in the park environment, due to the superposition of data dimensions and information volumes collected by multiple sensors, features often have high dimensions and significant redundancy. Among them, air quality and noise parameter data are processed through a convolutional neural network and a feature extraction module respectively, resulting in each input feature being converted into a high-dimensional feature map. These feature maps may have overlapping or redundant parts because there may be a high correlation between the measurement data of different sensors. Especially in similar time periods or locations, the air quality and noise levels may show certain synchronous changes, leading to information duplication. In addition, the features between different channels may have low correlation or redundancy because the specific data types obtained by each sensor are independent, but these redundancies are not completely eliminated during fusion. Moreover, traditional feature fusion methods usually ignore the relationship between the overall spatial context distributions in the feature map, which means they may only focus on local feature information and ignore the impact of the global context. This may result in the failure to correctly capture the overall distribution characteristics of the space during feature fusion. Therefore, when making early warning judgments on the environmental state, some potential key factors may be missed. To more accurately judge the park environment condition, the flow and correlation of global spatial information need to be considered during fusion. Therefore, in the technical solution of this application, a spatial constraint adaptation adjustment based on principal component spatial projection is performed on the classification feature map for environmental detection and early warning judgment in the park environment to obtain an optimized classification feature map for environmental detection and early warning judgment in the park environment.

[0058] Among them, performing a spatial constraint adaptation adjustment based on principal component spatial projection on the classification feature map for environmental detection and early warning judgment in the park environment to obtain an optimized classification feature map for environmental detection and early warning judgment in the park environment includes: performing feature deconstruction along the channel dimension on the classification feature map for environmental detection and early warning judgment in the park environment to obtain a set of environmental detection and early warning feature vectors in the park; performing eigenvalue-based feature decomposition on each environmental detection and early warning feature vector in the set of environmental detection and early warning feature vectors to obtain a set of environmental detection and early warning principal component eigen-encoding vectors; calculating the spatial distance entropy between each corresponding pair of environmental detection and early warning feature vectors and environmental detection and early warning principal component eigen-encoding vectors in the set of environmental detection and early warning feature vectors and the set of environmental detection and early warning principal component eigen-encoding vectors to obtain a spatial constraint vector composed of multiple spatial distance entropies; calculating the product between the spatial constraint vector and its transposed vector to obtain a spatial constraint matrix; inputting the spatial constraint matrix into a spatial constraint activation unit based on the Sigmoid activation function to obtain a spatial constraint feature matrix; and performing feature constraint on the classification feature map for environmental detection and early warning judgment in the park environment based on the spatial constraint feature matrix to obtain the optimized classification feature map for environmental detection and early warning judgment in the park environment.

[0059] Among them, each of the park environmental detection and early warning feature vectors in the set of park environmental detection and early warning feature vectors is subjected to eigenvalue-based eigen-decomposition to obtain a set of park environmental detection and early warning principal component eigen-coding vectors, which is represented by the following eigen-decomposition formula:

[0060]

[0061] Among them, V i represents the i-th park environmental detection and early warning feature vector in the set of park environmental detection and early warning feature vectors, and PCA(V i ) represents performing eigenvalue-based eigen-decomposition on V i , U i is the sequence of the first eigen-decomposition vectors corresponding to V i , Λ i is the first diagonal matrix corresponding to V i , U i T is the transpose of U i , v i1 , v i2 , v im are the first, second, and m-th eigenvectors in the sequence of the first eigen-decomposition vectors corresponding to V i , λ i1 , λ im are respectively the eigenvalues at the first position and the m-th position of the first diagonal matrix corresponding to V i , and V i ′ represents the park environmental detection and early warning principal component eigen-coding vector corresponding to V i .

[0062] Among them, the spatial distance entropy between each pair of corresponding park environmental detection and early warning feature vectors and park environmental detection and early warning principal component eigen-coding vectors in the set of park environmental detection and early warning feature vectors and the set of park environmental detection and early warning principal component eigen-coding vectors is calculated to obtain a spatial constraint vector composed of multiple spatial distance entropies, which is represented by the following spatial constraint formula:

[0063]

[0064] Among them, V i represents the i-th park environmental detection and early warning feature vector in the set of park environmental detection and early warning feature vectors, V i ′ represents the park environmental detection and early warning principal component eigen-coding vector corresponding to V i , ||·|| 2 2 is the square of the Euclidean norm of the vector, arccosh is the inverse hyperbolic cosine function, and d P (Vi , V i ′) represents the spatial distance entropy between V i and V i ′, and e i represents the i-th spatial distance entropy in the spatial constraint vector.

