Risk identification method and system in construction process of cement fiberboard

Abstract

The invention discloses a risk identification method and system in a cement fiberboard construction process, and the method comprises the steps: extracting a construction personnel wearing reference image which is marked as a wearing standard from a database, and taking the image as a detection reference; and carrying out multi-scale feature extraction and fine-grained semantic matching analysis on the construction personnel wearing reference image and the construction personnel wearing image acquired on site by adopting an image processing technology based on deep learning so as to intelligently judge whether the construction personnel wear normatively or not and carry out corresponding early warning prompt. In this way, the computing resource demand can be effectively reduced, and the efficiency and accuracy of construction site safety management are improved.

Description

Technical Field

[0001] The present application relates to the field of risk identification, and more specifically, to a method and system for risk identification during the construction process of cement fiberboard. Background Art

[0002] In the construction industry, cement fiberboard is widely used in various construction projects due to its excellent performance. However, with the continuous development of the construction industry, the safety management of construction sites has become a focus of attention. Among them, the implementation and supervision of construction workers' wearing regulations are the key links to ensure the safety of construction sites. However, traditional construction site safety management mostly relies on manual supervision and video surveillance, which is not only inefficient, but also prone to lax supervision due to human factors. Especially in the case of blind spots, traditional manual supervision and video surveillance often cannot effectively cover all risk points, leaving safety hazards on the construction site.

[0003] In this regard, the invention patent with publication number CN114936799A provides a risk identification method and system during the construction of cement fiberboard. It uses a target detection model to perform target detection and category recognition on each object in the image of the construction workers' clothing collected when the personnel enter the site, so as to determine whether the construction protective equipment worn by the workers is complete. When it is detected that the equipment is not complete, a prompt alarm is issued, thereby effectively improving the level of construction safety management. However, since the target detection model requires a large amount of computing resources and labeled data for training to achieve a high recognition accuracy, in a resource-constrained environment such as a construction site, a highly complex model not only increases the hardware cost, but may also affect the effect of real-time warning due to computing delays.

[0004] Therefore, an optimized risk identification method and system during the construction process of cement fiberboard is expected. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a risk identification method and system during the construction process of cement fiberboard, which extracts reference images of construction workers marked as wearing standards from the database as the detection benchmark, and uses deep learning-based image processing technology to perform multi-scale feature extraction and fine-grained semantic matching analysis on the reference images of construction workers and the images of construction workers collected on site, so as to intelligently judge whether the construction workers are wearing standards and make corresponding early warning prompts. In this way, the demand for computing resources can be effectively reduced, and the efficiency and accuracy of construction site safety management can be improved.

[0006] According to one aspect of the present application, a method for identifying risks during the construction of cement fiberboard is provided, which includes: collecting facial images, comparing them with facial information in a database, and collecting images of construction personnel after facial recognition; setting electronic tags in different work areas of the construction site, and after the personnel enter the site, the reading and writing devices they wear sense the electronic tags to obtain the personnel's work area information, and at the same time setting a positioning base station at the construction site to interact with the locator worn by the construction personnel to obtain response information, and obtaining personnel positioning data through algorithm processing, wherein the personnel positioning data includes personnel location information and movement trajectory; analyzing and processing the images of the construction personnel wearing, detecting whether the construction personnel are wearing specified types of construction protective equipment, and issuing an alarm if not all of them are worn; processing the personnel positioning data, and issuing an alarm if they are close to a dangerous area or are in the wrong work area for a long time, characterized in that the images of the construction personnel wearing are analyzed and processed to detect whether the construction personnel are wearing specified types of construction protective equipment, and issuing an alarm if not all of them are worn, including: Extracting construction workers' clothing reference images marked as wearing standards from the database; Performing multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map; Performing feature decoupling and feature screening on the multi-scale feature map of the wearing state of the construction workers and the reference multi-scale feature map of the wearing state of the construction workers to obtain a sequence of selected local feature vectors of the wearing state of the construction workers and a sequence of selected local reference feature vectors of the wearing state of the construction workers; Calculating the inter-sequence semantic matching coefficient between the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel; Based on the comparison between the semantic matching coefficient between the sequences and a preset threshold, it is determined whether to generate an alarm prompt for improper wearing.

[0007] According to another aspect of the present application, a risk identification system during the construction process of cement fiberboard is provided, which comprises: A reference image extraction module is used to extract reference images of construction workers wearing clothes marked as wearing standard from a database; A multi-scale feature extraction module, used to perform multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map; A feature decoupling and feature screening module, used for performing feature decoupling and feature screening on the multi-scale feature map of the wearing state of the construction personnel and the reference multi-scale feature map of the wearing state of the construction personnel to obtain a sequence of selected local feature vectors of the wearing state of the construction personnel and a sequence of selected local reference feature vectors of the wearing state of the construction personnel; An inter-sequence semantic matching coefficient calculation module, used to calculate the inter-sequence semantic matching coefficient between the sequence of the selected construction personnel wearing state local feature vectors and the sequence of the selected construction personnel wearing state local reference feature vectors; The alarm prompt determination module is used to determine whether to generate an alarm prompt for improper wearing based on the comparison between the semantic matching coefficient between the sequences and a preset threshold.

