A construction operation environment active perception monitoring method and system based on a dynamic environment risk field

Abstract

The application discloses a kind of based on dynamic environment risk field's construction operation environment initiative perception monitoring method and system, it is related to construction environment monitoring technical field, technical scheme is, including obtaining the multi-modal environment data of construction operation site, construction state data and monitoring equipment state data, after time-space registration, mapping to space monitoring unit, the dynamic environment risk value of each unit and monitoring uncertainty are calculated, and dynamic environment risk field is formed;According to risk value and monitoring uncertainty, determine target monitoring area, generate initiative perception strategy, obtain supplementary monitoring data and update dynamic environment risk field, output monitoring result.The beneficial effects of the present application are: it can represent the risk space distribution and monitoring reliability degree of construction site, reduce the false alarm, false alarm caused by single-point threshold alarm, and improve monitoring timeliness, accuracy, coverage and resource utilization efficiency through active supplement and closed-loop update.

Description

Technical Field

[0001] This invention relates to the field of construction environment monitoring technology, and in particular to a method and system for active sensing and monitoring of construction operation environment based on dynamic environmental risk field. Background Technology

[0002] With the development of smart construction sites, IoT sensing, edge computing and artificial intelligence technologies, the monitoring of construction environment is gradually evolving from manual inspection and single-point instrument detection to multi-source data acquisition, remote transmission, intelligent identification and platform-based management.

[0003] Existing technologies typically use fixed dust sensors, gas sensors, noise sensors, vibration sensors, video surveillance equipment, and mobile inspection equipment to collect data on dust, harmful gases, noise, vibration, temperature and humidity, smoke, and the status of personnel and equipment at construction sites. These data are then combined with threshold alarms, image recognition, multi-sensor fusion, or edge computing to identify and warn of abnormalities in the construction environment, playing a certain role in construction safety management and occupational health protection.

[0004] However, construction sites are characterized by frequent changes in work surfaces, overlapping construction procedures, dynamic movement of personnel and equipment, and spatial diffusion of environmental risks. Existing technologies are still insufficient to meet the needs of refined monitoring in dynamic scenarios.

[0005] First, existing monitoring methods mostly adopt fixed deployment and fixed sampling strategies. The sensor positions, camera orientations, and inspection routes usually remain relatively stable after deployment, making it difficult to dynamically adjust with changes in construction areas, movement of risk sources, and environmental diffusion trends. This can easily lead to monitoring blind spots or redundant data collection in low-risk areas.

[0006] Secondly, existing alarm methods mostly rely on single-point thresholds or static fusion results, lacking a unified expression of the spatial distribution of risks at the construction site, the trend of risk changes, and the propagation relationship between adjacent areas. It is difficult to accurately determine where the risk is formed, where it will spread, and which areas need to be prioritized for monitoring.

[0007] Furthermore, construction sites often experience issues such as visual obstruction, sensor offlineness, sampling delays, communication packet loss, and multimodal data conflicts. Existing technologies typically lack quantitative evaluation of the sufficiency of monitoring information and the reliability of risk assessment, leading to false alarms, missed alarms, and anomalies that are difficult to verify.

[0008] Finally, there is a lack of a collaborative scheduling mechanism between existing mobile inspection and fixed monitoring equipment based on real-time risk status. Supplementary data collection is also difficult to update the overall risk expression, making it impossible to form a closed loop between risk identification, proactive supplementary monitoring, and the next round of target area identification. Summary of the Invention

[0009] To address the aforementioned technical problems, this invention provides a method and system for active sensing and monitoring of construction operation environment based on dynamic environmental risk field.

[0010] The technical solution includes the following steps: Acquire multimodal environmental data, construction status data, and monitoring equipment status data from the construction site; The multimodal environmental data, construction status data, and monitoring equipment status data are spatiotemporally registered and mapped to pre-divided spatial monitoring units to obtain spatialized monitoring data for each spatial monitoring unit. The dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit are calculated based on the spatialized monitoring data to form the dynamic environmental risk field at the current moment. Based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field, the target monitoring area that needs enhanced monitoring is determined. With the goal of reducing the monitoring uncertainty of the target monitoring area, an active sensing strategy is generated based on the target monitoring area and the status data of the monitoring equipment. The active sensing strategy is executed to obtain supplementary monitoring data of the target monitoring area, the dynamic environmental risk field is updated based on the supplementary monitoring data, the construction operation environment monitoring results are output, and the updated dynamic environmental risk field is used as the basis for the next round of target monitoring area identification.

[0011] Preferably, the multimodal environmental data includes multiple types of image data, video data, dust concentration data, harmful gas concentration data, temperature and humidity data, wind speed and direction data, noise data, vibration data, illuminance data, and thermal imaging data. The construction status data includes multiple types of information such as construction procedure information, construction area information, construction progress information, personnel location information, construction equipment location information, construction equipment operation status information, material stacking information, and spatial structure information. The monitoring equipment status data includes multiple items from the following: monitoring equipment identifier, equipment type, installation location, current orientation, visible range, sampling frequency, communication status, remaining battery power, storage status, sensor working status, equipment fault status, mobility, reachable area, and current task status. The multimodal environmental data, construction status data, and monitoring equipment status data are preprocessed to obtain monitoring data to be processed containing data quality tags. The monitoring data to be processed is then used as input for subsequent spatiotemporal registration. The preprocessing includes at least one of the following: data format unification, data integrity check, anomaly marking, dimensional unification, range verification, and quality tag generation.

[0012] Preferably, the multimodal environmental data, construction status data, and monitoring equipment status data are spatiotemporally registered and mapped to pre-divided spatial monitoring units to obtain spatialized monitoring data for each spatial monitoring unit, including: Using a unified clock as the time reference, data from different sampling frequencies are aggregated into the same time window to obtain the monitoring data within the corresponding time window; Establish a unified spatial coordinate benchmark for the construction site, and divide the construction site into multiple spatial monitoring units based on the construction area, work surface, floor, room, foundation pit zone, tunnel mileage section, construction road, construction equipment operation area, material storage area, opening edge area, ventilation area, or regular grid. Based on the data acquisition location, data coverage area, acquisition device posture, movement trajectory, and corresponding construction area, the time-registered data is mapped to the corresponding spatial monitoring unit to form spatialized monitoring data for each spatial monitoring unit.

[0013] Preferably, the dynamic environmental risk value of each spatial monitoring unit is calculated based on spatialized monitoring data, including: For each spatial monitoring unit, environmental risk characteristics are extracted from the spatial monitoring data. These environmental risk characteristics include multiple features such as dust risk characteristics, harmful gas risk characteristics, noise risk characteristics, vibration risk characteristics, thermal environment risk characteristics, and visual recognition risk characteristics. The construction sequence corresponding to the spatial monitoring unit is determined based on the construction status data, and the impact of different environmental risk characteristics on the dynamic environmental risk value is adjusted according to the construction sequence. The dynamic environmental risk value is corrected based on the personnel distribution within the spatial monitoring unit, the operating status of construction equipment, spatial structure information, risk changes in adjacent spatial monitoring units, and historical risk change records.

[0014] Preferably, the monitoring uncertainty of each space monitoring unit is calculated based on the spatialized monitoring data, including: For each spatial monitoring unit, obtain its corresponding data coverage, data timeliness, data quality markers, multimodal data consistency, and monitoring equipment status; When a space monitoring unit has missing data modes, data acquisition time exceeds the preset time range, abnormal data quality marking, conflicting multimodal data, or abnormal monitoring equipment status, the monitoring uncertainty of the space monitoring unit shall be increased. When a space monitoring unit has multiple valid modal data, and the data acquisition time meets the preset timeliness requirements, the data quality label is normal, the multimodal data are consistent with each other, and the monitoring equipment is in normal condition, the monitoring uncertainty of the space monitoring unit is reduced.

[0015] Preferably, based on the dynamic environmental risk values ​​and monitoring uncertainties of each spatial monitoring unit in the dynamic environmental risk field, the target monitoring areas requiring enhanced monitoring are determined, including: The spatial monitoring units whose dynamic environmental risk values ​​reach the early warning conditions, whose dynamic environmental risk values ​​continuously increase within a continuous time window, whose monitoring uncertainty reaches the preset conditions, and whose spatial monitoring units are on the risk diffusion path determined based on the risk change status and spatial location relationship in the dynamic environmental risk field are identified as candidate spatial monitoring units. The priority of candidate spatial monitoring units is determined based on their dynamic environmental risk values, monitoring uncertainties, risk change status, personnel distribution, construction procedures, and spatial location relationships. Candidate spatial monitoring units that are spatially adjacent and have the same or similar main risk types are merged to form target monitoring areas, and target area description information is generated. The target area description information includes target area identifier, spatial location, main risk type, monitoring uncertainty, target area priority, and monitoring data types that need to be supplemented.

[0016] Preferably, with the goal of reducing the monitoring uncertainty of the target monitoring area, an active sensing strategy is generated based on the target monitoring area and the status data of the monitoring equipment, including: Candidate monitoring devices are determined from the monitoring device status data based on the main risk types and the types of monitoring data that need to be supplemented in the target monitoring area. Based on the equipment type, data acquisition capabilities, installation location, visibility range, mobility, and reachable area of ​​the candidate monitoring equipment, capability matching is performed on the candidate monitoring equipment. Active sensing tasks are generated based on the capability matching results. These active sensing tasks include at least one of the following: inspection tasks of mobile monitoring equipment, sampling and adjustment tasks of fixed monitoring equipment, and combined supplementary data collection tasks executed collaboratively by multiple monitoring equipment.

[0017] Preferably, when generating an active sensing task, the active sensing task is optimized based on the monitoring device status data, task execution cost, and security constraints, specifically including: Based on the communication status, remaining power, sensor operating status, current task status, and reachable area of ​​the monitoring equipment, the availability of candidate monitoring equipment is screened. The execution sequence of proactive sensing tasks is determined based on the dynamic environmental risk value, monitoring uncertainty, and target area priority of the target monitoring area. Based on the travel distance, estimated time, equipment power consumption, communication load, construction restricted areas, construction equipment operation danger zones and mobile monitoring equipment accessibility conditions, the active sensing task is constrained and conflict checked. When a conflict is detected in the proactive detection task, adjustments are made to the execution device, data collection location, data collection duration, or task sequence.

[0018] Preferably, the active sensing strategy is executed to acquire supplementary monitoring data of the target monitoring area, and the dynamic environmental risk field is updated based on the supplementary monitoring data, including: The supplementary monitoring data is mapped to the corresponding target monitoring area and its associated spatial monitoring unit, and the spatialized monitoring data of the corresponding spatial monitoring unit is corrected using the supplementary monitoring data; Based on the revised spatialized monitoring data, the dynamic environmental risk value and monitoring uncertainty of the target monitoring area and its associated spatial monitoring units are recalculated. Based on the recalculated dynamic environmental risk value and monitoring uncertainty, the main risk types, spatial range and risk change status of the target monitoring area are updated to form an updated dynamic environmental risk field. The updated dynamic environmental risk field outputs the construction operation environment monitoring results, which include various aspects such as the location of the risk area, risk level, main risk types, monitoring uncertainty status, data source description, risk diffusion trend, supplementary monitoring results, and early warning information.

