An ultra-dangerous large project automatic monitoring and management method, system and computer device

Abstract

The application relates to the field of image data recognition processing, and discloses an automatic monitoring and management method and system for super-dangerous large projects and computer equipment, wherein the monitoring images of a construction site are preprocessed first to adapt to the complex construction environment of the super-dangerous large project, and accurate monitoring under insufficient brightness or in the presence of smoke and haze is ensured; a double-path scene perception network module is designed to improve the quality of basic image features, improve the overall monitoring precision, guide and enhance the wide-area relationship and context relationship between image features, realize accurate monitoring of complex multi-targets in a wide area, and realize the adjustability of monitoring through a management and judgment module, thereby avoiding false positives or false negatives caused by different construction standards and construction conditions among various super-dangerous large projects. The application realizes automatic monitoring and management of super-dangerous large projects, and greatly improves the monitoring precision under wide-area monitoring conditions.

Description

Technical Field

[0001] This invention relates to the field of image data recognition and processing, and in particular to an automated monitoring and management method, system, and computer equipment for extremely dangerous and large-scale engineering projects. Background Technology

[0002] With the rapid development of engineering technology, the management of engineering projects has gradually received attention, especially for large-scale and high-risk projects. Due to their large scale, the large number of people involved, and the extreme risks during construction, these projects have become an extremely important and difficult target in engineering management.

[0003] In existing technologies, the monitoring and management of ultra-dangerous and large-scale projects mainly relies on on-site inspections and records by construction personnel and safety supervisors. However, traditional manual management methods are still difficult to fully inspect when faced with extremely large and high-risk construction environments. Furthermore, they cannot provide real-time management around the clock during construction and can only rely on sampling inspections or multiple inspections in different areas. The labor costs required to achieve real-time management are extremely high, further increasing the management difficulty of ultra-dangerous and large-scale projects. Moreover, the monitoring accuracy and scope of existing automated monitoring and management methods cannot meet the monitoring and management requirements of ultra-dangerous and large-scale projects, which involve a large number of personnel.

[0004] Therefore, how to design an automated monitoring and management method to meet the requirements of automated monitoring and management for extremely dangerous and large-scale projects has become an urgent problem to be solved. Summary of the Invention

[0005] Based on this, the purpose of this invention is to provide an automated monitoring and management method, system, and computer equipment for ultra-dangerous and large-scale engineering projects. Firstly, the monitoring images from the construction site are preprocessed to adapt to the complex construction environment of ultra-dangerous and large-scale engineering projects, ensuring accurate monitoring even in low light conditions or under conditions of smoke, dust, or haze. Then, a dual-path context-aware network module is designed, including a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule, improving the quality of basic image features and thus enhancing overall monitoring accuracy. Simultaneously, the wide-area and contextual relationships between image features are guided and enhanced, achieving accurate monitoring of complex multi-target situations over a wide area. Furthermore, a management judgment module is designed to achieve adjustable monitoring, avoiding false alarms or misreports caused by differences in construction standards and conditions among various ultra-dangerous and large-scale engineering projects, further improving monitoring accuracy. This invention achieves automated monitoring and management of ultra-dangerous and large-scale engineering projects, significantly improving monitoring accuracy in wide-area monitoring situations.

[0006] This invention proposes an automated management method for extremely dangerous large-scale engineering projects, comprising: The acquisition module acquires construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. The dual-path context-aware network module includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain an attention-enhanced image, which is then transmitted to the management and judgment module. The management judgment module performs construction standard judgment on the attention enhancement image based on a preset construction database. The preset construction database includes construction drawing data, construction personnel data, and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

