Construction dynamic monitoring system and method based on oblique photography and multi-source data

Abstract

The invention provides a construction dynamic monitoring system based on oblique photography and multi-source data, and the system comprises an unmanned plane collection module which is used for obtaining a construction site oblique image according to a preset rule and generating a three-dimensional real scene model; the sensor network module is deployed on the construction machinery and is used for acquiring the running state and position data of the machinery in real time; the AI analysis module is used for aligning the real scene model and the BIM plan model based on a convolutional neural network, detecting a difference region and calculating an engineering quantity; and the data fusion platform is used for integrating the meteorological data, the historical efficiency library and the construction log and generating a dynamic progress report and optimization suggestions. According to the method, multi-dimensional data are fused, and the accuracy of progress analysis and the decision support capability are improved; the whole process from data collection to report generation is automatic, the labor cost is greatly reduced, and the method has high practical significance.

Description

Technical Field

[0001] The present invention relates to the technical field of engineering construction progress management, and in particular to a construction dynamic monitoring system and method based on oblique photography and multi-source data. Background Art

[0002] Before calculating the earthwork volume, it is necessary to collect various data at the calculation site. Collecting earthwork volume-related information through UAV oblique photography is also one of the various data collection methods before earthwork volume calculation. In the existing earthwork volume calculation methods, data collection is easily affected by factors such as the environment, resulting in inaccurate collected data, which makes the final earthwork volume calculation result deviate greatly, bringing a certain impact on the use of the earthwork volume calculation method.

[0003] 1. Comparative document CN110287536A proposes a method for calculating the construction progress index of high slopes based on sensor technology and oblique photography technology. Its protected rights are: "Before the construction of the high slope, use UAV oblique photography technology to accurately measure the terrain data, use real-scene modeling technology to construct a 3D real-scene model of the terrain, input the 3D real-scene model into the BIM design software civil3D, load the construction drawing design data of the high slope, and calculate the total excavation and support engineering quantities of the high slope; during the slope construction process, install a three-axis acceleration sensor on the construction equipment, and use the mobile phone terminal to collect the construction machinery operation data sent back by the three-axis acceleration sensor; after the construction is fully launched, count the climate data such as sunny and rainy days of the current month from the construction log; comprehensively calculate the construction progress index of a single machine shift under different climate conditions by integrating climate data, engineering quantity data, and mechanical operation data." However, the data collection cycle of this solution is long, and it is impossible to achieve dynamic tracking and real-time feedback of the construction progress. It does not integrate multi-dimensional data such as mechanical operation, climate, and sensors, resulting in one-sided progress analysis.

[0004] 2. The comparative document CN115526450A proposes a construction progress monitoring method, system and medium based on the combination of oblique photography and BIM. Its protected rights are as follows: "Construct a three-dimensional real-scene model using the drone oblique photography technology; construct a BIM three-dimensional model and a BIM four-dimensional model using the BIM technology; match the three-dimensional real-scene model with the BIM three-dimensional model through the Boolean algorithm. On the basis of the match, cut the BIM three-dimensional model with the three-dimensional real-scene model, calculate and analyze the cutting results to obtain a construction progress report; incorporate the construction progress report into the BIM four-dimensional model to achieve construction progress monitoring. The present invention enables relevant personnel to more intuitively and visually understand the progress deviation during the construction progress monitoring process, and realizes the automated and intelligent monitoring management of the construction progress; it can not only solve the problem that relying solely on the BIM three-dimensional model is prone to deviate from the actual on-site situation, but also avoid the subjective errors caused by manually matching the three-dimensional real-scene model and the BIM three-dimensional model." However, there are deviations in the automatic matching of the three-dimensional real-scene model and the BIM model in a complex environment, and manual intervention is still required in links such as model difference detection and engineering quantity statistics, relying on manual correction, which is inefficient and error-prone.

[0005] Therefore, a construction dynamic monitoring system and method based on oblique photography and multi-source data are proposed to solve the above problems. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above deficiencies and provide a construction dynamic monitoring system and method based on oblique photography and multi-source data to solve the problems raised in the background technology.

