Safety assessment method and equipment for construction data

Abstract

According to the construction data safety assessment method and device, H.265 compression coding can be directly carried out on a hardware layer through FPGA acceleration, the compression time is greatly shortened, meanwhile, the video quality is kept, and the requirements for storage and transmission bandwidth are reduced. According to the method, the efficiency of field real-time processing is greatly improved by utilizing a fog calculation lightweight processing method, and the delay is extremely low. Only lightweight data is transmitted, a large amount of bandwidth cost is saved, and the method is more suitable for real-time monitoring and early warning of safety risks of multiple construction sites. According to the cross-chain retrieval method, the retrieval speed can be obviously increased according to stage fragmentation, fast table and weight routing; a traditional block chain is fixed in fragmentation and inflexible in expansibility, but the expansibility of the block chain in the method can be flexibly adjusted according to service requirements.

Description

Technical Field

[0001] The present invention relates to a construction data safety assessment method and equipment. Background Art

[0002] 1. Inefficient data on-chaining: Traditional blockchain solutions (such as Ethereum and Hyperledger) directly store raw construction data (such as BIM models and video streams), resulting in excessive on-chain storage burdens and TPS (transactions per second) generally below 500, which cannot meet the real-time requirements of construction scenarios. When used to store heterogeneous multi-source construction data (a mix of structured and unstructured data), this method lacks an efficient pre-processing mechanism, resulting in on-chain latency of up to minutes.

[0003] 2. Insufficient trustworthy assessment capabilities: Existing technologies rely on manual review or single sensor data to determine the compliance of construction activities. They lack multi-dimensional joint analysis of time, space, equipment, personnel, and environment, resulting in low credibility.

[0004] 3. Limitations of Sharding Technology: The general blockchain sharding solution divides data into random shards. When used in the construction industry, this will cause the construction data sharding to be disconnected from the business logic, resulting in low efficiency in cross-stage data retrieval and difficulty in adapting to the dynamic characteristics of the construction stages (foundation, main structure, and decoration). Summary of the Invention

[0005] The purpose of the present invention is to provide a construction data safety assessment method and equipment.

[0006] To solve the above problems, the present invention provides a construction data security assessment method, comprising:

[0007] Step S1: collecting and preprocessing multi-source heterogeneous data from the construction site to obtain preprocessed data; wherein the multi-source heterogeneous data includes: sensor, camera and BIM model data;

[0008] Step S2: constructing a fog computing layer, and performing lightweight processing on the pre-processed data by retaining key information and removing redundant data through the fog computing layer to obtain lightweight processed data; wherein the computing power is transferred from the cloud to the smart devices on the construction site;

[0009] Step S3: Use TensorFlow Lite to deploy a lightweight AI model. This involves compressing the traditional AI model and deploying it to fog computing nodes. On the fog computing nodes, based on metadata tags corresponding to key inspections during the construction phase, real-time feature extraction is performed on the lightweight processed data. Based on the extracted features, the corresponding alarm rule engine is loaded.

[0010] Step S4: The lightweight data is uploaded to the blockchain layer and dynamically sharded for storage;

[0011] Step S5: Perform dynamic trustworthy security assessment and early warning on the data on the main chain and side chain.

[0012] Furthermore, in step S1 of the above method, the multi-source heterogeneous data of the construction site are preprocessed to obtain preprocessed data, including:

[0013] Step S11: Supporting multi-protocol access of MQTT, RTSP, and IFC through edge layer devices, and converting them into standardized data streams. FPGA is used to accelerate the protocol parsing and data conversion process, and converting them into standardized data streams.

[0014] Step S12: Based on Beidou or GPS positioning and NTP time synchronization protocol, the standardized data stream is time-space aligned to generate original data packets with time and space stamps as pre-processed data.

[0015] Furthermore, in step S2 of the above method, lightweight processing is performed on the pre-processed data to retain key information and remove redundant data, including:

[0016] Step S21: For sensor abnormal values, use the isolation forest algorithm to automatically detect abnormal values ​​and eliminate noise data;

[0017] Step S22, extracting key frames from the 30pfs video (and compressing them to about 10% of their original size using H.265 encoding to obtain compressed data;

[0018] Step S23: creating a metadata tag for the compressed data to obtain data with the metadata tag;

[0019] Step S24: perform LOD (level of detail) classification on the BIM model in the data with metadata tags, extract only the component information of LOD level 300 and above, ignore the decorative details, and obtain the lightweight modeled data.