[0008] Compared with the prior art, the risk identification method and system provided by the present application in the construction process of cement fiberboard extracts reference images of construction workers marked as wearing standard from the database as the detection benchmark, and uses image processing technology based on deep learning to perform multi-scale feature extraction and fine-grained semantic matching analysis on the reference images of construction workers and the images of construction workers collected on site, so as to intelligently judge whether the construction workers are wearing standard and give corresponding early warning prompts. In this way, the demand for computing resources can be effectively reduced, and the efficiency and accuracy of construction site safety management can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 is a flow chart of a method for identifying risks during the construction of cement fiberboard according to an embodiment of the present application; Figure 2 A data flow diagram of a risk identification method during the construction process of cement fiberboard according to an embodiment of the present application; Figure 3 It is a flowchart of sub-step S3 of the risk identification method during the construction process of cement fiber board according to an embodiment of the present application; Figure 4 It is a block diagram of a risk identification system during the construction process of cement fiberboard according to an embodiment of the present application. DETAILED DESCRIPTION

[0011] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0012] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0013] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0014] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0015] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0016] Traditional construction site safety management mostly relies on manual supervision and video surveillance, which is not only inefficient, but also prone to lax supervision due to human factors. Especially in the case of blind spots, traditional manual supervision and video surveillance often cannot effectively cover all risk points, thus leaving safety hazards on the construction site. Therefore, an optimized risk identification method and system for the construction process of cement fiberboard is expected.

[0017] In the technical solution of the present application, a method for risk identification during the construction of cement fiberboard is proposed. Figure 1 The present invention is a flow chart of a method for risk identification during the construction of cement fiberboard according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of the risk identification method in the construction process of cement fiber board according to the embodiment of the present application. Figure 1 and Figure 2As shown, according to the embodiment of the present application, the risk identification method during the construction of cement fiberboard includes the following steps: S1, collecting facial images and comparing them with facial information in a database. After facial recognition is passed, collecting images of construction workers; S2, setting electronic tags in different work areas of the construction site. After the personnel enter the site, the reading and writing devices they wear sense the electronic tags to obtain the personnel's work area information. At the same time, a positioning base station is set at the construction site to interact with the locators worn by the construction personnel to obtain response information, and the personnel positioning data is obtained through algorithm processing, wherein the personnel positioning data includes personnel location information and movement trajectory; S3, analyzing and processing the images of the construction personnel to detect whether the construction personnel are wearing specified types of construction protective equipment, and issuing an alarm if they are not wearing all of them; S4, processing the personnel positioning data, and issuing an alarm if they are close to a dangerous area or have been in the wrong work area for a long time.

[0018] In particular, in S1, a face image is collected and compared with the face information in the database. After the face recognition is passed, an image of the construction worker's clothing is collected. It should be understood that before identifying whether the construction worker is wearing standard clothing, the face recognition technology can be used to determine whether the person is a construction worker recorded in the database, and then further determine whether the clothing is standard, thereby improving the safety management of the construction site.

[0019] In particular, the S2 and S3 set electronic tags in different working areas of the construction site. After the personnel enter the site, the reading and writing devices they wear sense the electronic tags to obtain the personnel's working area information. At the same time, a positioning base station is set at the construction site to interact with the locator worn by the construction personnel to obtain response information, and the personnel positioning data is obtained through algorithm processing, wherein the personnel positioning data includes personnel location information and movement trajectory; and the images worn by the construction personnel are analyzed and processed to detect whether the construction personnel are wearing the specified types of construction protective equipment, and if not all of them are worn, an alarm is issued. It should be understood that when the above scheme performs the standardization detection of personnel wearing, the target detection model is used to perform target detection and category recognition on each object in the image of the construction personnel wearing collected when the personnel enter the site, so as to determine whether the construction protective equipment worn by the operator is complete. However, since the target detection model requires a large amount of computing resources and labeled data for training to achieve a high recognition accuracy, in a resource-constrained environment such as a construction site, a high-complexity model not only increases the hardware cost, but also may affect the effect of real-time warning due to computing delays. In view of the above technical problems, the present application provides a lightweight model. In the detection of the standardization of personnel wearing, it is not necessary to confirm the specific type of the equipment worn, but only to confirm whether the personnel wear the equipment in a standard manner. In a specific example of the present application, Figure 3As shown, the S3 includes: S31, extracting reference images of construction workers marked as wearing standards from a database; S32, performing multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a multi-scale feature map of the construction worker wearing state and a reference multi-scale feature map of the construction worker wearing state; S33, performing feature decoupling and feature screening on the multi-scale feature map of the construction worker wearing state and the reference multi-scale feature map of the construction worker wearing state to obtain a sequence of selected local feature vectors of the construction worker wearing state and a sequence of selected local reference feature vectors of the construction worker wearing state; S34, calculating the inter-sequence semantic matching coefficient of the sequence of the selected local feature vectors of the construction worker wearing state and the sequence of the selected local reference feature vectors of the construction worker wearing state; S35, determining whether to generate an alarm prompt for improper wearing based on the comparison between the inter-sequence semantic matching coefficient and a preset threshold.

