Construction site dust fall and pollution prevention and control method and system based on multi-mode perception

Abstract

The invention discloses a construction site dust fall and pollution prevention and control method and system based on multi-modal perception, and relates to the technical field of construction management, and the method comprises the steps: collecting PM data at a boundary in real time through a boundary monitoring device preset at the boundary of a construction site, and meanwhile, using an unmanned plane carrying the monitoring device to monitor the PM data at the boundary; performing flight monitoring on the interior of the construction site according to a preset route, and collecting PM data of different areas of the site; performing preliminary analysis on the multi-modal sensing data in real time by using an edge calculation module, and identifying dust abnormal conditions; and the data processing center analyzes a current pollution scene of the construction site, matches an optimal dust falling strategy for the current pollution scene, controls surrounding dust falling equipment to perform dust falling operation, constructs a self-organizing communication network among the dust falling equipment, and adjusts a dust falling operation plan by using dust falling feedback information. According to the invention, real-time data collected by the boundary monitoring equipment and the unmanned aerial vehicle are fused, and environment parameters and construction working conditions are combined, so that precise regulation and control of the dust falling equipment are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental protection and construction management, and more specifically, to a method and system for dust reduction and pollution prevention and control at a construction site based on multimodal perception. Background Art

[0002] In the current construction environment, dust pollution has become a problem of great concern. The traditional dust reduction methods in construction sites are relatively simple and ineffective. Relying solely on occasional manual watering, it is impossible to perform precise dust reduction operations according to changes in dust concentration in real time. With the increasing requirements for the construction environment, it is urgent to create a dust-free site. The existing factory boundary spraying system and sprinkler truck spraying system, due to the lack of effective monitoring and intelligent control, have excessive spraying resulting in waste of water resources, or insufficient spraying resulting in unsatisfactory dust reduction effects. At the same time, there is a lack of comprehensive and real-time dust monitoring methods at the boundaries and inside the construction site, making it difficult to accurately grasp the dust concentration, and thus unable to initiate dust reduction measures in a timely manner, which can easily lead to secondary pollution and affect the surrounding environment and the health of construction workers.

[0003] In recent years, with the development of the Internet of Things (IoT) and drone technology, multimodal environmental monitoring methods have gradually been applied to construction pollution control. For example, by deploying particulate matter (PM2.5, PM10) sensors, weather stations and other equipment at the site boundary, dust concentration and diffusion trends can be monitored in real time; and drones equipped with environmental monitoring modules can dynamically collect high-altitude and blind spot data to make up for the shortcomings of fixed monitoring points. Therefore, how to build a management system that can comprehensively utilize the data collected by boundary monitoring equipment and drones equipped with monitoring equipment, analyze and automatically control dust reduction equipment to achieve energy-saving and efficient prevention and control of secondary pollution is an urgent problem that needs to be solved. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for dust reduction and pollution control at construction sites based on multimodal perception. By integrating real-time data collected by boundary monitoring equipment and drones, combining environmental parameters and construction conditions, a dynamic pollution diffusion model is constructed, and an adaptive algorithm is used to achieve precise regulation of dust reduction equipment, thereby achieving efficient dust reduction and secondary pollution prevention and control.

[0005] The first aspect of the present invention provides a construction site dust reduction and pollution prevention and control method based on multimodal perception, comprising the following steps: The PM data at the construction site boundary is collected in real time through the preset boundary monitoring equipment. At the same time, the drone equipped with the monitoring equipment is used to conduct flight monitoring inside the construction site according to the preset route to collect PM data from different areas of the site. The PM data obtained by monitoring is combined with the environment-related data obtained by the temperature and humidity sensor to generate multimodal perception data, and the edge computing module is used to perform preliminary analysis on the multimodal perception data in real time to identify abnormal dust conditions; The multimodal sensing data is marked according to the dust anomaly, and is transmitted to the data processing center in a priority and rapid manner, a dynamic pollution diffusion model is constructed to analyze the current pollution scene of the construction site, and the best dust reduction strategy is matched for the current pollution scene; Based on the optimal dust reduction strategy, a dust reduction equipment control signal is generated to control the surrounding dust reduction equipment to perform dust reduction operations, a self-organizing communication network is constructed between the dust reduction equipment, and the dust reduction feedback information of the dust reduction equipment is used to adjust the dust reduction operation plan.

[0006] In this solution, PM data of the construction site is collected, specifically: Deploy boundary monitoring equipment at preset intervals at the construction site boundary, collect PM data at the boundary based on PM indicators through the deployed boundary monitoring equipment, deploy drones equipped with monitoring equipment, conduct flight monitoring inside the construction site according to preset routes, and collect PM data and environment-related data inside the construction site; Performing grid processing on the construction site, obtaining construction data inside the construction site, marking grid blocks using the construction data, clustering the grid blocks according to the marking information, and dividing the construction site into regions using the clustering results; Identify the construction content and construction stage according to the marking information of the area, use the construction content and construction stage to identify the key construction area, and add attention feedback to the drone for intensive monitoring and attention of the key construction area; In different areas, multimodal sensing data corresponding to PM concentration is generated according to PM data at the boundary and PM data inside the area, and the multimodal sensing data is preprocessed.