[0020] Furthermore, the metadata tag in the above method includes:

[0021] Basic attribute tags represent static features, including objects such as employee race, equipment type, and environmental climate;

[0022] The status tag represents dynamic features, including real-time data on personnel wear, pump truck pressure, and construction phase;

[0023] The security specification tag represents the business rules and is used for compliance checking;

[0024] Behavior tags represent data used to analyze object behavior, including: data on equipment operation time and personnel violations used to analyze object behavior;

[0025] Association relationship tags represent association rules and interaction logic between objects and are used for complex scenario analysis;

[0026] Evaluation and prediction tags are used to safely assess risks.

[0027] Furthermore, in the above method, step S4, performing layered blockchain uploading and dynamic sharding storage on the lightweight processed data, includes:

[0028] Step S41: Establishing a main chain-side chain coordination mechanism for the lightweight processed data;

[0029] Step S42: establishing a dynamic sharding technology for the lightweight processed data;

[0030] Step S43: Establish a cross-chain retrieval method for the lightweight processed data.

[0031] Furthermore, in the above method, step S41, establishing a main chain-side chain coordination mechanism, includes:

[0032] The main chain stores small pieces of important data, including feature summaries and device identification hash values. Using an improved PBFT consensus algorithm, dynamic weighted voting improves consensus efficiency, allowing reputable devices to vote on whether data is uploaded to the main chain.

[0033] Store unimportant data on the sidechain, including raw data (video streams, BIM models) and extended metadata; combine with IPFS shard storage and only upload IPFS hashes to the chain;

[0034] Step S42: establishing a dynamic sharding technology for the lightweight processed data, including:

[0035] The unimportant data is divided into data shards according to the construction stage, including: the foundation stage shard, which stores geological exploration data and pile foundation construction records; the main stage shard, which stores concrete pouring parameters and steel structure welding records;

[0036] Step S43: Establish a cross-chain search method for the lightweight data, including:

[0037] Based on the improved Kademlia algorithm, a distributed hash table is used to attach "hash tags" to the data to achieve cross-shard data positioning; then the two optimization technologies of fast table technology and weight routing technology are combined.

[0038] Furthermore, in step S5 of the above method, dynamic trustworthy security assessment and early warning are performed on the data on the main chain and the side chain, including:

[0039] Step S51, establishing a four-dimensional feature model;

[0040] Step S52: establishing a four-dimensional feature association between the main chain and the side chain data based on the four-dimensional feature model to obtain four-dimensional feature association data;

[0041] Step S53: Based on the cross-chain search method, the four-dimensional feature correlation data of the construction site is collected from the main chain and the side chain;

[0042] Step S54, based on the collected four-dimensional feature association data, using the Bayesian network to calculate the probability value of the known risk;

[0043] Step S55: The SVM classifier combines the probability value calculated by the Bayesian network with the corresponding four-dimensional feature association data to form a feature vector, and obtains an SVM classification score based on the feature vector and using a preset classification rule;

[0044] Step S56: The SVM classifier fuses the probability value and the SVM classification score, and outputs a comprehensive credibility index = Bayesian probability × weight + SVM classification score × weight;

[0045] Step S57, based on the comprehensive credibility index, execute the preset warning mechanism.

[0046] Furthermore, in the above method, the four-dimensional feature model includes:

[0047] Spatiotemporal dimension: GPS coordinates (x, y, z) + timestamp to construct the spatiotemporal trajectory of construction behavior;

[0048] Equipment dimension: Real-time monitoring and threshold analysis of operating parameters such as tower crane inclination angle and pump truck pressure;

[0049] Personnel dimension: safety equipment wearing status, work qualifications and health data;

[0050] Environmental dimension: Assessment of the impact of PM2.5, noise, temperature and humidity on construction safety.

[0051] According to another aspect of the present invention, a computer-readable storage medium is provided, on which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, the processor is enabled to execute any one of the methods described above.

[0052] According to another aspect of the present invention, there is further provided a computer device, comprising:

[0053] processor; and

[0054] A memory arranged to store computer executable instructions, which when executed cause the processor to: perform any of the methods described above.

[0055] Compared with the prior art, the present invention has the following significant advantages:

[0056] When processing video streams, the traditional method is to compress the video using H.265 encoding through software (10% of the original video size). However, through FPGA acceleration, H.265 compression encoding can be performed directly at the hardware layer, significantly reducing compression time while maintaining video quality and reducing storage and transmission bandwidth requirements.

[0057] Compared to the traditional edge computing + cloud computing model, this invention leverages fog computing's lightweight processing approach to significantly improve on-site real-time processing efficiency with minimal latency. Transmitting only lightweight data saves significant bandwidth costs, making it more suitable for real-time monitoring and early warning of safety risks at multiple construction sites.