[0019] A construction operation environment active sensing and monitoring system based on a dynamic environmental risk field, characterized in that it includes: The data acquisition module is used to acquire multimodal environmental data, construction status data, and monitoring equipment status data at the construction site. The spatiotemporal registration module is used to perform spatiotemporal registration on the multimodal environmental data, construction status data and monitoring equipment status data, and map them to pre-divided spatial monitoring units to obtain spatialized monitoring data of each spatial monitoring unit. The risk field generation module is used to calculate the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit based on the spatialized monitoring data, and form the dynamic environmental risk field at the current moment. The target area determination module is used to determine the target monitoring area that needs enhanced monitoring based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field. The active sensing strategy generation module is used to generate an active sensing strategy based on the target monitoring area and the status data of the monitoring equipment, with the goal of reducing the monitoring uncertainty of the target monitoring area. The risk field update and result output module is used to execute the active perception strategy to obtain supplementary monitoring data of the target monitoring area, update the dynamic environmental risk field according to the supplementary monitoring data, output the construction operation environment monitoring results, and use the updated dynamic environmental risk field as the basis for the next round of target monitoring area identification.

[0020] The beneficial effects of the technical solution provided by the embodiments of the present invention are: 1. By performing spatiotemporal registration of multimodal environmental data, construction status data and monitoring equipment status data, and mapping them to a spatial monitoring unit, a dynamic environmental risk field reflecting the distribution of on-site risks and the reliability of monitoring can be formed; 2. By simultaneously calculating dynamic environmental risk values ​​and monitoring uncertainties, the problems of false alarms and missed alarms caused by relying solely on single-point thresholds can be avoided; 3. By identifying target monitoring areas and generating proactive sensing strategies, it is possible to dynamically schedule fixed and mobile monitoring equipment to conduct supplementary data collection in high-risk, high-uncertainty, or risk-spreading areas. 4. By supplementing monitoring data and updating the dynamic environmental risk field in reverse, and using it as the basis for the next round of identification, a closed loop of risk identification, proactive supplementary sampling, risk correction and continuous monitoring is achieved, thereby improving the timeliness, accuracy, coverage and resource utilization efficiency of environmental monitoring during construction operations. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0022] Example 1 See Figure 1 This invention provides a method for active sensing and monitoring of construction operation environment based on dynamic environmental risk field, comprising the following steps: S1. Acquire multimodal environmental data, construction status data, and monitoring equipment status data at the construction site; The construction site can be a building construction site, a foundation pit construction site, an underground space construction site, a tunnel construction site, a pipe gallery construction site, or other construction engineering work areas where personnel, equipment, materials, and environmental risks interact and affect each other. The site is equipped with fixed monitoring equipment, mobile monitoring equipment, construction status acquisition equipment, site gateways, and edge computing nodes. Among them, fixed monitoring equipment collects environmental information at fixed locations, mobile monitoring equipment collects environmental information along the movement path, construction status acquisition equipment collects construction procedures, personnel distribution, and construction equipment operation information, site gateways aggregate data from different sources, and edge computing nodes perform preprocessing, spatiotemporal registration, risk field generation, and active perception strategy generation on the aggregated data.

[0023] Multimodal environmental data refers to multi-source raw data that reflects the environmental conditions of a construction site, including multiple types of data such as image data, video data, dust concentration data, harmful gas concentration data, temperature and humidity data, wind speed and direction data, noise data, vibration data, illuminance data, and thermal imaging data. Image and video data are collected by cameras mounted on fixed cameras, dome cameras, or mobile monitoring equipment to identify dust, smoke, water accumulation, personnel gathering, construction equipment operation status, material stacking status, and work surface obstruction. Dust concentration data is collected by particulate matter sensors to reflect changes in airborne particulate matter caused by operations such as cutting, crushing, mixing, and earthwork excavation. Harmful gas concentration data... Temperature data, collected by gas sensors, reflects the accumulation of harmful gases in basements, pipe corridors, foundation pits, confined spaces, welding areas, or spraying areas; temperature and humidity data reflect heat stress risk, humid conditions, condensation conditions, and abnormal ventilation conditions; wind speed and direction data reflect the diffusion conditions of dust, smoke, and gases; noise data reflects changes in the acoustic environment caused by cutting, crushing, hoisting, and impact operations; vibration data reflects the operation of construction equipment, disturbance of support structures, ground impact, and abnormal vibration; illuminance data reflects the visibility conditions of nighttime construction, underground construction, or obstructed areas; and thermal imaging data reflects localized temperature rise, equipment overheating, welding heat sources, and early signs of fire.

[0024] Construction status data refers to data used to describe the construction background corresponding to the current environmental data. This includes construction procedure information, construction area information, construction progress information, personnel location information, construction equipment location information, construction equipment operation status information, material stacking information, and spatial structure information in the building information model. Construction procedure information can come from the construction planning system, team operation records, construction equipment operation records, video recognition results, or sound recognition results. It is used to determine whether the corresponding area is in the cutting, welding, spraying, concrete pouring, earthwork excavation, formwork erection, formwork dismantling, hoisting, or other construction states. Personnel location information can be obtained from location tags, access control records, video human body detection results, or mobile terminal location results. It is used to determine whether personnel are in areas affected by environmental risks. Construction equipment location information and construction equipment operation status information can be obtained from equipment location tags, equipment controllers, power consumption monitoring modules, or mechanical operation status acquisition modules. It is used to identify construction equipment that may be associated with environmental risks. Spatial structure information in the building information model includes spatial objects such as floors, rooms, openings, fences, passages, underground spaces, edge areas, enclosed areas, and ventilation openings.

[0025] Monitoring equipment status data refers to data used to describe the working status and dispatchability of each monitoring device, including monitoring equipment identification, equipment type, installation location, current orientation, line of sight, sampling frequency, communication status, remaining power, storage status, sensor working status, equipment fault status, mobility, reachable area, and current task status. The status data of fixed monitoring equipment reflects whether it is online, whether there is obstruction, whether the sampling frequency meets the requirements, whether the collected data is stable, and whether the sampling parameters need to be adjusted. The status data of mobile monitoring equipment reflects its current location, remaining power, mobile range, communication quality, current task status, and whether it can go to the target monitoring area to perform supplementary data collection tasks.

[0026] Fixed monitoring equipment uploads environmental data according to a preset sampling cycle, mobile monitoring equipment uploads data collected along the route according to the current inspection task, and construction status acquisition equipment uploads construction status data according to the update frequency of the construction management system or changes in the operation of construction equipment; the site gateway receives data from different communication protocols and adds data source identifiers, acquisition time identifiers, equipment identifiers, and initial location identifiers to the received data, so that each piece of data can be traced back to its acquisition source, acquisition time, and acquisition location.

[0027] After acquiring the above data, preliminary preprocessing was performed on the multimodal environmental data, construction status data, and monitoring equipment status data to facilitate subsequent spatiotemporal registration and risk analysis. The preliminary preprocessing included data format unification, data integrity check, anomaly marking, unit unification, range verification, and quality mark generation. Videos, images, sensor values, monitoring equipment status information, and construction status information of different formats were converted into a unified data structure that edge computing nodes could recognize, while retaining basic fields such as original data source, acquisition time, equipment number, and data type.

[0028] When checking data integrity, determine if there are missing fields, sampling interruptions, duplicate uploads, abnormal null values, or communication packet loss. For missing or abnormal data, retain the original data content and add anomaly or low-confidence markers to differentiate and handle them when calculating dynamic environmental risk values ​​and monitoring uncertainties. For sensor data, convert environmental parameters uploaded by sensors from different manufacturers and models to a unified unit, and identify data that clearly exceeds the physically reasonable range by combining equipment calibration information, sensor range, and historical stability range. For abnormal data caused by short-term spikes, communication jitter, or instantaneous sensor disturbances, preliminary noise reduction can be performed using sliding window smoothing, comparison of adjacent sampling points, comparison of historical baselines, or cross-validation of similar sensors. For data whose authenticity cannot be determined, retain the original values ​​and add data quality markers.

[0029] For image and video data, frame extraction, sharpness detection, occlusion detection, brightness detection, and image stability detection are performed. If an image is too dark, the camera is obstructed, the image is severely shaky, or the lens is contaminated, a low-quality label is added to the corresponding visual data. If the image quality meets the requirements, basic visual features for subsequent environmental risk assessment are extracted. These basic visual features include candidate information for smoke and dust areas, open flame areas, water accumulation areas, areas where people gather, areas where construction equipment operates, and areas where materials are piled up.

[0030] For construction status data, it is categorized and processed according to construction area, construction procedure, personnel, and construction equipment. Data from the construction planning system, construction equipment status acquisition module, personnel positioning module, and site management system are converted into a unified construction status record. When multiple construction status sources exist in the same area at the same time, data with real-time acquisition time is retained first, and planned data is used as auxiliary data. When real-time construction status is missing, the most recent valid construction status is used as a temporary construction status and a timeliness mark is added.

[0031] For monitoring device status data, the validity of the device's online status, communication status, remaining power, available sampling capacity, and current task status is checked; offline device records are marked as unavailable, and devices with low power, unstable communication, or abnormal sampling are marked as restricted devices; for monitoring devices that are still usable but whose data quality has deteriorated, their status data is retained, and their task allocation priority is reduced when the subsequent active sensing strategy is generated.

[0032] After the above processing, a set of monitoring data to be processed is generated, which includes multimodal environmental data, construction status data, monitoring equipment status data, and corresponding data quality labels.

[0033] S2. Perform spatiotemporal registration on the multimodal environment data, construction status data, and monitoring equipment status data, and map them to pre-divided spatial monitoring units to obtain spatialized monitoring data for each spatial monitoring unit. In this embodiment, spatiotemporal registration is used to unify data from different sources, at different collection times, in different spatial locations, and in different data formats under the same time and spatial reference, so that subsequent environmental risk analysis can be performed with spatial monitoring units as the object. Since the data sampling frequencies of cameras, dust sensors, gas sensors, noise sensors, vibration sensors, wind speed and direction sensors, personnel positioning devices, construction equipment positioning devices, mobile monitoring devices, and construction management systems at the construction site are different, and their installation locations and data expression methods are different, it is necessary to perform time alignment and spatial attribution processing on various types of data first.

[0034] When performing time registration, edge computing nodes use a unified clock as the time reference to perform time stamp correction on the data uploaded by each monitoring device. The unified clock can come from the field gateway, edge computing nodes, network time server, or construction site unified management platform. For data that carries the acquisition time, the edge computing node reads the acquisition time and corrects it according to the device reporting delay, network transmission delay, and device clock deviation. For data that does not carry the precise acquisition time, the edge computing node can use the time received by the field gateway as a temporary time stamp and estimate the acquisition time by combining it with the device upload cycle.

[0035] In one implementation, time registration is performed according to a preset time window; edge computing nodes aggregate multimodal environmental data, construction status data, and monitoring equipment status data collected within the same time window into monitoring data for the same moment or time period; for video data, vibration data, and noise data with high sampling frequency, representative segments, statistical features, or event features can be extracted within the time window; for data such as gas concentration, dust concentration, temperature and humidity, wind speed and direction with low sampling frequency, the monitoring value corresponding to the current time window can be formed by using the most recent valid value, window average value, or time interpolation; for construction status data and monitoring equipment status data, the latest valid status within the current time window can be selected as the status data for that time window.

