An AI vision recognition-based building dust intelligent monitoring and suppression system and method

Abstract

The application provides a kind of building dust intelligent monitoring and suppression system and method based on AI vision recognition, related to building construction environmental protection technical field.The method synchronously collects dust scene image data and multi-dimensional environmental parameters through visual collection equipment and environmental sensors deployed in the building construction area;Based on the improved YOLOv8 algorithm and optical flow method, the static visual features and dynamic visual features of dust are extracted;A multi-source data fusion dust concentration estimation model is constructed, and the dust concentration is accurately calculated combined with the environmental parameters;Through the LSTM network enhanced by attention mechanism, the diffusion trend of dust is predicted;Based on the concentration level and diffusion trend, the intelligent suppression strategy of dynamic partition and grading is generated, and the closed-loop suppression control of the linkage spray device, dust reduction system and transport vehicle is carried out.The application breaks through the limitation of traditional single sensor monitoring, realizes real-time, accurate and intelligent monitoring and suppression of dust, and improves the efficiency and environmental protection standard rate of dust control in building construction.

Description

Technical Field

[0001] This invention relates to the field of environmental protection technology in building construction, specifically to an intelligent monitoring and suppression system and method for building dust based on AI visual recognition. Background Technology

[0002] Construction dust is a major source of air pollution. The PM2.5 and PM10 particles it contains not only harm human health but also affect the surrounding ecological environment and residents' lives. Current methods for controlling construction dust mainly rely on traditional monitoring and manual suppression, which have many limitations.

[0003] The monitoring method is limited: Existing technologies mostly use fixed-point laser dust sensors to monitor dust concentration, which can only obtain single-point data and cannot reflect the spatial distribution and diffusion trend of dust. In addition, the sensors are easily affected by environmental humidity and vibration, resulting in a decrease in accuracy.

[0004] Insufficient recognition accuracy: Some visual recognition schemes only extract the static features of dust (such as grayscale and area), ignoring the dynamic characteristics of dust diffusion, making it difficult to distinguish dust from interference such as fog and haze, resulting in low recognition accuracy.

[0005] Delayed suppression strategies: Traditional suppression methods are mostly timed spraying or manual activation, which lack linkage with dust concentration and diffusion trend, resulting in problems of "over-suppression" or "under-suppression", causing waste of water resources or poor treatment effect;

[0006] Lack of closed-loop control: Monitoring data is disconnected from suppression operations, making it impossible to dynamically adjust suppression strategies according to real-time changes in dust, resulting in low governance efficiency.

[0007] Therefore, how to achieve comprehensive and high-precision monitoring of construction dust and generate intelligent and coordinated suppression strategies based on diffusion trends has become a pressing technical challenge in the field of environmental protection during construction. Summary of the Invention

[0008] The purpose of this invention is to provide an intelligent monitoring and suppression system and method for construction dust based on AI visual recognition, which aims to solve the problems of single monitoring, insufficient accuracy and delayed suppression in existing technologies, and to achieve real-time, accurate and intelligent management of dust.

[0009] To achieve the above objectives, the present invention is implemented through the following technical solution:

[0010] Firstly, a method for intelligent monitoring and suppression of construction dust based on AI visual recognition is provided, including the following steps:

[0011] Simultaneously collect image data of dust scenes in the construction area and multi-dimensional environmental parameters, including wind speed, relative humidity, temperature and baseline dust concentration;

[0012] The image data of the dust scene is preprocessed, and the static and dynamic visual features of the dust are extracted based on AI visual recognition algorithms.

[0013] A multi-source data fusion model is constructed, and the static visual features, dynamic visual features and multi-dimensional environmental parameters are input. The real-time dust concentration is calculated through weighted fusion and nonlinear mapping.

[0014] Based on the time series prediction model, combined with the real-time dust concentration, dynamic visual characteristics and environmental parameter change trends, the dust diffusion trend within a preset time window is predicted.

[0015] Based on the pollution level corresponding to the real-time dust concentration and the dust diffusion trend, a dynamic suppression strategy is generated, and the suppression execution equipment is linked to perform dust suppression operations, forming a closed-loop control.

[0016] As a further improvement to the technical solution of this invention, the extraction of static and dynamic visual features of dust specifically includes:

[0017] Superpixel segmentation is performed on the preprocessed dust scene image to filter candidate dust regions;

[0018] An improved HOG algorithm is used to extract the static visual features of the dust candidate region. The static visual features include gray-level mean, texture entropy, particle size equivalent area ratio, and gray-level gradient histogram features.

[0019] based on The optical flow method is used to calculate the pixel motion vectors of the dust candidate region and extract dynamic visual features, including dust diffusion speed, diffusion direction angle, particle size distribution dynamic variance, and motion consistency features.

[0020] As a further improvement to the technical solution of the present invention, the real-time dust concentration calculation process of the multi-source data fusion model satisfies:

[0021] ;

[0022] Where C is the real-time dust concentration (unit: ); , For visual feature weight coefficients, satisfying It is obtained through optimization using the gradient descent algorithm; Normalized fusion value of static visual features ( ); For dynamic visual feature normalized fusion value ( ); Environmental correction factor (range of values) ); Wind speed (unit: ); RH is relative humidity (unit: ); Sensor data calibration coefficients (range of values) ); Reference dust concentration (unit: (), collected by a laser dust sensor; For the error compensation term, satisfying .

[0023] As a further improvement to the technical solution of this invention, predicting the dust diffusion trend within a preset time window specifically includes:

[0024] A bidirectional LSTM network with enhanced attention mechanism is constructed to form a time-series dataset from the real-time dust concentration, dynamic visual features, and environmental parameters.

[0025] ;

[0026] The attention layer assigns weights to key features of the time series data. satisfy:

[0027] ;

[0028] in, Here is the attention weight matrix. For bias terms, This is the output of the hidden layer of the LSTM network. The cell state is defined; attention-weighted time-series data is input into a bidirectional LSTM network, which outputs a preset time window. ( Dust concentration prediction sequence within ) By combining diffusion speed and direction, a diffusion trend heatmap is generated.