[0007] In summary, based on the aforementioned automated management method for ultra-dangerous and large-scale engineering projects, the following steps are taken: First, the monitoring images of the construction site are preprocessed to adapt to the complex construction environment of ultra-dangerous and large-scale engineering projects, ensuring accurate monitoring even in low light conditions or under conditions of smoke, dust, and haze. Then, a dual-path context-aware network module is designed, including a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule, which improves the quality of basic image features and enhances the overall monitoring accuracy. Simultaneously, the wide-area and contextual relationships between image features are guided and enhanced, enabling accurate monitoring of complex multi-target situations over a wide area. Furthermore, a management judgment module is designed to achieve adjustable monitoring, avoiding false alarms or misreports caused by differences in construction standards and conditions among various ultra-dangerous and large-scale engineering projects, further improving monitoring accuracy. This invention achieves automated monitoring and management of ultra-dangerous and large-scale engineering projects, significantly improving monitoring accuracy in wide-area monitoring situations.Specifically, the acquisition module obtains construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed before being transmitted to the dual-path scene perception network module, eliminating the influence of dim lighting and smog, thus avoiding a decrease in monitoring accuracy caused by the complex construction environment of high-risk projects. The dual-path scene perception network module includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule obtains the basic image features of the construction site image data, performing targeted extraction of basic features from global wide-area scale, multi-local detail scale, and deep scale, improving the quality of basic features and thus improving overall monitoring accuracy. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain corresponding attention. The attention-weighted image features of the mechanism enhance the feature quality of individual target personnel and target areas in a large-scale multi-target situation, thereby improving the monitoring accuracy of single targets. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain attention-enhanced images, which are then transmitted to the management judgment module. This enhances the feature relationships of multiple target personnel and target areas in a wide area, thereby improving the monitoring accuracy in a wide area. The management judgment module performs construction standard judgment on the attention-enhanced images based on a preset construction database, which includes construction drawing data, construction personnel data, and construction safety standard data. If the current construction status is determined to be inconsistent with the construction standards, a construction problem alarm is triggered, realizing the adjustability of monitoring and avoiding false alarms or misreports caused by different construction standards and conditions among various high-risk projects, further improving the monitoring accuracy. This invention realizes automated monitoring and management of high-risk projects, greatly improving the monitoring accuracy in a wide-area monitoring situation.

[0008] Furthermore, the acquisition module acquires construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The preprocessing steps for the construction site image data include: The acquisition module acquires construction site data in real time and performs image enhancement preprocessing. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. Based on the brightness data of the construction site, it is determined whether the brightness of the construction site is less than a preset brightness threshold. If the brightness of the construction site is determined to be less than the preset brightness threshold, brightness enhancement preprocessing is performed on the construction site image data. The pixel values ​​of the red, green, and blue channels in the construction site image data are extracted, and weighted grayscale processing is performed on the construction site image data according to the following formula: in, This represents the weighted grayscale value of each pixel, where 'a' represents the total number of pixels. This represents the maximum value among the three channels of each pixel. The grayscale weight value, This represents the minimum value among the three channels of each pixel. The grayscale weight values ​​are given, where R, G, and B represent the pixel values ​​of the red, green, and blue channels of each pixel, respectively. After linear normalization of the construction site image data, the image grayscale mean is then processed according to the following formula: in, The grayscale value represents the average grayscale value of a pixel, and this average grayscale value is used as the image grayscale value of the construction site image data. Based on the haze intensity data at the construction site, it is determined whether the haze intensity at the construction site exceeds a preset haze intensity threshold. If the haze intensity exceeds the preset threshold, haze removal preprocessing is performed on the construction site image data. Atmospheric light value A is estimated based on the image grayscale values ​​of the construction site image data. The estimated atmospheric light value A is then used for calculation and guided filtering to obtain the transmittance. Based on the atmospheric light value A and the transmittance The preprocessed image is obtained using the following formula: in, Indicates the preprocessed image. Indicates an unprocessed image. This represents the lower limit threshold of transmittance, and max(·) represents obtaining the maximum value.

[0009] Furthermore, the basic feature extraction submodule's steps for acquiring the basic image features of the construction site image data include: The preprocessed construction site image data is input into the basic feature extraction submodule, which includes a first convolutional layer, a second convolutional layer, and a third convolutional layer. The first convolutional layer includes two identical convolutional branches, each of which includes two 5x5 convolutional blocks. The first convolutional layer acquires global basic features. The second convolutional layer includes a bottleneck structure branch, an inverse bottleneck structure branch, and a residual branch. The bottleneck structure branch includes two 1x1 convolutional blocks and one 3x3 convolutional block, and the inverse bottleneck structure branch includes one 1x1 convolutional block and two 3x3 convolutional blocks. The second convolutional layer acquires depth basic features. The third convolutional layer includes a multi-scale convolutional branch and a residual branch. The multi-scale convolutional branch includes one 1x1 convolutional block, one 2x2 convolutional block, and one 3x3 convolutional block. The third convolutional layer acquires multi-scale basic features. The global basic features, the depth basic features, and the multi-scale basic features are fused to obtain basic image features.

[0010] Furthermore, the step of the dual-path attention weighting submodule performing attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism includes: The basic image features are input into the spatial attention mechanism path of the dual-path attention weighting submodule. The basic image features are then subjected to spatial attention weighting processing according to the spatial attention mechanism. The spatial attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, a 3x3 convolutional block, a 5x5 convolutional block, and a 7x7 convolutional block. The formula for the spatial attention weighting processing is as follows: in, Represents basic image features, This represents spatial attention-weighted image features. This represents the activation function. Represents a convolution block. Indicates splicing, This represents the global max pooling layer. Indicates a global average pooling layer; To obtain spatial attention-weighted image features.