[0036] When performing spatial registration, a unified spatial coordinate benchmark is first established for the construction site. This benchmark can be determined based on building information model, construction layout plan, floor plan, foundation pit zoning plan, tunnel mileage coordinates, on-site measurement control points, or a pre-established 3D scene model. For monitoring equipment with fixed installation locations, its spatial location, installation height, orientation, field of view, and corresponding construction area are recorded during the equipment installation or initialization phase and registered in the equipment spatial information table. For mobile monitoring equipment, the current location is obtained through positioning modules, inertial measurement modules, odometers, visual positioning, ultra-wideband positioning, Bluetooth positioning, or on-site marker identification, and the collected data is associated with the corresponding spatial location.

[0037] For sensor data, edge computing nodes determine the spatial influence range of the sensor data based on the sensor's installation location, effective monitoring radius, sensor type, and on-site obstruction relationships. For example, data from dust and gas sensors can be mapped to the air environment area near their installation point, noise sensor data can be mapped to the effective acoustic coverage area, vibration sensor data can be mapped to the structural components, ground area, or equipment foundation area, and wind speed and direction sensor data can be mapped to the spatial unit where it is located and the spatial units in adjacent diffusion directions. For visual data, edge computing nodes map information such as smoke, open flames, water accumulation, personnel, construction equipment, and material stacking identified in images or videos to corresponding spatial locations based on the camera's installation location, shooting direction, focal length parameters, field of view, and the location of the target area in the image.

[0038] In one implementation, if the camera has been calibrated with the spatial coordinate reference of the construction site, the target area in the image is converted into the spatial location of the site according to the camera calibration parameters and the target position in the image. If the camera has not been accurately calibrated, the visual data can be mapped to the corresponding spatial monitoring unit according to the camera installation area, shooting range, manually marked reference points, floor number, component number, or area number. For video or image data collected by mobile monitoring equipment, the corresponding spatial monitoring unit can be determined by combining the position, posture, and shooting direction at the time of collection.

[0039] Before spatial mapping, the construction site is pre-divided into multiple spatial monitoring units. These spatial monitoring units are the basic spatial objects that carry multimodal monitoring data and dynamic environmental risk values. They can be divided according to construction floors, rooms, work surfaces, foundation pit zones, tunnel mileage sections, construction roads, construction equipment operation areas, material storage areas, opening edge areas, ventilation areas, or regular grids. For indoor spaces, basements, pipe corridors, tunnels, etc., spatial monitoring units can be divided according to rooms, passages, zones, or mileage sections in the building information model. For open construction sites, spatial monitoring units can be divided according to planar grids of preset length and width. For multi-story building construction sites, three-dimensional spatial monitoring units can be divided based on the planar grid and floor height.

[0040] The granularity of spatial monitoring units can be determined based on the scale of the construction site, the density of sensor deployment, the accuracy requirements of risk monitoring, and computing resources. For areas with strong diffuse environmental risks such as dust, gas, temperature, and humidity, a finer granularity can be used to more accurately describe the risk diffusion. For areas where risks change slowly or where sensor coverage is sparse, a coarser granularity can be used to reduce the computational burden. For areas with dense personnel operations, welding, cutting, spraying, deep foundation pits, hoisting, or underground confined space operations, the accuracy of spatial monitoring unit division can be improved.

[0041] After completing the spatial monitoring unit division, the edge computing nodes map the time-registered and spatially registered data to the corresponding spatial monitoring units. For each piece of multimodal environmental data, the edge computing nodes determine one or more corresponding spatial monitoring units based on the acquisition time, acquisition device location, effective coverage area, and construction area, and write the data into the corresponding spatial monitoring unit's data record. For data that simultaneously affects multiple spatial monitoring units, such as wind speed and direction data, smoke and dust area data within the camera's field of view, or data collected along the way by mobile monitoring equipment, the data can be allocated to multiple spatial monitoring units according to data coverage, distance relationship, field of view, or acquisition trajectory, and the data source and applicable scope are recorded.

[0042] For construction status data, edge computing nodes map it to the corresponding construction area or spatial monitoring unit. For example, when a construction plan or construction equipment operation record indicates that a cutting operation is underway in a certain area, the cutting process information is associated with the spatial monitoring unit corresponding to that area. When personnel location data indicates that personnel are in a certain work area, the number of personnel, personnel stay time, or personnel density information is associated with the spatial monitoring unit corresponding to that work area. When construction equipment location data indicates that a certain construction equipment is in a certain spatial monitoring unit and is in operation, the equipment type, operation status, and work intensity information are associated with that spatial monitoring unit.

[0043] For the status data of monitoring equipment, the edge computing node establishes an association between it and the spatial monitoring units covered or accessible by the monitoring equipment; the status data of fixed monitoring equipment is associated with its installation location and the spatial monitoring units within its coverage area, which is used to explain the data collection source and equipment availability status within these spatial monitoring units; the status data of mobile monitoring equipment is associated with its current location, accessible area and current task range, which is used to subsequently determine whether the mobile monitoring equipment is suitable for performing supplementary data collection tasks for a certain target monitoring area.

[0044] During the mapping process, if multiple data sources exist within the same spatial monitoring unit within the same time window, the edge computing nodes merge the data according to data type, acquisition time, monitoring equipment status, data quality markers, and spatial distance. For data of the same type, multiple data sources and corresponding markers can be retained, or a comprehensive data record representing the current status of the spatial monitoring unit can be generated. For data of different modalities, instead of simple overwriting, they are stored separately in the corresponding data fields of the spatial monitoring unit for subsequent multimodal fusion. For conflicting data, such as large differences in readings between adjacent sensors, inconsistencies between visual data and sensor data, or inconsistencies between the status of construction equipment and the construction plan, the edge computing nodes retain conflict markers and use these conflict markers as the basis for subsequent calculation of monitoring uncertainty.

[0045] Through the above processing, each spatial monitoring unit generates spatialized monitoring data corresponding to the current time window. The spatialized monitoring data includes environmental parameters, visual characteristics, construction procedures, personnel distribution, construction equipment operation status, monitoring equipment status, and data quality markers within or affecting the spatial monitoring unit. This spatialized monitoring data is used to describe the environmental status, construction status, and data acquisition status of the corresponding spatial monitoring unit at the current moment, and serves as the basis for subsequent calculations of dynamic environmental risk values ​​and monitoring uncertainties.

[0046] This step transforms data that was originally scattered across different devices, systems, times, and spatial locations into a unified data representation format using spatial monitoring units as carriers. Subsequent steps enable risk value calculation, monitoring uncertainty assessment, target monitoring area determination, and active sensing strategy generation for each spatial monitoring unit, thereby ensuring that the dynamic environmental risk field has a clear spatial basis and data source.

[0047] S3. Calculate the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit based on the spatialized monitoring data to form the dynamic environmental risk field at the current moment; In this embodiment, the dynamic environmental risk value is used to characterize the degree of environmental risk of construction operation occurring in the corresponding spatial monitoring unit at the current moment, and the monitoring uncertainty is used to characterize the reliability and information sufficiency of the current monitoring results of the corresponding spatial monitoring unit. The edge computing node calculates the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit based on the spatial monitoring data obtained in step S2, and combines the calculation results of each spatial monitoring unit according to the on-site spatial location to form the dynamic environmental risk field at the current moment.

[0048] For any spatial monitoring unit, the edge computing node reads the spatialized monitoring data of that spatial monitoring unit within the current time window. The spatialized monitoring data includes multimodal environmental data, construction status data, monitoring equipment status data, and data quality markers. Multimodal environmental data is used to reflect the environmental change status within the spatial monitoring unit, construction status data is used to describe the current operational background, monitoring equipment status data is used to describe the availability status of the corresponding monitoring data source, and data quality markers are used to describe whether there are any missing, abnormal, delayed, conflicting, or low-quality data.

[0049] When calculating dynamic environmental risk values, edge computing nodes extract risk features according to different environmental risk types. For dust risk, the degree of dust risk is determined based on dust concentration data, dust areas in visual images, current wind speed and direction, work procedures, and personnel distribution. For hazardous gas risk, the degree of gas risk is determined based on gas concentration data, ventilation status, underground or enclosed space attributes, personnel residence status, and gas changes in adjacent space monitoring units. For noise risk, the degree of noise risk is determined based on noise intensity, duration, construction equipment operating status, personnel distance, and current construction procedures. For vibration risk, the degree of vibration risk is determined based on vibration intensity, duration, trend, construction equipment operating status, and spatial structural attributes. For thermal environment risk, the degree of heat stress risk is determined based on temperature, humidity, thermal imaging information, ventilation status, work intensity, and personnel residence time. For visual recognition risk, the degree of risk is determined based on visual features identified in images or videos, such as smoke, open flames, water accumulation, obstruction, personnel gathering, abnormal operation of construction equipment, and abnormal material stacking.

[0050] In one implementation, edge computing nodes first convert different types of raw environmental data into a unified risk level or risk score. For sensor numerical data, it can be converted into low-risk, medium-risk, or high-risk levels based on the safety limits, warning limits, historical baselines, and current process requirements of the corresponding environmental parameters. For image and video data, it can be converted into a corresponding visual risk level based on the identified risk target type, target area, duration, location relationship, and distance from personnel or construction equipment. For continuous signals such as sound and vibration, it can be converted into a corresponding noise and vibration risk level based on intensity changes, duration, frequency band characteristics, and whether it matches the current construction process. Through this processing, data of different modalities, different units, and different sampling methods can be uniformly expressed for risk within the same spatial monitoring unit.

[0051] After identifying various risk characteristics, edge computing nodes combine construction status data to perform process-specific adaptation of these characteristics. Different construction processes correspond to different environmental risk priorities; therefore, the same environmental parameter has different risk implications under different processes. For example, in cutting, crushing, or earthwork excavation processes, dust concentration, visual characteristics of dust, and wind direction changes have a significant impact on dynamic environmental risk values. In welding processes, smoke, temperature rise, harmful gases, sparks, and the state of surrounding combustibles have a significant impact. In spraying or waterproofing operations, harmful gas concentration, ventilation status, and personnel dwell time have a significant impact. In basement, tunnel, or pipe gallery operations, oxygen content, harmful gases, humidity, and ventilation status have a significant impact. In hoisting or large construction equipment operations, wind speed, noise, vibration, the location of construction equipment, and personnel distribution have a significant impact. Edge computing nodes assign different levels of importance to different risk characteristics based on the construction process corresponding to the current spatial monitoring unit, enabling dynamic environmental risk values ​​to reflect the real environmental risks under specific operational contexts.

[0052] Edge computing nodes can also adjust dynamic environmental risk values ​​based on the distribution of personnel and construction equipment within a spatial monitoring unit. If there are many workers or personnel staying in a spatial monitoring unit for an extended period, the risk level can be increased even if environmental parameters have not yet reached a severely abnormal state. If there are operating cutting machines, welding machines, spraying equipment, air compressors, crushing equipment, hoisting equipment, or other construction equipment that may cause environmental disturbances within a spatial monitoring unit, the corresponding environmental risk impact level can be increased based on the equipment type and operating status. If a spatial monitoring unit is located in a closed space, semi-closed space, ventilation dead corner, near an edge opening, a densely stacked material area, or a personnel passageway, the dynamic environmental risk value can be adjusted according to the spatial attributes.