[0007] The present invention proposes a construction dynamic monitoring system based on oblique photography and multi-source data for the above problems, including: a drone acquisition module for obtaining oblique images of the construction site according to a preset rule and generating a three-dimensional real-scene model; a sensor network module deployed on construction machinery to collect mechanical operation status and position data in real time; an AI analysis module for aligning the real-scene model and the BIM planned model based on a convolutional neural network, detecting the difference area and calculating the engineering quantity; a data fusion platform for integrating meteorological data, historical efficiency library and construction logs to generate a dynamic progress report and optimization suggestions.

[0008] Preferably, the drone acquisition module is equipped with a lidar and a multispectral camera to synchronously obtain terrain point cloud and vegetation coverage data.

[0009] Preferably, the sensor network module transmits data through the LoRa wireless protocol, supporting collaborative computing between edge nodes and cloud servers.

[0010] Preferably, the AI analysis module is built-in with an adaptive threshold segmentation algorithm to automatically distinguish the construction completed area, uncompleted area and abnormal area.

[0011] Preferably, the data fusion platform provides an AR visualization interface to support construction workers to view real-time model differences and warning information through intelligent terminals.

[0012] In addition, the present invention also discloses a monitoring method for a construction dynamic monitoring system based on oblique photography and multi-source data, including the following steps: Step 1: Collect on-site construction image data through the drone oblique photography technology according to a preset cycle or event trigger mode, and generate an incrementally updated three-dimensional real scene model in combination with edge computing; Step 2: Real-time obtain the sensor data of construction machinery, including three-axis acceleration, GPS positioning and operation duration, and associate the machinery status with the construction area; Step 3: Input the three-dimensional real scene model and the BIM plan model into the AI matching module, extract feature points through deep learning algorithms and calculate the model differences, and output the deviation area and the change in the quantity of work; Step 4: Integrate climate data, the historical library of machinery efficiency and construction logs, dynamically correct the progress indicators, and generate a visual progress report and warning information.

[0013] Preferably, the event trigger mode in Step 1 includes: the completion signal of construction key nodes, the alarm of abnormal machinery status or the sudden change of meteorological conditions.

[0014] Preferably, the deep learning algorithm in Step 3 adopts an improved U-Net network to improve the model matching accuracy through multi-scale feature fusion. The loss function of the improved U-Net network is defined as: ; Where: is the cross-entropy loss, and its expression is: ; is the reverse structural similarity loss, and its expression is: ; Where: is the mean of the feature map, is the variance, is the covariance, is the stability constant; is the L2 regularization term, and its expression is: ; The dynamic weight coefficient satisfies: ; ; ; Among them, is the current training round, is the total number of rounds, is the initial regularization strength, is the anti-zero constant.

[0015] Preferably, when dynamically correcting the progress index in step four, a Markov chain model is introduced to predict the remaining construction period, and sensitivity analysis is carried out in combination with the fluctuation of mechanical efficiency. Preferably, the three-dimensional real-scene model adopts voxelization compression technology to reduce the data volume by more than 50% while retaining key geometric features. The present invention has the following beneficial effects: 1. The present invention can perform real-time dynamic monitoring: update the construction progress in real time, and reduce the plan deviation caused by information lag; 2. In the present invention, the AI algorithm is used to greatly reduce the model matching error, reduce manual intervention, and achieve high-precision matching; 3. The present invention conducts comprehensive data integration, integrates multi-dimensional data, and improves the accuracy of progress analysis and decision-making support capabilities; 4. The present invention conducts automated control, and the whole process from data collection to report generation is automated, greatly reducing the labor cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is the system control logic diagram of the present invention; Figure 2 is the schematic diagram of real-time data interaction of the present invention; Figure 3 is the construction flow chart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0017] The present invention will be further described below with reference to the drawings and embodiments: Refer to Figures 1 to 3 , a construction dynamic monitoring system and method based on oblique photography and multi-source data to solve the problems proposed in the background technology.