[0020] Specifically, S31 extracts reference images of construction workers wearing clothing marked as wearing standard from the database. It should be understood that the reference images of construction workers wearing clothing are images that have been reviewed by professionals and confirmed to meet safety standards. Using them as a detection benchmark can effectively reduce the computational complexity and more efficiently and accurately evaluate whether on-site construction workers are wearing standard clothing.

[0021] Specifically, in S32, multi-scale feature extraction is performed on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map. In a specific example of the present application, the construction worker wearing reference image and the construction worker wearing image are respectively input into a wearing state multi-scale feature extractor based on a dilated pyramid network to obtain the construction worker wearing state multi-scale feature map and the construction worker wearing state reference multi-scale feature map. Considering that different types of construction protective equipment may differ in size, shape and color. Therefore, in order to more comprehensively capture the detailed features of the construction worker wearing state, the present application uses a wearing state multi-scale feature extractor based on a dilated pyramid network to perform feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively. The dilated pyramid network (Dilated Pyramid Network) can effectively capture the detailed information at different scales in the image by using convolution kernels of different scales to perform sliding convolution operations on the image, thereby improving the accuracy and comprehensiveness of feature representation, and generating a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map.

[0022] Specifically, the S33 performs feature decoupling and feature screening on the multi-scale feature map of the construction worker's wearing state and the reference multi-scale feature map of the construction worker's wearing state to obtain a sequence of selected local feature vectors of the construction worker's wearing state and a sequence of selected local reference feature vectors of the construction worker's wearing state. In a specific example of the present application, first, the multi-scale feature map of the construction worker's wearing state and the reference multi-scale feature map of the construction worker's wearing state are input into a feature decoupling module to obtain a sequence of local feature vectors of the construction worker's wearing state and a sequence of local reference feature vectors of the construction worker's wearing state; considering that the construction worker's wearing reference image and the construction worker's wearing image usually also contain some background information. Therefore, in order to avoid misjudgment due to inconsistency of background information at the construction site, the present application further performs feature screening and optimization processing on the multi-scale feature map of the construction worker's wearing state and the reference multi-scale feature map of the construction worker's wearing state. Specifically, first, the multi-scale feature map of the wearing state of the construction workers and the multi-scale feature map of the reference wearing state of the construction workers are input into the feature decoupling module for feature map splitting processing, and the feature map is feature dispersed along the spatial dimension to form a feature representation of pixel granularity, thereby obtaining a sequence of local feature vectors of the wearing state of the construction workers and a sequence of local reference feature vectors of the wearing state of the construction workers. Further, the sequence of local feature vectors of the wearing state of the construction workers and the sequence of local reference feature vectors of the wearing state of the construction workers are input into the sequence screening module based on feature analysis to obtain the sequence of local feature vectors of the wearing state of the construction workers after selection and the sequence of local reference feature vectors of the wearing state of the construction workers after selection. By introducing the sequence screening module based on feature analysis, the sequence of local feature vectors of the wearing state of the construction workers and the sequence of local reference feature vectors of the wearing state of the construction workers are feature screened, so as to filter out irrelevant background information and retain effective features related to the wearing state of the construction workers.

[0023] In an embodiment of the present application, the sequence of local feature vectors of the construction workers' wearing states and the sequence of local reference feature vectors of the construction workers' wearing states are input into a sequence screening module based on feature analysis to obtain the sequence of local feature vectors of the selected construction workers' wearing states and the sequence of local reference feature vectors of the selected construction workers' wearing states, including: first, calculating the characteristic energy distribution spectrum vector of each local feature vector of the construction workers' wearing states in the sequence of local feature vectors of the construction workers' wearing states to obtain the sequence of local characteristic energy distribution spectrum vectors of the construction workers' wearing states; calculating the characteristic energy distribution spectrum vector of each local feature vector in the input sequence to reveal the characteristic coordinated distribution between each local feature and other local features; then, calculating the positional mean vector of the sequence of local characteristic energy distribution spectrum vectors of the construction workers' wearing states as the center vector of the energy spectrum distribution; calculating The mean vector is used to reveal the overall energy distribution trend of the feature graph, and the energy spectrum distribution center vector is obtained; then, the Mahalanobis distance between the energy spectrum distribution center vector and each local characteristic energy distribution spectrum vector of the construction worker's wearing state in the sequence of the local characteristic energy distribution spectrum vector of the construction worker's wearing state is calculated as the energy distribution spectrum span factor to obtain the sequence of energy distribution spectrum span factors; here, by calculating the difference between the energy spectrum distribution center vector and each characteristic energy distribution spectrum vector, the proximity of each local feature to the overall distribution is evaluated, and the corresponding energy distribution spectrum span factor is generated; finally, based on the comparison between each energy distribution spectrum span factor in the sequence of energy distribution spectrum span factors and a preset span threshold, the local characteristic vector of the construction worker's wearing state corresponding to the energy distribution spectrum span factor less than the preset span threshold is selected to form the sequence of the selected local characteristic vector of the construction worker's wearing state. That is, the sequence of original local feature vectors is screened based on the energy distribution spectrum span factor to remove redundant information that deviates from the overall distribution and retain useful features close to the overall distribution, so as to obtain a sequence of local feature vectors of the selected wearing status of construction workers and a sequence of local reference feature vectors of the selected wearing status of construction workers.