[0007] In this solution, an edge computing module is used to perform preliminary analysis on the multimodal sensing data in real time to identify dust anomalies, specifically: Integrate an edge computing module on the boundary monitoring equipment and drones at the construction site, import the multimodal sensing data into the edge computing module for preliminary analysis, and obtain PM concentration change sequences corresponding to different areas; Retrieving construction examples by similarity calculation based on the construction content and construction stage in different areas, obtaining historical dust monitoring data in the construction examples, performing data preprocessing and averaging on the historical dust monitoring data, and obtaining a baseline PM concentration change sequence; An autoencoder model is constructed, and the baseline PM concentration change sequences corresponding to different regions are used for training. The features are extracted by the random forest algorithm, and the extracted features are divided into feature subsets using hierarchical clustering. The feature subsets are mapped to the autoencoder for feature reconstruction. Calculate the root mean square value of each feature subset, calculate the anomaly score according to the root mean square value, test the trained autoencoder model using test data, and import the PM concentration change sequence corresponding to different regions into the trained autoencoder network; The anomaly score is calculated based on the root mean square value of feature reconstruction, and the dust anomaly in different areas is judged using the Softmax function.

[0008] In this solution, the multimodal sensing data is marked according to the dust anomaly and transmitted to the data processing center quickly and preferentially, specifically: When abnormal dust conditions are identified, the multimodal sensing data corresponding to the PM concentration is marked as "emergency pollution warning data" and transmitted to the data processing center as soon as possible; Positioning is performed based on the dust anomaly, and the surrounding dust reduction equipment is selected based on the positioning information. The feature subset corresponding to the PM concentration change sequence of the dust anomaly is used as the anomaly feature. Based on the call instance data of the simple dust reduction strategy stored locally, the similarity between the anomaly feature and each call instance data is calculated. Acquire call instance data whose similarity meets the preset standards, count the call frequency of the simple dust reduction strategy in the filtered call instance data, and autonomously control the surrounding dust reduction equipment to perform preliminary dust reduction operations based on the simple dust reduction strategy with the highest call frequency.

[0009] In this solution, a dynamic pollution diffusion model is constructed to analyze the current pollution scenario of the construction site, specifically: Obtain multimodal sensing data corresponding to PM concentration at the construction site, dust suppression equipment operation data, construction activity data, site geological condition data, surrounding traffic flow data, and environment-related data as a multi-source data sequence; Based on the multi-source data series of different areas of the construction site, the dynamic time warping algorithm is used to quantitatively judge the correlation between different areas. Different areas of the construction site are taken as nodes, and the correlation between different areas is used to build an edge structure to generate a correlation graph corresponding to the construction area. A dynamic pollution diffusion model is constructed based on the graph attention network and the gated recurrent unit, and the multi-source data sequence corresponding to the dust anomaly is imported into the dynamic pollution diffusion model to obtain the adjacency matrix of the nodes in the dust anomaly area, and the adjacency matrix is ​​imported into the graph attention network for representation learning; Graph convolution is used for information propagation, and a multi-head attention mechanism is introduced. In the attention layer, the feature representation of the regional nodes is output according to the weight of the regional nodes. The feature representation of the regional nodes output by multiple attention heads is spliced ​​and integrated through the average operation in the last layer to achieve the neighbor aggregation of the regional nodes and obtain the updated feature representation of the regional nodes of the dust anomaly. The node feature representation of the dust abnormality area is imported into the gated recurrent unit to obtain the feature time dependency, and the spatiotemporal characteristics of the PM concentration change in the dust abnormality area are output through the fully connected layer. According to the spatiotemporal characteristics of the PM concentration change in the dust abnormality area, the dust generation trend is obtained to characterize the current pollution scene.

[0010] In this solution, the best dust reduction strategy is matched for the current pollution scene, specifically: Predicting PM concentration distribution based on the current pollution scene of the construction site, updating the dust abnormality area in the construction site according to the predicted PM concentration distribution, and obtaining dust reduction equipment within a preset distance according to the updated dust abnormality area, wherein the dust reduction equipment includes a sprinkler, a fog cannon, and a drone equipped with a spray system; Read historical dust reduction instances according to the dust reduction equipment category, select Euclidean distance, Manhattan distance, Fréchet distance and Jaccard distance to calculate four distances between historical dust reduction instances and the current working conditions of the dust reduction equipment, and obtain distance feature vectors; The Stacking strategy is introduced to build a dust reduction strategy recommendation model. The KNN model is trained separately based on each distance feature vector, and a preset number of strategy candidate sets are output. The candidate strategies are scored based on the historical dust reduction effects. The strategy candidate sets and scores corresponding to different distances are spliced ​​into a meta-feature vector, and the meta-feature vector is used to train the neural network model, learn the weights of each distance metric, and iteratively train and output the dust reduction strategy recommendation model; The current pollution scenario is obtained according to the updated dust abnormality area, and the current pollution scenario is imported into the dust reduction strategy recommendation model for strategy screening and scoring to obtain the best dust reduction strategy.