[0011] Furthermore, the step of the dual-path attention weighting submodule performing attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism also includes: The basic image features are input into the channel attention mechanism path of the dual-path attention weighting submodule. Channel attention weighting processing is then applied to the basic image features according to the channel attention mechanism. The channel attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, and a multilayer perceptron. The formula for the channel attention weighting processing is as follows: in, Represents basic image features, This represents channel attention-weighted image features. This represents the activation function. This represents a multilayer perceptron. This represents the global max pooling layer. Indicates a global average pooling layer; To obtain channel attention-weighted image features.

[0012] Furthermore, the step of inputting the attention-weighted image features and the basic image features into the multi-head linear self-attention submodule to obtain an attention-enhanced image and transmitting it to the management and judgment module includes: Spatial attention-weighted image features, channel attention-weighted image features, and basic image features are input into the linear self-attention submodule, and used as the linear self-attention submodule's... Q matrix, K Matrix and V The linear self-attention submodule uses a matrix to obtain attention-enhanced image features, and its mechanism is as follows: in, express Q Matrix weights express K Matrix weights express V Matrix weights Represents depthwise convolution. express Q Matrix and K The dimension of a matrix Represents the transpose matrix; The acquired attention-enhanced image is transmitted to the management and judgment module.

[0013] Furthermore, the step of the management judgment module performing construction standard judgment on the attention enhancement image based on the preset construction database includes: After the management judgment module obtains the attention-enhanced image features, it determines whether the current ultra-dangerous major construction project complies with construction standards based on the data tags in the preset construction database. The construction data tags include construction area tags and construction structure tags in the construction drawing data, construction personnel quantity tags and construction personnel equipment tags in the construction personnel data, and construction hazard area tags and construction personnel safety equipment tags in the construction safety standard data. The construction area tags, construction structure tags, construction personnel quantity tags, and construction personnel equipment tags are safety tags, while the construction hazard area tags and construction personnel safety equipment tags are hazard tags. If the attention-enhanced image lacks safety tags or contains hazard tags, an alarm is triggered for ultra-dangerous major construction project issues.

[0014] This invention proposes an automated management system for extremely dangerous large-scale engineering projects, comprising: The acquisition module is used to acquire construction site data in real time. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. A dual-path context-aware network module is provided, comprising a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain an attention-enhanced image, which is then transmitted to the management and judgment module. The management judgment module is used to judge the construction standards of the attention enhancement image based on a preset construction database. The preset construction database includes construction drawing data, construction personnel data, and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

[0015] The present invention also provides a storage medium, including one or more programs stored in the storage medium, which, when executed by a processor, implement the above-described automated monitoring and management method for extremely dangerous large-scale engineering projects.

[0016] The present invention also provides a computer device, the computer device including a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the above-described automated monitoring and management method for extremely dangerous large-scale engineering projects. Attached Figure Description

[0017] Figure 1 The flowchart is a process for the automated monitoring and management method for extremely dangerous large-scale engineering projects proposed in the first embodiment of the present invention; Figure 2 The flowchart is a second embodiment of the automated monitoring and management method for extremely dangerous large-scale engineering projects proposed in this invention. Figure 3 This is a schematic diagram of the structure of the automated monitoring and management system for extremely dangerous and large-scale engineering projects proposed in the third embodiment of the present invention.

[0018] The following detailed description, in conjunction with the accompanying drawings, will further illustrate the present invention. Detailed Implementation

[0019] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Several embodiments of the invention are illustrated in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete.