[0053] In one implementation, edge computing nodes spatially correct the dynamic environmental risk value of the current spatial monitoring unit based on data changes from neighboring spatial monitoring units. For environmental risks with diffusion characteristics, such as dust, smoke, harmful gases, and heat, if the risk of the upwind spatial monitoring unit or adjacent spatial monitoring units shows an increasing trend, the potential risk level of the current spatial monitoring unit is increased. If the direct monitoring data of the current spatial monitoring unit is insufficient, but there is a clear risk diffusion trend in the neighboring area, the estimated risk value of the current spatial monitoring unit is generated based on the risk status of the adjacent spatial monitoring units, avoiding risk omissions caused by relying solely on a single monitoring point.

[0054] In one implementation, edge computing nodes perform trend correction on the current dynamic environmental risk value based on historical data. For each spatial monitoring unit, the edge computing node stores its environmental risk change records within multiple historical time windows. When the environmental parameters of a spatial monitoring unit have not yet exceeded the warning limit, but the dust concentration, harmful gas concentration, noise intensity, vibration intensity, or temperature and humidity conditions continue to rise in a short period of time, the upward trend is taken as a risk amplification factor. When the environmental parameters of a spatial monitoring unit continue to decline after ventilation, spraying, shutdown of construction equipment, or evacuation of personnel, the downward trend is taken as a risk mitigation factor. By introducing trend changes, the dynamic environmental risk value can reflect the current risk status and short-term development trend.

[0055] After completing the above processing, the edge computing node generates a dynamic environmental risk value for each spatial monitoring unit at the current moment. This dynamic environmental risk value can be stored in the form of risk level, risk score, color code, or warning status. The risk level can include multiple levels such as normal, attention, warning, and danger. The risk score can be used to represent the risk differences between different spatial monitoring units. The color code can be used to display the risk heat distribution in the construction site plan, 3D model, or management platform interface. The warning status can be used to trigger subsequent target monitoring area identification and active perception strategy generation.

[0056] While calculating the dynamic environmental risk value, the edge computing nodes calculate the monitoring uncertainty of each space monitoring unit. The monitoring uncertainty is used to describe the credibility of the current risk assessment result of the space monitoring unit. The higher the monitoring uncertainty, the less sufficient the monitoring information of the space monitoring unit, the more unstable the data quality, or the more the risk assessment needs to be supplemented with observations. The lower the monitoring uncertainty, the more sufficient the data source, the more stable the data quality, and the more reliable the risk assessment.

[0057] Edge computing nodes can determine monitoring uncertainty based on data coverage. If a spatial monitoring unit has multiple valid modal data and the data acquisition time is relatively recent, the monitoring uncertainty is low. If only a single modal data is available, or some key modal data are missing, the monitoring uncertainty is high. For example, for welding operation areas, the monitoring uncertainty increases if gas concentration data or visual smoke data is missing; for cutting operation areas, the monitoring uncertainty increases if dust data or wind speed and direction data is missing; and for underground enclosed spaces, the monitoring uncertainty increases if harmful gas data or ventilation status data is missing.

[0058] Edge computing nodes can also determine monitoring uncertainty based on data timeliness, data quality markers, multimodal consistency, and monitoring equipment status. If the data comes from the current time window, and multiple data sources are normal, different modalities are consistent, and the monitoring equipment covering the spatial monitoring unit is online and in stable communication, the monitoring uncertainty decreases. If the data comes from historical data, data from adjacent areas, or estimated data, or if there are sensor anomalies, communication packet loss, image blurring, image occlusion, lens contamination, sensor drift, low device battery, unstable communication, or multimodal data conflicts, the monitoring uncertainty increases. For example, if there is obvious dust in the visual image, and the dust sensor reading increases, and cutting operations are underway in the area, the risk assessment is highly reliable. If there is obvious smoke in the visual image but the gas sensor shows no significant change, or the sensor reading increases significantly but no corresponding operation or environmental change is observed in the video, the edge computing node increases the monitoring uncertainty so that the area can be considered as a candidate for enhanced monitoring.

[0059] In one implementation, the edge computing node stores the dynamic environmental risk value and monitoring uncertainty in the data record of the corresponding spatial monitoring unit. For each spatial monitoring unit, the data record includes at least the spatial unit identifier, spatial location, current time window, dynamic environmental risk value, risk level, monitoring uncertainty, main risk type, main data source, and data quality label. The main risk type is used to describe the current main risks, such as dust risk, gas risk, noise risk, vibration risk, thermal environment risk, or visual recognition risk. The main data source is used to describe which sensor data, visual data, or construction status data are mainly used to generate the risk judgment.

[0060] After each spatial monitoring unit completes the calculation of dynamic environmental risk value and monitoring uncertainty, the edge computing node combines the calculation results into a dynamic environmental risk field at the current moment according to the positional relationship between the spatial monitoring units. The dynamic environmental risk field is used to characterize the spatial distribution of environmental risks and the distribution of monitoring reliability at the construction site at the current moment. It can be stored and displayed in the form of two-dimensional planar heat map, three-dimensional spatial grid map, building information model risk layer, floor risk distribution map or construction area risk list.

[0061] In one implementation, each spatial monitoring unit in the dynamic environmental risk field has an environmental risk attribute and a monitoring uncertainty attribute; the environmental risk attribute indicates the level of environmental risk at that location, and the monitoring uncertainty attribute indicates whether the monitoring information at that location is sufficient; thus, subsequent steps can not only identify areas with high risk values, but also identify areas where the risk is not yet clear but the monitoring information is insufficient, enabling the active sensing strategy to supplement monitoring for high-risk and high-uncertainty areas.

[0062] This step transforms spatial monitoring data into a dynamic environmental risk field with spatial distribution characteristics. This dynamic environmental risk field is not a single alarm result, but a comprehensive expression of the environmental risk status and monitoring reliability of each spatial monitoring unit at the construction site. Subsequent steps can determine the target monitoring area based on the dynamic environmental risk field and generate corresponding active sensing strategies, thereby realizing proactive environmental monitoring driven by risk status.

[0063] S4. Based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field, determine the target monitoring area that needs enhanced monitoring. In this embodiment, the target monitoring area refers to the spatial area in the current dynamic environmental risk field where there is a high environmental risk, an upward trend in risk, a possibility of risk diffusion, or insufficient monitoring information. The target monitoring area is used to indicate the location where subsequent active sensing strategies should prioritize supplementary data collection, so that monitoring resources can be transferred from low-risk, low-uncertainty areas to areas that require more supplementary sensing.

[0064] After forming the dynamic environmental risk field at the current moment, the edge computing node reads the dynamic environmental risk value and monitoring uncertainty corresponding to each spatial monitoring unit. The dynamic environmental risk value reflects the current environmental risk level, and the monitoring uncertainty reflects the sufficiency of the current monitoring information and the reliability of the risk judgment. The edge computing node determines the target monitoring area based on both, rather than judging solely on whether a single environmental parameter exceeds the threshold.

[0065] In one implementation, the edge computing node first classifies the risk status of each spatial monitoring unit; when the dynamic environmental risk value of a spatial monitoring unit reaches the warning condition, it is identified as a high-risk candidate unit; when the dynamic environmental risk value has not yet reached the warning condition, but continues to rise within a continuous time window, it is identified as a risk-rising candidate unit; when the monitoring uncertainty is high, it is identified as an uncertain candidate unit; when the spatial monitoring unit is located in the adjacent area of ​​the high-risk candidate unit, or on the upwind or downwind diffusion path, it is identified as a diffusion-related candidate unit.

[0066] For high-risk candidate units, edge computing nodes determine whether the source of risk is clear and whether the risk assessment is reliable. If the dynamic environmental risk value is high and the monitoring uncertainty is low, it indicates that the risk in the area has been sufficiently confirmed and can be identified as a key tracking area, and warning information will be output subsequently. If the dynamic environmental risk value is high and the monitoring uncertainty is also high, it indicates that the area has both significant environmental risks and lacks sufficient monitoring information. Edge computing nodes will prioritize identifying it as a target monitoring area so as to confirm the risk type, risk intensity and risk range through supplementary data collection.

[0067] For candidate units with rising risks, edge computing nodes combine historical risk change processes to determine whether enhanced monitoring is needed. If a spatial monitoring unit continuously experiences increased dust concentration, increased gas concentration, increased noise, increased vibration, abnormal temperature and humidity, or enhanced visual risk characteristics in recent time windows, it can be identified as a target monitoring area even if the current risk level has not yet reached the danger level. Through this process, the system can supplement the perception of risk precursor areas before the risk is fully formed.

[0068] For uncertain candidate units, the edge computing node determines the cause of the increased monitoring uncertainty. If the increased uncertainty is caused by sensor offline, visual obstruction, outdated sampling time, missing modal data, multimodal data conflict, or abnormal monitoring equipment status, then the spatial monitoring unit is selected as a candidate target monitoring area. If the spatial monitoring unit is also located in the vicinity of densely populated areas, key construction process areas, or high-risk areas, then its priority in being identified as a target monitoring area is increased.

[0069] For diffusion-related candidate units, edge computing nodes determine whether enhanced monitoring is needed based on spatial relationships in the dynamic environmental risk field. When there are wind speed and direction indicators indicating the diffusion direction of dust, smoke, or harmful gases around a high-risk spatial monitoring unit, adjacent spatial monitoring units in the diffusion direction are included in the candidate range. When the risk values ​​between adjacent spatial monitoring units show a continuous increase or a spatial distribution that decreases from the high-risk area outwards, spatial monitoring units that may be affected by risk propagation are included in the candidate range. Thus, the target monitoring area includes not only areas where risks have already occurred, but also areas that may be affected by risks.

[0070] In one implementation, edge computing nodes determine the priority of target monitoring areas based on the combination of dynamic environmental risk values ​​and monitoring uncertainties. Spatial monitoring units with both high dynamic environmental risk values ​​and high monitoring uncertainties have the highest priority, indicating that the area requires immediate supplementary monitoring. Spatial monitoring units with high dynamic environmental risk values ​​but low monitoring uncertainties have the next highest priority, indicating that the risk in the area is clear but still requires continuous monitoring. Spatial monitoring units with low dynamic environmental risk values ​​but high monitoring uncertainties are given higher priority if they are located on critical work surfaces, densely populated areas, or risk diffusion paths. Spatial monitoring units with both low dynamic environmental risk values ​​and low monitoring uncertainties are not considered target monitoring areas or are given lower priority.

[0071] Edge computing nodes can also merge candidate spatial monitoring units based on spatial continuity. When multiple adjacent spatial monitoring units meet the target monitoring area conditions or have the same or similar main risk types, these spatial monitoring units are merged into a single target monitoring area. When a single spatial monitoring unit exhibits an isolated anomaly, but adjacent spatial monitoring units do not show an increase in risk or monitoring uncertainty, the edge computing node can mark the unit as a unit to be verified and determine whether to merge it into the target monitoring area based on subsequent data. Through area merging, the fragmentation of the target area can be reduced, facilitating the subsequent generation of continuous inspection paths, sampling adjustment strategies, or supplementary data collection tasks.