[0011] In this scheme, a self-organizing communication network is constructed between dust suppression equipment, and the dust suppression feedback information of dust suppression equipment is used to adjust the dust suppression operation plan. Specifically: A self-organizing communication network is built between dust suppression equipment through data blockchain. When a dust suppression equipment finds that the dust concentration in its coverage area exceeds the preset concentration threshold, it sends a help signal to the surrounding dust suppression equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan to coordinate dust suppression. During the operation of the dust reduction equipment, the data collected by the dust reduction equipment is stored on the chain, providing a reliable data source for pollution diffusion analysis and real-time monitoring of the working status of the dust reduction equipment. When the dust reduction equipment is in an abnormal working state, the dust concentration in the area covered by the abnormal dust reduction equipment itself is obtained. When it exceeds the preset concentration threshold, a help signal is sent to the surrounding dust reduction equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan.

[0012] The second aspect of the present invention provides a construction site dust reduction and pollution prevention and control system based on multimodal perception, the system comprising a data acquisition module, an edge computing module, a data processing center module, and a dust reduction equipment management module; The data collection module collects PM data at the boundary in real time through the boundary monitoring equipment preset at the boundary of the construction site, and uses the drone equipped with the monitoring equipment to perform flight monitoring inside the construction site according to the preset route to collect PM data from different areas of the site; The edge computing module integrates PM data and environment-related data to generate multimodal perception data, performs preliminary analysis on the multimodal perception data, identifies dust anomalies, and marks the multimodal perception data; The data processing center module analyzes the current pollution scene of the construction site according to the dust generation characteristics of different construction stages and multimodal perception data, and matches the best dust reduction strategy for the current pollution scene. The dust reduction equipment management module generates a dust reduction equipment control signal based on the optimal dust reduction strategy, controls the surrounding dust reduction equipment to perform dust reduction operations, builds a self-organizing communication network between dust reduction equipment, and uses the dust reduction feedback information of the dust reduction equipment to adjust the dust reduction operation plan.

[0013] Compared with the prior art, the present invention has the following beneficial effects: The present invention realizes comprehensive, real-time and accurate dust concentration monitoring of construction sites and timely grasps pollution conditions through multimodal data collection of boundary monitoring equipment and drones, combined with preliminary analysis of edge computing and abnormal data processing. Intelligent coordination is achieved for the factory boundary spraying system, the sprinkler truck spraying system, the fog cannon and the spraying system carried by drones, and the dust reduction operation is accurately started and adjusted according to the change of dust concentration to improve the dust reduction efficiency. By using intelligent adaptive spraying technology, drone group collaboration, flexible dust reduction strategy and self-organizing network communication between devices, the dust reduction effect can be further optimized, the water resource utilization efficiency can be improved, and the dust reduction range can be expanded.

[0014] When the dust concentration exceeds the standard, the dust reduction equipment is started in time, and stopped in time after reaching the standard, to avoid excessive spraying and equipment idling, to achieve energy saving, and to effectively prevent and control secondary pollution, to create a dust-free construction site. Historical data learning and prediction models are used to achieve preventive dust reduction, and strategies are adjusted in combination with weather changes to further improve the dust reduction effect. The entire dust reduction process is automatically analyzed and decided by the data processing center based on monitoring data, and equipment operation is controlled to reduce manual intervention and improve management efficiency. Multi-system linkage and data sharing optimize the overall management process of the construction site, and blockchain technology ensures data security and the standardization of equipment maintenance management. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments or exemplary embodiments of the present invention, the drawings required for use in the embodiments or exemplary descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained according to the drawings without paying creative work.

[0016] Figure 1 A flow chart of a construction site dust reduction and pollution prevention and control method based on multimodal perception is shown; Figure 2 A flow chart for preliminary analysis of multimodal sensing data to identify dust anomalies is shown; Figure 3 A flow chart is shown for matching the best dust reduction strategy for the current pollution scenario; Figure 4 A block diagram of a construction site dust reduction and pollution control system based on multimodal perception is shown. DETAILED DESCRIPTION

[0017] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0018] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0019] Figure 1 A flow chart of a construction site dust reduction and pollution control method based on multimodal perception is shown.

[0020] like Figure 1 As shown, this embodiment provides a construction site dust reduction and pollution prevention and control method based on multimodal perception, including: S102, using the boundary monitoring equipment preset at the boundary of the construction site to collect PM data at the boundary in real time, and using a drone equipped with the monitoring equipment to perform flight monitoring inside the construction site according to a preset route to collect PM data from different areas of the site; S104, combining the PM data obtained by monitoring with the environment-related data obtained by the temperature and humidity sensor to generate multimodal sensing data, and using the edge computing module to perform preliminary analysis on the multimodal sensing data in real time to identify dust abnormalities; S106, marking the multimodal sensing data according to the dust abnormality, and transmitting it to the data processing center in a priority and rapid manner, building a dynamic pollution diffusion model to analyze the current pollution scene of the construction site, and matching the best dust reduction strategy for the current pollution scene; S108, generating a dust reduction device control signal based on the optimal dust reduction strategy, controlling the surrounding dust reduction devices to perform dust reduction operations, building a self-organizing communication network between the dust reduction devices, and using the dust reduction feedback information of the dust reduction devices to adjust the dust reduction operation plan.