[0024] Among them, the process of calculating the characteristic energy distribution spectrum vector of each construction worker's wearing state local feature vector in the sequence of the construction workers' wearing state local feature vector to obtain the sequence of the construction workers' wearing state local feature energy distribution spectrum vector includes: calculating the characteristic distribution energy synergy representation vector between the i-th construction worker's wearing state local feature vector in the sequence of the construction workers' wearing state local feature vector and each other construction workers' wearing state local feature vector in a manner of taking the maximum eigenvalue element by element to obtain the sequence of characteristic distribution energy synergy representation vectors; calculating the characteristic distribution energy synergy factor of each characteristic distribution energy synergy representation vector in the sequence of characteristic distribution energy synergy representation vectors to obtain the characteristic energy distribution spectrum vector of the i-th construction worker's wearing state local feature vector composed of multiple characteristic distribution energy synergy factors. More specifically, the characteristic distribution energy synergy factor of each characteristic distribution energy synergy representation vector in the sequence of the characteristic distribution energy synergy representation vector is calculated to obtain the characteristic energy distribution spectrum vector of the local characteristic vector of the wearing state of the i-th construction worker composed of multiple characteristic distribution energy synergy factors, including: calculating the square of the difference between the maximum eigenvalue of the characteristic distribution energy synergy representation vector and its characteristic mean, the characteristic variance of the characteristic distribution energy synergy representation vector, and the weighted sum of the bias terms, and then adding the variance of the characteristic distribution energy synergy representation vector and the bias term and dividing the result by the weighted sum to obtain the characteristic distribution energy synergy factor.

[0025] In summary, in the above embodiment, taking the sequence of local feature vectors of the wearing state of the construction workers as an example, the sequence of local feature vectors of the wearing state of the construction workers is input into the sequence screening module based on feature analysis to obtain the sequence of local feature vectors of the wearing state of the construction workers after selection, including: processing the sequence of local feature vectors of the wearing state of the construction workers with the following feature selection optimization formula to obtain the sequence of local feature vectors of the wearing state of the construction workers after selection, wherein the feature selection optimization formula is:

[0026]

[0027]

[0028]

[0029]

[0030]

[0031] in, and The first and second local feature vectors of the construction workers' wearing status are represented in the sequence of and The local feature vector of the wearing status of the construction workers, It means taking the maximum eigenvalue of two eigenvectors element by element. Indicates the The local feature vector of the wearing state of the construction workers and the The feature distribution energy synergistic representation vector between the local feature vectors of the wearing status of the construction workers is , and represent the characteristic mean, characteristic variance and maximum eigenvalue of the characteristic distribution energy collaborative representation vector respectively, is the bias term, , , and The first and second local feature vectors of the construction workers' wearing status are represented in the sequence of The local feature vector of the wearing state of the construction workers is related to the first, second, and The characteristic distribution energy synergy factor between the local characteristic vectors of the wearing status of the construction workers, The value of is the number of local feature vectors of the construction workers’ wearing status, Indicates The local characteristic energy distribution spectrum vector of the wearing state of each construction worker, The sequence of local characteristic energy distribution spectrum vectors representing the wearing status of construction workers, represents the center vector of energy spectrum distribution, Indicates The covariance matrix between the local characteristic energy distribution spectrum vector of the wearing state of each construction worker and the center vector of the energy spectrum distribution, represents the distance metric function, Indicates The energy distribution spectrum span factor, Indicates the preset span threshold, Represents the sequence of local feature vectors of the selected wearing status of the construction workers.

[0032] Specifically, the S34 calculates the inter-sequence semantic matching coefficient of the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel. In a specific example of the present application, the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel are input into the inter-sequence matching network based on dense interactive distribution to obtain the inter-sequence semantic matching coefficient. That is, taking the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel as feature basis, by matching and analyzing the two, it is intelligently judged whether the construction personnel are wearing in accordance with the specification. In order to improve the accuracy of the matching analysis, the present application introduces an inter-sequence matching network based on dense interactive distribution to perform this matching analysis process.

[0033] In an embodiment of the present application, the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel are input into a sequence matching network based on dense interactive distribution to obtain the semantic matching coefficient between the sequences, including: first, calculating the energy synergy response value between any pair of the local feature vectors of the wearing state of the construction personnel and the local reference feature vectors of the wearing state of the construction personnel in the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel to obtain a feature synergy energy fine-grained global distribution matrix composed of multiple energy synergy response values; Through the energy synergy response value, the fine-grained semantic association between feature vectors is captured, and the matching degree between two local features is quantified to form a feature synergy energy fine-grained global distribution matrix; then, the feature synergy energy fine-grained global distribution matrix is ​​subjected to hole convolution coding to obtain a feature synergy energy fine-grained global distribution feature matrix; here, by performing hole convolution coding on the feature synergy energy fine-grained global distribution matrix, deeper feature matching pattern information is extracted to capture feature interactions at a longer distance; finally, the feature synergy energy fine-grained global distribution feature matrix is ​​input into the inter-sequence matching degree measurement network for decoding processing to obtain the inter-sequence semantic matching coefficient. Here, the feature synergy energy fine-grained global distribution matrix after hole convolution processing is input into the inter-sequence matching degree measurement network for feature decoding, so as to comprehensively analyze the inter-sequence global interaction matching information in the matrix, and perform quantitative measurement of the semantic similarity between sequences, thereby generating the inter-sequence semantic matching coefficient.

