Construction safety information management method and system

Abstract

The present invention discloses a construction safety information management method and system, which belongs to the field of construction safety information management. It includes: an information collection unit, an information processing unit, an information upload unit and a user management unit. With the help of two methods, manual inspection and intelligent technology, the original construction safety information is collected; with the help of deep learning technology, key construction safety information, such as text keywords, image key information, etc., is automatically extracted; the original information is hashed, and the original information hash value and key information are stored in the blockchain network; the local system front end reads the original information hash value and key information on the chain, and visualizes its statistics, and its related statistical information helps managers quickly identify key issues in safety management, take countermeasures in time, and prevent risks. The present invention not only realizes the automatic extraction and chain storage of key construction safety information, but also improves the information retrieval efficiency and safety performance of the system.

Description

Technical Field

[0001] The present invention belongs to the technical field of construction safety information management, and more specifically, relates to a construction safety information management method and system. Background Art

[0002] In order to ease the pressure of housing, my country's construction industry is showing a rapid growth trend. Construction information construction is an important idea to ensure the safe construction of intelligent construction, among which safety information collection and management are the main contents of information construction and the core of the safety information management system. Due to the current decentralized storage of construction safety information and the lack of certification by third-party authoritative organizations, the authenticity of data is difficult to guarantee and there is a risk of tampering. Real-time information management of construction sites before accidents occurs has become a difficult task. This problem has persisted in the construction industry for decades and has seriously hindered the improvement of construction safety management. In order to improve the trust of all parties, it is necessary to form a system with traceability and transparency, which depends not only on the integrity of data collection, but also on the unforgeability and transparency of data transmission and storage. However, in the construction industry, various stakeholders cannot easily capture equipment information, and it is difficult to transmit and store it securely. Once accountability is involved, the responsible party has the motivation to tamper with the data to avoid responsibility.

[0003] With the development of "Internet" technology, safety information management system has become a powerful tool to assist the development of information-based construction. However, the current traditional safety information management system still has some shortcomings and defects, especially it cannot effectively solve the current problems of decentralized storage, poor reliability and poor traceability of construction safety information. Therefore, a unified, secure and transparent system jointly managed by stakeholders is urgently needed to manage safety information; obviously, the traditional centralized information management system is not an adaptable solution. Summary of the invention

[0004] The integration of "deep learning + blockchain technology" provides a new solution to the above problems. Deep learning uses its mechanism of imitating the human brain to interpret data instead of manually acquiring features, thereby automatically, quickly and efficiently mining and analyzing information. Blockchain uses a series of complex mathematical algorithms such as hash encryption algorithms, timestamps, distributed consensus, and smart contracts to build a credible technical system to ensure the openness, transparency, full traceability and non-tamperability of information, and establish a "trust" network between machines. The integration of deep learning technology and blockchain technology can not only realize the automatic extraction and chaining of construction safety information, but also improve the information retrieval efficiency and safety performance of the system.

[0005] In view of the defects of the related technologies, the purpose of the present invention is to provide a construction safety information management method and system, which aims to manage the construction safety data based on deep learning and blockchain technology, ensure the reliability, traceability and transparency of construction safety information, and improve the retrieval performance and practical performance of the management system.

[0006] To achieve the above object, one aspect of the present invention provides a construction safety information management method, comprising the following steps:

[0007] S1. Collect original information on construction safety during the construction process;

[0008] S2. Classify the format of the original information, use a deep learning algorithm to process the original information according to different formats to automatically extract key construction safety information, and convert the original information of all formats into hash values;

[0009] S3. Broadcast the hash value of the original information and the key information processed by deep learning to each node in the blockchain network, and upload all the original information to the cloud server to complete the sorting and statistics;

[0010] S4. During the construction process, call the blockchain network interface and query and analyze the construction safety management information from the blockchain network.

[0011] Furthermore, the step S2 specifically includes:

[0012] The original information is classified into data formats, the classification types include text data, image data and video data, and the text data is divided into training text samples and test text samples, and the image data is divided into training image samples and test image samples;

[0013] Construct a text keyword extraction model based on the KEYBERT algorithm to identify keywords in construction safety text data;

[0014] Build a construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm to identify text descriptions in construction safety image data;

[0015] The text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are trained using training text samples and training picture samples respectively;

[0016] The test text samples and test picture samples are respectively input into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm, and the error value between the output test results and the actual results is analyzed. If the error value is within the preset range, the result obtained after deep learning processing is sent to the blockchain network; otherwise, the samples are collected again for training until the error value is within the preset range.