[0021] It should be noted that boundary monitoring equipment is deployed at preset intervals at the construction site boundary, equipped with high-precision PM sensors. The deployed boundary monitoring equipment collects PM data at the boundary based on PM indicators, including PM2.5, PM10 and other indicators. The deployment interval of the boundary monitoring equipment is no greater than the effective monitoring radius of the PM concentration of a single device × the coverage factor (the coverage factor ranges from 0.8 to 1.0, and is dynamically adjusted according to the fluctuation amplitude of the historical dust concentration of the site). Referring to the requirement of "the interval between fixed monitoring equipment shall not exceed 50 meters" in the "Technical Standard for Dust Monitoring in Construction" (JGJ / T 393-2016), combined with the distribution density of dust pollution sources on this site, the formula is used to calculate: interval distance = reference distance × correction factor, where the correction factor is determined based on parameters such as the length-to-width ratio of the site and the frequency of the dominant wind direction. Deploy drones equipped with monitoring equipment, conduct flight tests and calibration on the drones, set flight routes to ensure that they can fly stably and collect data accurately, conduct flight monitoring inside the construction site according to the preset routes, and collect PM data inside the construction site. In addition to PM sensors, drone monitoring equipment can also be equipped with temperature and humidity sensors to obtain environmental-related data for auxiliary analysis. The drone route can be adaptively adjusted according to real-time data. The initial preset route is a grid traversal route covering the entire site (such as an S-shaped path). When the edge computing module detects that the PM concentration of a grid block exceeds the threshold by 1.5 times, the route optimization algorithm is triggered: the route density in the area is increased, the route spacing is reduced to 60% of the original spacing, the number of circling circles is increased, and the single monitoring time is extended by 20%~30%.

[0022] The construction site is gridded to obtain construction data inside the construction site, the construction data is used to mark the grid blocks, the grid blocks are clustered according to the marking information, and the construction site is divided into regions using the clustering results; the construction content and construction stage are identified according to the marking information of the region, the construction content and construction stage are used to identify key construction areas, and attention feedback is added to the key construction areas to the drone for intensive monitoring and attention; in different areas, multimodal perception data corresponding to the PM concentration is generated according to the PM data at the boundary and the PM data inside, and the multimodal perception data is preprocessed.

[0023] Figure 2 A flow chart for preliminary analysis of multimodal sensing data to identify dust anomalies is shown.

[0024] According to an embodiment of the present invention, an edge computing module is used to perform a preliminary analysis on the multimodal sensing data in real time to identify dust abnormalities, specifically: S202, integrating an edge computing module on the boundary monitoring equipment and the drone at the construction site, importing the multimodal sensing data into the edge computing module for preliminary analysis, and obtaining PM concentration change sequences corresponding to different areas; S204, searching for construction examples by similarity calculation based on construction contents and construction stages in different areas, obtaining historical dust monitoring data in the construction examples, performing data preprocessing and averaging on the historical dust monitoring data, and obtaining a baseline PM concentration change sequence; S206, constructing an autoencoder model, using the baseline PM concentration change sequence corresponding to different regions for training, extracting features using a random forest algorithm, dividing the extracted features into feature subsets using hierarchical clustering, mapping the feature subsets to the autoencoder, and performing feature reconstruction; S208, calculating the root mean square value of each feature subset, calculating the anomaly score according to the root mean square value, testing the trained autoencoder model using the test data, and importing the PM concentration change sequence corresponding to different regions into the trained autoencoder network; S210, calculating the anomaly score according to the root mean square value of the feature reconstruction, and judging the dust anomaly in different areas through the Softmax function.

[0025] It should be noted that the autoencoder model is constructed, and the autoencoder model consists of two layers of autoencoders with the same network structure. The first layer of autoencoders is composed of autoencoders with the same structure in parallel, which are used to reconstruct feature subsets. The second layer of autoencoders calculates the anomaly score according to the root mean square value of the feature reconstruction for anomaly detection. Each tree in the random forest randomly selects m features from the baseline PM concentration change sequence corresponding to different regions to build a tree, calculates the score according to the importance of each feature and arranges them in descending order, selects a preset number of features for hierarchical clustering, and obtains a feature subset with strong correlation. The PM concentration change sequence corresponding to different regions is imported into the trained autoencoder network to obtain the estimated value of the PM concentration change sequence corresponding to different regions under normal conditions. The estimated value is judged by comparing the root mean square representation difference between the estimated value and the actual value. If the root mean square value increases sharply and deviates from the normal value, it is judged as abnormal.

[0026] It should be noted that the data processing center is built and the hardware and software environment required for the data analysis algorithm is configured to ensure that the data processing center can efficiently process and analyze the received data. At the same time, a stable communication link is established between the data processing center and the border monitoring equipment, drones, and dust reduction equipment, blockchain nodes are deployed, and interfaces with other management systems and meteorological systems are integrated. The equipment operating status and communication links are regularly checked to ensure the stability and accuracy of data collection.

[0027] When an abnormal dust situation is identified, the multimodal perception data corresponding to the PM concentration will be marked as "emergency pollution warning data" and transmitted to the data processing center quickly and preferentially; the abnormal dust situation will be positioned according to the abnormal dust situation, and the surrounding dust reduction equipment will be selected based on the positioning information; the feature subset corresponding to the PM concentration change sequence of the abnormal dust situation will be used as the abnormal feature, and the similarity between the abnormal feature and each calling instance data will be calculated based on the calling instance data of the simple dust reduction strategy stored locally; the calling instance data whose similarity meets the preset standard will be obtained, and the calling frequency of the simple dust reduction strategy will be counted in the screened calling instance data, and the surrounding dust reduction equipment will be autonomously controlled to perform preliminary dust reduction operations according to the simple dust reduction strategy with the highest calling frequency.