[0020] It should be noted that when a component is said to be "fixed to" another component, it can be directly on the other component or there may be an intervening component. When a component is said to be "connected to" another component, it can be directly connected to the other component or there may be an intervening component. The terms "vertical," "horizontal," "left," "right," and similar expressions used in this document are for illustrative purposes only.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0022] Please see Figure 1 The diagram shows a flowchart of the automated monitoring and management method for extremely dangerous and large-scale engineering projects proposed in the first embodiment of the present invention. This automated monitoring and management method for extremely dangerous and large-scale engineering projects includes steps S01 to S03, wherein: Step S01: The acquisition module acquires construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. It should be noted that the preprocessing described in this embodiment includes image preprocessing for dimly lit construction environments and preprocessing for construction environments with smoke and haze. The image preprocessing process for dimly lit construction environments is as follows: based on the brightness data of the construction site, it is determined whether the brightness of the construction site is less than a preset brightness threshold. If it is determined that the brightness of the construction site is less than the preset brightness threshold, then brightness enhancement preprocessing is performed on the construction site image data, extracting the pixel values ​​of the red, green, and blue channels from the construction site image data, and performing weighted grayscale processing on the construction site image data according to the following formula: in, This represents the weighted grayscale value of each pixel, where 'a' represents the total number of pixels. This represents the maximum value among the three channels of each pixel. The grayscale weight value, This represents the minimum value among the three channels of each pixel. The grayscale weight values ​​are given, where R, G, and B represent the pixel values ​​of the red, green, and blue channels of each pixel, respectively. After linear normalization of the construction site image data, the image grayscale mean is then processed according to the following formula: in, The grayscale value represents the average grayscale value of a pixel. This average grayscale value is used as the image grayscale value of the construction site image data. The preprocessing process for the construction environment with smoke and haze involves: determining whether the haze level at the construction site exceeds a preset haze threshold based on the haze level data; if the haze level exceeds the preset threshold, then performing haze removal preprocessing on the construction site image data; estimating the atmospheric light value A based on the image grayscale value; and performing calculations and guided filtering based on the estimated atmospheric light value A to obtain the transmittance. Based on the atmospheric light value A and the transmittance The preprocessed image is obtained using the following formula: in, Indicates the preprocessed image. Indicates an unprocessed image. This represents the lower limit threshold of transmittance, and max(·) represents obtaining the maximum value.

[0023] Step S02: The dual-path context-aware network module includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule obtains the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain the attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain the attention-enhanced image and transmit it to the management judgment module. It should be noted that the basic feature extraction submodule in this embodiment includes a first convolutional layer, a second convolutional layer, and a third convolutional layer. The first convolutional layer includes two identical convolutional branches, each of which includes two 5x5 convolutional blocks. The first convolutional layer acquires global basic features. The second convolutional layer includes a bottleneck structure branch, an inverse bottleneck structure branch, and a residual branch. The bottleneck structure branch includes two 1x1 convolutional blocks and one 3x3 convolutional block. The inverse bottleneck structure branch includes one 1x1 convolutional block and two 3x3 convolutional blocks. The second convolutional layer acquires depth basic features. The third convolutional layer includes a multi-scale convolutional branch and a residual branch. The multi-scale convolutional branch includes one 1x1 convolutional block, one 2x2 convolutional block, and one 3x3 convolutional block. The third convolutional layer acquires multi-scale basic features. The global basic features, the depth basic features, and the multi-scale basic features are fused to obtain basic image features.

[0024] It should be noted that in the spatial attention mechanism path of the dual-path attention weighting submodule in this embodiment, the basic image features are subjected to spatial attention weighting processing according to the spatial attention mechanism. The spatial attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, a 3x3 convolutional block, a 5x5 convolutional block, and a 7x7 convolutional block. The formula for the spatial attention weighting processing is as follows: in, Represents basic image features, This represents spatial attention-weighted image features. This represents the activation function. Represents a convolution block. Indicates splicing, This represents the global max pooling layer. The global average pooling layer is used to obtain spatial attention-weighted image features. In the channel attention mechanism path of the dual-path attention weighting submodule in this embodiment, the basic image features are processed by channel attention weighting according to the channel attention mechanism. The channel attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, and a multilayer perceptron. The formula for the channel attention weighting process is as follows: in, Represents basic image features, This represents channel attention-weighted image features. This represents the activation function. This represents a multilayer perceptron. This represents the global max pooling layer. This represents a global average pooling layer used to obtain channel attention-weighted image features.

[0025] It should be noted that in this embodiment, the linear self-attention submodule takes spatial attention-weighted image features, channel attention-weighted image features, and basic image features as inputs to the linear self-attention submodule, and uses them as the linear self-attention submodule's... Q matrix, K Matrix and V The linear self-attention submodule uses a matrix to obtain attention-enhanced image features, and its mechanism is as follows: in, express Q Matrix weights express K Matrix weights express V Matrix weights Represents depthwise convolution. express Q Matrix and K The dimension of a matrix This represents the transpose of the matrix.

[0026] Step S03: The management judgment module performs construction standard judgment on the attention enhancement image according to the preset construction database. The preset construction database includes construction drawing data, construction personnel data and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

[0027] It should be noted that, after the management judgment module in this embodiment obtains the attention-enhanced image features, it determines whether the current ultra-dangerous major construction project meets the construction standards based on the data tags in the preset construction database. The construction data tags include the construction area tag and construction structure tag in the construction drawing data, the number of construction personnel tag and the equipment used by construction personnel tag in the construction personnel data, and the construction hazard area tag and the safety equipment tag for construction personnel tag in the construction safety standard data. The construction area tag, construction structure tag, number of construction personnel tag, and equipment used by construction personnel tag are safety tags, and the construction hazard area tag and the safety equipment tag for construction personnel tag are hazard tags. If the attention-enhanced image lacks safety tags or contains hazard tags, an alarm is triggered for ultra-dangerous major construction project problems.

