Construction scheme optimization method and system for carbon emission assessment based on deep learning

Abstract

The invention provides a construction scheme optimization method and system for carbon emission assessment based on deep learning, and belongs to the technical field of green construction, and the method comprises the steps: S1, obtaining the image data of a construction site through an unmanned aerial vehicle, carrying out the time sequence consistency analysis, illumination normalization processing and mixed denoising, extracting image features, and carrying out the registration, generating a live-action three-dimensional model by adopting deep learning; s2, performing segmentation based on point cloud data of the live-action three-dimensional model, extracting different construction areas, comparing the models before and after reconstruction to obtain an earth excavation volume and a vegetation damage volume, and checking carbon emission in a construction process by combining multi-dimensional carbon emission factor data and adaptive dynamic factors; and S3, performing comprehensive evaluation and adjustment on equipment configuration, an energy structure and a construction process by adopting a multi-objective optimization method, and generating a low-carbon construction scheme. The advanced three-dimensional modeling technology and the intelligent analysis means are utilized, and carbon emission management of the construction site and optimization of the green construction scheme are achieved.

Description

Technical Field

[0001] The present invention relates to the fields of green construction technology and deep learning technology, and in particular to a construction scheme optimization method and system based on deep learning-based carbon emission assessment. Background Art

[0002] With rapid economic development and the continuous advancement of urbanization, the construction industry faces unprecedented challenges. Especially during the construction process, ensuring project quality, progress, and environmental impact has become a critical issue that needs to be addressed. Traditional construction management methods often rely on manual inspections and measurement tools to monitor construction site progress and environmental conditions. This method is not only inefficient but also susceptible to human factors and lacks real-time, accurate data support. At the same time, carbon emissions in construction are increasingly becoming a focus of environmental protection and sustainable development. Real-time assessment and reduction of carbon emissions has become a key component of green construction.

[0003] In recent years, 3D reconstruction technology, especially drone-based 3D model reconstruction, has been widely used in the construction field. Point cloud data acquired by drones can accurately reproduce the geometric form of the construction site, providing high-precision data support for construction progress monitoring and quality assessment. However, there are some technical challenges in exploring the potential and applying point cloud data. Among them, the processing and analysis of point cloud data often involves the collection, simplification, registration, and analysis of large amounts of data, which may introduce inevitable errors. Especially when processing large-scale point cloud data, traditional data processing methods and conventional computing tools may not meet the requirements of high precision and high efficiency. These errors will affect the accuracy of 3D reconstruction and carbon emission calculations. Therefore, how to process and utilize these point cloud data to ensure their quality and reliability has become a technical problem that needs to be solved urgently. Summary of the Invention

[0004] To address the challenges of inadequate carbon emission control, inaccurate data management, and low green implementation in existing construction site management, this paper provides a construction plan optimization method and system based on deep learning-based carbon emission assessment. Leveraging advanced 3D modeling technology and intelligent analysis, this paper achieves carbon emission management and green construction plan optimization at construction sites.

[0005] To achieve the above object, the present invention adopts the following technical solutions: The construction plan optimization method based on deep learning for carbon emission assessment includes: S1. Acquire construction site image data using drones, perform temporal consistency analysis, illumination normalization, and hybrid denoising on the image data, extract image features, perform image registration, and generate a 3D model of the real scene using deep learning. S2. Segment the point cloud data of the real-world 3D model to extract different construction areas. Compare the models before and after reconstruction to determine the amount of earthwork excavation and vegetation destruction. Combined with multi-dimensional carbon emission factor data and adaptive dynamic factors, calculate the carbon emissions during the construction process. S3. Use a multi-objective optimization approach to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

[0006] In this specification, S1 includes: S11. Perform temporal consistency analysis and illumination normalization on the construction site image data, and remove noise using a hybrid denoising strategy that combines median filtering and edge-preserving bilateral filtering. S12. Image matching is performed using a dynamic matching threshold function based on local image density. Feature points are extracted using a feature extraction algorithm guided by local geometric consistency. Registration and global stitching are performed using a partitioned parallel incremental reconstruction method. S13. In the dense point cloud reconstruction stage, a depth-guided multi-view stereo method is used to reconstruct the hole areas of the point cloud, and a lightweight 3D model is generated through a model compression framework.

[0007] In this manual, the point cloud data segmentation steps in S2 include: S21. Use an airborne visual camera to acquire multi-view 3D images and video data to construct an original spatial scene dataset; S22. Perform local neighborhood voxelization on the original point cloud of the original spatial scene dataset, and perform feature encoding through the coordinate attention mechanism and high-order semantic fusion module to obtain local features at each scale; S23. Use the multi-scale feature aggregation module to fuse local features at each scale and send them to the semantic classification layer to predict regional labels for functional area division, including the main structure area, excavation area, processing area, and living area.