[0029] As a further improvement to the technical solution of this invention, the dynamic suppression strategy specifically includes:

[0030] Preset dust concentration pollution level: Level 1 ( No need to suppress), secondary ( Mild inhibition), Grade III ( Moderate inhibition), Level IV ( (severe inhibition)

[0031] Based on the heat map of pollution level and diffusion trend, determine the suppression area, suppression intensity, and suppression method:

[0032] Mild suppression: Activate the fixed perimeter spray system, spray water pressure Spray frequency Second-rate;

[0033] Moderate suppression: Activate the fixed + mobile spray system, spray water pressure Spray frequency Next, the speed limit for joint transport vehicles. ;

[0034] Severe suppression: Activate the full-area spray system, spray water pressure Continuous spraying was initiated, and construction machinery and transport vehicles were suspended.

[0035] As a further improvement to the technical solution of the present invention, the preprocessing of the dust scene image data specifically includes:

[0036] Adaptive median filtering is used to remove salt-and-pepper noise from the image, with the filter window size satisfying the following:

[0037] ;

[0038] in, For pixels grayscale contrast;

[0039] pass The algorithm enhances the image, corrects uneven lighting, and improves the contrast between the dusty area and the background. It also uses image morphological opening operations (erosion followed by dilation) to remove small noise points while preserving the main dusty area.

[0040] As a further improvement to the technical solution of this invention, the AI ​​visual recognition algorithm is an improved YOLOv8 algorithm. By introducing a CBAM attention module into the Neck layer, the ability to extract dust features is enhanced. The channel attention output of the CBAM attention module satisfies the following:

[0041] ;

[0042] Spatial attention output satisfies: ;

[0043] in, The activation function is Sigmoid, and MLP is a multilayer perceptron. Convolution operation.

[0044] A second aspect of the present invention provides an intelligent monitoring and suppression system for building dust based on AI visual recognition, comprising:

[0045] The acquisition unit includes a high-definition infrared camera (resolution). Frame rate ), laser dust sensor, wind speed sensor, temperature and humidity sensor, used to simultaneously collect image data of dust scene and multi-dimensional environmental parameters;

[0046] The processing unit includes an edge computing gateway and a cloud server. The edge computing gateway is used for image preprocessing, feature extraction and real-time concentration calculation, and the cloud server is used for diffusion trend prediction, suppression strategy generation and model optimization.

[0047] The control unit is used to receive suppression strategy instructions from the processing unit and coordinate with the suppression execution device to perform operations.

[0048] The suppression execution unit includes a fixed fence spraying device, a mobile sprinkler spraying device, a construction machinery control module, and a transport vehicle linkage module;

[0049] Communication unit, adopts Dual-mode communication enables data transmission between the acquisition unit, processing unit, control unit, and suppression execution unit.

[0050] A third aspect of the present invention provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described intelligent monitoring and suppression method for building dust based on AI visual recognition.

[0051] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for intelligent monitoring and suppression of building dust based on AI visual recognition.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] This invention simultaneously collects image data and multi-dimensional environmental parameters of dust scenes in construction areas, and combines them with AI visual recognition algorithms to comprehensively extract the static and dynamic visual features of dust. This overcomes the limitations of traditional single sensors, which can only achieve single-point monitoring and cannot capture the spatial distribution and dynamic diffusion characteristics of dust. Furthermore, it accurately estimates the real-time dust concentration through a multi-source data fusion model, and predicts the dust diffusion trend in advance with an attention-enhanced time series prediction model, solving the problems of insufficient accuracy and lack of foresight in traditional monitoring. Finally, it generates a graded and zoned dynamic closed-loop suppression strategy based on concentration level and diffusion trend, and links various suppression execution devices to achieve precise control. This avoids the water waste and the drawbacks of "insufficient suppression" or "over-suppression" caused by traditional timed spraying, and significantly improves the real-time, accuracy, and intelligence level of construction dust control. It effectively improves the air pollution compliance rate, reduces labor and resource consumption, and provides efficient, economical, and sustainable technical support for environmental protection in construction. Attached Figure Description

[0054] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0055] Figure 1 is an exemplary flowchart of the intelligent monitoring and suppression method for building dust based on AI visual recognition in an embodiment of the present invention;

[0056] Figure 2 is a schematic diagram of the intelligent monitoring and suppression system for building dust based on AI visual recognition in an embodiment of the present invention;

[0057] Figure 3 is a schematic diagram of the computer device structure for implementing the method in an embodiment of the present invention. Detailed Implementation

[0058] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0059] The present invention will be further described in detail below with reference to the accompanying drawings.

[0060] Reference Figure 1 Firstly, a method for intelligent monitoring and suppression of construction dust based on AI visual recognition is provided, comprising the following steps:

[0061] Simultaneously collect image data of dust scenes in the construction area and multi-dimensional environmental parameters, including wind speed, relative humidity, temperature and baseline dust concentration;

[0062] The image data of the dust scene is preprocessed, and the static and dynamic visual features of the dust are extracted based on AI visual recognition algorithms.

[0063] A multi-source data fusion model is constructed, and the static visual features, dynamic visual features and multi-dimensional environmental parameters are input. The real-time dust concentration is calculated through weighted fusion and nonlinear mapping.

[0064] Based on the time series prediction model, combined with the real-time dust concentration, dynamic visual characteristics and environmental parameter change trends, the dust diffusion trend within a preset time window is predicted.

[0065] Based on the pollution level corresponding to the real-time dust concentration and the dust diffusion trend, a dynamic suppression strategy is generated, and the suppression execution equipment is linked to perform dust suppression operations, forming a closed-loop control.