[0028] In summary, based on the aforementioned automated management method for ultra-dangerous and large-scale engineering projects, the following steps are taken: First, the monitoring images of the construction site are preprocessed to adapt to the complex construction environment of ultra-dangerous and large-scale engineering projects, ensuring accurate monitoring even in low light conditions or under conditions of smoke, dust, and haze. Then, a dual-path context-aware network module is designed, including a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule, which improves the quality of basic image features and enhances the overall monitoring accuracy. Simultaneously, the wide-area and contextual relationships between image features are guided and enhanced, enabling accurate monitoring of complex multi-target situations over a wide area. Furthermore, a management judgment module is designed to achieve adjustable monitoring, avoiding false alarms or misreports caused by differences in construction standards and conditions among various ultra-dangerous and large-scale engineering projects, further improving monitoring accuracy. This invention achieves automated monitoring and management of ultra-dangerous and large-scale engineering projects, significantly improving monitoring accuracy in wide-area monitoring situations.Specifically, the acquisition module obtains construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed before being transmitted to the dual-path scene perception network module, eliminating the influence of dim lighting and smog, thus avoiding a decrease in monitoring accuracy caused by the complex construction environment of high-risk projects. The dual-path scene perception network module includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule obtains the basic image features of the construction site image data, performing targeted extraction of basic features from global wide-area scale, multi-local detail scale, and deep scale, improving the quality of basic features and thus improving overall monitoring accuracy. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain corresponding attention. The attention-weighted image features of the mechanism enhance the feature quality of individual target personnel and target areas in a large-scale multi-target situation, thereby improving the monitoring accuracy of single targets. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain attention-enhanced images, which are then transmitted to the management judgment module. This enhances the feature relationships of multiple target personnel and target areas in a wide area, thereby improving the monitoring accuracy in a wide area. The management judgment module performs construction standard judgment on the attention-enhanced images based on a preset construction database, which includes construction drawing data, construction personnel data, and construction safety standard data. If the current construction status is determined to be inconsistent with the construction standards, a construction problem alarm is triggered, realizing the adjustability of monitoring and avoiding false alarms or misreports caused by different construction standards and conditions among various high-risk projects, further improving the monitoring accuracy. This invention realizes automated monitoring and management of high-risk projects, greatly improving the monitoring accuracy in a wide-area monitoring situation.

[0029] Please see Figure 2 The diagram shows a flowchart of the automated monitoring and management method for extremely dangerous and large-scale engineering projects proposed in the second embodiment of the present invention. This automated monitoring and management method for extremely dangerous and large-scale engineering projects includes steps S11 to S16, wherein: Step S11: The acquisition module acquires construction site data in real time and performs image enhancement preprocessing. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. Based on the construction site brightness data, it is determined whether the construction site brightness is less than a preset brightness threshold. If it is determined that the construction site brightness is less than the preset brightness threshold, the construction site image data is subjected to brightness enhancement preprocessing. Based on the construction site haze data, it is determined whether the construction site haze is greater than a preset haze threshold. If it is determined that the construction site haze is greater than the preset haze threshold, the construction site image data is subjected to haze removal preprocessing. It should be noted that the preset light intensity threshold in this embodiment adopts the construction environment light visibility requirements of the emergency plan for super-dangerous and large-scale projects, and the preset haze threshold adopts the dust standard for super-dangerous and large-scale projects and the construction environment smoke and haze requirements of the weather emergency plan for super-dangerous and large-scale projects.

[0030] Step S12: Input the preprocessed construction site image data into the basic feature extraction submodule. The basic feature extraction submodule includes a first convolutional layer, a second convolutional layer and a third convolutional layer. The first convolutional layer obtains global basic features, the second convolutional layer obtains deep basic features, and the third convolutional layer obtains multi-scale basic features. The global basic features, deep basic features and multi-scale basic features are fused to obtain basic image features. Step S13: Input the basic image features into the spatial attention mechanism path of the dual-path attention weighting submodule, and perform spatial attention weighting processing on the basic image features according to the spatial attention mechanism to obtain spatial attention weighted image features; Step S14: Input the basic image features into the channel attention mechanism path of the dual-path attention weighting submodule, and perform channel attention weighting processing on the basic image features according to the channel attention mechanism to obtain channel attention weighted image features; Step S15: Input the spatial attention-weighted image features, channel attention-weighted image features, and basic image features into the linear self-attention submodule, and use them as the linear self-attention submodule's... Q matrix, K Matrix and V The matrix is ​​used to obtain attention-enhanced image features, and the obtained attention-enhanced image is transmitted to the management and judgment module; Step S16: After the management judgment module obtains the attention-enhanced image features, it determines whether the current super-dangerous major construction project meets the construction standards based on the data tags in the preset construction database. The construction area tag, construction structure tag, number of construction personnel tag, and equipment used by construction personnel tag are safety tags, while the construction danger area tag and construction personnel safety equipment tag are danger tags. If the attention-enhanced image lacks safety tags or contains danger tags, an alarm is triggered for super-dangerous major construction project problems.