[0008] In this manual, the calculation steps for earthwork excavation and vegetation destruction in S2 include: S24. Obtain a high-density point cloud model before and after construction, and perform model registration using multi-scale rigid registration and Gaussian error field weight adjustment strategy; S25. Perform grid projection difference analysis on the registered model, calculate grid cell height difference, and obtain earth excavation volume; S26. Extract the area and type of vegetation damage based on the before and after comparison of the 3D model.

[0009] In this manual, the carbon emissions accounting steps in S2 include: S27. Based on regional geometric characteristics and emission factor method, an adaptive dynamic factor is introduced to calculate the carbon emissions of excavation projects; S28. Calculate carbon emissions from vegetation destruction by combining the area of vegetation destruction, biomass per unit area, carbon content, and respiration rate; S29. Summarize the carbon emissions from excavation and vegetation destruction to arrive at the total carbon emissions.

[0010] In this manual, the steps for optimizing device configuration in S3 include: S31. With the goal of minimizing carbon emissions, determine the optimal number of electric excavators, loaders, and dump trucks to meet the total workload requirements without exceeding the maximum number of equipment; S32. Optimize the ratio of electric equipment to diesel equipment through genetic algorithms to reduce the use of diesel equipment.

[0011] In this specification, the energy structure optimization steps in S3 include: S33. Design the photovoltaic system coverage area and energy storage capacity, calculate the daily power generation of the photovoltaic system, and achieve low-carbon substitution for construction electricity demand; S34. Determine the optimal photovoltaic coverage and energy storage configuration based on sunshine hours, photovoltaic efficiency, and construction electricity demand.

[0012] In this manual, the construction process optimization steps in S3 include: S35. Combine the 3D model to predict excavation progress and dynamically adjust equipment scheduling strategies to reduce equipment no-load energy consumption; S36. Optimize transportation route planning to reduce carbon emissions during material transportation.

[0013] In this specification, S1 also includes: S14. When preprocessing the data collected by the drone, use high-pass filters and median filters to suppress noise, and use edge detection and contour extraction to retain important features; S15. Use principal component analysis to achieve data dimensionality reduction and improve the efficiency of feature extraction and registration.

[0014] A construction scheme optimization system for carbon emission assessment based on deep learning, applying any of the above-mentioned construction scheme optimization methods for carbon emission assessment based on deep learning, and a construction scheme optimization system for carbon emission assessment based on deep learning, comprising: The 3D model reconstruction module is used to acquire construction site image data through drones, perform time-series consistency analysis, illumination normalization, and hybrid denoising on the construction site image data, extract image features and perform registration, and use deep learning to generate a real-scene 3D model; The carbon emission calculation module is used to segment the point cloud data of the real-life 3D model, extract different construction areas, compare the models before and after reconstruction to determine the amount of earth excavation and vegetation destruction, and combine multi-dimensional carbon emission factor data and adaptive dynamic factors to calculate the carbon emissions during the construction process; The construction plan optimization module uses a multi-objective optimization method to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

[0015] In summary, the present invention has at least the following beneficial effects: This invention utilizes deep learning-based large-scale scene reconstruction technology to efficiently process and analyze point cloud data, generating accurate, large-field-of-view 3D models and model details, which can then be used for further analysis and evaluation. By optimizing point cloud segmentation and region recognition, this technology can precisely delineate different areas of a construction site (such as the main structure, vegetation areas, and excavation areas), providing accurate baseline data for subsequent carbon emission assessment and construction progress monitoring. Furthermore, deep learning-based large-scale scene reconstruction technology offers higher data processing accuracy and efficiency, significantly improving computational efficiency while ensuring data quality, thus overcoming the limitations of traditional methods in processing large-scale construction site data. Consequently, a deep learning-based large-scale scene 3D reconstruction and carbon reduction assessment method, as well as a green construction plan optimization method and carbon emission assessment method, has emerged. By integrating 3D reconstruction, point cloud segmentation, and carbon emission calculation and assessment technologies, this method can significantly improve the level of green construction and reduce carbon emissions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 This is a system flow diagram of the construction plan optimization method for carbon emission assessment based on deep learning involved in the present invention.

[0018] Figure 2 Schematic diagram of the principle of the three-dimensional reconstruction method involved in the present invention.

[0019] Figure 3 Schematic diagram of data collection for the three-dimensional reconstruction model involved in the present invention.