[0066] In practice, the process begins by simultaneously collecting image data of dust pollution scenes and multi-dimensional environmental parameters from the construction area. The image data captures the visual characteristics of dust, while the multi-dimensional environmental parameters provide an environmental reference for dust concentration estimation. Next, an AI visual recognition algorithm extracts static and dynamic visual features of the dust from the pre-processed images, comprehensively depicting the dust's morphology and diffusion characteristics. Then, the static and dynamic visual features, along with the multi-dimensional environmental parameters, are input into a multi-source data fusion model. Weighted fusion and nonlinear mapping are used to calculate the real-time dust concentration. Based on a time-series prediction model, combined with the real-time dust concentration, dynamic visual features, and environmental parameter trends, the dust diffusion trend within a preset time window is predicted. Finally, based on the pollution level and diffusion trend corresponding to the real-time dust concentration, a dynamic suppression strategy is generated, triggering the suppression execution equipment to perform dust suppression operations, forming a closed-loop process of "collection-analysis-prediction-control." This invention overcomes the limitations of traditional single sensors, which can only achieve single-point monitoring and cannot capture the spatial distribution and dynamic diffusion characteristics of dust. It solves the problems of insufficient monitoring accuracy and lagging suppression strategies in traditional methods. By using multi-source data fusion and AI visual recognition technology, it improves the accuracy of dust concentration estimation. With the help of trend prediction, it realizes the forward-looking adjustment of suppression strategies. Finally, through hierarchical and zoned closed-loop suppression control, it avoids the drawbacks of water waste and inappropriate suppression, and significantly improves the real-time performance, accuracy and intelligence of construction dust control, effectively improving the air pollution compliance rate and reducing labor and resource consumption.

[0067] In some embodiments, extracting the static and dynamic visual features of dust specifically includes:

[0068] Superpixel segmentation is performed on the preprocessed dust scene image to filter candidate dust regions;

[0069] An improved HOG algorithm is used to extract the static visual features of the dust candidate region. The static visual features include gray-level mean, texture entropy, particle size equivalent area ratio, and gray-level gradient histogram features.

[0070] based on The optical flow method is used to calculate the pixel motion vectors of the dust candidate region and extract dynamic visual features, including dust diffusion speed, diffusion direction angle, particle size distribution dynamic variance, and motion consistency features.

[0071] In practice, the collected dust scene images are first segmented into superpixels, and candidate dust regions are selected through pixel clustering to eliminate irrelevant background interference. For the candidate dust regions, an improved HOG algorithm is used to extract static visual features such as gray-level mean, texture entropy, particle size equivalent area ratio, and gray-level gradient histogram. These features can reflect the static morphology and concentration-related attributes of dust. Simultaneously, based on... Optical flow is used to calculate pixel motion vectors in dust candidate regions. Vector clustering removes background noise, allowing for the extraction of dynamic visual features such as dust diffusion velocity, diffusion direction angle, particle size distribution dynamic variance, and motion consistency. This enables a dynamic depiction of the dust diffusion process, ultimately yielding a comprehensive set of static and dynamic visual features reflecting dust characteristics. This invention accurately locates dust regions through superpixel segmentation, providing precise targets for feature extraction. The collaborative extraction of static and dynamic visual features overcomes the shortcomings of traditional methods that rely solely on static feature extraction, which struggle to distinguish dust from interfering substances like fog and haze. This significantly improves the accuracy and comprehensiveness of dust feature recognition, providing high-quality data support for subsequent dust concentration estimation and diffusion trend prediction, indirectly enhancing the overall system's effectiveness.

[0072] In some embodiments, the dust real-time concentration calculation process of the multi-source data fusion model satisfies:

[0073] ;

[0074] Where C is the real-time dust concentration (unit: ); , For visual feature weight coefficients, satisfying It is obtained through optimization using the gradient descent algorithm; Normalized fusion value of static visual features ( ); For dynamic visual feature normalized fusion value ( ); Environmental correction factor (range of values) ); Wind speed (unit: ); RH is relative humidity (unit: ); Sensor data calibration coefficients (range of values) ); Reference dust concentration (unit: (), collected by a laser dust sensor; For the error compensation term, satisfying .

[0075] It should be noted that the multi-source data fusion model uses preset weighting coefficients. , Normalized fusion value of static visual features Normalized fusion value with dynamic visual features A weighted summation is performed, where the weight coefficients are optimized using a gradient descent algorithm to ensure that the contribution of visual features to concentration estimation matches the actual scene; an environment correction term is introduced. The effects of wind speed and relative humidity on dust dispersion were quantified; the higher the wind speed and the lower the relative humidity, the more significant the adjustment effect of this correction term on the concentration estimation results. The baseline dust concentration collected by the laser dust sensor was adjusted using the sensor data calibration coefficient ε. Perform calibration to correct inherent sensor errors; finally, add an error compensation term. To further reduce model estimation errors, the final result is achieved through the formula:

[0076] The invention calculates accurate real-time dust concentration. By organically integrating multi-source data, it fully leverages the complementary advantages of visual features, environmental parameters, and sensor data, avoiding the limitations of single data types. The scientific setting and optimization of various parameters and coefficients keep the dust concentration estimation error at a low level. Compared with traditional concentration measurement methods that rely solely on sensors, the accuracy is significantly improved, providing a reliable quantitative basis for pollution level classification and suppression strategy generation.

[0077] In some embodiments, predicting the dust diffusion trend within a preset time window specifically includes:

[0078] A bidirectional LSTM network with enhanced attention mechanism is constructed to form a time-series dataset from the real-time dust concentration, dynamic visual features, and environmental parameters.

[0079] ;

[0080] The attention layer assigns weights to key features of the time series data. satisfy:

[0081] ;

[0082] in, Here is the attention weight matrix. For bias terms, This is the output of the hidden layer of the LSTM network. The cell state is defined; attention-weighted time-series data is input into a bidirectional LSTM network, which outputs a preset time window. ( Dust concentration prediction sequence within ) By combining diffusion speed and direction, a diffusion trend heatmap is generated.