[0031] In summary, based on the aforementioned automated management method for ultra-dangerous and large-scale engineering projects, the following steps are taken: First, the monitoring images of the construction site are preprocessed to adapt to the complex construction environment of ultra-dangerous and large-scale engineering projects, ensuring accurate monitoring even in low light conditions or under conditions of smoke, dust, and haze. Then, a dual-path context-aware network module is designed, including a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule, which improves the quality of basic image features and enhances the overall monitoring accuracy. Simultaneously, the wide-area and contextual relationships between image features are guided and enhanced, enabling accurate monitoring of complex multi-target situations over a wide area. Furthermore, a management judgment module is designed to achieve adjustable monitoring, avoiding false alarms or misreports caused by differences in construction standards and conditions among various ultra-dangerous and large-scale engineering projects, further improving monitoring accuracy. This invention achieves automated monitoring and management of ultra-dangerous and large-scale engineering projects, significantly improving monitoring accuracy in wide-area monitoring situations.Specifically, the acquisition module obtains construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed before being transmitted to the dual-path scene perception network module, eliminating the influence of dim lighting and smog, thus avoiding a decrease in monitoring accuracy caused by the complex construction environment of high-risk projects. The dual-path scene perception network module includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule obtains the basic image features of the construction site image data, performing targeted extraction of basic features from global wide-area scale, multi-local detail scale, and deep scale, improving the quality of basic features and thus improving overall monitoring accuracy. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain corresponding attention. The attention-weighted image features of the mechanism enhance the feature quality of individual target personnel and target areas in a large-scale multi-target situation, thereby improving the monitoring accuracy of single targets. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain attention-enhanced images, which are then transmitted to the management judgment module. This enhances the feature relationships of multiple target personnel and target areas in a wide area, thereby improving the monitoring accuracy in a wide area. The management judgment module performs construction standard judgment on the attention-enhanced images based on a preset construction database, which includes construction drawing data, construction personnel data, and construction safety standard data. If the current construction status is determined to be inconsistent with the construction standards, a construction problem alarm is triggered, realizing the adjustability of monitoring and avoiding false alarms or misreports caused by different construction standards and conditions among various high-risk projects, further improving the monitoring accuracy. This invention realizes automated monitoring and management of high-risk projects, greatly improving the monitoring accuracy in a wide-area monitoring situation.

[0032] Please see Figure 3 The diagram shown is a structural schematic of the automated monitoring and management system for extremely dangerous and large-scale engineering projects proposed in the third embodiment of the present invention. The system includes: The acquisition module 10 is used to acquire construction site data in real time. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. The dual-path context-aware network module 20 includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain an attention-enhanced image and transmit it to the management judgment module. The management judgment module 30 is used to judge the construction standards of the attention enhancement image according to the preset construction database. The preset construction database includes construction drawing data, construction personnel data and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

[0033] Furthermore, the acquisition module 10 includes: The acquisition and preprocessing unit 101 is used to acquire construction site data in real time by the acquisition module. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. The acquisition and preprocessing unit 101 is used to acquire construction site image data in real time and then transmit it to the dual-path scene perception network module.

[0034] Furthermore, the dual-path context-aware network module 20 includes: The basic feature extraction unit 201 is used in the dual-path scene-aware network module, which includes a basic feature extraction submodule, a dual-channel attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting unit 202 is used by the dual-path attention weighting submodule to perform attention weighting processing on the basic image features according to the attention mechanism, so as to obtain the attention weighted image features corresponding to the attention mechanism; The multi-head linear self-attention unit 203 is used to input the attention-weighted image features and the basic image features into the multi-head linear self-attention submodule to obtain an attention-enhanced image and transmit it to the management and judgment module.

[0035] Furthermore, the management judgment module 30 includes: The management judgment unit 301 is used by the management judgment module to perform construction standard judgment on the attention enhancement image according to the preset construction database. The preset construction database includes construction drawing data, construction personnel data and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

[0036] The present invention also proposes a computer storage medium storing one or more programs, which, when executed by a processor, implement the aforementioned automated monitoring and management method for extremely dangerous large-scale engineering projects.