[0020] Figure 4 It is a schematic diagram of the process of implementing carbon emission involved in the present invention. DETAILED DESCRIPTION

[0021] Hereinafter, only certain exemplary embodiments are briefly described. As will be appreciated by those skilled in the art, the described embodiments may be modified in various ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and description are to be regarded as illustrative in nature and not restrictive.

[0022] The disclosure below provides many different embodiments or examples for implementing different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the components and configurations of specific examples are described below. Of course, these are merely examples and are not intended to limit the embodiments of the present invention. In addition, the embodiments of the present invention may repeat reference numerals and / or reference letters in different examples. Such repetition is for the purpose of simplicity and clarity and does not in itself indicate the relationship between the various embodiments and / or configurations discussed.

[0023] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0024] like Figure 1 As shown, this embodiment provides a construction plan optimization method for carbon emission assessment based on deep learning, including: S1. Acquire construction site image data using drones, perform temporal consistency analysis, illumination normalization, and hybrid denoising on the image data, extract image features, perform image registration, and generate a 3D model of the real scene using deep learning. S2. Segment the point cloud data of the real-world 3D model to extract different construction areas. Compare the models before and after reconstruction to determine the amount of earthwork excavation and vegetation destruction. Combined with multi-dimensional carbon emission factor data and adaptive dynamic factors, calculate the carbon emissions during the construction process. S3. Use a multi-objective optimization approach to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

[0025] In some embodiments, S1 includes: S11. Perform temporal consistency analysis and illumination normalization on the construction site image data, and remove noise using a hybrid denoising strategy that combines median filtering and edge-preserving bilateral filtering. S12. Image matching is performed using a dynamic matching threshold function based on local image density. Feature points are extracted using a feature extraction algorithm guided by local geometric consistency. Registration and global stitching are performed using a partitioned parallel incremental reconstruction method. S13. In the dense point cloud reconstruction stage, a depth-guided multi-view stereo method is used to reconstruct the hole areas of the point cloud, and a lightweight 3D model is generated through a model compression framework.

[0026] In some embodiments, the point cloud data segmentation step in S2 includes: S21. Use an airborne visual camera to acquire multi-view 3D images and video data to construct an original spatial scene dataset; S22. Perform local neighborhood voxelization on the original point cloud of the original spatial scene dataset, and perform feature encoding through the coordinate attention mechanism and high-order semantic fusion module to obtain local features at each scale; S23. Use the multi-scale feature aggregation module to fuse local features at each scale and send them to the semantic classification layer to predict regional labels for functional area division, including the main structure area, excavation area, processing area, and living area.

[0027] In some embodiments, the step of calculating the earth excavation volume and vegetation destruction volume in S2 includes: S24. Obtain a high-density point cloud model before and after construction, and perform model registration using multi-scale rigid registration and Gaussian error field weight adjustment strategy; S25. Perform grid projection difference analysis on the registered model, calculate grid cell height difference, and obtain earth excavation volume; S26. Extract the area and type of vegetation damage based on the before and after comparison of the 3D model.

[0028] In some embodiments, the carbon emissions accounting step in S2 includes: S27. Based on regional geometric characteristics and emission factor method, an adaptive dynamic factor is introduced to calculate the carbon emissions of excavation projects; S28. Calculate carbon emissions from vegetation destruction by combining the area of vegetation destruction, biomass per unit area, carbon content, and respiration rate; S29. Summarize the carbon emissions from excavation and vegetation destruction to arrive at the total carbon emissions.

[0029] In some embodiments, the device configuration optimization step in S3 includes: S31. With the goal of minimizing carbon emissions, determine the optimal number of electric excavators, loaders, and dump trucks to meet the total workload requirements without exceeding the maximum number of equipment; S32. Optimize the ratio of electric equipment to diesel equipment through genetic algorithms to reduce the use of diesel equipment.

[0030] In some embodiments, the energy structure optimization step in S3 includes: S33. Design the photovoltaic system coverage area and energy storage capacity, calculate the daily power generation of the photovoltaic system, and achieve low-carbon substitution for construction electricity demand; S34. Determine the optimal photovoltaic coverage and energy storage configuration based on sunshine hours, photovoltaic efficiency, and construction electricity demand.

[0031] In some embodiments, the construction process optimization step in S3 includes: S35. Combine the 3D model to predict excavation progress and dynamically adjust equipment scheduling strategies to reduce equipment no-load energy consumption; S36. Optimize transportation route planning to reduce carbon emissions during material transportation.

[0032] In some embodiments, S1 further includes: S14. When preprocessing the data collected by the drone, use high-pass filters and median filters to suppress noise, and use edge detection and contour extraction to retain important features; S15. Use principal component analysis to achieve data dimensionality reduction and improve the efficiency of feature extraction and registration.