[0058] Compared with the storage and retrieval methods of traditional blockchains, the storage cost of the present invention can be reduced by 80% (only storing the key + shards); the cross-chain retrieval method is based on stage sharding + quick table + weight routing, which can significantly improve the retrieval speed; the traditional blockchain has fixed sharding and inflexible scalability, while the blockchain scalability in this method can be flexibly adjusted according to business needs.

[0059] Advantages of this assessment method: Traditional methods rely on manual inspections or single sensor alarms, which are prone to missed judgments and ignore the probability of accidents caused by the combination of behavioral and environmental factors. This assessment method can update scores based on real-time data, has non-fixed rules, and considers the integration of multiple factors. The monitoring, assessment, and early warning at the construction site are explainable, facilitating targeted rectification.

[0060] In response to the above-mentioned problems in the prior art, the present invention proposes a fog chain layered uploading and dynamic security assessment system for construction data, which can effectively solve the shortcomings of the prior art:

[0061] 1. Realize lightweight processing and efficient chain-up of multi-source heterogeneous data (IoT sensors, BIM models, video streams) in construction scenarios.

[0062] 2. Build a dynamic and credible assessment model for construction behavior to solve the problems of low efficiency and high misjudgment rate of manual review.

[0063] 3. Improve the scalability of the blockchain system and the traceability of construction data through layered storage and dynamic sharding mechanisms. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 This is an overall architecture diagram of a construction data security assessment method and device according to an embodiment of the present invention;

[0065] Figure 2 is a flowchart of data layering processing according to an embodiment of the present invention;

[0066] Figure 3This is a schematic diagram of the main chain-side chain storage structure of an embodiment of the present invention;

[0067] Figure 4 This is a four-dimensional trustworthy evaluation model diagram according to an embodiment of the present invention;

[0068] Figure 5 This is a schematic diagram of dynamic shard storage and cross-chain retrieval according to an embodiment of the present invention;

[0069] Figure 6 This is a timing diagram of fog computing node hardware deployment and data flow according to an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The present invention is further described in detail below with reference to the accompanying drawings.

[0071] In a typical configuration of the present application, the terminal, the device of the service network and the trusted party all include one or more processors (CPUs), input / output interfaces, network interfaces and memories.

[0072] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0073] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include non-transitory media such as modulated data signals and carrier waves.

[0074] like Figures 1 to 6 As shown, the present invention provides a method for safety assessment of construction data, the method comprising:

[0075] Step S1: collecting and preprocessing multi-source heterogeneous data from the construction site to obtain preprocessed data; wherein the multi-source heterogeneous data includes: sensor, camera and BIM model data;

[0076] Multi-source heterogeneous data: In the construction of smart sites, construction sites involve a variety of data sources, including sensors, cameras, BIM models, and more. This data not only comes from diverse sources but also in different formats, potentially using different communication protocols such as MQTT (for sensor data), RTSP (for video streaming), and IFC (for BIM models). This can lead to the following difficulties:

[0077] Problem 1) Inconsistent data formats: Data with different protocols and formats are difficult to directly integrate and process;

[0078] Problem 2) Difficulty in temporal and spatial alignment: Data from different devices may be deviated in time or space, affecting the accuracy of subsequent analysis;

[0079] Problem 3) Processing efficiency: Real-time processing of large amounts of data requires efficient hardware and software support.

[0080] Preferably, in step S1, preprocessing the multi-source heterogeneous data of the construction site to obtain preprocessed data includes:

[0081] Step S11: Using a hybrid protocol adapter: Supporting multi-protocol access such as MQTT (sensor data), RTSP (video stream), and IFC (BIM model) through edge layer devices (such as construction site smart gateways), and converting them into standardized data streams for subsequent processing and analysis;

[0082] Regarding the need to convert various types of heterogeneous data into the same format mentioned in question 1), the traditional method is to perform protocol parsing and data conversion through software, which may result in slower processing speeds, especially when the amount of data is large. Preferably, the present application adopts a hardware acceleration method: using FPGA (a programmable hardware) to accelerate the protocol parsing and data conversion process, and uniformly convert them into standardized data streams. FPGA can perform these tasks directly at the hardware layer, reducing software-level overhead, thereby significantly improving processing speed, and theoretically supporting the processing of tens of thousands of data points per second.