[0018] The present invention proposes a construction dynamic monitoring system based on oblique photography and multi-source data for the above problems, including: a drone acquisition module for obtaining oblique images of the construction site according to a preset rule and generating a three-dimensional real scene model; a sensor network module deployed on construction machinery to collect mechanical operation status and position data in real time; an AI analysis module for aligning the real scene model and the BIM plan model based on a convolutional neural network, detecting the difference area and calculating the engineering quantity; a data fusion platform for integrating meteorological data, historical efficiency library and construction logs to generate a dynamic progress report and optimization suggestions. In this embodiment, the drone can be a DJI M300 RTK drone equipped with a Zenmuse L1 lidar, whose scanning frequency can reach 240 kHz, generating a real scene model with centimeter-level accuracy; the sensor network module configuration can select the built-in three-axis accelerometer and dual-frequency GPS of the CAT® intelligent construction machinery, with an accuracy of ±2 cm; the AI analysis module can be based on the 3D ResNet-50 network of the PyTorch framework, with an input size of 256×256×32 voxels, and output a difference heat map; the data fusion platform can adopt an Apache Kafka real-time data pipeline to integrate meteorological data from the Weather.com API; this technical solution can achieve a timeliness leap of construction progress monitoring from "daily level" to "hourly level", and the cross-verification of multi-source data makes the engineering quantity calculation error <3%, compared with about 8-12% of the traditional method; it greatly reduces the overall system cost.

[0019] Preferably, the drone acquisition module is equipped with a lidar and a multispectral camera to synchronously obtain terrain point cloud and vegetation coverage data. In the embodiment, Parrot Sequoia+ green / red / red edge / near-infrared bands are used, with a resolution of 1280×960, and data fusion: through the NDVI index, the vegetation coverage area is identified, and the area to be cleared is automatically marked. Its expression is: ; In this embodiment, terrain and vegetation data can be obtained synchronously to avoid secondary aerial photography; the vegetation recognition accuracy rate >90%, reducing the calculation error of earthwork volume.

[0020] Preferably, the sensor network module transmits data through the LoRa wireless protocol, supporting collaborative computing between edge nodes and cloud servers. In the embodiment, the LoRa configuration is: Frequency band: CN470 MHz; Transmission power: 20 dBm; Transmission distance: line of sight 3 km; Edge computing: The NVIDIA Jetson Xavier NX node processes 50% of the sensor data; compared with 4G transmission, the power consumption is reduced by 85%, and the edge filters invalid data, greatly reducing the cloud bandwidth requirement.

[0021] Preferably, the AI analysis module is built with an adaptive threshold segmentation algorithm to automatically distinguish the construction completed area, uncompleted area, and abnormal area. In this embodiment, the calculation expression of the adaptive threshold is as follows: ; where: μ is the local mean, σ is the standard deviation, and k = 0.5 is an empirical coefficient.

[0022] Preferably, the data fusion platform provides an AR visualization interface, supporting construction workers to view real-time model differences and warning information through intelligent terminals. In this embodiment, the AR device can adopt Microsoft HoloLens 2. Through this solution, the positioning efficiency of construction workers is greatly improved, and the design change communication time becomes real-time.

[0023] In addition, the present invention also discloses a monitoring method for a construction dynamic monitoring system based on oblique photography and multi-source data, including the following steps: Step 1: Collect on-site construction image data through the drone oblique photography technology according to a preset cycle or event trigger mode, and generate an incrementally updated three-dimensional real scene model in combination with edge computing; Step 2: Real-time obtain the sensor data of construction machinery, including three-axis acceleration, GPS positioning, and operation duration, and associate the mechanical state with the construction area; Step 3: Input the three-dimensional real scene model and the BIM plan model into the AI matching module, extract feature points through the deep learning algorithm and calculate the model difference, and output the deviation area and the change in the quantity of work; Step 4: Integrate climate data, the historical library of mechanical efficiency, and construction logs, dynamically correct the progress indicators, and generate a visual progress report and warning information.

[0024] Preferably, the event trigger mode in Step 1 includes: the completion signal of construction key nodes, the alarm of abnormal mechanical states, or the sudden change of meteorological conditions.

[0025] Preferably, the deep learning algorithm in Step 3 adopts an improved U-Net network to improve the model matching accuracy through multi-scale feature fusion. The loss function of the improved U-Net network is defined as: ; where: is the cross-entropy loss, and its expression is: ; is the structural similarity reverse loss, and its expression is: ; where: is the mean of the feature map, is the variance, is the covariance, is the stability constant; is the L2 regularization term, and its expression is: ; The dynamic weight coefficient satisfies: ; ; ; wherein, is the current training round, is the total number of rounds, is the initial regularization strength, is the anti-zero constant.

[0026] Preferably, when dynamically correcting the progress index in step four, a Markov chain model is introduced to predict the remaining construction period, and sensitivity analysis is carried out in combination with the fluctuation of mechanical efficiency. In this embodiment, the Markov model includes a state space of ahead, normal, and lag, and a transition matrix; the sensitivity analysis can specifically perform 1000 Monte Carlo simulations for data statistics.