[0083] In practice, a bidirectional LSTM network with enhanced attention mechanism is first constructed to form a time-series dataset consisting of real-time dust concentration, dynamic visual features, wind speed, and relative humidity over a continuous time period. Attention weights for each time step are then calculated using an attention layer. ,in From attention weight matrix LSTM network hidden layer output With cell state Obtained through linear transformation. pass Function normalization enables the network to automatically focus on key moment data that significantly impacts the diffusion trend. Attention-weighted time-series data is input into a bidirectional LSTM network. The forward LSTM captures historical trends, while the backward LSTM captures dependencies in future data, jointly outputting a dust concentration prediction sequence within a preset time window. Combined with dust diffusion speed and direction, a diffusion trend heatmap that intuitively reflects the spatial diffusion range of dust is generated. The introduction of the attention mechanism allows the network to accurately identify key information in the time series, avoiding interference from irrelevant data and improving the targeting of trend prediction. The bidirectional LSTM network fully utilizes the dependencies between time-series data, resulting in higher prediction accuracy and stronger foresight compared to traditional time-series prediction models, enabling it to predict in advance. The ability to predict dust dispersion trends within minutes allows ample time for adjustments to control strategies, effectively preventing the problem of dust dispersion becoming difficult to control and enhancing proactive governance.

[0084] In some embodiments, generating a dynamic suppression strategy specifically includes:

[0085] Preset dust concentration pollution level: Level 1 ( No need to suppress), secondary ( Mild inhibition), Grade III ( Moderate inhibition), Level IV ( (severe inhibition)

[0086] Based on the heat map of pollution level and diffusion trend, determine the suppression area, suppression intensity, and suppression method:

[0087] Mild suppression: Activate the fixed perimeter spray system, spray water pressure Spray frequency Second-rate;

[0088] Moderate suppression: Activate the fixed + mobile spray system, spray water pressure Spray frequency Next, the speed limit for joint transport vehicles. ;

[0089] Severe suppression: Activate the full-area spray system, spray water pressure Continuous spraying was initiated, and construction machinery and transport vehicles were suspended.

[0090] In practice, four levels of dust concentration pollution are pre-set, and the corresponding pollution level is determined based on the real-time dust concentration C. Combining the pollution level with a heat map of diffusion trends, the severity and spatial diffusion range of the dust are clarified, thereby determining the suppression area, suppression intensity, and suppression method. For light pollution, fixed perimeter spray devices are activated, using a low-pressure, low-frequency spray mode to focus on covering the initial dust diffusion area. For moderate pollution, fixed and mobile spray devices are activated, increasing water pressure and spray frequency, while simultaneously limiting the speed of transport vehicles to reduce secondary dust. For heavy pollution, full-area spray devices are activated, using a high-pressure continuous spray mode, while simultaneously halting construction machinery and transport vehicles to prevent further dust diffusion. After suppression is implemented, the concentration change is monitored in real-time by the data acquisition unit. If the level threshold is not reached, the suppression intensity is automatically increased, forming a closed-loop feedback control. This invention employs a dynamic suppression strategy based on pollution levels and diffusion trends, achieving "on-demand suppression" and avoiding the water waste and drawbacks of "insufficient suppression" or "over-suppression" caused by traditional timed spraying. The flexible combination and closed-loop feedback adjustment of different suppression methods ensure the accuracy and effectiveness of suppression operations, improving dust control while minimizing the impact on construction progress, thus achieving a balance between environmental protection and construction efficiency.

[0091] In some embodiments, preprocessing the dust scene image data specifically includes:

[0092] Adaptive median filtering is used to remove salt-and-pepper noise from the image, with the filter window size satisfying the following:

[0093] ;

[0094] in, For pixels grayscale contrast;

[0095] pass The algorithm enhances the image, corrects uneven lighting, and improves the contrast between the dusty area and the background. It also uses image morphological opening operations (erosion followed by dilation) to remove small noise points while preserving the main dusty area.

[0096] In practice, adaptive median filtering is used to remove salt-and-pepper noise from the image, and the filter window size is... According to pixels grayscale contrast Dynamic adjustments are made to remove noise while preserving dust details to the maximum extent possible; through The algorithm decomposes the illumination and reflection components of an image, corrects the impact of uneven illumination on image quality, and enhances the contrast between dusty areas and the background; it employs image morphological opening operations, first through... The erosion kernel removes small noise points from the image, and then... The expansion kernel restores the morphology of the main dust region, further purifying the image and laying the foundation for subsequent dust candidate region selection and feature extraction. Compared with fixed window filtering, adaptive median filtering can dynamically adjust the filtering intensity according to local image features, balancing noise removal and detail preservation. The algorithm effectively solves the problem of low image contrast caused by changes in lighting in construction scenarios; morphological opening operations further remove small interferences and improve image purity; the image quality is significantly improved after multiple preprocessing steps, providing high-quality input data for subsequent feature extraction, concentration estimation and other steps, indirectly improving the operating accuracy and stability of the entire system.

[0097] In some embodiments, the AI ​​visual recognition algorithm is an improved YOLOv8 algorithm. By introducing a CBAM attention module in the Neck layer, the ability to extract dust features is enhanced. The channel attention output of the CBAM attention module satisfies the following:

[0098] ;

[0099] Spatial attention output satisfies: ;

[0100] in, The activation function is Sigmoid, and MLP is a multilayer perceptron. Convolution operation.

[0101] It should be noted that the improved YOLOv8 algorithm is used as the AI ​​visual recognition algorithm, and the CBAM attention module is introduced into the Neck layer of YOLOv8. The CBAM attention module includes a channel attention branch and a spatial attention branch. The channel attention branch performs global average pooling and global max pooling on the feature map, inputs the results into the MLP for processing, and obtains the channel attention weights through the Sigmoid activation function. This is used to enhance important feature channels and suppress irrelevant feature channels; the spatial attention branch performs global average pooling and global max pooling on the feature maps, and then concatenates them... Convolution operations and Activation function yields spatial attention weights The CBAM attention module is used to enhance dust-related feature maps and suppress background areas. Through the synergistic effect of channel and spatial dual attention mechanisms, the algorithm's ability to extract dust features is enhanced. The introduction of the CBAM attention module enables the YOLOv8 algorithm to accurately focus on dust-related feature channels and spatial regions, effectively suppressing background interference. Compared to the original YOLOv8 algorithm, the targeting and effectiveness of dust feature extraction are significantly improved. The improved algorithm can more accurately identify dust and other interfering substances such as fog and haze, achieving a higher accuracy rate. This provides high-quality feature data for subsequent concentration estimation and trend prediction, further improving the core performance of the entire system.