[0083] Step S12, using a time-space alignment engine: based on Beidou or GPS positioning and NTP time synchronization protocol, the standardized data stream is time-space aligned (error <10ms), and the original data packet with time and space stamps is generated as the pre-processed data;

[0084] Step S2: constructing a fog computing layer, and performing lightweight processing on the pre-processed data by retaining key information and removing redundant data through the fog computing layer to obtain lightweight processed data; wherein the computing power is transferred from the cloud to the smart devices on the construction site;

[0085] Fog computing shifts computing power from the cloud to smart devices on-site (such as construction site gateways or edge servers). This lightweight data processing allows for on-site processing, retaining critical information and eliminating redundant data, making subsequent analysis and storage more efficient. This approach reduces data transmission latency and cloud load, making it suitable for scenarios requiring real-time response, such as construction site safety monitoring.

[0086] Preferably, step S2, performing lightweight processing on the pre-processed data to retain key information and remove redundant data, includes:

[0087] Step S21, performing data cleaning on the preprocessed data to obtain cleaned data: for sensor abnormal values ​​(such as a sudden change of 360° in the inclination angle of a tower crane), the isolation forest algorithm (statistically, in the data space, if there are only sparsely distributed points in an area, it means that the probability of data points falling in this area is very low, so the points in these areas can be considered abnormal) is used to automatically detect abnormal values, eliminate noise data, and reduce the false positive rate (false positive rate <5%, for example, at most 5 normal data are mistakenly deleted in 100 abnormal data) to obtain cleaned data.

[0088] Step S22: Perform video stream frame compression on the cleaned data to obtain compressed data: extract key frames (1fps, such as a moment of a worker working at height) from the 30pfs video and compress them to about 10% of the original size using H.265 encoding to obtain compressed data;

[0089] Step S23: Create metadata tags for the compressed data to obtain data with metadata tags. Preferably, six categories of metadata tags can be defined:

[0090] ①, basic attribute tags represent static features, including objects such as employee race, equipment type, and environmental climate;

[0091] ②, the status tag represents dynamic features, including real-time data such as personnel clothing, pump truck pressure, and construction stage;

[0092] ③ Safety regulation labels represent business rules and are used for compliance checks, such as whether PM2.5 exceeds the standard;

[0093] ④,Behavior tags represent data used to analyze object behavior, including equipment operation time and personnel illegal operation data used to analyze object behavior;

[0094] ⑤, association relationship tags represent the association rules and interaction logic between objects, which are used for complex scenario analysis, such as process dependency and equipment-personnel binding;

[0095] ⑥, Assessment and prediction tags are used for safety assessment risks, such as comprehensive credibility scores, equipment failure predictions, construction delay warnings and other data.

[0096] Metadata tags are used to label the basic data obtained from the previous steps (monitoring data cleaning, model lightweighting, and video frame compression). Data such as outdoor temperature and PM2.5 concentration are inherently static. Once these data are labeled, they can be organized, described, and retrieved.

[0097] The system automatically tags data as it's generated. For example, if the system collects on-site temperature and humidity data every five minutes, then through lightweight data processing, the final data obtained is the hourly average temperature and humidity. As soon as the data is generated, it can be tagged with environmental climate metadata.

[0098] Step S24, lightweight modeling is performed on the data with metadata tags to obtain lightweight modeled data: LOD (Level of Detail) classification is performed on the BIM model in the data with metadata tags, only component information of LOD300 and above (such as steel density, concrete strength) is extracted, and decorative details (such as tile patterns) are ignored to obtain lightweight modeled data.

[0099] In step S3, feature extraction is performed on the lightweight modeled data according to the metadata tags: the lightweight AI model is deployed using TensorFlow Lite, including compressing the traditional AI model (such as pruning and quantization), and deployed to the fog computing node. In the fog computing node, based on the metadata tags of the key inspections corresponding to the construction phase, feature extraction is performed on the lightweight processed data in real time; based on the extracted features, the corresponding alarm rule engine is loaded.

[0100] Here, the alarm rule engine can be dynamically loaded: based on the construction stage (foundation / main structure / decoration, etc.), the generation rules corresponding to the result tags derived from metadata tags (determining high risk or safe, or issuing an alarm) can be switched, reducing redundant calculations. For example, during the foundation stage, metadata tags such as "pile foundation offset" and "geological settlement" can be checked, while during the main structure stage, metadata tags such as "high-altitude work" and "crane overload" can be checked.