[0072] When determining the target monitoring area, edge computing nodes can also perform redundant screening of candidate areas; if a candidate area has been stably covered by fixed monitoring equipment and the current monitoring uncertainty is low, its enhanced monitoring priority is reduced; if a candidate area has a high risk but there is already a mobile monitoring task being executed, the duplicate dispatch of supplementary monitoring tasks is avoided; if the risk status of a candidate area is highly similar to that of an adjacent area and the adjacent area already has a target monitoring task with a higher priority, the candidate area is merged into the adjacent target monitoring area for unified processing.

[0073] In one implementation, the edge computing node generates target area description information for each target monitoring area. The target area description information includes the target area identifier, the corresponding spatial monitoring unit range, spatial location, main risk type, dynamic environmental risk value, monitoring uncertainty, risk change status, target area priority, and data types of monitoring that need to be supplemented. The main risk type indicates that the target area mainly has dust risk, gas risk, noise risk, vibration risk, thermal environment risk, or visual recognition risk. The risk change status indicates that the target area is in a state of increased risk, stable risk, risk diffusion, or risk pending verification. The data types of monitoring that need to be supplemented indicate which data should be prioritized for subsequent active sensing strategies.

[0074] For example, when the dust risk in a cutting operation area increases, and the camera view in that area is obstructed and the dust sensor sampling time is old, the edge computing node identifies that area as the target monitoring area and marks the data types of monitoring that need to be supplemented as dust concentration data, visual image data, and wind speed and direction data. As another example, when a monitoring unit in a basement space lacks gas concentration data, but the risk of harmful gases in its adjacent area increases and there are people staying there, the edge computing node merges the monitoring unit and its adjacent area to identify the target monitoring area and marks the data types of monitoring that need to be supplemented as harmful gas concentration data and ventilation status data.

[0075] If there are no high-risk areas in the current dynamic environmental risk field, but there are multiple spatial monitoring units with high monitoring uncertainty, the edge computing nodes can still determine the target monitoring area based on the monitoring uncertainty, personnel distribution, construction procedures, and spatial location relationships. As a result, the system can not only enhance monitoring after the risk increases, but also actively supplement data in monitoring blind spots, data missing areas, or key operation areas, thereby reducing the overall uncertainty of the dynamic environmental risk field.

[0076] After determining the target monitoring area, the edge computing node writes the target area description information into the target monitoring task queue. The target monitoring task queue is called when the subsequent active sensing strategy is generated, so that the scheduling method of mobile monitoring equipment, the sampling adjustment method of fixed monitoring equipment, or other active sensing methods can be determined according to the location, priority, required supplementary data type, and monitoring equipment status data of the target area.

[0077] This step allows us to select the spatial regions most in need of supplementary observation from the dynamic environmental risk field. This selection process considers both the dynamic environmental risk value and the monitoring uncertainty, ensuring that the enhanced monitoring targets are not limited to high-risk areas that have exceeded the warning conditions, but also include areas with rapidly changing risks, areas with risk diffusion and related risks, and areas with insufficient monitoring information. This provides clear spatial objectives and task basis for the subsequent generation of proactive sensing strategies.

[0078] S5. With the goal of reducing the monitoring uncertainty of the target monitoring area, generate an active sensing strategy based on the target monitoring area and the status data of the monitoring equipment. In this embodiment, the proactive sensing strategy refers to the supplementary monitoring execution plan automatically generated by the edge computing node based on the target monitoring area determined in step S4, combined with the working status, spatial location, acquisition capability, and task constraints of the currently available monitoring equipment. The proactive sensing strategy is used to guide fixed monitoring equipment, mobile monitoring equipment, or other schedulable monitoring resources to perform targeted supplementary acquisition of the target monitoring area, thereby reducing the monitoring uncertainty of the target monitoring area and improving the accuracy of the dynamic environmental risk field.

[0079] When generating an active sensing strategy, the edge computing node first reads the target area description information from the target monitoring task queue. The target area description information includes the target area identifier, spatial location, corresponding spatial monitoring unit range, main risk type, dynamic environmental risk value, monitoring uncertainty, risk change status, target area priority, and the data types of monitoring that need to be supplemented. The data types of monitoring that need to be supplemented are used to indicate what kind of monitoring information is currently missing in the target monitoring area, such as visual image data, dust concentration data, harmful gas concentration data, wind speed and direction data, noise data, vibration data, thermal imaging data, or personnel distribution data.

[0080] Edge computing nodes read the status data of monitoring devices and determine candidate monitoring devices that can be used to perform active sensing tasks based on the status data. For fixed monitoring devices, edge computing nodes determine whether they are online, whether communication is stable, whether sampling is normal, whether the sampling frequency is adjustable, whether they have directional adjustment capabilities, whether the visible range covers the target monitoring area, and whether there are any obstructions or malfunctions. For mobile monitoring devices, edge computing nodes determine their current location, remaining battery power, communication quality, mobile range, reachable path, type of sensor they carry, current task status, and whether they have the conditions to enter the target monitoring area or adjacent areas for supplementary data collection.

[0081] In one implementation, the edge computing node performs capability matching on candidate monitoring devices based on the types of monitoring data required to supplement the target monitoring area. When visual image data is required, fixed cameras, rotatable cameras, dome cameras, or mobile monitoring devices equipped with cameras are preferred. When dust concentration data is required, mobile monitoring devices equipped with particulate matter sensors and capable of reaching the target area or its upwind or downwind location are preferred. When harmful gas concentration data is required, mobile monitoring devices equipped with gas sensors and capable of entering underground spaces, enclosed spaces, or locations adjacent to the target area are preferred. When wind speed, wind direction, noise, vibration, or thermal imaging data is required, fixed monitoring devices or mobile monitoring devices with corresponding acquisition capabilities are preferred.

[0082] After completing the device capability matching, the edge computing node performs availability screening on the candidate monitoring devices. If a monitoring device is offline, faulty, low in power, has communication abnormalities, abnormal sampling, or is executing a high-priority task, its task allocation priority is reduced, or it is excluded from this round of active sensing tasks. If a monitoring device is online, has stable communication, sufficient power, normal sampling, and has the data collection capability required by the target area, it is retained as an executable device. If a monitoring device has low power or unstable communication, but is closest to the target monitoring area and can complete supplementary collection in a short time, it can be retained as a limited available device, and a shorter collection duration or lower task load is set in the active sensing strategy.

[0083] Edge computing nodes determine the execution order of proactive sensing tasks based on the priority of target monitoring areas. For target monitoring areas with high dynamic environmental risk values ​​and high monitoring uncertainty, supplementary monitoring tasks are generated first. For target monitoring areas with high dynamic environmental risk values ​​but low monitoring uncertainty, continuous tracking tasks are generated. For target monitoring areas with low dynamic environmental risk values ​​but high monitoring uncertainty, and located in densely populated areas, key process areas, or risk diffusion paths, verification monitoring tasks are generated. For areas with low dynamic environmental risk values ​​and low monitoring uncertainty, no proactive sensing tasks are generated, or only routine monitoring is maintained.

[0084] In one implementation, edge computing nodes generate inspection tasks for mobile monitoring equipment based on the location of the target monitoring area, risk type, and monitoring equipment status. The inspection task includes the location of the target monitoring area, movement path, arrival sequence, data collection points, data collection direction, data collection duration, data type, and task completion conditions. When generating the movement path, the path that can safely reach the target monitoring area or a nearby observation location is selected by considering factors such as construction site passages, floors, adjacent areas, enclosed areas, restricted areas, construction equipment operating areas, densely populated areas, and the reachability of the mobile monitoring equipment. When there is a risk of dust, smoke, or harmful gas diffusion in the target monitoring area, the data collection points of the mobile monitoring equipment can be arranged inside the target area, at the boundary of the target area, upwind, downwind, or near the personnel exposure path to obtain supplementary data that can reduce the uncertainty of risk assessment.

[0085] In one implementation, the edge computing node generates a sampling adjustment task for the fixed monitoring equipment based on the target monitoring area and the status of the fixed monitoring equipment. The sampling adjustment task includes increasing the sampling frequency, extending the sampling duration, adjusting the acquisition direction, adjusting the shooting focal length, activating specific sensors, increasing the data upload frequency, or changing the acquisition time window. For example, when the target monitoring area is at the edge of the field of view of a fixed camera with directional adjustment capability, the edge computing node controls the camera to turn towards the target monitoring area and adjusts the focal length to obtain clearer visual data. When the dust risk in a certain area rises rapidly but the dust sensor sampling period is long, the edge computing node increases the sampling frequency of the dust sensor within a preset time. When there is an uncertain gas risk in a certain underground space, the edge computing node increases the data upload frequency of the gas sensor in that area and simultaneously calls the ventilation status data of the adjacent area.

[0086] In one implementation, edge computing nodes can also generate combined proactive sensing strategies. When a single monitoring device cannot sufficiently reduce the monitoring uncertainty of the target monitoring area, the edge computing node can simultaneously schedule multiple monitoring devices to participate in supplementary data collection. For example, for dust risk areas caused by cutting operations, fixed cameras can be controlled to turn towards the area, while the sampling frequency of dust sensors is increased, and mobile monitoring devices can be scheduled to move to downwind positions to collect dust concentration and wind speed and direction data. For hazardous gas risk areas in basements, the sampling frequency of fixed gas sensors can be increased, while mobile monitoring devices can be scheduled to enter the periphery of the target area to collect gas concentration, and personnel location data can be called to confirm the personnel exposure range.

[0087] When generating an active sensing strategy, edge computing nodes also consider task execution costs and security constraints. Task execution costs include travel distance, estimated time, device power consumption, communication load, data processing load, and impact on existing inspection tasks. Security constraints include construction restricted areas, areas affected by hoisting operations, areas near openings, dangerous areas for construction equipment operations, access conditions for enclosed spaces, and conditions for mobile equipment to pass through. If a target monitoring area is high-risk but mobile monitoring equipment cannot safely enter, the edge computing node can choose to perform alternative data collection at the target area boundary, adjacent space monitoring units, ventilation openings, or visual observation locations, and mark it as an indirect supplementary data collection task in the active sensing strategy.

[0088] In one implementation, the edge computing node determines the priority of supplementary data collection based on the sources of monitoring uncertainty in the target monitoring area. When the monitoring uncertainty mainly stems from missing data, the proactive sensing strategy prioritizes supplementing the data corresponding to the missing modes. When the monitoring uncertainty mainly stems from insufficient data timeliness, the latest data collection is prioritized. When the monitoring uncertainty mainly stems from multimodal data conflicts, data collection that can verify the cause of the conflict is prioritized. When the monitoring uncertainty mainly stems from monitoring equipment malfunctions, other equipment is prioritized for replacement data collection, or the sampling parameters of the original monitoring equipment are adjusted for verification.