[0017] Furthermore, building a text keyword extraction model based on the KEYBERT algorithm includes:

[0018] Embed the document description based on the BERT model to obtain the document vector;

[0019] Use the bag-of-words technique to extract keywords and expressions from the same document, then embed the keywords based on the BERT model to obtain the keyword vector;

[0020] Based on the cosine similarity between the document vector and the keyword vector, extracting the most similar keywords according to the highest cosine similarity score;

[0021] The parameters of the text keyword extraction model based on the KEYBERT algorithm are automatically adjusted regularly according to actual construction safety information.

[0022] Furthermore, the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm is constructed, including:

[0023] Input the image to be detected, and extract features of the input image through the encoder;

[0024] The image semantic information and word sequence features are captured and processed through the decoder;

[0025] Output statement description about image scene information and global information;

[0026] The parameters of the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are automatically adjusted regularly according to actual construction safety information.

[0027] Furthermore, the step S3 specifically includes:

[0028] Wirelessly upload construction safety information data to the cloud server;

[0029] Calling the deep learning algorithm in step S2 to further process and store the data in the cloud, sending the key construction safety information processed by deep learning and the hash value of the original construction safety information to the blockchain network, and uploading all the original information to the cloud server;

[0030] In the blockchain network, data is put into a block according to a certain structure, and legal block header information is generated. Then the generated block is verified by members to generate a new block.

[0031] Smart contracts execute transaction commands, and transaction information is broadcast to each node of the blockchain network.

[0032] Another aspect of the present invention further provides a construction safety information management system, comprising:

[0033] Information collection unit, collecting original information on construction safety during the construction process;

[0034] An information processing unit, which classifies the format of the original information, processes the original information accordingly according to different formats using a deep learning algorithm to automatically extract key construction safety information, and converts the original information of all formats into hash values;

[0035] The information upload unit broadcasts the hash value of the original information and the key information processed by deep learning to each node in the blockchain network, and uploads all the original information to the cloud server to complete the sorting and statistics;

[0036] User management unit, during the construction process, users call the blockchain network interface and query and analyze construction safety management information from the blockchain network.

[0037] Furthermore, the information processing unit specifically includes:

[0038] A classification module classifies the original information into data formats, the classification types include text data, image data and video data, and divides the text data into training text samples and test text samples, and divides the image data into training image samples and test image samples;

[0039] The first building module builds a text keyword extraction model based on the KEYBERT algorithm to identify keywords in construction safety text data;

[0040] The second building block is to build a construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm, which is used to identify text descriptions in construction safety image data;

[0041] A training module, which uses training text samples and training picture samples to train the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm;

[0042] The test module inputs the test text samples and the test picture samples into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model of the encoding-decoding architecture based on the ATTENTION algorithm, respectively, and analyzes the error value between the output test results and the actual results. If the error value is within a preset range, the result obtained after deep learning processing is sent to the blockchain network; otherwise, the samples are collected again for training until the error value is within the preset range.

[0043] Compared with the prior art, the above technical solution conceived by the present invention can achieve the following beneficial effects:

[0044] (1) The present invention discloses a construction safety information management method and system. The method designs a novel information storage scheme, based on which the key construction safety information and the hash value of the original construction safety information processed by deep learning are stored in the blockchain network, and all the original construction safety information is stored in the cloud server. This scheme can automatically extract key construction safety information based on models such as keyword extraction and construction scene image description based on deep learning, and can effectively solve the problem of highly redundant information storage in the blockchain system, thereby improving the information retrieval efficiency of the blockchain system. At the same time, blockchain technology can maintain the data sharing / service exchange conducted by users in an open and cooperative environment, and the large amount of data accumulated in the blockchain can promote the learning ability and generalization ability of the deep learning model. In summary, the integration of deep learning technology and blockchain technology not only realizes the automatic extraction and chaining of construction safety information, but also improves the information retrieval efficiency and safety performance of the system. Finally, with the help of intuitive and timely key construction safety information, managers can quickly understand the on-site construction situation and make reliable and credible data-assisted decisions;

[0045] (2) Compared with traditional shallow machine learning methods, deep learning does not require complex manually defined vocabulary, syntax and semantic feature engineering, and can automatically extract and identify text-related features. Based on models such as keyword extraction and image semantic recognition based on deep learning, information features are automatically extracted and identified from massive construction safety information to express the construction safety information in an intuitive and visual structured manner, thereby improving the efficiency of construction safety information monitoring and safety management decision-making, and promoting intelligent and professional retrieval services for massive construction safety information resources;