[0101] The six categories of metadata tags defined represent the specific descriptions of the data that need to be uploaded to the chain. For example:

[0102] Worker: Wang; Job Type: Tower Crane Driver; Associated with a Tower Crane T01;

[0103] Crane 01: GPS coordinates (x, y, z), maintenance date: 2025-07-01

[0104] The alarm rule engine establishes the following two rules:

[0105] 1) If it is detected that the device is out of the working range for a long time or in the non-working range for a long time, the response processing plan will be activated;

[0106] Here, if Wang is captured by the camera at work away from the range of tower crane T01 or in the working range of other types of work, the system will handle this incident.

[0107] 2) If "overtime operation" is detected, the response processing plan is activated

[0108] Here, Wang worked on T01 for a long time, exceeding the continuous working time range, so the system will also handle this event.

[0109] Metadata tags are used to label monitored data, categorize various data types, and establish a one-to-many relationship between individual data items and metadata tags. For example, consider worker Wang, whose job title is "crane driver"; whose clothing label is "wearing a hard hat and high-altitude work safety rope"; and whose behavior label is "complying with XXX crane operating regulations."

[0110] For example, real-time data (such as ambient temperature and humidity obtained by sensors) > metadata tags (meteorological data) > outdoor temperature higher than 35°C > triggering the alarm rule for outdoor high temperature.

[0111] Real-time data (such as the working hours of worker Wang on a tower crane) > metadata tags (equipment operating specifications) > working hours exceeding 4 hours > triggering the alarm rule for overtime work.

[0112] Step S4: The lightweight processed data is stored in a blockchain layered manner and dynamically in shards.

[0113] In order to solve the problem of too much and too complicated construction site data, directly storing it in the blockchain will cause congestion and waste of resources. For example, if the mobile phone storage is full, it will be stuck if all the photos and videos are stored in it. Therefore, the data needs to be classified: important data is stored on the mobile phone (main chain), and unimportant data can be stored on the network disk (side chain + IPFS). Then, hash tags are attached to the data to make it easier to find.

[0114] Step S41: Establish a main chain-side chain collaboration mechanism for the lightweight processed data:

[0115] ① Main chain storage: Important data is stored on the main chain, such as feature summaries (e.g., equipment identification: crane load exceeds limit) and equipment identification hash values. An improved PBFT consensus algorithm is used to improve consensus efficiency through dynamic weighted voting (the higher the node reputation, the greater the voting weight). This allows "reliable nodes" (devices with high reputation) to vote on whether data should be uploaded to the main chain, giving them a higher voting weight and preventing malicious nodes from disrupting the process.

[0116] ② Sidechain storage: Store unimportant data on the sidechain. Unimportant data, such as raw data (video streams, BIM models) and extended metadata, is cut into small pieces and stored dispersedly across multiple nodes (similar to a network disk). Combined with IPFS sharded storage, only the IPFS hash is uploaded to the chain for storage optimization, which can reduce storage costs. Specifically, videos or models are cut into small pieces and dispersed across multiple nodes (similar to a network disk). Only these "network disk links" (IPFS hashes) are stored on the sidechain, not the complete files.

[0117] Step S42: establishing dynamic sharding technology for the lightweight processed data:

[0118] Non-critical data is sharded by construction phase. For example, sharding for the foundation phase stores geological exploration data and pile foundation construction records; sharding for the main structure phase stores concrete pouring parameters and steel structure welding records. Traditional blockchains randomly shard data, making it difficult to find a needle in a haystack. This dynamic sharding technology is similar to how libraries divide books into specialized sections, such as literature, science, and art. This makes data retrieval faster and eliminates the need for a full network scan.

[0119] Step S43: Establish a cross-chain search method for the lightweight processed data:

[0120] Traditional cross-chain retrieval method: Searching for side chain data from the main chain is like using a search engine to search for network disk files. It requires jumping through multiple links and is time-consuming. However, this invention is based on the improved Kademlia algorithm and uses a distributed hash table (DHT) to attach "hash tags" to the data to achieve cross-shard data positioning; it also combines two optimization technologies: ① Fast table technology: caches frequently accessed data (such as video surveillance of the past three days) in data hotspots for on-demand access; ② Weighted routing: prioritizes searching for "fast-responding nodes" (such as the nearest server) rather than randomly searching.

[0121] Step S5: Perform dynamic trustworthy security assessment and early warning on the data on the main chain and side chain;

[0122] Step S51, establishing a four-dimensional feature model:

[0123] Spatiotemporal dimension: GPS coordinates (x, y, z) + timestamp to construct the spatiotemporal trajectory of construction behavior.

[0124] Equipment dimension: Real-time monitoring and threshold analysis of operating parameters such as tower crane inclination angle and pump truck pressure.

[0125] Personnel dimension: safety equipment wearing status, work qualifications, and health data.