[0102] Reference Figure 2 The second aspect of the present invention provides an intelligent monitoring and suppression system for building dust based on AI visual recognition, comprising:

[0103] The acquisition unit includes a high-definition infrared camera (resolution). Frame rate ), laser dust sensor, wind speed sensor, temperature and humidity sensor, used to simultaneously collect image data of dust scene and multi-dimensional environmental parameters;

[0104] The processing unit includes an edge computing gateway and a cloud server. The edge computing gateway is used for image preprocessing, feature extraction and real-time concentration calculation, and the cloud server is used for diffusion trend prediction, suppression strategy generation and model optimization.

[0105] The control unit is used to receive suppression strategy instructions from the processing unit and coordinate with the suppression execution device to perform operations.

[0106] The suppression execution unit includes a fixed fence spraying device, a mobile sprinkler spraying device, a construction machinery control module, and a transport vehicle linkage module;

[0107] Communication unit, adopts Dual-mode communication enables data transmission between the acquisition unit, processing unit, control unit, and suppression execution unit.

[0108] It should be noted that the acquisition unit simultaneously collects image data and multi-dimensional environmental parameters of the dust storm scene through high-definition infrared cameras, laser dust sensors, wind speed sensors, and temperature and humidity sensors, ensuring the comprehensiveness and synchronization of the data. The processing unit consists of an edge computing gateway and a cloud server. The edge computing gateway is responsible for image preprocessing, feature extraction, and real-time concentration calculation, achieving rapid data processing and low-latency response. The cloud server is responsible for diffusion trend prediction, suppression strategy generation, and model optimization, utilizing powerful computing capabilities to complete complex algorithm calculations. The control unit receives suppression strategy instructions from the cloud server and controls the operating parameters and status of the suppression execution equipment through the Modbus protocol. The suppression execution unit executes specific suppression operations through fixed enclosure spraying devices, mobile water spraying devices, construction machinery control modules, and transport vehicle linkage modules. The communication unit adopts... Dual-mode communication: 5G enables high-speed data transmission, while LoRa achieves low-power, wide-coverage communication, ensuring stable and efficient data transmission between units. Beneficial effects: Clear division of labor and collaborative cooperation among system units enable end-to-end intelligent processing of dust monitoring and suppression; the edge computing + cloud collaborative architecture balances real-time performance with computing power requirements. Dual-mode communication ensures communication reliability in different scenarios; diverse acquisition and suppression execution devices enable the system to adapt to construction areas of different sizes and types, with flexible deployment and strong scalability; the entire system realizes full-process automation from data acquisition to suppression execution, greatly reducing manual intervention and improving the efficiency and intelligence level of dust control.

[0109] Reference Figure 3 The third aspect of the present invention provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the intelligent monitoring and suppression method for building dust based on AI visual recognition as described above.

[0110] The computer device's memory stores the code for implementing an AI-based visual recognition-based intelligent monitoring and suppression method for construction dust, including program code for each stage such as data acquisition, image preprocessing, feature extraction, concentration estimation, trend prediction, and suppression strategy generation. The processor retrieves the code stored in the memory via a communication bus, executes the computation and control instructions for each step according to the program logic, and calls upon relevant hardware resources (such as GPUs and communication interfaces) to complete data reception, processing, calculation, and instruction transmission, ultimately achieving intelligent monitoring and suppression of dust. This computer device provides a stable and efficient hardware platform for the AI-based visual recognition-based intelligent monitoring and suppression method for construction dust. The processor's powerful computing capabilities ensure the rapid execution of complex algorithms (such as attention-LSTM networks and improved YOLOv8 algorithms), meeting the needs of real-time monitoring and suppression. The large storage capacity of the memory supports the temporary storage of massive amounts of data and the stable storage of program code, ensuring the continuous and reliable operation of the method and providing solid hardware support for the implementation of the entire technical solution.

[0111] In some embodiments, the AI-based visual recognition-based intelligent monitoring and suppression method for building dust in the above embodiments can be implemented by a computer device, which includes at least one processor, a communication bus, a memory, and at least one communication interface.

[0112] A processor can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0113] A communication bus can be used to transmit information between the aforementioned components.

[0114] The memory can be read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions, random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, universal optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited to these. The memory can exist independently and be connected to the processor via a communication bus. The memory can also be integrated with the processor.

[0115] The memory stores program code for executing the solution of this application, and its execution is controlled by a processor. The processor executes the program code stored in the memory. The program code may include one or more software modules. In the above embodiments, the intelligent monitoring and suppression method for construction dust based on AI visual recognition can be implemented by a processor and one or more software modules in the program code in the memory.

[0116] A communication interface is a device that uses any transceiver or similar device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0117] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0118] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0119] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for intelligent monitoring and suppression of building dust based on AI visual recognition.

[0120] It should be noted that the computer-readable storage medium stores the computer program for the AI-based visual recognition-based intelligent monitoring and suppression method of construction dust through physical storage media (such as hard drives, flash memory, optical discs, etc.). When the storage medium is connected to a computer device, the computer device's processor can read the computer program in the storage medium, load the program into memory, and execute it. Through the step-by-step execution of program instructions, the entire process of dust monitoring and suppression is realized. Beneficial effects: The computer-readable storage medium provides a convenient storage and transmission medium for the method's program code, facilitating program installation, deployment, and upgrades; the universality of the storage medium allows the method to run on different types of computer devices, expanding the application scope of the method; the long-term storage of the storage medium ensures the security and reusability of the program code, reduces the cost of promoting and applying the technical solution, and facilitates the popularization of construction dust control technology.