[0037] The present invention also proposes a computer device, including a memory and a processor, wherein the memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory, so as to realize the above-mentioned automated monitoring and management method for extremely dangerous large-scale projects.

[0038] Those skilled in the art will understand that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can mean any means that can contain stored, communicated, propagated, or transmitted programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0039] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0040] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0041] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0042] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An automated management method for extremely dangerous and large-scale engineering projects, characterized in that, include: The acquisition module acquires construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. The dual-path context-aware network module includes a basic feature extraction submodule, a dual-path attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain an attention-enhanced image, which is then transmitted to the management judgment module. The steps of the basic feature extraction submodule to obtain the basic image features of the construction site image data include: The preprocessed construction site image data is input into the basic feature extraction submodule, which includes a first convolutional layer, a second convolutional layer, and a third convolutional layer. The first convolutional layer includes two identical convolutional branches, each of which includes two 5x5 convolutional blocks. The first convolutional layer obtains global basic features. The second convolutional layer includes a bottleneck structure branch, an inverse bottleneck structure branch, and a residual branch. The bottleneck structure branch includes two 1x1 convolutional blocks and one 3x3 convolutional block, and the inverse bottleneck structure branch includes one 1x1 convolutional block and two 3x3 convolutional blocks. The second convolutional layer obtains depth basic features. The third convolutional layer includes a multi-scale convolutional branch and a residual branch. The multi-scale convolutional branch includes one 1x1 convolutional block, one 2x2 convolutional block, and one 3x3 convolutional block. The third convolutional layer obtains multi-scale basic features. The global basic features, the depth basic features, and the multi-scale basic features are fused to obtain basic image features. The management judgment module performs construction standard judgment on the attention enhancement image based on a preset construction database, which includes construction drawing data, construction personnel data, and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

2. The automated management method for ultra-dangerous and large-scale engineering projects according to claim 1, characterized in that, The acquisition module acquires construction site data in real time, including construction site image data, construction site brightness data, and construction site haze data. The preprocessing steps for the construction site image data include: The acquisition module acquires construction site data in real time and performs image enhancement preprocessing. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. Based on the brightness data of the construction site, it is determined whether the brightness of the construction site is less than a preset brightness threshold. If the brightness of the construction site is determined to be less than the preset brightness threshold, brightness enhancement preprocessing is performed on the construction site image data. The pixel values ​​of the red, green, and blue channels in the construction site image data are extracted, and weighted grayscale processing is performed on the construction site image data according to the following formula: in, This represents the weighted grayscale value of each pixel, where 'a' represents the total number of pixels. This represents the maximum value among the three channels of each pixel. The grayscale weight value, This represents the minimum value among the three channels of each pixel. The grayscale weight values ​​are given, where R, G, and B represent the pixel values ​​of the red, green, and blue channels of each pixel, respectively. After linear normalization of the construction site image data, the image grayscale mean is then processed according to the following formula: in, The grayscale value represents the average grayscale value of a pixel, and this average grayscale value is used as the image grayscale value of the construction site image data. Based on the haze intensity data at the construction site, it is determined whether the haze intensity at the construction site exceeds a preset haze intensity threshold. If the haze intensity exceeds the preset threshold, haze removal preprocessing is performed on the construction site image data. Atmospheric light value A is estimated based on the image grayscale values ​​of the construction site image data. The estimated atmospheric light value A is then used for calculation and guided filtering to obtain the transmittance. Based on the atmospheric light value A and the transmittance The preprocessed image is obtained using the following formula: in, Indicates the preprocessed image. Indicates an unprocessed image. This represents the lower limit threshold of transmittance, and max(·) represents obtaining the maximum value.

3. The automated management method for ultra-dangerous large-scale engineering projects according to claim 1, characterized in that, The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The steps include: The basic image features are input into the spatial attention mechanism path of the dual-path attention weighting submodule. The basic image features are then subjected to spatial attention weighting processing according to the spatial attention mechanism. The spatial attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, a 3x3 convolutional block, a 5x5 convolutional block, and a 7x7 convolutional block. The formula for the spatial attention weighting processing is as follows: in, Represents basic image features, This represents spatial attention-weighted image features. This represents the activation function. Represents a convolution block. Indicates splicing, This represents the global max pooling layer. Indicates a global average pooling layer; To obtain spatial attention-weighted image features.

4. The automated management method for ultra-dangerous and large-scale engineering projects according to claim 1, characterized in that, The step of the dual-path attention weighting submodule performing attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism further includes: The basic image features are input into the channel attention mechanism path of the dual-path attention weighting submodule. Channel attention weighting processing is then applied to the basic image features according to the channel attention mechanism. The channel attention mechanism path of the dual-path attention weighting submodule includes a global average pooling layer, a global max pooling layer, and a multilayer perceptron. The formula for the channel attention weighting processing is as follows: in, Represents basic image features, This represents channel attention-weighted image features. This represents the activation function. This represents a multilayer perceptron. This represents the global max pooling layer. Indicates a global average pooling layer; To obtain channel attention-weighted image features.