[0033] In some embodiments, drone image data is preprocessed. The original image sequence is subjected to temporal consistency analysis and illumination normalization. A hybrid denoising strategy combining median filtering and edge-preserving bilateral filtering is used to preserve texture details and remove noise. During the image matching phase, a dynamic matching threshold function based on the local density of the image is used: ; in, is the initial matching threshold, For image pairs The distribution density of feature points, is the average density of the image set, is the adjustment coefficient, It is the base threshold of the dynamic matching threshold function.

[0034] In some embodiments, feature extraction and registration uses a feature extraction algorithm guided by local geometric consistency, taking into account the texture gradient direction and spatial adjacency: ; in, For the The feature vector of a point, For the The texture gradient of the point, For the Point and j The spatial distance of the points, is the weight parameter, N(i) For the The neighborhood of a point, For the The texture gradient of each point.

[0035] The registration stage uses a partitioned parallel incremental reconstruction method to divide the image into sub-images according to overlapping areas. After local reconstruction, global splicing is performed through pose optimization: ; in, For subgraph The reconstructed point cloud, The image is divided into multiple sub-images according to the overlapping area. is the local reconstruction function.

[0036] Feature registration introduces a local consistency scoring function: ; in, is the gradient angle of the matching feature point pair, is the neighborhood strength weight, which is used to enforce geometric consistency constraints. is the local consistency score, which reflects the weighted sum of the angle consistency of the matching feature point pairs; The total number of matching pairs for calculating the score.

[0037] In some embodiments, when reconstructing a three-dimensional model, in the dense point cloud reconstruction stage, a depth-guided multi-view stereo method is used to reconstruct the hole areas of the point cloud: ; in To reconstruct the point cloud, For the reconstruction The coordinates of the points; The first point in the original point cloud The coordinates of the points; express The sum of the L1 norms of the gradients of is a hyperparameter that controls the weight between the two loss terms.

[0038] A model compression framework that combines knowledge distillation and structural sparsification is adopted to guide the sub-models to learn key geometric feature response maps. During the merging phase, redundant structures are sparsely pruned to form a lightweight 3D model. The optimization objectives are: ; in, is the geometric reconstruction error, is the characteristic distillation loss, is the sparse term of the sub-model parameter, 、 is the regulating factor.

[0039] In some embodiments, point cloud data segmentation uses a visual camera on an airborne platform to acquire 3D images and video data from multiple perspectives and fields of view, constructing a raw spatial scene dataset. Automated segmentation and region label generation utilizes a point cloud semantic segmentation network based on spatial context embedding. The network incorporates a coordinate attention mechanism and a high-level semantic fusion module. The calculation formula for the specific steps is as follows: Original point cloud input: ; in, is the spatial coordinate, is the reflection intensity value. is a 3D point cloud dataset containing reflection intensity; N is the number of points in the point cloud dataset.

[0040] Perform local neighborhood voxelization and construct local regions using radius neighborhood queries: ; in, Indicates is the center and the radius is The neighborhood point set of ; Represents the position vector of the current query point, Represented in a point cloud collection The position vector of any point in ; Indicates a point with dot The Euclidean distance between is the neighborhood radius parameter.

[0041] In the feature encoding stage, the model maps the spatial coordinates of the point into position features through frequency encoding, and after splicing with the local geometric tensor, generates local features through multi-layer perceptron fusion: ; in Encodes spatial location; is the local geometry tensor; represents vector concatenation, MLP is a multi-layer perceptron, is the local eigenvector.

[0042] The coordinate attention mechanism is used to calculate the global attention weight, aggregate neighborhood information, enhance feature expression, obtain fused point features, and improve scene perception. Global scene perception is introduced through the coordinate attention mechanism: ; ; in, This is the first multi-layer perceptron; for point The offset relative to the center point of its neighborhood; is the Euclidean distance between the point and the center point; To introduce enhanced features after attention; is the second multi-layer perceptron; for point The normalized global center coordinates of ; The function is the Sigmoid activation function; is the global eigenvector.

[0043] The multi-scale feature aggregation module extracts local features of point clouds at different scales to obtain multiple sets of feature vectors reflecting geometric information within different neighborhoods. Each scale corresponds to a set of feature vectors extracted, representing the local geometric information of the points at that scale. The local features of each scale are: ; in For the The point in The local feature vector extracted at each scale; is the index of the point; is the scale index; d is the dimension of each scale feature; is the set of real numbers.