[0046] (3) Blockchain is considered to be a disruptive technology that opens a "new era of trust". Compared with traditional information technology, the decentralized distributed structure in blockchain can save a lot of intermediary agency costs; the tamper-proof timestamp can solve the problem of data tracking and information anti-counterfeiting; blockchain helps users to control data, thereby solving the problem of key data protection and authorized access; and the flexible programmable characteristics help to regulate the existing market order. By uploading construction safety information to the blockchain network, construction safety information can be transparently monitored and tracked, thereby solving the problems of decentralized storage, poor reliability, and poor traceability of construction safety information, and ultimately providing participants with a safe and stable information interaction environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 is a functional architecture diagram of a system in an embodiment of the present invention;

[0048] Figure 2 It is a logical architecture diagram of the system of the embodiment of the present invention;

[0049] Figure 3 This is an information collection architecture diagram of an embodiment of the present invention;

[0050] Figure 4 This is a diagram of an information processing architecture of an embodiment of the present invention;

[0051] Figure 5 This is a diagram of the KEYBERT deep learning algorithm architecture based on an embodiment of the present invention;

[0052] Figure 6 This is an architecture diagram of the ATTENTION deep learning algorithm based on an embodiment of the present invention;

[0053] Figure 7 This is a diagram of the information uplink architecture of an embodiment of the present invention;

[0054] Figure 8 The figure is a flow chart of the uplink solution of an embodiment of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0056] This method mainly relies on deep learning and blockchain technology. Deep learning uses its mechanism of imitating the human brain to interpret data instead of manually obtaining features, thereby automatically, quickly and efficiently mining key construction safety information. This process greatly reduces manpower and material resources. Blockchain technology uses a series of complex mathematical algorithms such as hash encryption algorithms, timestamps, distributed consensus, and smart contracts to build a trusted technical system to ensure the openness, transparency, full traceability and non-tamperability of information, and establish a "trust" network between machines. With the key construction safety information stored on the blockchain network, managers will be able to make faster and smarter decisions on how to ensure the safety of on-site employees.

[0057] See also Figure 1 and Figure 2 , is a construction safety information management method and system function and logic architecture diagram in the preferred embodiment of the present invention, including an information collection unit, an information processing unit, an information upload unit and a user management unit.

[0058] A. Information Collection Unit

[0059] A1. The information collection unit is mainly used to collect relevant information related to safety during the construction process, such as Figure 3 As shown in the figure, the on-site safety information is collected. Taking the monitoring of tower crane safety information as an example, during the on-site construction process, the behavior of people related to the tower crane, the status of objects, the interaction information between people and objects, the operation status of large equipment and temporary facilities, and environmental factors are collected mainly by means of manual and intelligent equipment. The specific methods 1 and 2 are as follows:

[0060] Method 1: During the construction process of the tower crane, the inspector will check the tower crane operator's work status and the operator's operation status of the tower crane on a daily basis, and fill in the tower crane daily inspection record, etc.;

[0061] Method 2: With the help of sensor equipment installed on the tower crane, collect the working status information of the tower crane, such as the weight, angle, wind speed, amplitude, height and other data of the tower crane; with the help of cameras and other equipment, monitor the status of the tower crane and the surrounding environment of the tower crane.

[0062] A2. The person in charge shall check and verify the daily inspection records of the tower crane and sign;

[0063] A3. Transmit security information to the server via the Internet.

[0064] According to the actual project content, the data types and processes collected above can be increased or decreased, and the specific security information can also be adjusted for the security information of a certain device;

[0065] B. Information processing unit

[0066] The information processing unit mainly processes the collected raw information with the help of deep learning algorithms to automatically identify valid information, such as Figure 4 As shown, with the help of deep learning and other algorithms, the behavior of relevant people, the status of objects, the interaction information between people and objects, the operating status of large equipment and temporary facilities, and environmental factors are converted into text key information, picture and video description information, picture information base64 code, person in charge information, original information hash value, and other key information. Taking tower crane monitoring as an example, the specific operation steps are as follows;

[0067] B1. Based on the daily inspection records of the tower crane collected by A and the sensor data such as the weight, angle, wind speed, amplitude, height of the tower crane monitored by the intelligent device, as well as the monitoring video and other data, the above data are archived and classified in the server. The classification types are mainly classified according to data formats such as text, picture, video, etc. Specifically, the daily inspection records of the tower crane and the sensor data will be stored in the form of text, and the monitoring video will be stored in the form of video or picture;