[0126] Environmental dimension: Assessment of the impact of PM2.5, noise, temperature and humidity on construction safety.

[0127] Here, a four-dimensional feature model comprehensively assesses construction safety using four major categories of data at the construction site: time and space, equipment, personnel, and environment. This model is based on weighting the probability of safety incidents in each category of historical data to form a trained assessment model.

[0128] Step S52: Establish a four-dimensional feature association between the main chain and the side chain data based on the four-dimensional feature model to obtain four-dimensional feature association data. Specifically, the association logic of the four-dimensional feature is as follows:

[0129] Dimensions Data Source Evaluation weight Typical risk scenarios Space and Time GPS positioning + timestamp 30％ Temporal and spatial conflict between tower cranes and workers' activity areas equipment Sensors (tilt / pressure / vibration) 35％ Tower crane overload / pump truck hydraulic abnormality personnel Camera + OCR electronic certificate 25％ Not wearing a safety rope / operating without a license environment Environmental sensor (PM2.5 / temperature and humidity) 10％ Risk of heat stroke during high temperature work

[0130] Evaluation Methodology:

[0131] The Bayesian network combined with the SVM classifier forms a "safety scoring system." The Bayesian network calculates the probability of known risks, while the SVM classifier integrates multiple factors to determine the risk level. The goal is to dynamically assign safety scores to construction sites based on historical and real-time data, and trigger different levels of warning (yellow, orange, or red) based on the scores.

[0132] Specific steps: Step S53, based on the cross-chain search method, collect four-dimensional feature association data of the four major categories (personnel, equipment, environment, time and space) of the construction site from the main chain and side chain;

[0133] Step S54: Based on the collected four-dimensional feature association data, the Bayesian network is used to calculate the probability value of the known risk. For example, if 35 of the past 100 accidents were caused by "not wearing a helmet", then the prior probability of "not wearing a helmet" is 0.35. When the crane angle exceeds the standard, the Bayesian network will update the relevant probabilities, for example, P(accident) = P(not wearing a helmet) * P(crane exceeding the standard) * adjustment coefficient, and finally a basic risk probability is obtained, which increases from 0.35 to 0.6.

[0134] In step S55, the SVM classifier combines the probability value calculated by the Bayesian network with the corresponding four-dimensional feature association data (real-time data) to form a "feature vector", and obtains the SVM classification score based on the feature vector and using the preset classification rules;

[0135] For example, the trained model can be used to determine whether the current situation is "safe", "normal", or "dangerous", and to judge "not wearing a helmet (accident probability 0.6) + high temperature (35°C) -> SVM output classification score 45".

[0136] Step S56: The SVM classifier fuses the probability value and the SVM classification score, and outputs a comprehensive credibility index (0-100 points) = Bayesian probability × weight + SVM classification score × weight;

[0137] Step S57: Based on the comprehensive credibility index, the preset three-level warning mechanism is executed:

[0138] Yellow warning (60-80 points): Push message to administrator.

[0139] Orange warning (40-60 minutes): Automatically triggers on-site sound and light alarms.

[0140] Red alert (<40 points): Forcefully shut down relevant equipment and report to the supervision platform.

[0141] Here, in the evaluation of the four-dimensional feature model, combined with historical data on the main chain and side chain, there have been many accidents in which tower crane drivers have worked overtime in hot weather, resulting in heatstroke. In this case, the system will calculate that the risk probability of accidents in this situation has increased significantly, and the comprehensive credibility index will be reduced to the "dangerous" level.

[0142] Optimizing the evaluation model through federated learning iteration (training an evaluation model through multi-party collaboration without sharing the original data, achieving data stability while the model moves, and data availability without visibility) and using a multi-site joint iterative evaluation model. This protects each site's private sensitive data while enabling multi-party joint model training, breaking down data silos and reducing reliance on central servers and transmission costs through distributed computing.

[0143] In summary, the present invention has the following significant advantages:

[0144] When processing video streams, the traditional method is to compress the video using H.265 encoding through software (10% of the original video size). However, through FPGA acceleration, H.265 compression encoding can be performed directly at the hardware layer, significantly reducing compression time while maintaining video quality and reducing storage and transmission bandwidth requirements.

[0145] Compared to the traditional edge computing + cloud computing model, this invention leverages fog computing's lightweight processing approach to significantly improve on-site real-time processing efficiency with minimal latency. Transmitting only lightweight data saves significant bandwidth costs, making it more suitable for real-time monitoring and early warning of safety risks at multiple construction sites.