[0065] In the technical solution of this application, a spatial constraint adaptation adjustment based on principal component spatial projection is performed on the classification feature map of the park environment detection and early warning judgment. This process begins with the feature deconstruction of the original feature map along the channel dimension, so as to expand its high-dimensional information into a set of feature vector sets represented along different channels. The core idea of this operation is to independently extract the features of different channels by decomposing the feature map of multiple channels (usually a three-dimensional tensor: height, width, and channel), so as to explicitly represent the feature capabilities characterized by each channel in the form of one-dimensional feature vectors. In this solution, the goal is to separate and abstract the significant features hidden in the high-dimensional feature map and further adapt them to the requirements of spatial constraint modeling and optimization.

[0066] After the feature deconstruction is completed, the first set of separated feature vector sets needs to be subjected to eigenvalue-based feature decomposition to generate a corresponding set of principal component eigen-encoding vectors. The theoretical basis of this step stems from principal component analysis, and its mathematical principle is to find the most important direction of the data in the feature space by calculating the covariance matrix of the feature vectors and using the eigen-decomposition method. In this way, the most representative and discriminative components of a specific set of feature vectors, that is, the principal component eigen-encoding vectors, can be extracted. This process not only effectively reduces the feature dimension but also compresses feature redundancy, avoids the influence of noise, and provides a more concentrated and concise representation for subsequent distance calculation and spatial constraint.

[0067] With the above principal component representation, it is necessary to further calculate the spatial distance entropy between each pair of corresponding vectors in the feature vector set and the principal component eigen-encoding vector set. This step combines distance measure and entropy theory to quantify the spatial consistency and distribution difference between each pair of feature vectors and their principal components. Usually, the calculation of spatial distance entropy will use typical distance metric methods such as Euclidean distance or cosine similarity to capture the separability or aggregativity between different features. All calculation results are aggregated into a set of spatial constraint vectors, which generally reflect the overall complexity of the feature distribution and the degree of feature structure. By introducing entropy, an information-theoretic metric, it is possible to effectively characterize the uniformity of the feature distribution and the projection adaptability of the data in the principal component space. Especially in high-dimensional feature learning, spatial distance entropy modeling provides a statistically meaningful global constraint for describing the behavior of feature distribution, so that redundant features can be quantitatively controlled and significant parts can be emphasized.

[0068] Next, perform a multiplication operation on the above spatial constraint vector and its transposed vector to generate a spatial constraint matrix. This process is based on the principle of outer product calculation in tensor algebra, expanding the spatial quantization result of a single vector into a global spatial constraint framework represented in matrix form. The element values of the spatial constraint matrix represent the correlation weights between the feature vectors on the feature channel dimension, revealing the global relationships in the entire feature space. This global dependence modeling in matrix form not only enables the capture of local feature relationships in a larger scope but also provides a fundamental input for the subsequent activation and optimization operations. Through matrix operation, the constraint of atomic feature pairs can be extended to a more extensive local and global spatial dependence modeling, which is particularly important for high-dimensional multi-channel feature analysis.

[0069] To further standardize the weight distribution in the spatial constraint matrix, this matrix is input into a spatial constraint activation unit based on the Sigmoid activation function to generate an activated spatial constraint feature matrix. The role of the activation unit is to normalize the matrix element values through a non-linear function, limiting their range between 0 and 1. The non-linear compression property of the Sigmoid activation function can not only smooth the variation range of feature values but also differentially enhance the saliency of high-weight and low-weight features. During this compression process, noise components are explicitly suppressed, while feature information with obvious distribution patterns or significant spatial correlations is highlighted. Therefore, spatial constraint activation not only achieves excellent regularization effects and characteristic normalization but also enhances the network's ability to learn significant features.

[0070] Finally, after obtaining the spatial constraint feature matrix, perform a feature constraint operation on the original park environment detection and early warning judgment classification feature map to generate an optimized park environment detection and early warning judgment classification feature map. This step uses the constraint feature matrix as a weighting factor to control different spatial information in the feature map element by element. Specifically, the weight values of feature regions with strong saliency are further amplified, while weakly correlated regions are reduced or even masked, so that the optimized feature map mainly focuses on specific regions or characteristic signals with semantic meanings. The final effect of the optimization step is to obtain a sparser, more efficient, and more expressive feature representation.