[0034] Among them, the process of calculating the energy synergistic response value between any pair of local feature vectors of the construction worker's wearing state and the local reference feature vector of the construction worker's wearing state in the sequence of the selected local feature vectors of the construction worker's wearing state and the sequence of the selected local reference feature vectors of the construction worker's wearing state to obtain a feature synergistic energy fine-grained global distribution matrix composed of multiple energy synergistic response values ​​includes: calculating the positional division between the selected local feature vector of the construction worker's wearing state and the selected local reference feature vector of the construction worker's wearing state to obtain a construction worker wearing state detection-reference local feature synergistic response vector; calculating the weighted sum between the feature variance and the bias parameter of the construction worker wearing state detection-reference local feature synergistic response vector to obtain a first intermediate term, and calculating the square of the difference between the maximum eigenvalue and the feature mean of the construction worker wearing state detection-reference local feature synergistic response vector, the feature variance of the construction worker wearing state detection-reference local feature synergistic response vector and the weighted sum of the bias term to obtain a second intermediate term; calculating the first intermediate term divided by the second intermediate term to obtain the energy synergistic response value.

[0035] In summary, in the above embodiment, the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel are input into the sequence matching network based on dense interactive distribution to obtain the semantic matching coefficient between sequences, including: processing the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel with the following sequence matching metric formula to obtain the semantic matching coefficient between sequences, wherein the sequence matching metric formula is:

[0036]

[0037]

[0038]

[0039]

[0040]

[0041] in, The first in the sequence of local feature vectors of the selected construction worker's wearing state The local feature vector of the construction workers’ wearing status after selection, The first in the sequence of the local reference feature vectors of the selected construction worker's wearing state The local reference feature vector of the selected construction workers’ wearing status, Indicates the The local feature vector of the wearing state of the construction workers after the selection and the The construction worker wearing state detection-reference local feature collaborative response vector between the selected construction worker wearing state local reference feature vectors, , and They respectively represent the maximum eigenvalue, feature mean and feature variance of the construction worker wearing status detection-reference local feature collaborative response vector, represents the energy synergistic response value of the construction worker wearing state detection-reference local feature synergistic response vector, represents the fine-grained global distribution matrix of the feature synergy energy, , , , and They respectively represent the energy synergy response values ​​of the corresponding positions in the fine-grained global distribution matrix of the feature synergy energy, The value of is the number of feature vectors in the sequence of local feature vectors of the wearing status of the construction workers after the selection, The value of is the number of feature vectors in the sequence of the selected local reference feature vectors of the wearing status of the construction workers, represents the dilated convolution processing, represents the fine-grained global distribution feature matrix of the feature synergy energy, represents the sequence matching measurement network, represents the semantic matching coefficient between the sequences.

[0042] In particular, in this specific example, the feature collaborative energy fine-grained global distribution feature matrix represents the global vector granularity association pattern characteristics between the sequence of the selected construction worker wearing state local feature vectors and the sequence of the selected construction worker wearing state local reference feature vectors. However, the feature collaborative energy fine-grained global distribution feature matrix has a selective feature interaction state space representation under the distribution field due to the fine-grained interaction encoding difference between the construction worker wearing state local feature vector and the construction worker wearing state local reference feature vector, resulting in regression comprehension imbalance and affecting the accuracy of the decoding result.

[0043] Based on this, in a specific example, the feature collaborative energy fine-grained global distribution feature matrix is ​​input into the inter-sequence matching degree measurement network for decoding processing to obtain the inter-sequence semantic matching coefficient, including: Determine the number of zero eigenvalues ​​in the characteristic synergistic energy fine-grained global distribution characteristic matrix, and respectively calculate the reciprocal of the logarithm of the number of zero eigenvalues ​​with base two and the exponential value of the reciprocal of the number of zero eigenvalues ​​with base natural constants to obtain a first characteristic synergistic energy fine-grained global distribution representation value and a second characteristic synergistic energy fine-grained global distribution representation value; Expanding the feature collaborative energy fine-grained global distribution feature matrix into a feature collaborative energy fine-grained global distribution feature vector; Calculate a power function of each eigenvalue of the characteristic synergistic energy fine-grained global distribution eigenvector with the inverse of the number of zero eigenvalues ​​as an exponent to obtain a characteristic synergistic energy fine-grained global distribution structural state vector; Performing a dot multiplication of the feature collaborative energy fine-grained global distribution structure state vector and the first feature collaborative energy fine-grained global distribution representation value to obtain a feature collaborative energy fine-grained global distribution information representation vector; Performing dot multiplication of the autocorrelation matrix of the feature synergy energy fine-grained global distribution structure state vector with the second feature synergy energy fine-grained global distribution representation value and weight hyperparameter respectively to obtain a feature synergy energy fine-grained global distribution regression understanding matrix; Performing matrix multiplication of the feature synergy energy fine-grained global distribution information representation vector and the feature synergy energy fine-grained global distribution regression understanding matrix to obtain an optimized feature synergy energy fine-grained global distribution feature vector; The optimized feature synergy energy fine-grained global distribution feature vector is decoded through the inter-sequence matching degree measurement network to obtain the inter-sequence semantic matching coefficient.