[0121] To provide a clearer understanding of the invention, the invention is further described below:

[0122] Reference Figure 1 In a first aspect, the present invention provides a method for intelligent monitoring and suppression of construction dust based on AI visual recognition, the method comprising the following steps:

[0123] Data acquisition steps: Simultaneously acquire image data and multi-dimensional environmental parameters of dust pollution scenes in the construction area. High-definition infrared cameras deployed at high points in the construction area (such as tower cranes or the top of construction barriers) acquire real-time video stream data of dust pollution scenes. Image resolution... Frame rate This ensures the capture of dynamic changes in dust; simultaneously, laser dust sensors, wind speed sensors, and temperature and humidity sensors are deployed around the camera to synchronously collect baseline dust concentrations. ), wind speed ( ), relative humidity ( ),temperature( ), sampling frequency is It timestamps the image data to ensure synchronization of multi-source data.

[0124] Image preprocessing and feature extraction steps: Preprocess the image data to extract static and dynamic visual features of dust.

[0125] Preprocessing: First, adaptive median filtering is used to remove salt-and-pepper noise from the image. The size of the filtering window is dynamically adjusted based on the pixel grayscale contrast, using the following formula: ,in For pixels The grayscale contrast is adjusted to ensure that dust details are preserved while removing noise; then through... The algorithm corrects uneven lighting and enhances the contrast between the dusty area and the background; finally, morphological opening operations (erosion kernels) are used. Expanding nucleus Remove minor noise points while retaining the main dust-generating area.

[0126] Static feature extraction: Superpixel segmentation is performed on the preprocessed image ( The algorithm filters out candidate dust areas; an improved HOG algorithm is used to extract static features of the candidate areas, including mean gray level (a direct feature reflecting dust concentration), texture entropy (reflecting the uniformity of dust particles), particle size equivalent area ratio (the ratio of the dust area to the total image area), and gray level gradient histogram features (reflecting the edge contour of dust). These features are then normalized to form a static feature vector. (dimension is) ).

[0127] Dynamic feature extraction: based on Optical flow is used to calculate pixel motion vectors of dust candidate regions. Background interference is removed through vector clustering, and dynamic features are extracted: dust diffusion velocity (average magnitude of motion vectors), diffusion direction angle (average direction of motion vectors), particle size distribution dynamic variance (reflecting the variation in dust particle size), and motion consistency features (mean cosine similarity of motion vectors). These features are then normalized to form a dynamic feature vector. (dimension is) ).

[0128] Dust concentration estimation steps: Calculate real-time concentration based on a multi-source data fusion model.

[0129] A multi-source data fusion model is constructed, integrating visual features and environmental parameters to accurately estimate real-time dust concentration. The core formula of the model is:

[0130] ;

[0131] in:

[0132] , The visual feature weight coefficients are obtained through gradient descent algorithm, with an initial value of 0.5. During the iteration process, the mean square error between the predicted concentration and the sensor-measured concentration is minimized.

[0133] This is an environmental correction factor, calibrated based on historical data of the construction area, with a value range of... The higher the wind speed and the lower the humidity, The higher the value, the greater the impact of the environment on dust dispersion;

[0134] These are sensor data calibration coefficients used to correct sensor errors, with a value range of... It was obtained through laboratory calibration;

[0135] This is the error compensation term, obtained from the residual statistics during model training, and satisfies... To ensure the accuracy of concentration estimation .

[0136] Steps for predicting dust dispersion trends: Predicting dust dispersion trends based on attention-LSTM networks.

[0137] Construct a bidirectional LSTM network with enhanced attention mechanism to predict dust diffusion trends within a preset time window:

[0138] Data preprocessing: The dust concentration (C) and dynamic feature vector (C) for 5 consecutive minutes were processed. ), environmental parameters ( The time series sample consists of n = 300 (RH) units. );

[0139] Attention layer computation: Assigning attention weights to the data at each time step in the time series sample. The formula is:

[0140] ;

[0141] ;

[0142] in for Weight matrix (hidden layer dimension 256) Cell state dimension 128). For bias terms, For the output of the LSTM hidden layer, For cellular states, key moment data that significantly influence diffusion trends are highlighted through attention mechanisms.

[0143] Trend prediction: Input attention-weighted samples into a bidirectional LSTM network (forward LSTM captures historical trends, backward LSTM captures future dependencies), and output a preset time window. ( Dust concentration prediction sequence within ) By combining diffusion speed and direction, a diffusion trend heat map is generated, which intuitively shows the spatial diffusion range of dust.

[0144] Intelligent suppression control steps: Generate a closed-loop suppression strategy based on concentration level and diffusion trend.

[0145] Pollution level classification: Four preset dust concentration pollution levels, Level 1 ( No need to suppress), secondary ( Mild inhibition), Grade III ( Moderate inhibition), Level IV ( (severe inhibition)

[0146] Suppression strategy generation: Based on the heat map of pollution level and diffusion trend, determine the suppression area (high concentration area on the heat map), suppression intensity (water pressure, frequency), and suppression method:

[0147] Mild suppression: Activate the fixed perimeter spray system, spray water pressure Spray frequency Secondly, the focus is on covering the initial dust diffusion area;

[0148] Moderate suppression: Activate the fixed + mobile spray system (water truck), spray water pressure Spray frequency Next, the speed limit for joint transport vehicles. Reduce secondary dust;

[0149] Severe suppression: Activate the full-area spray system (enclosure). tower crane (Sprinkler truck), spray water pressure Continuous spraying, suspension of construction machinery and cessation of transport vehicles to prevent further spread of dust;

[0150] Closed-loop feedback: After suppression is implemented, the dust concentration change is monitored in real time through the acquisition unit. If the concentration does not drop below the corresponding level threshold within 5 minutes, the suppression intensity is automatically increased to form a closed-loop control.

[0151] Reference Figure 2 Secondly, the present invention provides an intelligent monitoring and suppression system for building dust based on AI visual recognition. This system includes a data acquisition unit, a processing unit, a control unit, a suppression execution unit, and a communication unit.