[0071] Furthermore, the optimized classification feature map has integrated multi-dimensional environmental information such as air quality and noise level in the park. By classifying through a deep learning model or other machine learning techniques, different environmental states can be effectively identified, and based on this, it is determined whether to trigger the warning system. Specifically, the task of the classifier is to identify which category the input feature map belongs to according to the patterns learned during the training process. When training the classifier, it is first necessary to use the labeled environmental data set for training so that the classifier can learn the feature patterns under different environmental conditions. After training, the classifier can output a classification result based on the optimized feature map obtained in real time. If the classification result indicates that the current environment in the park is in a state of excessive pollution or high health risk, the system will trigger a warning to prompt relevant management personnel to take measures, such as adjusting industrial production, restricting traffic flow, or activating the emergency response mechanism. By inputting the optimized park environmental detection and warning judgment classification feature map into the classifier, according to the classification result, the warning system can automatically determine whether to issue a warning signal, thereby improving the efficiency of environmental monitoring and the timeliness of response, and ensuring the safety of the park and the health of residents.

[0072] In summary, in the embodiment of the present application, first, the park air quality parameter data collected by multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor are obtained, and then deep learning technology is used to extract features and perform correlation analysis on the two. Finally, through the classifier, it is determined whether to issue a park environmental detection warning, so as to timely discover air pollution problems, help relevant personnel respond quickly, and protect the health of park residents.

[0073] As described above, the Internet of Things-based intelligent park management system 100 according to the embodiment of the present application can be implemented in various terminal devices. In one example, the Internet of Things-based intelligent park management system 100 can be integrated into the terminal device as a software module and / or a hardware module. For example, the Internet of Things-based intelligent park management system 100 can be a software module in the operating system of the terminal device, or can be an application program developed for the terminal device; of course, the Internet of Things-based intelligent park management system 100 can also be one of the many hardware modules of the terminal device.

[0074] Alternatively, in another example, the Internet of Things-based intelligent park management system 100 and the terminal device can also be separate devices, and the Internet of Things-based intelligent park management system 100 can be connected to the terminal device through a wired and / or wireless network and transmit interaction information according to a predefined data format.

[0075] Figure 5 It is a flowchart of the Internet of Things-based intelligent park management method according to the embodiment of the present application. As Figure 5As shown, the IoT-based intelligent park management method according to an embodiment of the present application includes: S110, obtaining park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by a sound sensor; S120, extracting a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by the multiple sensors and the park noise parameter data at the multiple predetermined time points collected by the sound sensor; S130, based on the park air quality word granularity feature map and the park noise parameter global feature map, determining whether to issue a park environmental detection warning.

[0076] Here, those skilled in the art can understand that the specific operations of each step in the above IoT-based intelligent park management method have been described in detail in the description of the IoT-based intelligent park management system above Figures 1 to 4 and thus, the repeated description thereof will be omitted.

[0077] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0078] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0079] In addition, in each embodiment of the present invention, the functional modules can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated unit can be implemented in the form of hardware, or in the form of hardware plus software functional modules.

[0080] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms.

[0081] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, it is intended to cover all changes falling within the meaning and scope of the equivalent elements of the claims in the present invention. Any reference signs attached to the claims should not be regarded as limiting the claimed rights.

[0082] In addition, it is obvious that the term "including" does not exclude other units or steps, and the singular does not exclude the plural. A plurality of units or devices stated in the system claims can also be implemented by one unit or device through software or hardware. The terms such as "second" are used to denote names and do not denote any particular order.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit of the technical solutions of the present invention.

Claims

1. A smart park management system based on the Internet of Things, characterized in that: include: A park environment data acquisition module, used to acquire park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by sound sensors; A park environment data extraction module, used to extract a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by the multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor; The park environment early warning module is used to determine whether to issue a park environment detection early warning based on the park air quality word granularity feature map and the park noise parameter global feature map.

2. The smart park management system based on the Internet of Things according to claim 1 is characterized in that: The park environment data extraction module includes: A park air quality parameter feature extraction unit, used for extracting features from the park air quality parameter data collected by the multiple sensors to obtain a park air quality word granularity feature map; The park noise parameter feature extraction unit is used to extract features from the park noise parameter data at multiple predetermined time points collected by the sound sensor to obtain a global feature map of the park noise parameters.

3. The smart park management system based on the Internet of Things according to claim 2 is characterized in that: The park air quality parameter feature extraction unit comprises: A park air quality parameter preprocessing subunit, used for performing data preprocessing on the park air quality parameter data collected by the plurality of sensors to obtain a park air quality word embedding matrix; The park air quality word embedding semantic encoding subunit is used to pass the park air quality word embedding matrix through a park air quality word convolutional neural network based on a convolutional neural network to obtain the park air quality word granularity feature map.