[0089] For example, when a visual image shows obvious smoke in a target area, but the gas sensor data does not show a corresponding change, the edge computing node can dispatch a mobile monitoring device equipped with a gas sensor to collect gas concentration data around the target area, while increasing the image acquisition frequency of the fixed camera to confirm the source and spread range of the smoke. When a dust sensor shows that the dust concentration in the target area has increased, but the video footage cannot identify the dust situation due to lens contamination, the edge computing node can dispatch a mobile monitoring device to the target area to obtain new visual data and mark the original fixed camera as a device that needs maintenance or is of low reliability.

[0090] In one implementation, edge computing nodes can sort and merge tasks for multiple target monitoring areas. When multiple target monitoring areas are adjacent and require the same or similar data types for supplementation, they are merged into a single continuous inspection task to reduce the round-trip path of mobile monitoring equipment. When multiple target monitoring areas have different risk types but are located near the same inspection path, multi-point inspection tasks are generated according to the priority of the target areas and the order of the paths. When multiple target monitoring areas simultaneously require fixed monitoring equipment to adjust the sampling frequency, the order of sampling resource allocation is determined based on the risk level of each area and the processing capacity of the equipment to avoid a single device undertaking too many collection tasks at the same time.

[0091] In one implementation, the proactive sensing strategy may include task triggering conditions and task stopping conditions. Task triggering conditions may include the dynamic environmental risk value of the target monitoring area reaching a preset condition, the monitoring uncertainty reaching a preset condition, the dynamic environmental risk value continuously increasing, the data quality declining, the key modality being missing, or multimodal data conflict. Task stopping conditions may include the target monitoring area obtaining sufficient supplementary data, the monitoring uncertainty decreasing to the target range, the risk status being confirmed, the mobile monitoring device's battery level falling below the safety return requirement, the construction site being in a restricted area, or the task execution time reaching a preset upper limit. By setting task triggering conditions and task stopping conditions, the proactive sensing task can avoid occupying monitoring resources for a long time.

[0092] After the proactive sensing strategy is generated, the edge computing node generates task instructions for each proactive sensing task. The task instructions include task number, target area identifier, executing device identifier, task type, data type, data collection location, data collection time, sampling frequency, data collection duration, path information, execution priority, and return data format. For fixed monitoring devices, task instructions can be used to adjust the sampling frequency, data collection direction, focal length, or data upload frequency. For mobile monitoring devices, task instructions can be used to issue inspection paths, data collection points, dwell time, data collection angle, and return paths. For field gateways, task instructions can be used to increase the receiving priority or upload priority of relevant data in the target area.

[0093] Before issuing the proactive sensing strategy, the edge computing node performs conflict checks on the task instructions. The conflict check includes determining whether the same monitoring device is assigned multiple tasks with overlapping time, whether the path of the mobile monitoring device passes through a restricted area, whether the task execution time exceeds the allowable range of the device's remaining power, whether the adjustment of the fixed monitoring device will cause other high-risk areas to lose coverage, and whether the communication bandwidth can support the data upload required for task execution. If a task conflict is detected, the edge computing node adjusts the task order, device selection, collection location, or collection duration until the proactive sensing strategy meets the execution conditions.

[0094] Through the above processing, the edge computing node generates an active sensing strategy corresponding to the target monitoring area. This active sensing strategy is not a fixed inspection plan, but is dynamically generated based on the risk status of the target monitoring area in the dynamic environmental risk field, the sources of monitoring uncertainty, the required supplementary data types, and the current status of the monitoring equipment. As a result, monitoring resources can be preferentially allocated to areas with higher risks, faster risk changes, greater potential for risk diffusion, or insufficient monitoring information, thereby improving the timeliness, accuracy, and resource utilization efficiency of construction operation environment monitoring.

[0095] S6. Execute the active sensing strategy to obtain supplementary monitoring data of the target monitoring area, update the dynamic environmental risk field according to the supplementary monitoring data, output the construction operation environment monitoring results, and use the updated dynamic environmental risk field as the basis for the next round of target monitoring area identification.

[0096] In this embodiment, after generating an active perception strategy, the edge computing node converts the active perception strategy into a corresponding execution instruction and sends it to fixed monitoring equipment, mobile monitoring equipment, field gateway, or other data acquisition equipment participating in the monitoring. The execution instruction is used to control the corresponding equipment to perform supplementary monitoring according to the specified target area, acquisition type, acquisition time, sampling frequency, acquisition direction, inspection path, or task priority, thereby obtaining supplementary monitoring data for correcting the dynamic environmental risk field.

[0097] For fixed monitoring equipment, edge computing nodes issue sampling adjustment commands based on proactive sensing strategies. These commands can include increasing sampling frequency, extending sampling duration, adjusting camera shooting direction, adjusting camera focal length, activating specific sensors, increasing data upload frequency, or switching monitoring modes. For example, when there is a dust risk in the target monitoring area and the dust data is not timely enough, the edge computing node controls the dust sensors near the target area to increase their sampling frequency within a preset time. When there is a lack of visual information in the target monitoring area, the edge computing node controls the camera to turn towards the target area and re-acquire images or videos. When the target monitoring area is located in an underground space and the gas risk is uncertain, the edge computing node controls the adjacent gas sensors to increase their data upload frequency to obtain more timely gas concentration change data.

[0098] For mobile monitoring devices, edge computing nodes issue inspection execution instructions based on proactive sensing strategies. These instructions may include the target monitoring area location, movement path, data collection points, collection sequence, dwell time, collection direction, data type, and safe return conditions. Upon receiving the inspection execution instructions, the mobile monitoring device moves to the target monitoring area or a nearby observation location according to the planned path and collects images, videos, dust concentration, harmful gas concentration, temperature and humidity, wind speed and direction, noise, vibration, or thermal imaging data at the designated data collection points. When access to the target monitoring area is not possible, the mobile monitoring device can perform indirect supplementary data collection at the target area boundary, adjacent spatial monitoring units, ventilation openings, downwind locations, or visual observation locations.

[0099] During the execution of the proactive sensing strategy, edge computing nodes or field gateways continuously receive task status information returned by the executing devices. The task status information may include task start time, current location, task progress, working status of the acquisition device, communication status, remaining battery power, acquisition completion status, and abnormal event status. If communication interruption, low battery power, path blockage, construction access restrictions, dense crowds, equipment failure, or increased security risks occur during execution, the edge computing node can pause the current task, adjust the acquisition point, replace the executing device, or re-add the task to the target monitoring task queue to ensure that the proactive sensing task can be executed under safe and feasible conditions.

[0100] After the supplementary monitoring data collection is completed, the fixed or mobile monitoring equipment will upload the supplementary monitoring data to the field gateway or edge computing node. The supplementary monitoring data includes the data newly collected after being triggered by the active sensing strategy, as well as the corresponding collection time, collection location, execution device identifier, target area identifier, collection task number, and data quality tag. After receiving the supplementary monitoring data, the edge computing node will perform format checks, integrity checks, time stamp correction, spatial location confirmation, and quality tag updates on the supplementary monitoring data to ensure that the supplementary monitoring data can be accurately mapped back to the corresponding target monitoring area and spatial monitoring unit.

[0101] In one implementation, the preprocessing method for supplementary monitoring data corresponds to the preliminary preprocessing method in step S1. For sensor-based supplementary monitoring data, the edge computing node performs unit unification, range verification, outlier identification, and short-term noise processing. For image or video-based supplementary monitoring data, the edge computing node performs sharpness detection, occlusion detection, brightness detection, image stability detection, and risk candidate region extraction. For data collected by mobile monitoring devices, the edge computing node also determines its effective observation range by combining the location, posture, movement trajectory, and acquisition direction during acquisition, thereby reducing data errors caused by mobile acquisition, communication delays, or changes in device status.

[0102] After completing the preprocessing of supplementary monitoring data, the edge computing nodes map the supplementary monitoring data to the corresponding spatial monitoring units. If the supplementary monitoring data comes directly from within the target monitoring area, it is written into the data record of the corresponding spatial monitoring unit in the target monitoring area. If the supplementary monitoring data comes from the boundary of the target area, adjacent areas, upwind location, downwind location, or visible observation location, it is associated with the target monitoring area and related spatial monitoring units according to the collection location, data type, effective range of the sensor, wind speed and direction, field of view, or spatial connectivity. Thus, the supplementary monitoring data can participate in the dynamic environmental risk value and monitoring uncertainty update of the target monitoring area.

[0103] When updating the dynamic environmental risk field, edge computing nodes first use supplementary monitoring data to correct the spatialized monitoring data of the corresponding spatial monitoring units. When supplementary monitoring data fills in the missing key modes in the target area, the key mode is added to the data record of the corresponding spatial monitoring unit, and the risk status is reassessed. When supplementary monitoring data improves the timeliness of data, the latest acquisition results are used to replace or correct the original outdated data records. When supplementary monitoring data is used to verify multimodal data conflicts, the source of the original conflict is determined based on the newly acquired data, and the credibility label of the relevant data is adjusted. When supplementary monitoring data comes from alternative monitoring equipment, the impact of the original abnormal monitoring equipment is reduced, and the role of alternative data in the current risk assessment is improved.

[0104] For updating dynamic environmental risk values, edge computing nodes recalculate the risk status of the target monitoring area and its associated spatial monitoring units based on supplementary monitoring data. If the supplementary monitoring data confirms that the target area has risks such as increased dust concentration, smoke diffusion, accumulation of harmful gases, increased noise, abnormal vibration, abnormal thermal environment, or visual recognition risks, the dynamic environmental risk value or risk level of the corresponding spatial monitoring unit is increased. If the supplementary monitoring data indicates that the original risk judgment was caused by instantaneous sensor disturbances, image misidentification, communication anomalies, or insufficient data timeliness, the dynamic environmental risk value of the corresponding spatial monitoring unit is reduced or the corresponding risk marker is removed. If the supplementary monitoring data indicates that the risk is spreading to adjacent spatial monitoring units, the risk status of the spatial monitoring units along the diffusion path is updated synchronously.

[0105] For updating monitoring uncertainty, edge computing nodes reassess the sufficiency of monitoring information in the target monitoring area based on supplementary monitoring data. When supplementary monitoring data fills in missing modes, improves data timeliness, eliminates multimodal data conflicts, verifies risk sources, or replaces abnormal monitoring equipment data, the monitoring uncertainty of the corresponding spatial monitoring unit is reduced. When supplementary monitoring data still has missing data, low quality, acquisition failure, acquisition location deviation, abnormal equipment status, or continues to conflict with the original data, the monitoring uncertainty is maintained or increased. For target areas where uncertainty cannot be effectively reduced through this round of supplementary acquisition, edge computing nodes can mark them as areas requiring continuous enhancement of monitoring and reallocate acquisition tasks in the next round of active sensing.

[0106] In one implementation, the edge computing node updates the primary risk type of the target monitoring area based on supplementary monitoring data. For example, if the original target area was identified as an area with uncertain dust risk, when the images collected by the mobile monitoring device show obvious smoke and dust and the dust sensor readings increase synchronously, the primary risk type is confirmed as dust risk. As another example, if the original target area was identified as a candidate area for gas risk, when the supplementary gas concentration data does not show any abnormalities, but the video data confirms the presence of welding fumes, the primary risk type can be adjusted from hazardous gas risk to smoke risk. By updating the risk type, the relevance of subsequent early warning information and handling suggestions can be improved.