[0044] That is, the first feature collaborative energy fine-grained global distribution representation value and the second feature collaborative energy fine-grained global distribution representation value are expressed as: ; in, is the number of zero eigenvalues ​​in the fine-grained global distribution eigenvector of the feature synergy energy, is the first feature synergy energy fine-grained global distribution representation value, is the second feature synergy energy fine-grained global distribution representation value, It is an exponential function with a natural constant as its base.

[0045] And the optimized feature synergy energy fine-grained global distribution feature vector is expressed as:

[0046] in, is the fine-grained global distribution feature vector of the feature synergy energy, is the weight hyperparameter, is the feature synergy energy fine-grained global distribution regression understanding matrix, is the feature synergy energy fine-grained global distribution information representation vector, is the number of zero eigenvalues ​​in the fine-grained global distribution eigenvector of the feature synergy energy, is the first feature synergy energy fine-grained global distribution representation value, is the second feature synergy energy fine-grained global distribution representation value, represents a power function that calculates each eigenvalue of the fine-grained global distribution eigenvector of the characteristic cooperative energy with the inverse of the number of zero eigenvalues ​​as the exponent, represents the transpose of a vector, It is the point multiplication by position. is matrix multiplication, is the optimized feature synergy energy fine-grained global distribution feature vector.

[0047] Based on this, in the above-mentioned preferred example, by modeling the vector field zero dimension of the feature synergistic energy fine-grained global distribution feature vector after the feature synergistic energy fine-grained global distribution feature matrix is ​​expanded as the global field redundant dependency, the effective long-range modeling of the feature synergistic energy fine-grained global distribution feature matrix under the field linear complexity is realized, so that in the feature regression process of the feature synergistic energy fine-grained global distribution feature matrix, the structured correlation self-distillation between information representation and decoding regression understanding is performed in the selective feature enhancement state space based on the feature distribution, so as to maintain the overall information complexity of the feature distribution of the feature synergistic energy fine-grained global distribution feature matrix while promoting the long-range dynamic regression understanding balance of its feature convergence, so as to improve the regression understanding consistency between the feature regression and feature extraction processes, and improve the accuracy of the inter-sequence semantic matching coefficient obtained by the decoding processing of the inter-sequence matching degree measurement network of the feature synergistic energy fine-grained global distribution feature matrix.

[0048] Specifically, in S35, based on the comparison between the semantic matching coefficient between the sequences and the preset threshold, it is determined whether to generate an alarm prompt for irregular wearing. In a specific example, in response to the semantic matching coefficient between the sequences being equal to or higher than the preset threshold, an alarm prompt for irregular wearing is generated; otherwise, an alarm prompt for irregular wearing is not generated.

[0049] In particular, the S4 processes the personnel location data and issues an alarm if the personnel are close to a dangerous area or are in a wrong working area for a long time. By ensuring that personnel comply with safety regulations and work area restrictions, the efficiency and accuracy of construction site safety management are improved.

[0050] In summary, the risk identification method during the construction of cement fiberboard according to the embodiment of the present application is explained, which extracts reference images of construction workers marked as wearing standard from the database as the detection benchmark, and uses image processing technology based on deep learning to perform multi-scale feature extraction and fine-grained semantic matching analysis on the reference images of construction workers and the images of construction workers collected on site, so as to intelligently judge whether the construction workers are wearing standard and give corresponding early warning prompts. In this way, the demand for computing resources can be effectively reduced, and the efficiency and accuracy of construction site safety management can be improved.

[0051] Furthermore, a risk identification system during the construction process of cement fiberboard is also provided.

[0052] Figure 4 FIG. 1 is a block diagram of a risk identification system during the construction process of cement fiberboard according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the risk identification system 300 in the construction process of cement fiberboard includes: a reference image extraction module 310, which is used to extract the construction worker wearing reference image marked as wearing standard from the database; a multi-scale feature extraction module 320, which is used to perform multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain the construction worker wearing state multi-scale feature map and the construction worker wearing state reference multi-scale feature map; a feature decoupling and feature screening module 330, which is used to perform feature decoupling and feature screening on the construction worker wearing state multi-scale feature map and the construction worker wearing state reference multi-scale feature map to obtain a sequence of selected construction worker wearing state local feature vectors and a sequence of selected construction worker wearing state local reference feature vectors; a sequence semantic matching coefficient calculation module 340, which is used to calculate the sequence semantic matching coefficient of the sequence of the selected construction worker wearing state local feature vectors and the sequence of the selected construction worker wearing state local reference feature vectors; an alarm prompt determination module 350, which is used to determine whether to generate an alarm prompt of irregular wearing based on the comparison between the sequence semantic matching coefficient and the preset threshold.