[0027] Preferably, the three-dimensional real scene model adopts voxelization compression technology to reduce the data volume by more than 50% while retaining key geometric features. In this embodiment, octree coding is adopted: the initial resolution is 1 cm, and empty voxels are skipped for storage; the retention of geometric features follows the following principles: surfaces with a curvature > 0.05 are not downsampled, and RGB values are retained for edge voxels.

[0028] The specific implementation manners of the above embodiments are only the preferred technical solutions of the present invention and should not be regarded as limitations on the present invention. The protection scope of the present invention should be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims, that is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A construction dynamic monitoring system based on oblique photography and multi-source data, characterized in that: include: The drone acquisition module is used to obtain oblique images of the construction site according to preset rules and generate a three-dimensional real-life model. The sensor network module is deployed on construction machinery to collect the machine's operating status and location data in real time. The AI ​​analysis module aligns the real-life model with the BIM planning model based on a convolutional neural network, detects different areas and calculates the project volume. The data fusion platform integrates meteorological data, historical efficiency libraries and construction logs to generate dynamic progress reports and optimization suggestions.

2. A construction dynamic monitoring system based on oblique photography and multi-source data according to claim 1, characterized in that: The drone acquisition module is equipped with a lidar and a multispectral camera to simultaneously acquire terrain point cloud and vegetation coverage data.

3. The construction dynamic monitoring system based on oblique photography and multi-source data according to claim 1 is characterized in that: The sensor network module transmits data via the LoRa wireless protocol and supports collaborative computing between edge nodes and cloud servers.

4. The construction dynamic monitoring system based on oblique photography and multi-source data according to claim 1 is characterized in that: The AI ​​analysis module has a built-in adaptive threshold segmentation algorithm to automatically distinguish between completed construction areas, unfinished areas and abnormal areas.

5. The construction dynamic monitoring system based on oblique photography and multi-source data according to claim 1, characterized in that: The data fusion platform provides an AR visualization interface, supporting construction personnel to view real-time model differences and early warning information through smart terminals.

6. A monitoring method for a construction dynamic monitoring system based on oblique photography and multi-source data according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Use drone oblique photography technology to collect construction site image data according to a preset cycle or event trigger mode, and combine edge computing to generate an incrementally updated 3D real scene model; Step 2: Obtain sensor data of construction machinery in real time, including three-axis acceleration, GPS positioning and running time, and associate the machinery status with the construction area; Step 3: Input the 3D real-life model and the BIM plan model into the AI ​​matching module, extract feature points and calculate model differences through a deep learning algorithm, and output the deviation area and engineering quantity changes; Step 4: Integrate climate data, mechanical efficiency history database and construction logs, dynamically correct progress indicators, and generate visual progress reports and early warning information.

7. The monitoring method of a construction dynamic monitoring system based on oblique photography and multi-source data according to claim 6 is characterized in that: The event triggering modes in step one include: completion signal of key construction nodes, alarm of abnormal mechanical status or sudden change of meteorological conditions.

8. The monitoring method of a construction dynamic monitoring system based on oblique photography and multi-source data according to claim 6, characterized in that: In step 3, the deep learning algorithm uses an improved U-Net network to improve the model matching accuracy by fusion of multi-scale features. The loss function of the improved U-Net network is defined as: ； in: is the cross entropy loss, which is expressed as: ； is the structural similarity inverse loss, and its expression is: ； in: is the feature map mean, is the variance, is the covariance, is the stability constant; is the L2 regularization term, and its expression is: ； The dynamic weight coefficient satisfy: ； ； ； in, is the current training round, For the total rounds, is the initial regularization strength, To prevent zero constant.

9. The monitoring method of a construction dynamic monitoring system based on oblique photography and multi-source data according to claim 6, characterized in that: When dynamically correcting the progress indicators in step 4, the Markov chain model is introduced to predict the remaining construction period, and a sensitivity analysis is performed in combination with the fluctuation of mechanical efficiency.

10. The monitoring method of a construction dynamic monitoring system based on oblique photography and multi-source data according to claim 6, characterized in that: The 3D reality model uses voxel compression technology to reduce the data volume by more than 50% while retaining key geometric features.

Images