[0107] In one implementation, edge computing nodes update the spatial extent of the target monitoring area based on supplementary monitoring data. When the supplementary monitoring data indicates that the actual impact range of the risk is greater than the original target monitoring area, adjacent affected spatial monitoring units are incorporated into the updated risk area. When the supplementary monitoring data indicates that the risk is concentrated only in a part of the original target area, the target area is narrowed. When the supplementary monitoring data indicates that the risk propagates along wind direction, passageways, stairwells, pipe corridors, openings, or ventilation paths, the relevant spatial monitoring units along the propagation path are included in the risk-associated area. Thus, the updated dynamic environmental risk field can more accurately reflect the spatial distribution of the risk.

[0108] After updating the dynamic environmental risk values ​​and monitoring uncertainties of the target monitoring area and its associated spatial monitoring units, the edge computing nodes write the updated results into the dynamic environmental risk field. For spatial monitoring units not affected by the supplementary monitoring data, the original risk status can be maintained, or regular updates can be performed based on time decay, changes in adjacent areas, and new construction status. For spatial monitoring units affected by the supplementary monitoring data, the edge computing nodes update their risk level, main risk types, monitoring uncertainties, data sources, update time, and task status. Thus, the updated dynamic environmental risk field is formed.

[0109] In one implementation, the updated dynamic environmental risk field can be simultaneously displayed on the construction site management terminal, mobile terminal, or large visualization screen; the displayed content may include a construction area plan risk map, floor risk distribution map, three-dimensional spatial risk map, target monitoring area list, risk level color coding, monitoring uncertainty coding, main risk types, risk change trends, and supplementary monitoring task status; through visualization, on-site management personnel can intuitively understand the current concentrated areas of environmental risks, risk change trends, and whether the monitoring information is sufficient.

[0110] Edge computing nodes output construction operation environment monitoring results based on the updated dynamic environmental risk field. The construction operation environment monitoring results may include the location of the risk area, risk level, main risk types, risk change status, monitoring uncertainty status, data source description, personnel exposure status, construction equipment association status, risk diffusion trend, supplementary monitoring results, and early warning information. Among them, the location of the risk area is used to indicate the spatial range where environmental risks occur or that need attention, the risk level is used to indicate the severity of the risk, the main risk types are used to indicate dust, gas, noise, vibration, thermal environment, or visual identification risks, and the monitoring uncertainty status is used to indicate whether the risk judgment has been confirmed by supplementary data.

[0111] In one implementation, when the updated dynamic environmental risk field shows that a certain spatial monitoring unit or target area has reached the warning conditions, the edge computing node generates corresponding warning information. The warning information may include the warning area, warning type, warning level, triggering reason, associated data source, risk change trend, and suggested objects of concern. For example, the warning information may indicate that the dust risk in a certain cutting operation area has increased, and the triggering reasons may include increased dust sensor readings, video identification of dust areas, wind direction pointing to adjacent work surfaces, and the presence of personnel in the area. As another example, the warning information may indicate that the gas risk uncertainty in a certain underground space is high, and the triggering reasons may include the absence of gas sensors, increased gas concentration in adjacent areas, and personnel location data showing the presence of personnel.

[0112] In one implementation, the edge computing node can also generate response prompts based on the monitoring results of the construction operation environment. The response prompts may include turning on the sprinklers, increasing ventilation, suspending related operations, adjusting personnel positions, checking monitoring equipment, adding temporary monitoring points, clearing obstructions, checking gas sensors, investigating abnormal construction equipment, or maintaining continuous monitoring. The response prompts are not a replacement for manual management measures, but rather provide on-site management personnel with auxiliary response references corresponding to the risk type based on the current dynamic environmental risk field and supplementary monitoring results.

[0113] After outputting the monitoring results of the construction operation environment, the edge computing node uses the updated dynamic environmental risk field as the basis for the identification of the target monitoring area in the next round. Specifically, the updated dynamic environmental risk value, monitoring uncertainty, main risk type, risk change status and data quality mark of each spatial monitoring unit are saved as the input data for the next time window. When entering the next monitoring cycle, the edge computing node re-executes the target monitoring area identification based on the updated dynamic environmental risk field to determine whether there are new high-risk areas, risk-increasing areas, diffusion-related areas or monitoring uncertainty areas.

[0114] In one implementation, if the monitoring uncertainty of a target monitoring area is significantly reduced after this round of supplementary monitoring, and the dynamic environmental risk value does not reach the warning condition, the edge computing node can remove the area from the target monitoring task queue and restore the normal sampling state of the corresponding monitoring equipment. If the dynamic environmental risk value of a target monitoring area is still high or continues to rise after this round of supplementary monitoring, it will remain in the target monitoring task queue for the next round and its priority will be increased. If an area is identified as a risk diffusion path after this round of supplementary monitoring, the candidate range of relevant areas will be expanded in the next round of target monitoring area identification.

[0115] In one implementation, the edge computing node also records the execution effect of the proactive sensing strategy. The execution effect may include whether the supplementary monitoring data was successfully acquired, whether the monitoring uncertainty of the target area was reduced, whether the dynamic environmental risk value was confirmed or corrected, whether the executing device completed the task on time, whether the task was interrupted, and whether the dynamic environmental risk field was effectively updated after the task was executed. The execution effect can be used to optimize the proactive sensing strategy generation process in the future, such as increasing the task allocation priority of monitoring devices with better execution effect, reducing the task allocation priority of devices that frequently fail, or adjusting the supplementary collection method corresponding to different risk types.

[0116] Through this step, the proactive sensing strategy no longer remains at the task generation stage, but is actually executed and generates supplementary monitoring data. The supplementary monitoring data further updates the dynamic environmental risk field, enabling the dynamic environmental risk field to be continuously corrected as construction status changes, environmental risks change, and monitoring data is supplemented. The updated dynamic environmental risk field continues to serve as the basis for identifying the target monitoring area in the next round, thus forming a closed-loop monitoring process of "risk field generation, target area determination, proactive sensing execution, supplementary data update, and next round of risk field driving," which improves the continuity, accuracy, and resource utilization efficiency of construction operation environmental monitoring.

[0117] Example 2 To verify the implementation effect of the present invention in a real construction operation environment, a construction site of a building construction project was selected as the test object. The construction site includes three typical risk areas: a steel bar processing shed, a basement electromechanical installation area, and a foundation pit support operation area. The main risks at the steel bar processing shed include cutting dust, noise, and intensive equipment operation; the main risks at the basement electromechanical installation area include welding fumes, accumulation of harmful gases, and insufficient ventilation; and the main risks at the foundation pit support operation area include dust diffusion, vibration disturbance, and personnel exposure. Fixed cameras, dust sensors, harmful gas sensors, temperature and humidity sensors, wind speed and direction sensors, noise sensors, vibration sensors, personnel positioning tags, construction equipment operation data acquisition modules, mobile monitoring equipment, and edge computing nodes were deployed at the test site.

[0118] During the experiment, the same construction site was divided into spatial monitoring units, and three control schemes and the present invention scheme were run simultaneously during four consecutive hours of construction. The control schemes included existing fixed threshold systems, existing fixed inspection systems, and existing multi-sensor fusion systems. The present invention scheme was executed according to the steps of claim 1, first acquiring multimodal environmental data, construction status data, and monitoring equipment status data, then completing spatiotemporal registration and spatial unit mapping, and subsequently calculating the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit to form the dynamic environmental risk field at the current moment.

[0119] Edge computing nodes identify target monitoring areas based on dynamic environmental risk fields. For areas with high risk values, continuously increasing risk, missing data, or high monitoring uncertainty, proactive sensing strategies are generated. For example, when dust concentration increases in the cutting area of ​​a rebar processing shed and the camera is obstructed by materials, the system dispatches mobile monitoring equipment to a downwind location to collect dust concentration and video data, while simultaneously increasing the sampling frequency of nearby dust sensors. When the gas sensor readings in the basement welding area are inconsistent with the video smoke identification results, the system identifies this area as a high-uncertainty target monitoring area and dispatches mobile monitoring equipment equipped with gas sensors to supplement data collection between ventilation openings and work points. When dust risk spreads along the wind direction in adjacent spatial units of the foundation pit support area, the system includes spatial units along the diffusion path in the target monitoring area and proactively adjusts the inspection points of the mobile monitoring equipment. Each scheme records indicators such as the first anomaly identification time, supplementary data collection response time, monitoring uncertainty, risk area coverage, number of false alarms, number of missed alarms, data upload volume, effective operation ratio of mobile monitoring equipment, early warning accuracy, and high-risk handling confirmation time, which are used to evaluate the technical effects of this invention compared to existing technologies.

[0120] Table 1 Experimental Data

[0121] As can be seen from the table data, this invention demonstrates significant advantages in anomaly detection speed, supplementary data collection in target areas, reduced monitoring uncertainty, improved risk coverage, and control of false alarms and missed alarms. Taking the initial anomaly identification time as an example, existing fixed threshold systems require 78.6 seconds to identify dust anomalies in processing shed A, existing fixed inspection systems require 126.4 seconds to identify gas or smoke anomalies in basement B, and existing multi-sensor fusion systems require 92.7 seconds in foundation pit C. In contrast, this invention reduces the anomaly identification time in the three areas to 32.8 seconds, 45.3 seconds, and 38.9 seconds, respectively. This indicates that the dynamic environmental risk field can capture changes in the spatial distribution of risks more quickly, rather than waiting for a single fixed point to reach a threshold before triggering an alarm.

[0122] Further examining the response time for supplementary data acquisition in the target area, existing fixed threshold systems lack the capability for proactive supplementary data acquisition. The response times for fixed inspection systems and ordinary multi-sensor fusion systems reach 182.5 seconds and 143.8 seconds, respectively. In contrast, this invention can complete the supplementary data acquisition response in the target area within approximately 40 to 60 seconds in all three areas, demonstrating the advantage of driving monitoring resource scheduling based on monitoring uncertainty.

[0123] Regarding the average monitoring uncertainty, existing solutions maintain a range of 0.38 to 0.51, while this invention reduces it to 0.19 to 0.23, indicating that supplementary monitoring data effectively improves issues such as data gaps, sampling lag, visual occlusion, and multimodal conflicts. In terms of effective coverage of risk areas, this invention exceeds 88% in all three scenarios, reaching a maximum of 93.1%, significantly higher than the 61.5% to 74.2% of existing systems. This demonstrates that the proactive sensing strategy can shift monitoring resources from low-value areas to high-risk and high-uncertainty areas.

[0124] The number of false alarms and missed alarms also decreased significantly. For example, the existing multi-sensor fusion system in pit C still had 3 missed alarms, while the present invention did not have any missed alarms in that area. This indicates that by using dynamic environmental risk values ​​and monitoring uncertainty together in the judgment, it is possible to simultaneously reduce excessive alarms and risk omissions. In terms of data upload volume, the present invention is lower than the existing solution, indicating that actively increasing the sampling of the target area is not equivalent to high-frequency acquisition across the entire field, but rather reduces invalid data uploads through area filtering and task scheduling.