[0146] Compared with the storage and retrieval methods of traditional blockchains, the storage cost of the present invention can be reduced by 80% (only storing the key + shards); the cross-chain retrieval method is based on stage sharding + quick table + weight routing, which can significantly improve the retrieval speed; the traditional blockchain has fixed sharding and inflexible scalability, while the blockchain scalability in this method can be flexibly adjusted according to business needs.

[0147] Advantages of this assessment method: Traditional methods rely on manual inspections or single sensor alarms, which are prone to missed judgments and ignore the probability of accidents caused by the combination of behavioral and environmental factors. This assessment method can update scores based on real-time data, has non-fixed rules, and considers the integration of multiple factors. The monitoring, assessment, and early warning at the construction site are explainable, facilitating targeted rectification.

[0148] In response to the above-mentioned problems in the prior art, the present invention proposes a fog chain layered uploading and dynamic security assessment system for construction data, which can effectively solve the shortcomings of the prior art:

[0149] 1. Realize lightweight processing and efficient chain-up of multi-source heterogeneous data (IoT sensors, BIM models, video streams) in construction scenarios.

[0150] 2. Build a dynamic and credible assessment model for construction behavior to solve the problems of low efficiency and high misjudgment rate of manual review.

[0151] 3. Improve the scalability of the blockchain system and the traceability of construction data through layered storage and dynamic sharding mechanisms.

[0152] The detailed contents of the various device embodiments of the present invention can be found in the corresponding parts of the various method embodiments, which will not be repeated here.

[0153] Obviously, those skilled in the art may make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalents, this application is intended to include these modifications and variations.

[0154] It should be noted that the present invention can be implemented in software and / or a combination of software and hardware, for example, can be implemented using an application specific integrated circuit (ASIC), a general purpose computer or any other similar hardware device. In one embodiment, the software program of the present invention can be executed by a processor to implement the steps or functions described above. Similarly, the software program of the present invention (including related data structures) can be stored in a computer-readable recording medium, for example, a RAM memory, a magnetic or optical drive or a floppy disk and similar devices. In addition, some steps or functions of the present invention can be implemented using hardware, for example, as a circuit that cooperates with a processor to perform each step or function.

[0155] In addition, a portion of the present invention may be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. The program instructions for calling the method of the present invention may be stored in a fixed or removable recording medium, and / or transmitted through a data stream in a broadcast or other signal-carrying medium, and / or stored in a working memory of a computer device that operates according to the program instructions. Here, according to one embodiment of the present invention, a device is included, which includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein, when the computer program instructions are executed by the processor, the device is triggered to operate based on the aforementioned methods and / or technical solutions according to multiple embodiments of the present invention.

[0156] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the invention is defined by the appended claims rather than the above description, and it is intended that all changes that fall within the meaning and scope of the equivalents of the claims be encompassed within the present invention. Any figure marks in the claims should not be regarded as limiting the claims involved. In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim may also be implemented by one unit or device through software or hardware. Words such as first and second are used to indicate names and do not indicate any particular order.

Claims

1. A construction data security assessment method, characterized in that: include: Step S1: collecting and preprocessing multi-source heterogeneous data from the construction site to obtain preprocessed data; wherein the multi-source heterogeneous data includes: sensor, camera and BIM model data; Step S2: constructing a fog computing layer, and performing lightweight processing on the pre-processed data by retaining key information and removing redundant data through the fog computing layer to obtain lightweight processed data; wherein the computing power is transferred from the cloud to the smart devices on the construction site; Step S3: Use TensorFlow Lite to deploy a lightweight AI model. This involves compressing the traditional AI model and deploying it to fog computing nodes. On the fog computing nodes, based on metadata tags corresponding to key inspections during the construction phase, real-time feature extraction is performed on the lightweight processed data. Based on the extracted features, the corresponding alarm rule engine is loaded. Step S4: The lightweight data is uploaded to the blockchain layer and dynamically sharded for storage; Step S5: Perform dynamic trusted security assessment and early warning on the data on the main chain and side chain.

2. The construction data safety assessment method according to claim 1, wherein: In step S1, the multi-source heterogeneous data of the construction site is preprocessed to obtain preprocessed data, including: Step S11: Supporting multi-protocol access of MQTT, RTSP, and IFC through edge layer devices, and converting them into standardized data streams. FPGA is used to accelerate the protocol parsing and data conversion process, and converting them into standardized data streams. Step S12: Based on Beidou or GPS positioning and NTP time synchronization protocol, the standardized data stream is time-space aligned to generate original data packets with time and space stamps as pre-processed data.