[0068] B2. Build a text keyword extraction model based on the KEYBERT algorithm to identify text keywords in construction safety information. Input the hidden danger text records in the daily inspection records of the tower crane into the trained model, such as Figure 5 As shown in the figure, the keyword extraction model based on KEYBERT automatically identifies the daily hidden dangers of tower cranes as "the distance between A# and B# tower cranes is not greater than 2 meters, the boom of A# tower crane directly collides with the standard section of B# tower crane during construction, and there are obstacles within the turning radius of multi-tower operations that interfere with the tower cranes without approved anti-collision measures" and other keywords such as "tower crane, distance, 2 meters, boom, collision, standard section, turning radius, obstacle, anti-collision measures". The specific operation steps of the keyword extraction model based on KEYBERT are: the pre-trained BERT model is embedded in the above description of daily hidden dangers of tower cranes to obtain a document vector; keywords and expressions are extracted from the same document based on TfidfVectorizer or CountVectorizer bag-of-words technology; then the pre-trained BERT model is embedded in the above identified keywords to obtain a keyword vector; finally, based on the cosine similarity between the above document vector and the keyword vector, the most similar keywords are extracted according to the highest cosine similarity score.

[0069] B3. Build a construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm to identify the semantic description of construction safety pictures; input the picture records in the daily inspection records of the tower crane into the trained model, such as Figure 6As shown, the construction scene image description model of the encoding-decoding architecture of the ATTENTION algorithm automatically recognizes the semantic description of "a worker is installing a template" in the template installation picture. The specific operation steps of the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are the encoding-decoding process. The encoding process includes extracting image visual features and extracting image semantic features based on the ATTENTION algorithm. The decoding process includes selecting image features based on the ATTENTION algorithm, extracting grammatical features, word position encoding and generating description sentences. Through the encoder and decoder, the constructed image description model can learn the mapping relationship between image features and sentence semantic features. At the same time, the attention mechanism is introduced to improve the ability to extract key information in construction images.

[0070] B4. Repeat B2 and B3 for the training samples to train the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm;

[0071] B5. Input the test text sample and the test image sample into step B4 for training, analyze the error value between the output test result and the actual result, and if the error value meets the expectation, end the training and output the trained model; otherwise, collect new training samples and test samples according to step B1, and repeat B2, B3 and B4 until the error value meets the expectation;

[0072] Preferably, assuming that the expected error value is 10%, the error value can be adjusted up or down according to actual needs;

[0073] B6. Convert all original construction safety information (text, pictures, and video information) into hash values.

[0074] The keyword extraction model based on the KEYBERT algorithm and the decoding and encoding image semantic recognition model based on the ATTENTION algorithm proposed in this invention are respectively models proposed for construction safety information data. According to the different actual engineering contents, different data types have different semantic characteristics, and the corresponding models should be adjusted accordingly with different data types. This paper aims to provide guidance for the effective organization and management of massive construction safety information resources in the industry and provide intelligent and professional retrieval and services.

[0075] C. Information upload unit

[0076] The information upload unit performs hash processing on the original information and stores the hash value of the original information and the key information processed by deep learning in the blockchain network, such as Figure 7 As shown, the details are as follows;

[0077] C1. Build a chain solution, taking tower crane monitoring as an example, such as Figure 8 The specific plan is as follows:

[0078] The first step is to upload the tower crane related safety information data (basic information of the tower crane and its construction personnel, daily inspection records of the tower crane, tower crane status information and tower crane surrounding environment information) to the local database based on wireless methods such as WIFI, Bluetooth, Zigbee, etc., and back it up to the cloud service platform;

[0079] In the second step, the deep learning algorithm in step B is called to further process the data in the local database. The manager decides to upload the key information of the crane processed by deep learning, the authorized signature of the relevant responsible person and the hash value of the original information to the blockchain network;

[0080] In the third step, after the original data in the cloud server is reviewed, the hash value of the original information is uploaded to the blockchain network;

[0081] In the fourth step, the data in the blockchain network is put into a block according to a certain structure, and the legal block header information is generated. Then the generated block is verified by members to generate a new block. Specifically, the data put into the blockchain network needs to be verified by the consensus algorithm. Once it fails, feedback will be sent to the blockchain network and the off-chain database.

[0082] C2. The smart contract executes the transaction command, reviews the data, obtains the review results and feeds back the review results to the relevant construction personnel;

[0083] C3. Transaction information is broadcast to each node of the blockchain network.