[0044] These local features of different scales are fused using direct splicing to integrate multi-scale information: ; in is the vector after multi-scale feature concatenation; symbol Represents a vector concatenation operation.

[0045] The fused features are further mapped and nonlinearly activated through a multi-layer perceptron (MLP) or linear transformation to enhance the representation capability. Finally, they are sent to the semantic classification layer for region label prediction: ; Output label set { }They correspond to the main structure area, excavation area, processing area, living area and other functional areas respectively. is the weight parameter matrix of the semantic classification layer, used to map features to category scores; is the bias vector of the semantic classification layer, which adjusts the offset of the mapping result; for point Feature vector enhanced by the coordinate attention mechanism.

[0046] In some embodiments, by comparing the 3D models before and after excavation, the amount of earthwork excavated is accurately calculated, and the area and type of vegetation damage are extracted. This includes data acquisition, model alignment, difference model generation, earthwork volume calculation, and vegetation damage data extraction. The specific calculation formula is as follows: Data acquisition: The visual camera on the airborne platform is used to obtain 3D images and video data from multiple perspectives and fields of view, and generate a high-density, uniform-scale point cloud model, which is recorded as: ; ; in, are the three-dimensional coordinates of the corresponding point in the space before construction; It is the point cloud collection before construction; It is the point cloud collection after construction; is the three-dimensional coordinate of the corresponding point in the space after construction; is the number of points in the point cloud before construction; is the number of points in the point cloud after construction.

[0047] Model registration optimization: A point cloud registration algorithm based on multi-scale rigid registration and Gaussian error field weight adjustment strategy. Coarse registration is used to estimate the initial rigid body transformation, and the following optimization formula is used for fine registration: ; in, is the standard deviation, is the rotation matrix, is the translation vector, ∈ , ∈ ; Difference model generation: Perform grid projection difference analysis on the registered model and define the height difference of each grid cell as: ; in, It indicates the height difference. A positive value indicates fill and a negative value indicates cut. It can be regarded as the earthwork change value approximately. Indicates the surface elevation before construction; Indicates the surface elevation after construction, The volume of earthwork is calculated using the following formula: ; ; in is the earthwork excavation area calculated based on the before-after comparison of the above 3D reconstruction model, is the number of boundary points in the 3D reconstructed model, is the coordinate of each point in the earthwork area of the 3D reconstructed model, The excavation volume calculated for the before and after comparison of the 3D reconstructed model, Indicates the difference in ground height before and after excavation in the 3D reconstructed model.

[0048] Vegetation destruction data extraction, the formula is as follows: ; in is the vegetation destruction area calculated based on the before-after comparison of the above 3D reconstruction model. is the number of boundary points in the 3D reconstructed model, is the coordinate of each point in the vegetation destruction area in the 3D reconstruction model.

[0049] In some embodiments, a dynamic carbon emissions accounting method based on regional geometric characteristics and emission factors combines multiple dimensions of carbon emission sources for accurate assessment. In the process of selecting carbon emission factors, an adaptive dynamic factor is introduced to calculate the carbon emissions from excavation and vegetation destruction to obtain the total carbon emissions. The specific calculation formula is as follows.

[0050] The carbon emissions from excavation are calculated using the following formula: ; in is the total carbon emissions of the excavation project calculated based on the before-and-after comparison of the 3D reconstruction model; According to the construction time stage and regional geometric characteristics Dynamically adjusted weight factors to reflect the variability of actual working conditions; For the Construction log data on equipment energy consumption, power consumption, vegetation distribution, etc. For the Emission factors for this type of equipment.

[0051] The carbon emissions from vegetation destruction are calculated using the following formula: ; ; ; in Carbon emissions due to carbon storage destruction; Carbon emissions released for vegetation respiration; is the total carbon emissions from vegetation destruction; is the number of boundary points in the 3D reconstructed model; The coordinates of each point in the vegetation destruction area in the 3D reconstruction model is the biomass per unit area; is the carbon content; is the respiratory rate.

[0052] The total carbon emissions are calculated using the following formula: .

[0053] In some embodiments, the optimization of low-carbon construction plans specifically includes: optimizing the electrification configuration of equipment, with the goal of minimizing carbon emissions, determining the optimal number of electric excavators, loaders and dump trucks; integrating renewable energy, designing the photovoltaic system coverage and energy storage capacity to meet the low-carbon replacement ratio for construction electricity demand; optimizing construction timing, combining three-dimensional models to predict excavation progress, and dynamically adjusting equipment scheduling to reduce no-load energy consumption.

[0054] Optimize the electrification configuration of equipment and optimize the equipment configuration with the goal of minimizing carbon emissions.