[0053] As described above, the risk identification system 300 during the construction of cement fiberboard according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a risk identification algorithm during the construction of cement fiberboard. In a possible implementation, the risk identification system 300 during the construction of cement fiberboard according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the risk identification system 300 during the construction of cement fiberboard can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the risk identification system 300 during the construction of cement fiberboard can also be one of the many hardware modules of the wireless terminal.

[0054] Alternatively, in another example, the risk identification system 300 during the construction of the cement fiber board and the wireless terminal may be separate devices, and the risk identification system 300 during the construction of the cement fiber board may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0055] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for identifying risks during the construction of cement fiberboard, comprising: Collect facial images and compare them with facial information in the database. After facial recognition is passed, collect images of the construction workers wearing the equipment. Electronic tags are set up in different working areas of the construction site. After the personnel enter the site, the reading and writing devices they wear sense with the electronic tags to obtain the personnel's working area information. At the same time, a positioning base station is set up at the construction site to interact with the locator worn by the construction personnel to obtain response information, and the personnel positioning data is obtained through algorithm processing, wherein the personnel positioning data includes the personnel position information and movement trajectory; the images of the construction personnel wearing are analyzed and processed to detect whether the construction personnel are wearing the specified types of construction protection equipment, and if not all of them are worn, an alarm is issued; the personnel positioning data is processed, and an alarm is issued if the personnel are close to the dangerous area or are in the wrong working area for a long time, which is characterized in that the images of the construction personnel wearing are analyzed and processed to detect whether the construction personnel are wearing the specified types of construction protection equipment, and if not all of them are worn, an alarm is issued, including: Extracting construction workers' clothing reference images marked as wearing standards from the database; Performing multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map; Performing feature decoupling and feature screening on the multi-scale feature map of the wearing state of the construction workers and the reference multi-scale feature map of the wearing state of the construction workers to obtain a sequence of selected local feature vectors of the wearing state of the construction workers and a sequence of selected local reference feature vectors of the wearing state of the construction workers; Calculating the inter-sequence semantic matching coefficient between the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel; Based on the comparison between the semantic matching coefficient between the sequences and a preset threshold, it is determined whether to generate an alarm prompt for improper wearing.

2. The risk identification method during the construction process of cement fiber board according to claim 1 is characterized in that: The multi-scale feature extraction is performed on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map, including: The construction worker wearing reference image and the construction worker wearing image are respectively input into a wearing state multi-scale feature extractor based on a dilated pyramid network to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map.

3. The risk identification method during the construction process of cement fiberboard according to claim 2 is characterized in that: Performing feature decoupling and feature screening on the multi-scale feature map of the wearing state of the construction workers and the reference multi-scale feature map of the wearing state of the construction workers to obtain a sequence of selected local feature vectors of the wearing state of the construction workers and a sequence of selected local reference feature vectors of the wearing state of the construction workers, including: Inputting the multi-scale feature map of the wearing state of the construction workers and the multi-scale feature map of the wearing state of the construction workers into a feature decoupling module to obtain a sequence of local feature vectors of the wearing state of the construction workers and a sequence of local reference feature vectors of the wearing state of the construction workers; The sequence of the local feature vectors of the wearing status of the construction workers and the sequence of the local reference feature vectors of the wearing status of the construction workers are input into a sequence screening module based on feature analysis to obtain the sequence of the local feature vectors of the wearing status of the selected construction workers and the sequence of the local reference feature vectors of the wearing status of the selected construction workers.

4. The method for identifying risks during the construction of cement fiberboard according to claim 3, characterized in that: Inputting the sequence of the local feature vectors of the wearing state of the construction workers and the sequence of the local reference feature vectors of the wearing state of the construction workers into a sequence screening module based on feature analysis to obtain the sequence of the local feature vectors of the wearing state of the construction workers after selection and the sequence of the local reference feature vectors of the wearing state of the construction workers after selection, including: Calculating the characteristic energy distribution spectrum vector of each construction worker wearing state local feature vector in the sequence of the construction worker wearing state local feature vectors to obtain a sequence of the construction worker wearing state local feature energy distribution spectrum vectors; Calculate the position mean vector of the sequence of local characteristic energy distribution spectrum vectors of the wearing state of the construction worker as the center vector of the energy spectrum distribution; Calculating the Mahalanobis distance between the energy spectrum distribution center vector and each construction worker wearing state local characteristic energy distribution spectrum vector in the sequence of the construction worker wearing state local characteristic energy distribution spectrum vector as the energy distribution spectrum span factor to obtain a sequence of energy distribution spectrum span factors; Based on the comparison between each energy distribution spectrum span factor in the sequence of energy distribution spectrum span factors and a preset span threshold, the local feature vectors of the construction workers' wearing status corresponding to the energy distribution spectrum span factors that are smaller than the preset span threshold are selected to form the sequence of the selected local feature vectors of the construction workers' wearing status.