[0084] D. User Management Unit

[0085] The local system front end reads the hash value and key information of the original information on the chain and visualizes its statistics. The relevant statistical information helps managers quickly identify key issues in security management, take timely response measures and prevent risks.

[0086] The construction safety information management method and system provided by the present invention can realize automatic mining and storage of the value of construction safety information, and combined with intuitive and vivid structured representation, it is conducive to the learning and retrieval of construction safety information.

[0087] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A construction safety information management method, characterized in that: The following steps are involved: S1. Collect original information on construction safety during the construction process; S2. Classify the format of the original information, use a deep learning algorithm to process the original information according to different formats to automatically extract key construction safety information, and convert the original information of all formats into hash values; S3. Broadcast the hash value of the original information and the key information processed by deep learning to each node in the blockchain network, and upload all the original information to the cloud server to complete the storage transaction; S4. During the construction process, call the blockchain network interface and query and analyze the construction safety management information from the blockchain network; The step S2 specifically includes: The original information is classified into data formats, the classification types include text data, image data and video data, and the text data is divided into training text samples and test text samples, and the image data is divided into training image samples and test image samples; Construct a text keyword extraction model based on the KEYBERT algorithm to identify keywords in construction safety text data; Build a construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm to identify text descriptions in construction safety image data; The text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are trained using training text samples and training picture samples respectively; The test text samples and test picture samples are respectively input into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm, and the error value between the output test result and the actual result is analyzed. If the error value is within the preset range, the result obtained after deep learning processing is sent to the blockchain network; otherwise, the samples are collected again for training until the error value is within the preset range; The construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm includes: Input the image to be detected, and extract features of the input image through the encoder; The image semantic information and word sequence features are captured and processed through the decoder; Output statement description about image scene information and global information; The parameters of the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are automatically adjusted regularly according to actual construction safety information.

2. The construction safety information management method according to claim 1, characterized in that: Building a text keyword extraction model based on the KEYBERT algorithm includes: Embed the document description based on the BERT model to obtain the document vector; Use the bag-of-words technique to extract keywords and expressions from the same document, then embed the keywords based on the BERT model to obtain the keyword vector; Based on the cosine similarity between the document vector and the keyword vector, the most similar keywords are extracted according to the highest cosine similarity score; The parameters of the text keyword extraction model based on the KEYBERT algorithm are automatically adjusted regularly according to actual construction safety information.

3. The construction safety information management method according to claim 1, characterized in that: The step S3 specifically includes: Wirelessly upload construction safety information data to the cloud server; Calling the deep learning algorithm in step S2 to further process and store the data in the cloud, sending the key construction safety information processed by deep learning and the hash value of the original construction safety information to the blockchain network, and uploading all the original information to the cloud server; In the blockchain network, data is put into a block according to a certain structure, and legal block header information is generated. Then the generated block is verified by members to generate a new block. Smart contracts execute transaction commands, and transaction information is broadcast to each node of the blockchain network.

4. A construction safety information management system, characterized in that: include: Information collection unit, collecting original information on construction safety during the construction process; An information processing unit, which classifies the format of the original information, processes the original information accordingly according to different formats using a deep learning algorithm to automatically extract key construction safety information, and converts the original information of all formats into hash values; The information upload unit broadcasts the hash value of the original information and the key information processed by deep learning to each node in the blockchain network, and uploads all the original information to the cloud server to complete the sorting and statistics; User management unit, during the construction process, users call the blockchain network interface and query and analyze construction safety management information from the blockchain network; The information processing unit specifically includes: A classification module classifies the original information into data formats, the classification types include text data, image data and video data, and divides the text data into training text samples and test text samples, and divides the image data into training image samples and test image samples; The first building module builds a text keyword extraction model based on the KEYBERT algorithm to identify keywords in construction safety text data; The second building block is to build a construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm, which is used to identify text descriptions in construction safety image data; A training module, which uses training text samples and training picture samples to train the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm; The test module inputs the test text sample and the test picture sample into the text keyword extraction model based on the KEYBERT algorithm and the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm, respectively, and analyzes the error value between the output test result and the actual result. If the error value is within the preset range, the result obtained after deep learning processing is sent to the blockchain network; otherwise, the sample is collected again for training until the error value is within the preset range; The construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm includes: Input the image to be detected, and extract features of the input image through the encoder; The image semantic information and word sequence features are captured and processed through the decoder; Output statement description about image scene information and global information; The parameters of the construction scene image description model based on the encoding-decoding architecture of the ATTENTION algorithm are automatically adjusted regularly according to actual construction safety information.

Images