5. The automated management method for ultra-dangerous large-scale engineering projects according to claim 1, characterized in that, The step of inputting the attention-weighted image features and the basic image features into the multi-head linear self-attention submodule to obtain an attention-enhanced image and transmitting it to the management and judgment module includes: Spatial attention-weighted image features, channel attention-weighted image features, and basic image features are input into the linear self-attention submodule, and used as the linear self-attention submodule's... Q matrix, K Matrix and V The linear self-attention submodule uses a matrix to obtain attention-enhanced image features, and its mechanism is as follows: in, express Q Matrix weights express K Matrix weights express V Matrix weights Represents depthwise convolution. express Q Matrix and K The dimension of a matrix Represents the transpose matrix; The acquired attention-enhanced image is transmitted to the management and judgment module.

6. The automated management method for ultra-dangerous and large-scale engineering projects according to claim 1, characterized in that, The management judgment module performs construction standard judgment on the attention enhancement image based on a preset construction database, including the following steps: After the management judgment module obtains the attention-enhanced image features, it determines whether the current ultra-dangerous and large-scale construction project complies with construction standards based on the data tags in the preset construction database. The construction data tags include construction area tags and construction structure tags in the construction drawing data, construction personnel quantity tags and construction personnel equipment tags in the construction personnel data, and construction hazard area tags and construction personnel safety equipment tags in the construction safety standard data. The construction area tags, construction structure tags, construction personnel quantity tags, and construction personnel equipment tags are safety tags, while the construction hazard area tags and construction personnel safety equipment tags are hazard tags. If the attention-enhanced image lacks safety tags or contains hazard tags, an alarm for ultra-dangerous and large-scale construction problems is triggered.

7. An automated management system for extremely dangerous and large-scale engineering projects, characterized in that, include: The acquisition module is used to acquire construction site data in real time. The construction site data includes construction site image data, construction site brightness data, and construction site haze data. The construction site image data is preprocessed and then transmitted to the dual-path scene perception network module. A dual-path context-aware network module is provided, comprising a basic feature extraction submodule, a dual-path attention weighting submodule, and a multi-head linear self-attention submodule. The dual-path attention weighting submodule includes a spatial attention mechanism path and a channel attention mechanism path. The basic feature extraction submodule acquires the basic image features of the construction site image data. The dual-path attention weighting submodule performs attention weighting processing on the basic image features according to the attention mechanism to obtain attention-weighted image features corresponding to the attention mechanism. The attention-weighted image features and the basic image features are input into the multi-head linear self-attention submodule to obtain an attention-enhanced image, which is then transmitted to the management and judgment module. The steps of the basic feature extraction submodule to obtain the basic image features of the construction site image data include: The preprocessed construction site image data is input into the basic feature extraction submodule, which includes a first convolutional layer, a second convolutional layer, and a third convolutional layer. The first convolutional layer includes two identical convolutional branches, each of which includes two 5x5 convolutional blocks. The first convolutional layer obtains global basic features. The second convolutional layer includes a bottleneck structure branch, an inverse bottleneck structure branch, and a residual branch. The bottleneck structure branch includes two 1x1 convolutional blocks and one 3x3 convolutional block, and the inverse bottleneck structure branch includes one 1x1 convolutional block and two 3x3 convolutional blocks. The second convolutional layer obtains depth basic features. The third convolutional layer includes a multi-scale convolutional branch and a residual branch. The multi-scale convolutional branch includes one 1x1 convolutional block, one 2x2 convolutional block, and one 3x3 convolutional block. The third convolutional layer obtains multi-scale basic features. The global basic features, the depth basic features, and the multi-scale basic features are fused to obtain basic image features. The management judgment module is used to judge the construction standards of the attention enhancement image based on a preset construction database. The preset construction database includes construction drawing data, construction personnel data, and construction safety standard data. If it is determined that the current construction status does not meet the construction standards, a construction problem alarm is triggered.

8. A storage medium, characterized in that, The storage medium stores one or more programs, which, when executed by a processor, implement the automated management method for ultra-dangerous and large-scale engineering projects as described in any one of claims 1-6.

9. A computer device, characterized in that, The computer device includes a memory and a processor, wherein: The memory is used to store computer programs; When the processor executes the computer program stored in the memory, it implements the automated management method for ultra-dangerous and large-scale engineering projects as described in any one of claims 1-6.

Images