5. The method for identifying risks during the construction of cement fiberboard according to claim 4, characterized in that: Calculating the characteristic energy distribution spectrum vector of each construction worker wearing state local feature vector in the sequence of the construction worker wearing state local feature vectors to obtain a sequence of the construction worker wearing state local feature energy distribution spectrum vectors, including: Calculate the feature distribution energy collaborative representation vector between the i-th construction worker wearing state local feature vector in the sequence of the construction worker wearing state local feature vector and each other construction worker wearing state local feature vector in a manner of taking the maximum eigenvalue element by element to obtain a sequence of feature distribution energy collaborative representation vectors; The characteristic distribution energy synergy factor of each characteristic distribution energy synergy representation vector in the sequence of the characteristic distribution energy synergy representation vector is calculated to obtain a characteristic energy distribution spectrum vector of the local characteristic vector of the wearing state of the i-th construction worker composed of multiple characteristic distribution energy synergy factors.

6. The method for identifying risks during the construction of cement fiberboard according to claim 5, characterized in that: Calculating the characteristic distribution energy synergy factor of each characteristic distribution energy synergy representation vector in the sequence of the characteristic distribution energy synergy representation vector to obtain the characteristic energy distribution spectrum vector of the local characteristic vector of the wearing state of the i-th construction worker composed of multiple characteristic distribution energy synergy factors, including: The square of the difference between the maximum eigenvalue of the feature distribution energy synergistic representation vector and its eigenmean, the feature variance of the feature distribution energy synergistic representation vector, and the weighted sum of the bias terms are calculated, and then the variance of the feature distribution energy synergistic representation vector and the bias term are added and divided by the weighted sum to obtain the feature distribution energy synergy factor.

7. The method for identifying risks during the construction of cement fiberboard according to claim 6, characterized in that: Calculating the inter-sequence semantic matching coefficient of the sequence of the selected local feature vectors of the wearing state of the construction personnel and the sequence of the selected local reference feature vectors of the wearing state of the construction personnel, including: The sequence of the selected local feature vectors of the wearing state of the construction workers and the sequence of the selected local reference feature vectors of the wearing state of the construction workers are input into a sequence matching network based on dense interaction distribution to obtain the semantic matching coefficient between the sequences.

8. The method for identifying risks during the construction of cement fiberboard according to claim 7, characterized in that: Inputting the sequence of the selected local feature vectors of the wearing state of the construction workers and the sequence of the selected local reference feature vectors of the wearing state of the construction workers into a sequence matching network based on dense interaction distribution to obtain the semantic matching coefficient between the sequences, including: Calculating the energy synergy response value between any pair of the local feature vectors of the construction worker's wearing state and the local reference feature vectors of the construction worker's wearing state in the sequence of the selected local feature vectors of the construction worker's wearing state and the sequence of the selected local reference feature vectors of the construction worker's wearing state to obtain a feature synergy energy fine-grained global distribution matrix composed of multiple energy synergy response values; Performing dilated convolution coding on the feature collaborative energy fine-grained global distribution matrix to obtain a feature collaborative energy fine-grained global distribution feature matrix; The feature collaborative energy fine-grained global distribution feature matrix is ​​input into the inter-sequence matching degree measurement network for decoding processing to obtain the inter-sequence semantic matching coefficient.

9. The method for identifying risks during the construction of cement fiberboard according to claim 8, characterized in that: Calculating the energy synergy response value between any pair of the local feature vectors of the construction worker's wearing state and the local reference feature vectors of the construction worker's wearing state in the sequence of the selected local feature vectors of the construction worker's wearing state and the sequence of the selected local reference feature vectors of the construction worker's wearing state to obtain a feature synergy energy fine-grained global distribution matrix composed of multiple energy synergy response values, including: Calculate the positional division between the selected local feature vector of the wearing state of the construction worker and the selected local reference feature vector of the wearing state of the construction worker to obtain a construction worker wearing state detection-reference local feature collaborative response vector; Calculate the weighted sum between the characteristic variance and the bias parameter of the construction worker wearing state detection-reference local feature collaborative response vector to obtain a first intermediate term, and calculate the weighted sum between the maximum eigenvalue and the characteristic mean of the construction worker wearing state detection-reference local feature collaborative response vector, the characteristic variance of the construction worker wearing state detection-reference local feature collaborative response vector, and the bias term to obtain a second intermediate term; The first intermediate term is divided by the second intermediate term to obtain the energy synergy response value.

10. A risk identification system during the construction process of cement fiberboard, characterized in that: include: A reference image extraction module is used to extract reference images of construction workers wearing clothes marked as wearing standard from a database; A multi-scale feature extraction module, used to perform multi-scale feature extraction on the construction worker wearing reference image and the construction worker wearing image respectively to obtain a construction worker wearing state multi-scale feature map and a construction worker wearing state reference multi-scale feature map; A feature decoupling and feature screening module, used for performing feature decoupling and feature screening on the multi-scale feature map of the wearing state of the construction personnel and the reference multi-scale feature map of the wearing state of the construction personnel to obtain a sequence of selected local feature vectors of the wearing state of the construction personnel and a sequence of selected local reference feature vectors of the wearing state of the construction personnel; An inter-sequence semantic matching coefficient calculation module, used to calculate the inter-sequence semantic matching coefficient between the sequence of the selected construction personnel wearing state local feature vectors and the sequence of the selected construction personnel wearing state local reference feature vectors; The alarm prompt determination module is used to determine whether to generate an alarm prompt for improper wearing based on the comparison between the semantic matching coefficient between the sequences and a preset threshold.

Images