[0125] The effective operation rate of mobile monitoring equipment has increased from 46.8% to 52.4% under the existing fixed inspection method to 72.9% to 79.5%, indicating that the present invention can reduce aimless inspections and increase the working ratio of equipment in high-value monitoring areas.

[0126] In summary, this invention does not simply add sensors or increase the sampling frequency, but rather identifies risk states and information gaps through a dynamic environmental risk field, then obtains supplementary monitoring data through an active sensing strategy, and finally updates the dynamic environmental risk field in reverse, forming a continuously iterative closed-loop monitoring mechanism.

[0127] This mechanism addresses the shortcomings of existing technologies, such as blind spots caused by fixed deployment, slow response of fixed inspections, lack of proactive data collection due to multi-sensor fusion, and difficulty in verifying abnormal data. It can improve the timeliness, accuracy, coverage, and resource utilization efficiency of construction operation environment monitoring, demonstrating strong creativity and practical application value.

[0128] Example 3 A construction operation environment active sensing and monitoring system based on a dynamic environmental risk field, characterized in that it includes: The data acquisition module is used to acquire multimodal environmental data, construction status data, and monitoring equipment status data at the construction site. The spatiotemporal registration module is used to perform spatiotemporal registration on the multimodal environmental data, construction status data and monitoring equipment status data, and map them to pre-divided spatial monitoring units to obtain spatialized monitoring data of each spatial monitoring unit. The risk field generation module is used to calculate the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit based on the spatialized monitoring data, and form the dynamic environmental risk field at the current moment. The target area determination module is used to determine the target monitoring area that needs enhanced monitoring based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field. The active sensing strategy generation module is used to generate an active sensing strategy based on the target monitoring area and the status data of the monitoring equipment, with the goal of reducing the monitoring uncertainty of the target monitoring area. The risk field update and result output module is used to execute the active perception strategy to obtain supplementary monitoring data of the target monitoring area, update the dynamic environmental risk field according to the supplementary monitoring data, output the construction operation environment monitoring results, and use the updated dynamic environmental risk field as the basis for the next round of target monitoring area identification.

[0129] 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, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for active sensing and monitoring of construction operation environment based on dynamic environmental risk field, characterized in that, Includes the following steps: Acquire multimodal environmental data, construction status data, and monitoring equipment status data at the construction site; The multimodal environmental data, construction status data, and monitoring equipment status data are spatiotemporally registered and mapped to pre-divided spatial monitoring units to obtain spatialized monitoring data for each spatial monitoring unit. The dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit are calculated based on the spatialized monitoring data to form the dynamic environmental risk field at the current moment. Based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field, the target monitoring area that needs enhanced monitoring is determined. With the goal of reducing the monitoring uncertainty of the target monitoring area, an active sensing strategy is generated based on the target monitoring area and the status data of the monitoring equipment. The active sensing strategy is executed to obtain supplementary monitoring data of the target monitoring area, the dynamic environmental risk field is updated based on the supplementary monitoring data, the construction operation environment monitoring results are output, and the updated dynamic environmental risk field is used as the basis for the next round of target monitoring area identification.

2. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 1, characterized in that, The multimodal environmental data includes multiple types of data such as image data, video data, dust concentration data, harmful gas concentration data, temperature and humidity data, wind speed and direction data, noise data, vibration data, illuminance data, and thermal imaging data. The construction status data includes multiple types of information such as construction procedure information, construction area information, construction progress information, personnel location information, construction equipment location information, construction equipment operation status information, material stacking information, and spatial structure information. The monitoring equipment status data includes multiple items from the following: monitoring equipment identifier, equipment type, installation location, current orientation, visible range, sampling frequency, communication status, remaining battery power, storage status, sensor working status, equipment fault status, mobility, reachable area, and current task status. The multimodal environmental data, construction status data, and monitoring equipment status data are preprocessed to obtain monitoring data to be processed containing data quality tags. The monitoring data to be processed is then used as input for subsequent spatiotemporal registration. The preprocessing includes at least one of the following: data format unification, data integrity check, anomaly marking, dimensional unification, range verification, and quality tag generation.

3. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 1, characterized in that, The multimodal environmental data, construction status data, and monitoring equipment status data are spatiotemporally registered and mapped to pre-divided spatial monitoring units to obtain spatialized monitoring data for each spatial monitoring unit, including: Using a unified clock as the time reference, data from different sampling frequencies are aggregated into the same time window to obtain the monitoring data within the corresponding time window; Establish a unified spatial coordinate benchmark for the construction site, and divide the construction site into multiple spatial monitoring units based on the construction area, work surface, floor, room, foundation pit zone, tunnel mileage section, construction road, construction equipment operation area, material storage area, opening edge area, ventilation area, or regular grid. Based on the data acquisition location, data coverage area, acquisition device posture, movement trajectory, and corresponding construction area, the time-registered data is mapped to the corresponding spatial monitoring unit to form spatialized monitoring data for each spatial monitoring unit.

4. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 1, characterized in that, The dynamic environmental risk value of each spatial monitoring unit is calculated based on spatial monitoring data, including: For each spatial monitoring unit, environmental risk characteristics are extracted from the spatial monitoring data. These environmental risk characteristics include multiple features such as dust risk characteristics, harmful gas risk characteristics, noise risk characteristics, vibration risk characteristics, thermal environment risk characteristics, and visual recognition risk characteristics. The construction sequence corresponding to the spatial monitoring unit is determined based on the construction status data, and the impact of different environmental risk characteristics on the dynamic environmental risk value is adjusted according to the construction sequence. The dynamic environmental risk value is corrected based on the personnel distribution within the spatial monitoring unit, the operating status of construction equipment, spatial structure information, risk changes in adjacent spatial monitoring units, and historical risk change records.

5. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 2, characterized in that, The monitoring uncertainty of each space monitoring unit is calculated based on the spatialized monitoring data, including: For each spatial monitoring unit, obtain its corresponding data coverage, data timeliness, data quality markers, multimodal data consistency, and monitoring equipment status; When a space monitoring unit has missing data modes, data acquisition time exceeds the preset time range, abnormal data quality marking, conflicting multimodal data, or abnormal monitoring equipment status, the monitoring uncertainty of the space monitoring unit shall be increased. When a space monitoring unit has multiple valid modal data, and the data acquisition time meets the preset timeliness requirements, the data quality label is normal, the multimodal data are consistent with each other, and the monitoring equipment is in normal condition, the monitoring uncertainty of the space monitoring unit is reduced.

6. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 1, characterized in that, Based on the dynamic environmental risk values ​​and monitoring uncertainties of each spatial monitoring unit in the dynamic environmental risk field, the target monitoring areas requiring enhanced monitoring are determined, including: The spatial monitoring units whose dynamic environmental risk values ​​reach the early warning conditions, whose dynamic environmental risk values ​​continuously increase within a continuous time window, whose monitoring uncertainty reaches the preset conditions, and whose spatial monitoring units are on the risk diffusion path determined based on the risk change status and spatial location relationship in the dynamic environmental risk field are identified as candidate spatial monitoring units. The priority of candidate spatial monitoring units is determined based on their dynamic environmental risk values, monitoring uncertainties, risk change status, personnel distribution, construction procedures, and spatial location relationships. Candidate spatial monitoring units that are spatially adjacent and have the same or similar main risk types are merged to form target monitoring areas, and target area description information is generated. The target area description information includes target area identifier, spatial location, main risk type, monitoring uncertainty, target area priority, and monitoring data types that need to be supplemented.

7. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 6, characterized in that, With the goal of reducing the monitoring uncertainty of the target monitoring area, an active sensing strategy is generated based on the target monitoring area and the status data of the monitoring equipment, including: Candidate monitoring devices are determined from the monitoring device status data based on the main risk types and the types of monitoring data that need to be supplemented in the target monitoring area. Based on the equipment type, data acquisition capabilities, installation location, visibility range, mobility, and reachable area of ​​the candidate monitoring equipment, capability matching is performed on the candidate monitoring equipment. Active sensing tasks are generated based on the capability matching results. These active sensing tasks include at least one of the following: inspection tasks of mobile monitoring equipment, sampling and adjustment tasks of fixed monitoring equipment, and combined supplementary data collection tasks executed collaboratively by multiple monitoring equipment.

8. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 7, characterized in that, When generating proactive sensing tasks, the tasks are optimized based on monitoring device status data, task execution costs, and security constraints. Specifically, this includes: Based on the communication status, remaining power, sensor operating status, current task status, and reachable area of ​​the monitoring equipment, the availability of candidate monitoring equipment is screened. The execution sequence of proactive sensing tasks is determined based on the dynamic environmental risk value, monitoring uncertainty, and target area priority of the target monitoring area. Based on the travel distance, estimated time, equipment power consumption, communication load, construction restricted areas, construction equipment operation danger zones and mobile monitoring equipment accessibility conditions, the active sensing task is constrained and conflict checked. When a conflict is detected in the proactive detection task, adjustments are made to the execution device, data collection location, data collection duration, or task sequence.

9. The active sensing and monitoring method for construction operation environment based on dynamic environmental risk field according to claim 1, characterized in that, Execute proactive sensing strategies to acquire supplementary monitoring data for the target monitoring area, and update the dynamic environmental risk field based on the supplementary monitoring data, including: The supplementary monitoring data is mapped to the corresponding target monitoring area and its associated spatial monitoring unit, and the spatialized monitoring data of the corresponding spatial monitoring unit is corrected using the supplementary monitoring data; Based on the revised spatialized monitoring data, the dynamic environmental risk value and monitoring uncertainty of the target monitoring area and its associated spatial monitoring units are recalculated. Based on the recalculated dynamic environmental risk value and monitoring uncertainty, the main risk types, spatial range and risk change status of the target monitoring area are updated to form an updated dynamic environmental risk field. The updated dynamic environmental risk field outputs the construction operation environment monitoring results, which include various aspects such as the location of the risk area, risk level, main risk types, monitoring uncertainty status, data source description, risk diffusion trend, supplementary monitoring results, and early warning information.

10. A construction operation environment active sensing and monitoring system based on a dynamic environmental risk field as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to acquire multimodal environmental data, construction status data, and monitoring equipment status data at the construction site. The spatiotemporal registration module is used to perform spatiotemporal registration on the multimodal environmental data, construction status data and monitoring equipment status data, and map them to pre-divided spatial monitoring units to obtain spatialized monitoring data of each spatial monitoring unit. The risk field generation module is used to calculate the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit based on the spatialized monitoring data, and form the dynamic environmental risk field at the current moment. The target area determination module is used to determine the target monitoring area that needs enhanced monitoring based on the dynamic environmental risk value and monitoring uncertainty of each spatial monitoring unit in the dynamic environmental risk field. The active sensing strategy generation module is used to generate an active sensing strategy based on the target monitoring area and the status data of the monitoring equipment, with the goal of reducing the monitoring uncertainty of the target monitoring area. The risk field update and result output module is used to execute the active perception strategy to obtain supplementary monitoring data of the target monitoring area, update the dynamic environmental risk field according to the supplementary monitoring data, output the construction operation environment monitoring results, and use the updated dynamic environmental risk field as the basis for the next round of target monitoring area identification.

Images