[0055] ; The constraints are: ≥ D； ≤ ; in For the The upper limit of the number of electric devices (maximum configurable number); To minimize overall carbon emissions; For the Number of electric equipment types (e.g. electric excavators, electric loaders, dump trucks); For the Carbon emissions per unit operating time of each type of equipment; For the Estimated operating time of the equipment; is the production capacity per unit time; D The total workload.

[0056] Renewable energy integration, design of photovoltaic coverage and energy storage capacity, and achieve the optimal low-carbon substitution ratio.

[0057] ; ; in for renewable energy coverage; is the photovoltaic system area; is the photovoltaic efficiency; is the number of sunshine hours; is the daily power generation of the photovoltaic system; is the energy storage capacity; To meet the electricity demand for construction.

[0058] A construction scheme optimization system for carbon emission assessment based on deep learning, applying any of the above-mentioned construction scheme optimization methods for carbon emission assessment based on deep learning, and a construction scheme optimization system for carbon emission assessment based on deep learning, comprising: The 3D model reconstruction module is used to acquire construction site image data through drones, perform time-series consistency analysis, illumination normalization, and hybrid denoising on the construction site image data, extract image features and perform registration, and use deep learning to generate a real-scene 3D model; The carbon emission calculation module is used to segment the point cloud data of the real-life 3D model, extract different construction areas, compare the models before and after reconstruction to determine the amount of earth excavation and vegetation destruction, and combine multi-dimensional carbon emission factor data and adaptive dynamic factors to calculate the carbon emissions during the construction process; The construction plan optimization module uses a multi-objective optimization method to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

[0059] The technical ideas of the present invention are as follows: The method of the present invention: The drone acquires data and uses deep learning-based large-scale scene reconstruction technology to perform high-precision 3D reconstruction of the construction site: drones are used to collect point cloud data of the construction site to ensure the comprehensiveness and high accuracy of the data. Subsequently, a 2D Gaussian model is used to perform probabilistic modeling on the collected point cloud data. Projection and differentiable rasterization technology are used to convert the 3D information into a differentiable image representation, enabling efficient rendering and optimization of complex scenes (such as Figure 2 (as shown). This process combines adaptive density control and extended filters to effectively improve model accuracy and stability. Finally, large-scale scene reconstruction technology based on deep learning efficiently models the processed data, generating an accurate 3D model of the construction site. This allows for pre- and post-excavation model comparisons, revealing the amount of earthwork excavated and vegetation destroyed during construction, providing foundational data for subsequent analysis and management.

[0060] The carbon emissions of the construction site are calculated based on the carbon emission factor coefficient. By analyzing the geometric characteristics of each construction site area (such as area and volume) and incorporating the carbon emission factor, carbon emissions during the construction process are calculated. A multi-dimensional assessment system is established, encompassing equipment energy consumption, material transportation, and vegetation damage. Furthermore, adaptive dynamic factors (such as construction intensity, equipment utilization, weather conditions, and construction period) are introduced to enable dynamic adjustment and real-time accounting of carbon emissions. The system can also integrate low-carbon parameters such as photovoltaic systems and electric equipment, and combined with dynamic optimization algorithms, supports real-time adjustment and optimization of construction plans.

[0061] The system of the present invention: Based on deep learning technology, this module employs a large-scale scene reconstruction method. Specific steps include: using the onboard visual camera to acquire 3D images and video data from multiple perspectives and fields of view; preprocessing the acquired drone data to remove noise and unnecessary information; using a convolutional neural network (CNN) to extract and register features from the preprocessed data; and generating a highly accurate 3D reconstruction model through multi-perspective fusion technology, thus forming a visual 3D model of the real scene. This module provides an efficient and accurate 3D modeling solution, laying a solid foundation for subsequent applications.

[0062] First, the generated 3D model undergoes point cloud segmentation, utilizing advanced point cloud processing technology to identify and delineate the different functional areas of the construction site, such as the main structure, excavation, and living areas. This provides a spatial foundation for subsequent carbon emissions analysis. Based on this foundation, adaptive dynamic factors (construction intensity, equipment utilization, meteorological conditions, and construction period) are introduced, drawing on multi-dimensional carbon emission factor data (including equipment energy consumption, material transportation, and vegetation damage). Carbon emissions during construction are dynamically calculated, and a carbon emission assessment model that can be updated in real time is constructed. The system further integrates construction logs and environmental parameters to support zoning analysis and time-series tracking of carbon emission levels during construction. By integrating low-carbon technology options such as photovoltaic systems and electric equipment, combined with dynamic optimization algorithms, it enables continuous adjustment of construction plans and full-process control of carbon emissions.