[0028] It should be noted that the data processing center analyzes the collected PM data and environmental data, and determines the trend of dust concentration changes based on the specific optimization of the complex working conditions of the construction site and the time and space factors. According to the environmental requirements and relevant standards of the construction site, a reasonable preset value of dust concentration is set. The historical data learning and prediction model is used to combine multi-source data such as construction process, weather, geology and traffic to predict the trend of dust generation.

[0029] The multimodal perception data corresponding to the PM concentration of the construction site, the operation data of the dust reduction equipment, the construction activity data, the site geological condition data, the surrounding traffic flow data and the environment-related data (temperature and humidity data, wind direction data and weather data, etc.) are obtained as multi-source data sequences. The dynamic time warping algorithm is used to quantitatively judge the correlation between different areas based on the multi-source data sequences of different areas of the construction site. The dynamic time warping algorithm calculates the distance between the aligned sequences, which can intuitively represent the similarity between the multi-source data sequences of the two areas, and thereby represent the correlation between the areas.

[0030] Different areas of the construction site are taken as nodes, and the correlation between different areas is used to construct an edge structure to generate a correlation graph corresponding to the construction area; a dynamic pollution diffusion model is constructed based on the graph attention network and the gated recurrent unit, and the multi-source data sequence corresponding to the dust abnormality is imported into the dynamic pollution diffusion model to obtain the adjacency matrix of the nodes in the dust abnormality area, and the adjacency matrix is ​​imported into the graph attention network for representation learning; graph convolution is used for information propagation, and a multi-head attention mechanism is introduced to achieve parallel computing through multiple attention heads to accelerate the calculation speed of the model. In the attention layer, the feature representation of the regional node is output according to the weight of the regional node, and multiple The regional node feature representations output by the attention head are spliced, and the splicing and integration are performed through the average operation in the last layer to realize the neighbor aggregation of the regional nodes, and obtain the updated regional node feature representation of the dust abnormality; the regional node feature representation of the dust abnormality is imported into the gated recurrent unit, and the feature time dependency is obtained by using the hidden state update. The gated recurrent unit overcomes the problems of gradient vanishing and gradient explosion in complex networks, and outputs the spatiotemporal characteristics of the PM concentration change in the dust abnormality area through the fully connected layer. According to the spatiotemporal characteristics of the PM concentration change in the dust abnormality area, the dust generation trend is obtained to characterize the current pollution scene. Set the dust concentration preset value. When the PM data monitored for a continuous period of time (such as 15 minutes) exceeds the preset value, the data processing center determines that dust reduction measures need to be initiated, and generates corresponding dust reduction instructions according to the area and degree of dust concentration exceeding the standard.

[0031] Figure 3 A flow chart for matching the best dust reduction strategy for the current pollution scenario is shown.

[0032] According to an embodiment of the present invention, the best dust reduction strategy is matched for the current pollution scene, specifically: S302, predicting PM concentration distribution based on the current pollution scene of the construction site, updating the dust abnormality area in the construction site according to the predicted PM concentration distribution, and acquiring dust reduction equipment that meets the preset distance according to the updated dust abnormality area, wherein the dust reduction equipment includes a sprinkler, a fog cannon, and a drone equipped with a spray system; S304, reading historical dust reduction instances according to the dust reduction equipment category, selecting Euclidean distance, Manhattan distance, Fréchet distance and Jaccard distance to respectively calculate four distances between the historical dust reduction instances and the current working condition of the dust reduction equipment, and obtaining a distance feature vector; S306, introducing the Stacking strategy to build a dust reduction strategy recommendation model, training a KNN model based on each distance feature vector separately, outputting a preset number of strategy candidate sets, and scoring the candidate strategies based on historical dust reduction effects; S308, splicing the strategy candidate sets and scores corresponding to different distances into a meta-feature vector, using the meta-feature vector to train a neural network model, learning the weights of each distance metric, and iteratively training to output a dust suppression strategy recommendation model; S310, obtaining the current pollution scene according to the updated dust abnormality area, importing the current pollution scene into the dust reduction strategy recommendation model for strategy screening and scoring, and obtaining the best dust reduction strategy.

[0033] It should be noted that historical dust reduction instances usually contain multi-dimensional environmental data (such as PM2.5, humidity, wind speed) and equipment operating parameters (such as fog cannon angle, water consumption). Stacking integration improves the robustness of strategy selection, which is suitable for intelligent decision-making on dust reduction in dynamic construction environments. After executing the selected optimal dust reduction strategy, the actual dust reduction effect is recorded and the instance library is updated to form a closed-loop learning. In the dust reduction strategy recommendation model, the contribution of each distance under specific working conditions is analyzed through a neural network model (such as the Fréchet distance is more important in strong winds), and the KNN network input is dynamically optimized.