4. The smart park management system based on the Internet of Things according to claim 3 is characterized in that: The park air quality parameter preprocessing subunit includes: Performing word segmentation processing on the park air quality parameter data collected by the multiple sensors to obtain multiple park air quality parameter data items; The plurality of park air quality parameter data items are arranged into the park air quality word embedding matrix.

5. The smart park management system based on the Internet of Things according to claim 4 is characterized in that: The park noise parameter feature extraction unit comprises: The park noise parameter data at a plurality of predetermined time points collected by the sound sensor are passed through a park noise parameter encoder to obtain a plurality of park noise parameter characteristic values; Constructing the plurality of park noise parameter characteristic values ​​into a park noise parameter characteristic matrix; The park noise parameter feature vector is passed through a park noise parameter convolutional neural network based on a multi-scale neighborhood feature extraction module to obtain a global feature map of the park noise parameter.

6. The smart park management system based on the Internet of Things according to claim 5 is characterized in that: The plurality of park noise parameter characteristic values ​​are constructed into a park noise parameter characteristic matrix, including: Create a project through SpringInitializr and select dependencies; Define the model NoiseParameter that represents the characteristic value of the park noise parameter; Create a service layer NoiseService to process multiple noise parameter feature values ​​and construct a park noise parameter feature matrix; Create a controller layer NoiseController to expose the REST interface, receive requests and return results; Create the main startup class Application to scan and start all necessary components; Use Maven to build and run the project.

7. The smart park management system based on the Internet of Things according to claim 6 is characterized in that: The park environment early warning module includes: A park environment parameter feature fusion unit, used to fuse the park air quality word granularity feature map and the park noise parameter global feature map to obtain a park environment detection early warning judgment classification feature map; A park environment parameter feature optimization unit is used to perform spatial constraint adaptation adjustment on the park environment detection, early warning and judgment classification feature map based on principal component space projection to obtain an optimized park environment detection, early warning and judgment classification feature map; The park environment detection and early warning classification judgment unit is used to pass the optimized park environment detection and early warning judgment classification feature map through a classifier to obtain a classification result, and the classification result is used to determine whether to issue a park environment detection early warning.

8. The smart park management system based on the Internet of Things according to claim 7 is characterized in that: The park environment parameter characteristic optimization unit includes: Performing feature deconstruction along the channel dimension on the park environment detection early warning judgment classification feature graph to obtain a set of park environment detection early warning feature vectors; Performing eigenvalue-based eigendecomposition on each park environment detection and early warning feature vector in the set of park environment detection and early warning feature vectors to obtain a set of park environment detection and early warning principal component eigencoding vectors; Calculate the spatial distance entropy between each group of corresponding park environment detection and warning feature vectors and park environment detection and warning principal component intrinsic coding vectors in the set of the park environment detection and warning feature vectors and the set of the park environment detection and warning principal component intrinsic coding vectors to obtain a spatial constraint vector composed of multiple spatial distance entropies; Calculating the product between the spatial constraint vector and its transposed vector to obtain a spatial constraint matrix; Inputting the spatial constraint matrix into a spatial constraint activation unit based on a Sigmoid activation function to obtain a spatial constraint feature matrix; Based on the spatial constraint feature matrix, feature constraints are performed on the park environment detection, early warning and judgment classification feature map to obtain the optimized park environment detection, early warning and judgment classification feature map.

9. A smart park management method based on the Internet of Things, characterized in that: include: Acquire park air quality parameter data collected by multiple sensors and park noise parameter data at multiple predetermined time points collected by sound sensors; Extracting a park air quality word granularity feature graph and a park noise parameter global feature graph from the park air quality parameter data collected by the multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor; Based on the park air quality word granularity feature map and the park noise parameter global feature map, determine whether to issue a park environment detection warning.

10. The method for managing a smart park based on the Internet of Things according to claim 9, characterized in that: Extracting a park air quality word granularity feature map and a park noise parameter global feature map from the park air quality parameter data collected by the multiple sensors and the park noise parameter data at multiple predetermined time points collected by the sound sensor, including: Performing feature extraction on the park air quality parameter data collected by the multiple sensors to obtain a park air quality word granularity feature map; Feature extraction is performed on the park noise parameter data at multiple predetermined time points collected by the sound sensor to obtain a global feature map of the park noise parameters.

Images