[0152] Data acquisition unit: includes a high-definition infrared camera, a laser dust sensor, a wind speed sensor, and a temperature and humidity sensor; the high-definition infrared camera is deployed at a high point in the construction area, with a coverage radius of [missing information]. 50 meters, supports nighttime infrared imaging; laser dust sensor measurement range Precision Wind speed sensor measurement range Precision Temperature and humidity sensor measurement range: temperature ,humidity RH, precision , RH.

[0153] Processing unit: Includes edge computing gateway and cloud server; the edge computing gateway uses NVIDIA Jetson Xavier NX chip, supports image preprocessing, feature extraction and real-time density calculation, and has low latency. 100ms; The cloud server uses a GPU cluster (NVIDIA A100) to run an attention-LSTM network to predict diffusion trends, generate suppression strategies, and optimize model parameters through an online learning mechanism (the weight coefficients are updated every 24 hours).

[0154] Control unit: Employs a PLC controller, receiving suppression strategy instructions from the cloud server, and controlling the start / stop and parameter adjustment (water pressure, frequency) of the suppression execution equipment via the Modbus protocol, with response delay... 500ms.

[0155] Suppression Execution Unit: Includes fixed fencing spray devices (deployed on top of the construction fencing, spaced 5 meters apart). (Unit), mobile sprinkler system (mounted on sprinkler truck, supporting...) Rotary spray), construction machinery control module (linking the start and stop of equipment such as excavators and bulldozers), and transport vehicle linkage module (sending speed limit / stop commands via 5G communication).

[0156] Communication unit: adopts Dual-mode communication; 5G is used for high-speed data transmission (images, video streams) between edge computing gateways and cloud servers, with bandwidth... 100Mbps; LoRa is used for low-power communication between the control unit and the suppression actuator, with a communication distance of 100Mbps. 1km, ensuring signal coverage throughout the entire construction area.

[0157] To enable those skilled in the art to more clearly understand the technical solution of this application, the following detailed description is provided in conjunction with specific embodiments:

[0158] Example 1: Method Implementation Flow

[0159] Data Acquisition: Two high-definition infrared cameras (covering the entire construction site) were deployed on the top of a tower crane at a construction site (approximately 50,000 square meters). Four sets of laser dust sensors, wind speed sensors, and temperature and humidity sensors were deployed around the perimeter fence to simultaneously collect image data and environmental parameters. Timestamp alignment accuracy was ensured. 10ms;

[0160] Preprocessing and feature extraction: The edge computing gateway performs adaptive median filtering on the image. Enhanced morphological opening preprocessing was performed, and static feature vectors of dust were extracted using an improved YOLOv8 algorithm (introducing the CBAM attention module). ( ) and dynamic feature vector ( ), processing delay 100ms;

[0161] Concentration estimation: Substitute into the multi-source data fusion model, , , , , The real-time dust concentration was calculated. (Level 3 pollution);

[0162] Trend prediction: A cloud server inputs 5 consecutive minutes of time-series data into an attention-LSTM network to predict dust concentration 15 minutes later. The diffusion direction is northeast, generating a diffusion trend heatmap;

[0163] Suppression Control: The control unit initiates a moderate suppression strategy, activating all fixed perimeter fencing spray devices (water pressure 0.6MPa, frequency...). Two mobile water trucks were dispatched to spray water in the northeastern area, and transport vehicles were restricted from speed. The concentration dropped to [a certain level] after 5 minutes. It automatically adjusts to a mild suppression strategy.

[0164] Example 2: Model Training and Optimization

[0165] Dataset Construction: Image data, environmental parameters, and sensor data were collected from three different construction sites (residential construction, road construction, and bridge construction) to construct a training dataset (100,000 samples), a validation dataset (20,000 samples), and a test dataset (20,000 samples).

[0166] Model Training: The improved YOLOv8 algorithm was trained for 100 iterations with a learning rate of 0.001, using the Adam optimizer. After training, the dust recognition accuracy reached [percentage missing]. The attention-LSTM network was trained for 200 epochs with a learning rate of 0.0001 and the MSE loss function was used. The concentration prediction error after training was... ;

[0167] Online optimization: During system operation, 1,000 new samples are added to the training dataset every 24 hours, and the model parameters are updated using incremental learning to ensure that the model adapts to the dust characteristics of different construction scenarios.

[0168] Example 3: System Deployment and Results

[0169] The system of this invention was deployed at a large residential construction site and operated for 3 months. Compared with traditional treatment methods:

[0170] Dust concentration compliance rate: from traditional methods Upgraded to (GB16297-1996 Integrated Emission Standard for Air Pollutants);

[0171] Water consumption: from traditional methods The sky descends Heaven, save ;

[0172] Labor costs: from the traditional 6 people Two people fell from the sky. Days (only requires regular equipment maintenance);

[0173] Response time: From 15 minutes in the traditional method (manual detection) Start-up time reduced to 1 minute (automatic monitoring + startup).

[0174] The above embodiments demonstrate that the intelligent monitoring and suppression system and method for construction dust based on AI visual recognition provided in this application can effectively improve the accuracy and efficiency of construction dust control, and has significant environmental and economic benefits.

[0175] The technical solutions provided by the embodiments of the present invention have the following significant beneficial effects:

[0176] High monitoring accuracy: It integrates static and dynamic feature extraction using AI visual recognition, combines multi-dimensional environmental parameters, and employs a multi-source data fusion model to estimate dust concentration, minimizing errors. It can also distinguish between dust and interference such as fog and haze, with a high accuracy rate. This solves the problems of traditional sensors, such as single-point monitoring and susceptibility to interference.

[0177] Foresight: Predicting dust dispersion trends in advance using attention-LSTM networks. The spatial spread range of dust can be predicted in minutes, allowing time for adjustment of suppression strategies and avoiding "suppression lag";

[0178] Intelligent Suppression: Based on dust concentration levels and diffusion trends, a zoned and graded closed-loop suppression strategy is generated, which is linked to fixed mechanisms. Mobile spraying devices, construction machinery, and transport vehicles enable "precision suppression," saving 30% of water compared to traditional timed spraying. The above results in a 40% improvement in governance efficiency. ;

[0179] Flexible deployment: Employing edge computing A cloud-based collaborative architecture enables edge processing of real-time data and cloud-based optimization of models and strategies, supporting... Dual-mode communication adapts to construction areas of different sizes (from several thousand square meters to hundreds of thousands of square meters), with low deployment costs and strong scalability.