3. The construction data safety assessment method according to claim 1 or 2, characterized in that: Step S2, performing lightweight processing on the pre-processed data to retain key information and remove redundant data, including: Step S21: For sensor abnormal values, use the isolation forest algorithm to automatically detect abnormal values ​​and eliminate noise data; Step S22: extract key frames from the 30pfs video and compress them to about 10% of their original size using H.265 encoding to obtain compressed data; Step S23: creating a metadata tag for the compressed data to obtain data with the metadata tag; In step S24, the BIM model in the data with metadata tags is LOD graded, and only the component information of LOD300 and above is extracted, and the decorative details are ignored to obtain the lightweight modeled data.

4. The construction data safety assessment method according to claim 3, wherein: The metadata tag includes: Basic attribute tags represent static features, including objects such as employee race, equipment type, and environmental climate; The status tag represents dynamic features, including real-time data on personnel wear, pump truck pressure, and construction phase; The security specification tag represents the business rules and is used for compliance checking; Behavior tags represent data used to analyze object behavior, including: data on equipment operation time and personnel violations used to analyze object behavior; Association relationship tags represent association rules and interaction logic between objects and are used for complex scenario analysis; Evaluation and prediction tags are used to safely assess risks.

5. The construction data safety assessment method according to claim 1, wherein: Step S4, performing layered blockchain uploading and dynamic sharding storage on the lightweight processed data, including: Step S41: Establishing a main chain-side chain coordination mechanism for the lightweight processed data; Step S42: establishing a dynamic sharding technology for the lightweight processed data; Step S43: Establish a cross-chain retrieval method for the lightweight processed data.

6. The construction data safety assessment method according to claim 1, wherein: Step S41, Establish a main chain-side chain coordination mechanism, including: The main chain stores small pieces of important data, including feature summaries and device identification hash values. Using an improved PBFT consensus algorithm, dynamic weighted voting improves consensus efficiency, allowing reputable devices to vote on whether data is uploaded to the main chain. Store unimportant data on the sidechain, including original data and extended metadata; combine with IPFS shard storage and only upload IPFS hashes to the chain; Step S42: establishing a dynamic sharding technology for the lightweight processed data, including: The unimportant data is divided into data shards according to the construction stage, including: the foundation stage shard, which stores geological exploration data and pile foundation construction records; the main stage shard, which stores concrete pouring parameters and steel structure welding records; Step S43: Establish a cross-chain search method for the lightweight data, including: Based on the improved Kademlia algorithm, a distributed hash table is used to attach "hash tags" to the data to achieve cross-shard data positioning; then the two optimization technologies of fast table technology and weight routing technology are combined.

7. The construction data safety assessment method according to claim 1, wherein: Step S5: Dynamically conduct trustworthy security assessment and early warning on the data on the main chain and side chain, including: Step S51, establishing a four-dimensional feature model; Step S52: establishing a four-dimensional feature association between the main chain and the side chain data based on the four-dimensional feature model to obtain four-dimensional feature association data; Step S53: Based on the cross-chain search method, the four-dimensional feature correlation data of the construction site is collected from the main chain and the side chain; Step S54, based on the collected four-dimensional feature association data, using the Bayesian network to calculate the probability value of the known risk; Step S55: The SVM classifier combines the probability value calculated by the Bayesian network with the corresponding four-dimensional feature association data to form a feature vector, and obtains an SVM classification score based on the feature vector and using a preset classification rule; Step S56: The SVM classifier fuses the probability value and the SVM classification score, and outputs a comprehensive credibility index = Bayesian probability × weight + SVM classification score × weight; Step S57: executing a preset early warning mechanism based on the comprehensive credibility index.

8. The construction data safety assessment method according to claim 1, wherein: The four-dimensional feature model includes: Spatiotemporal dimension: GPS coordinates (x, y, z) + timestamp to construct the spatiotemporal trajectory of construction behavior; Equipment dimension: Real-time monitoring and threshold analysis of operating parameters such as tower crane inclination angle and pump truck pressure; Personnel dimension: safety equipment wearing status, work qualifications and health data; Environmental dimension: Assessment of the impact of PM2.5, noise, temperature and humidity on construction safety.

9. A computer-readable storage medium having computer-executable instructions stored thereon, wherein: When the computer executable instructions are executed by a processor, the processor is caused to perform the method according to any one of claims 1 to 8.

10. A computer device, wherein: include: processor; as well as A memory arranged to store computer executable instructions which, when executed, cause the processor to: perform the method according to any one of claims 1 to 8.

Images