[0063] The following takes a tunnel construction project as an example to describe the implementation process of the present invention in detail. Figure 4 As shown: 1. Data collection: 1) UAVs acquire data and use deep learning-based large-scale scene reconstruction technology to perform high-precision 3D reconstruction of the construction site: Use drones or other remote sensing equipment to collect point cloud data of the construction site to ensure the comprehensiveness and high accuracy of the data (collection diagram as shown in the figure). Figure 3Using deep learning-based large-scale scene reconstruction technology, we efficiently modeled the acquired large-scale point cloud data, generating an accurate 3D model of the construction site. This allowed us to compare the models before and after excavation, determining the amount of earthwork excavated and vegetation destroyed during construction, providing foundational data for subsequent analysis and management.

[0064] 2) Synchronously collect construction log data, including: Equipment energy consumption: average daily fuel consumption of diesel loaders, transport distance of dump trucks; power consumption: average daily operating time of air compressors, transformer capacity of mixing stations, etc.; vegetation distribution: bamboo forest coverage rate of 65% in the construction area, and the damaged area is 145m 2 .

[0065] 2. 3D modeling: The present invention relates to an efficient three-dimensional modeling process, which specifically includes preprocessing of data acquired by drones, feature extraction and alignment, and high-precision model reconstruction. In the preprocessing stage, high-pass filters and median filters are used to suppress noise in the image, and statistical methods are used to remove outliers. At the same time, edge detection and contour extraction algorithms are applied to retain important features, and data accuracy is improved through spatial alignment and fusion. Subsequently, principal component analysis is used to achieve dimensionality reduction and optimize overall processing efficiency. In the feature extraction stage, deep learning-related neural networks are used to extract key feature points, and comparison is performed using an efficient feature matching algorithm; to improve matching efficiency, parallel computing and adaptive threshold strategies are used. After alignment, incremental optimization methods are used to minimize alignment errors to ensure data consistency and accuracy. Finally, in the process of generating a high-precision three-dimensional model, knowledge distillation technology is used to merge and compress the weights of sub-models to form a lightweight three-dimensional reconstruction model that can support the reconstruction of complex terrain scenes, significantly improving the overall performance of the model and surpassing existing technologies.

[0066] 3. Carbon emissions calculation: When calculating carbon emissions from site excavation, the first step is to identify the primary emission sources. These include equipment fuel consumption, such as the carbon emissions from fuel consumption by construction equipment; equipment electricity consumption, taking into account the differences in carbon emissions from different power generation methods; carbon emissions from material production and transportation; and vegetation destruction, which requires specifying the type and area of destroyed vegetation, as different vegetation types have different carbon sequestration capacities.

[0067] To accurately calculate carbon emissions, this paper collected a range of activity data. This data includes equipment operation records, such as operating hours and workload; energy consumption data, including fuel and electricity consumption; material data, including material usage, transportation distance, and method; personnel data, including carbon emissions from commuting and on-site work; earthwork volume data, using drone oblique photography to capture earthwork data before and after excavation; and vegetation destruction data, including vegetation type, destroyed area, biomass per unit area, respiration rate, and carbon content.

[0068] Regarding emission factors, energy emission factors have been clarified. For example, the diesel emission factor is approximately 2.68 kg CO2 / L, and the electricity emission factor is determined based on the local grid structure. Emission factors for material production and transportation have also been determined. For example, the explosives production emission factor is approximately 3 kg CO2 / kg, and the transportation emission factor is determined based on the mode of transportation.

[0069] The specific carbon emissions are calculated as follows: 3.1 Fuel consumption Loaders (ZL50): 2 units, each with a power of 160kW, a diesel consumption rate of approximately 0.2L / kWh, 8 hours of work per day, and a construction period of 30 days; Total fuel consumption = 2 units × 160kW × 0.2L / kWh × 8h × 30 days = 15,360 liters of diesel; Dump truck (20m 3 ): 10 vehicles, diesel consumption is about 30L / 100km, average daily driving distance is 50km, and the construction period is 30 days; Total fuel consumption = 10 vehicles × 30L / 100km × 50km / day × 30 days = 45,000 liters of diesel; Other equipment (excavators, air compressors, etc.): estimated total fuel consumption is approximately 5,000 liters of diesel; Total fuel consumption = 65,360 liters of diesel.

[0070] 3.2 Power consumption Main electrical equipment: transformers (630kVA x 2 units), electric drilling rigs, ventilation equipment, etc., with an average daily operation time of 10 hours and a construction period of 30 days; Total power consumption = 630kVA × 2 × 0.8 (load factor) × 10h × 30 days = 302,400 kWh.