[0034] After receiving the dust reduction command, the sprinkler, fog cannon and drone equipped with the spray system start working according to the command requirements. According to different construction stages, the operation mode and parameters of the dust reduction equipment are switched, and the size and shape of the spray droplets are intelligently adjusted to accurately match different pollution scenarios. For example, a large flow and low pressure spraying mode is used in the foundation construction stage, and a small flow and high pressure spraying mode is used in the main construction stage. The sprinkler adjusts the spraying flow and range according to the environmental perception matrix and the intelligent flow regulating device, and evenly sprays around the factory boundary. The fog cannon adjusts the angle and intensity according to the command to carry out targeted dust reduction in key areas. The drone flies to the designated area, turns on the spray system, and performs dust reduction operations in the air. At the same time, multiple drones work together to automatically plan the trajectory and spraying area according to the pollution area to form a collaborative operation network. The drone group exchanges real-time information through self-organizing network communication technology and dynamically adjusts the operation strategy according to the changes in dust concentration.

[0035] The concept of flexible dust reduction is introduced, and the operation mode and parameters of the dust reduction equipment are dynamically adjusted according to the dust generation characteristics and environmental conditions at different construction stages. In the foundation construction stage, the dust reduction equipment adopts a large flow, low pressure spray mode; in the main construction stage, it switches to a small flow, high pressure spray mode. In addition, microbial flora is added to the spray liquid, and intelligent dust reduction materials are used to adsorb and fix dust, converting organic dust into harmless substances, broadening the scope of application of the dust reduction system. Intelligent dust reduction materials have the characteristics of hygroscopic expansion, which can adsorb dust when the dust concentration is high, fix dust after the concentration is reduced, and reduce secondary dust.

[0036] During the dust reduction process, boundary monitoring equipment and drones continuously monitor PM data. The data processing center receives monitoring data in real time and compares it with the preset value. When the dust concentration index returns to the normal range and remains within it for a period of time, the data processing center sends a stop-work instruction to the dust reduction equipment. After receiving the instruction, the dust reduction equipment stops running to avoid unnecessary energy consumption and waste of resources. Integrate with other management systems of the construction site (such as the safety monitoring system and the construction progress management system) to obtain data from other systems and optimize the dust reduction plan. At the same time, establish a real-time connection with the weather forecast system of the meteorological department to adjust the dust reduction strategy in advance according to weather changes.

[0037] It should be noted that blockchain technology, due to its decentralization, data immutability and automatic execution of smart contracts, provides a new solution for building a self-organizing communication network between dust reduction devices. By combining blockchain with the Internet of Things (IoT), distributed decision-making and collaborative optimization of dust reduction equipment can be achieved. Each dust reduction device, as a blockchain network node, achieves data synchronization and decision-making collaboration through a consensus mechanism (such as PoA or PBFT) without relying on a central server, thereby improving the robustness and scalability of the system. Dust reduction strategies are set based on smart contracts. For example, when PM10 exceeds the standard in a certain area, nearby devices automatically negotiate the optimal dust reduction solution to avoid energy waste caused by multiple devices running at high power at the same time. Dynamically adjust the contract logic in combination with environmental data to achieve adaptive control. Blockchain encryption mechanisms (such as asymmetric encryption and hash verification) ensure the security of communication between dust reduction devices, prevent malicious nodes from forging data or launching attacks, and the identity of dust reduction devices is authenticated by digital certificates to ensure that only legitimate devices are allowed to access the network.

[0038] A self-organizing communication network between dust reduction equipment is constructed through data blockchain. Preferably, the help signal trigger mechanism based on smart contracts realizes communication between equipment, and uses hash algorithms (such as SHA-256) and storage structures to upload data to the chain to ensure data integrity. When the dust reduction equipment finds that the dust concentration in its own coverage area exceeds the preset concentration threshold, it sends a help signal to the surrounding dust reduction equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan to coordinate dust reduction; during the operation of the dust reduction equipment, the data collected by the dust reduction equipment is stored on the chain, providing a trusted data source for pollution diffusion analysis, and monitoring the working status of the dust reduction equipment in real time. When the dust reduction equipment is in an abnormal working state, the dust concentration in the area covered by the abnormal dust reduction equipment itself is obtained. When it exceeds the preset concentration threshold, a help signal is sent to the surrounding dust reduction equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan.

[0039] Regularly maintain and inspect the entire system, including edge computing modules, data processing centers, blockchain nodes, self-organizing network communication modules, and interfaces with other management systems and meteorological systems, to ensure that the system can continue to operate stably. Use blockchain smart contracts to manage equipment maintenance to ensure the timeliness and standardization of equipment maintenance. Record the equipment's maintenance cycle, maintenance content, and maintenance responsible party. When the equipment's operating time reaches the maintenance cycle, maintenance reminders are automatically triggered; after maintenance is completed, the maintenance records are uploaded to the blockchain for verification and confirmation.

[0040] Figure 4 A block diagram of a construction site dust reduction and pollution control system based on multimodal perception is shown.

[0041] The second embodiment of the present invention provides a construction site dust reduction and pollution prevention and control system 4 based on multimodal perception, which includes a data acquisition module 401, an edge computing module 402, a data processing center module 403, and a dust reduction equipment management module 404; The data collection module collects PM data at the boundary in real time through the boundary monitoring equipment preset at the boundary of the construction site, and uses the drone equipped with the monitoring equipment to perform flight monitoring inside the construction site according to the preset route to collect PM data from different areas of the site; The edge computing module integrates PM data and environment-related data to generate multimodal perception data, performs preliminary analysis on the multimodal perception data, identifies dust anomalies, and marks the multimodal perception data; The data processing center module analyzes the current pollution scene of the construction site according to the dust generation characteristics of different construction stages and multimodal perception data, and matches the best dust reduction strategy for the current pollution scene. The dust reduction equipment management module generates a dust reduction equipment control signal based on the optimal dust reduction strategy, controls the surrounding dust reduction equipment to perform dust reduction operations, builds a self-organizing communication network between dust reduction equipment, and uses the dust reduction feedback information of the dust reduction equipment to adjust the dust reduction operation plan.