[0180] In summary, this application achieves comprehensive monitoring, accurate prediction, and efficient suppression of construction dust through AI visual recognition, multi-source data fusion, and intelligent closed-loop control, providing a brand-new technical solution for environmental protection management in construction.

[0181] The technical solutions provided by the embodiments of the present invention have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of the embodiments of the present invention. The descriptions of the embodiments above are only for helping to understand the principles of the embodiments of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the embodiments of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for intelligent monitoring and suppression of construction dust based on AI visual recognition, characterized in that, Includes the following steps: Simultaneously collect image data of dust scenes in the construction area and multi-dimensional environmental parameters, including wind speed, relative humidity, temperature and baseline dust concentration; The image data of the dust scene is preprocessed, and the static and dynamic visual features of the dust are extracted based on AI visual recognition algorithms. A multi-source data fusion model is constructed, and the static visual features, dynamic visual features and multi-dimensional environmental parameters are input. The real-time dust concentration is calculated through weighted fusion and nonlinear mapping. Based on the time series prediction model, combined with the real-time dust concentration, dynamic visual characteristics and environmental parameter change trends, the dust diffusion trend within a preset time window is predicted. Based on the pollution level corresponding to the real-time dust concentration and the dust diffusion trend, a dynamic suppression strategy is generated, and the suppression execution equipment is linked to perform dust suppression operations, forming a closed-loop control.

2. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: Extracting the static and dynamic visual features of dust specifically includes: Superpixel segmentation is performed on the preprocessed dust scene image to filter candidate dust regions; An improved HOG algorithm is used to extract the static visual features of the dust candidate region. The static visual features include gray-level mean, texture entropy, particle size equivalent area ratio, and gray-level gradient histogram features. based on The optical flow method is used to calculate the pixel motion vectors of the dust candidate region and extract dynamic visual features, including dust diffusion speed, diffusion direction angle, particle size distribution dynamic variance, and motion consistency features.

3. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: The real-time dust concentration calculation process of the multi-source data fusion model satisfies: ； Where C represents the real-time dust concentration; , For visual feature weight coefficients, satisfying It is obtained through optimization using the gradient descent algorithm; This represents the normalized fusion value of static visual features; This represents the normalized fusion value of dynamic visual features; This is an environmental correction factor; Wind speed; RH is relative humidity; For sensor data calibration coefficients; The baseline dust concentration was collected by a laser dust sensor. This is the error compensation term.

4. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: The predicted dust dispersion trend within the preset time window specifically includes: A bidirectional LSTM network with enhanced attention mechanism is constructed to form a time-series dataset from the real-time dust concentration, dynamic visual features, and environmental parameters. ； The attention layer assigns weights to key features of the time series data. satisfy: ； in, Here is the attention weight matrix. For bias terms, This is the output of the hidden layer of the LSTM network. The cell state is defined; attention-weighted time-series data is input into a bidirectional LSTM network, which outputs a preset time window. Dust concentration prediction sequence within By combining diffusion speed and direction, a diffusion trend heatmap is generated.

5. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: The specific dynamic suppression strategies include: Preset dust concentration pollution levels: Level 1, Level 2, Level 3, Level 4; Based on the heat map of pollution level and diffusion trend, determine the suppression area, suppression intensity, and suppression method: Mild suppression: Activate the fixed perimeter spray system, spray water pressure Spray frequency Second-rate; Moderate suppression: Activate the fixed + mobile spray system, spray water pressure Spray frequency Next, the speed limit for joint transport vehicles. ; Severe suppression: Activate the full-area spray system, spray water pressure Continuous spraying was implemented, and construction machinery and transport vehicles were suspended.

6. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: The preprocessing of the dust scene image data specifically includes: Adaptive median filtering is used to remove salt-and-pepper noise from the image, with the filter window size satisfying the following: ； in, For pixels grayscale contrast; pass The algorithm enhances the image, corrects uneven lighting, and improves the contrast between the dusty area and the background; it also uses image morphological opening operations to remove small noise points while preserving the main dusty area.

7. The intelligent monitoring and suppression method for building dust based on AI visual recognition according to claim 1, characterized in that: The AI ​​visual recognition algorithm is an improved YOLOv8 algorithm. By introducing a CBAM attention module in the Neck layer, the ability to extract dust features is enhanced. The channel attention output of the CBAM attention module satisfies the following: ； Spatial attention output satisfies: ; in, The activation function is Sigmoid, and MLP is a multilayer perceptron. Convolution operation.

8. A smart monitoring and suppression system for construction dust based on AI visual recognition, used to implement the method described in any one of claims 1-7, characterized in that, include: The data acquisition unit includes a high-definition infrared camera, a laser dust sensor, a wind speed sensor, and a temperature and humidity sensor, which are used to simultaneously acquire image data of dusty scenes and multi-dimensional environmental parameters. The processing unit includes an edge computing gateway and a cloud server. The edge computing gateway is used for image preprocessing, feature extraction and real-time concentration calculation, and the cloud server is used for diffusion trend prediction, suppression strategy generation and model optimization. The control unit is used to receive suppression strategy instructions from the processing unit and coordinate with the suppression execution device to perform operations. The suppression execution unit includes a fixed fence spraying device, a mobile sprinkler spraying device, a construction machinery control module, and a transport vehicle linkage module; Communication unit, adopts Dual-mode communication enables data transmission between the acquisition unit, processing unit, control unit, and suppression execution unit.

9. A computer device comprising a memory and a processor, the memory storing code, characterized in that, The processor is configured to acquire the code and execute the intelligent monitoring and suppression method for building dust based on AI visual recognition as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the intelligent monitoring and suppression method for building dust based on AI visual recognition as described in any one of claims 1 to 7.

Images