[0042] The third aspect of the present invention provides a computer-readable storage medium, which includes a construction site dust reduction and pollution control method program based on multimodal perception. When the construction site dust reduction and pollution control method program based on multimodal perception is executed by a processor, the steps of the construction site dust reduction and pollution control method based on multimodal perception are implemented.

[0043] In the several embodiments provided in the present application, it should be understood that the disclosed methods and systems can be implemented in other ways. The system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as: multiple modules or components can be combined, or can be integrated into another system, or some features can be ignored, or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms. In addition, the functional modules in the various embodiments of the present invention can be all integrated into one processing module, or each module can be a separate module, or two or more modules can be integrated into one module; the above-mentioned integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0044] A person of ordinary skill in the art can understand that: all or part of the steps of implementing the above method embodiment can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiment; and the aforementioned storage medium includes: a mobile storage device, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0045] Alternatively, if the above-mentioned integrated module of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0046] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for dust reduction and pollution prevention and control at a construction site based on multimodal perception, characterized in that: The following steps are involved: The PM data at the construction site boundary is collected in real time through the preset boundary monitoring equipment. At the same time, the drone equipped with the monitoring equipment is used to conduct flight monitoring inside the construction site according to the preset route to collect PM data from different areas of the site. The PM data obtained by monitoring is combined with the environment-related data obtained by the temperature and humidity sensor to generate multimodal perception data, and the edge computing module is used to perform preliminary analysis on the multimodal perception data in real time to identify abnormal dust conditions; The multimodal sensing data is marked according to the dust anomaly, and is transmitted to the data processing center in a priority and rapid manner, a dynamic pollution diffusion model is constructed to analyze the current pollution scene of the construction site, and the best dust reduction strategy is matched for the current pollution scene; Based on the optimal dust reduction strategy, a dust reduction equipment control signal is generated to control the surrounding dust reduction equipment to perform dust reduction operations, a self-organizing communication network is constructed between the dust reduction equipment, and the dust reduction feedback information of the dust reduction equipment is used to adjust the dust reduction operation plan.

2. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: Collect PM data at the construction site, specifically: Deploy boundary monitoring equipment at preset intervals at the construction site boundary, collect PM data at the boundary based on PM indicators through the deployed boundary monitoring equipment, deploy drones equipped with monitoring equipment, conduct flight monitoring inside the construction site according to preset routes, and collect PM data and environment-related data inside the construction site; Performing grid processing on the construction site, obtaining construction data inside the construction site, marking grid blocks using the construction data, clustering the grid blocks according to the marking information, and dividing the construction site into regions using the clustering results; Identify the construction content and construction stage according to the marking information of the area, use the construction content and construction stage to identify the key construction area, and add attention feedback to the drone for intensive monitoring and attention of the key construction area; In different areas, multimodal sensing data corresponding to PM concentration is generated according to PM data at the boundary and PM data inside the area, and the multimodal sensing data is preprocessed.

3. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: Use the edge computing module to perform preliminary analysis on the multimodal sensing data in real time to identify abnormal dust conditions, specifically: Integrate an edge computing module on the boundary monitoring equipment and drones at the construction site, import the multimodal sensing data into the edge computing module for preliminary analysis, and obtain PM concentration change sequences corresponding to different areas; Retrieving construction examples by similarity calculation based on the construction content and construction stage in different areas, obtaining historical dust monitoring data in the construction examples, performing data preprocessing and averaging on the historical dust monitoring data, and obtaining a baseline PM concentration change sequence; An autoencoder model is constructed, and the baseline PM concentration change sequences corresponding to different regions are used for training. The features are extracted by the random forest algorithm, and the extracted features are divided into feature subsets using hierarchical clustering. The feature subsets are mapped to the autoencoder for feature reconstruction. Calculate the root mean square value of each feature subset, calculate the anomaly score according to the root mean square value, test the trained autoencoder model using test data, and import the PM concentration change sequence corresponding to different regions into the trained autoencoder network; The anomaly score is calculated based on the root mean square value of feature reconstruction, and the dust anomaly in different areas is judged using the Softmax function.

4. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: The multimodal sensing data is marked according to the dust anomaly and transmitted to the data processing center quickly and preferentially, specifically: When abnormal dust conditions are identified, the multimodal sensing data corresponding to the PM concentration is marked as "urgent pollution warning data" and transmitted to the data processing center as soon as possible; Positioning is performed based on the dust anomaly, and the surrounding dust reduction equipment is selected based on the positioning information. The feature subset corresponding to the PM concentration change sequence of the dust anomaly is used as the anomaly feature. Based on the call instance data of the simple dust reduction strategy stored locally, the similarity between the anomaly feature and each call instance data is calculated. Acquire call instance data whose similarity meets the preset standards, count the call frequency of the simple dust reduction strategy in the filtered call instance data, and autonomously control the surrounding dust reduction equipment to perform preliminary dust reduction operations based on the simple dust reduction strategy with the highest call frequency.

5. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: A dynamic pollution diffusion model is constructed to analyze the current pollution scenario of the construction site, specifically: Obtain multimodal sensing data corresponding to PM concentration at the construction site, dust suppression equipment operation data, construction activity data, site geological condition data, surrounding traffic flow data, and environment-related data as a multi-source data sequence; Based on the multi-source data series of different areas of the construction site, the dynamic time warping algorithm is used to quantitatively judge the correlation between different areas. Different areas of the construction site are taken as nodes, and the correlation between different areas is used to build an edge structure to generate a correlation graph corresponding to the construction area. A dynamic pollution diffusion model is constructed based on the graph attention network and the gated recurrent unit, and the multi-source data sequence corresponding to the dust anomaly is imported into the dynamic pollution diffusion model to obtain the adjacency matrix of the nodes in the dust anomaly area, and the adjacency matrix is ​​imported into the graph attention network for representation learning; Graph convolution is used for information propagation, and a multi-head attention mechanism is introduced. In the attention layer, the feature representation of the regional nodes is output according to the weight of the regional nodes. The feature representation of the regional nodes output by multiple attention heads is spliced ​​and integrated through the average operation in the last layer to achieve the neighbor aggregation of the regional nodes and obtain the updated feature representation of the regional nodes of the dust anomaly. The node feature representation of the dust abnormality area is imported into the gated recurrent unit to obtain the feature time dependency, and the spatiotemporal characteristics of the PM concentration change in the dust abnormality area are output through the fully connected layer. According to the spatiotemporal characteristics of the PM concentration change in the dust abnormality area, the dust generation trend is obtained to characterize the current pollution scene.

6. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: Match the best dust reduction strategy for the current pollution scene, specifically: Predicting PM concentration distribution based on the current pollution scene of the construction site, updating the dust abnormality area in the construction site according to the predicted PM concentration distribution, and obtaining dust reduction equipment within a preset distance according to the updated dust abnormality area, wherein the dust reduction equipment includes a sprinkler, a fog cannon, and a drone equipped with a spray system; Read historical dust reduction instances according to the dust reduction equipment category, select Euclidean distance, Manhattan distance, Fréchet distance and Jaccard distance to calculate four distances between historical dust reduction instances and the current working conditions of the dust reduction equipment, and obtain distance feature vectors; The Stacking strategy is introduced to build a dust reduction strategy recommendation model. The KNN model is trained separately based on each distance feature vector, and a preset number of strategy candidate sets are output. The candidate strategies are scored based on the historical dust reduction effects. The strategy candidate sets and scores corresponding to different distances are spliced ​​into a meta-feature vector, and the meta-feature vector is used to train the neural network model, learn the weights of each distance metric, and iteratively train and output the dust reduction strategy recommendation model; The current pollution scenario is obtained according to the updated dust abnormality area, and the current pollution scenario is imported into the dust reduction strategy recommendation model for strategy screening and scoring to obtain the best dust reduction strategy.

7. The method for dust reduction and pollution prevention and control at a construction site based on multimodal perception according to claim 1 is characterized in that: Build a self-organizing communication network between dust suppression equipment, and use the dust suppression feedback information of dust suppression equipment to adjust the dust suppression operation plan, specifically: A self-organizing communication network is built between dust suppression equipment through data blockchain. When a dust suppression equipment finds that the dust concentration in its coverage area exceeds the preset concentration threshold, it sends a help signal to the surrounding dust suppression equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan to coordinate dust suppression. During the operation of the dust reduction equipment, the data collected by the dust reduction equipment is stored on the chain, providing a reliable data source for pollution diffusion analysis and real-time monitoring of the working status of the dust reduction equipment. When the dust reduction equipment is in an abnormal working state, the dust concentration in the area covered by the abnormal dust reduction equipment itself is obtained. When it exceeds the preset concentration threshold, a help signal is sent to the surrounding dust reduction equipment through the self-organizing communication network, and the surrounding equipment automatically adjusts the operation plan.

8. A construction site dust reduction and pollution control system based on multimodal perception, characterized in that: Implementing the construction site dust reduction and pollution prevention and control method based on multimodal perception as described in any one of claims 1 to 7, the system includes a data acquisition module, an edge computing module, a data processing center module, and a dust reduction equipment management module; The data collection module collects PM data at the boundary in real time through the boundary monitoring equipment preset at the boundary of the construction site, and uses the drone equipped with the monitoring equipment to perform flight monitoring inside the construction site according to the preset route to collect PM data from different areas of the site; The edge computing module integrates PM data and environment-related data to generate multimodal perception data, performs preliminary analysis on the multimodal perception data, identifies dust anomalies, and marks the multimodal perception data; The data processing center module analyzes the current pollution scene of the construction site according to the dust generation characteristics of different construction stages and multimodal perception data, and matches the best dust reduction strategy for the current pollution scene. The dust reduction equipment management module generates a dust reduction equipment control signal based on the optimal dust reduction strategy, controls the surrounding dust reduction equipment to perform dust reduction operations, builds a self-organizing communication network between dust reduction equipment, and uses the dust reduction feedback information of the dust reduction equipment to adjust the dust reduction operation plan.

Images