[0071] 3.3 Material data 1. Material usage Anchor rods: 4m long Φ22 mortar anchor rods, with a total consumption of approximately 1,200 rods (estimated based on an anchor rod spacing of 1m×1m in the tunnel entrance section and a coverage area of 490m2); Explosives: V-level surrounding rock blasting unit consumption is about 0.5kg / m3 , the excavation volume of the tunnel entrance section is 861m 3 ; Explosive quantity = 861m 3 ×0.5kg / m 3 =430.5kg; Reinforcement: C35 reinforced concrete 254m 3 , steel bars account for about 150kg / m 3 ; Steel bar usage = 254m 3 ×150kg / m 3 =38,100kg.

[0072] 2. Material transportation Steel transportation: From the steel structure factory to the construction site, the transportation distance is 3km, using heavy-duty trucks (fuel consumption 30L / 100km); Transport energy consumption = 38100kg ÷ 20 tons / vehicle × 3km × 30L / 100km = 17.2 liters of diesel; Concrete transportation: 4 mixing stations to the construction site, 3.05 km, total concrete volume 254m 3 (about 600 trips); Transport energy consumption = 600 trips × 3.05km × 30L / 100km = 549 liters of diesel.

[0073] 3.4 Carbon emissions of relevant personnel 1. Personnel commuting Construction workers: 56 people, assuming an average daily commute of 20 km per person, using gasoline vehicles (fuel consumption 8L / 100km); Commuting fuel consumption = 56 people × 20km × 8L / 100km × 30 = 2688 liters of gasoline.

[0074] 2. On-site office Office electricity consumption: 50kWh per day, construction period 30 days; Office power consumption = 50kWh × 30 days = 1500 kWh.

[0075] 3.5 Vegetation destruction The total area of vegetation destruction can be obtained from the construction plan. Based on the topography surrounding the Xiyu Railway, the ratio of trees, shrubs, and herbs is approximately 5:3:2. To calculate the carbon emissions of trees, shrubs, and herbs, we need to use the previously provided data on biomass per unit area, carbon content, and respiration rate. The following are the calculation steps and results: 1. Calculate vegetation carbon storage ( C storage ) ; Arbor: 73m2×7.5kg / m2×0.475=264.375kg; Shrubs: 43m2×3.5kg / m2×0.425=64.7375kg; Herbaceous plants: 29m2×1.25kg / m2×0.425=15.3125kg; Total vegetation carbon storage: 264.375 + 64.7375 + 15.3125 = 344.425 kg; 2. Calculate the amount of carbon dioxide released by respiration ( C call ) ; Arbor: 73m2×0.2gCO2 / m2×24=345.6gCO2; Shrubs: 43m2×0.125gCO2 / m2×24=130.5gCO2; Herbaceous plants: 29m2×0.06gCO2 / m2×24=41.04gCO2; Total amount of carbon dioxide released by respiration: 345.6 + 130.5 + 41.04 = 517.14 g CO2; The calculation process of the optimized solution is the same as that of the original solution.

[0076] 4. Calculation results: The original scheme's total carbon emissions were calculated, with diesel accounting for 48.4% and electricity for 49.7%. The optimized scheme reduced diesel consumption by 33.4% and electricity consumption by 33.0%, resulting in an average daily reduction of 6,972 kg of CO₂ from the photovoltaic system. Table 1 shows the carbon emission calculation process for the original construction scheme, and Table 2 shows the carbon emission reductions after implementing the optimized scheme. After calculating carbon emissions, the present invention also calculated the capital expenditures for the three schemes, calculating fuel consumption, electricity-consuming equipment rental / usage fees, labor costs, material costs, and other expenses based on market unit prices. Finally, the construction costs for the three schemes are shown in Table 3. Taking both carbon emission and cost reductions into account, the first scheme required approximately 2.552 million yuan during the construction period, resulting in the highest carbon emissions. The second scheme reduced the total capital requirement to 2.088 million yuan by reducing the number of equipment, but did not introduce any new technologies. The third scheme, due to the higher initial investment in electric equipment and the photovoltaic system, increased the total cost to 3.029 million yuan, but it offered lower long-term operation and maintenance costs and a 34.3% reduction in carbon emissions. Based on the above carbon emissions calculations and the overall total capital expenditure, the general recommendation is that if the budget is sufficient and the focus is on long-term benefits, choose option 3; if you want to control costs in the short term, option 2 is better.

[0077] Table 1. Carbon emission calculation process of implementing the original construction plan

[0078] Table 2. Carbon emissions reduction after implementation of the optimized construction plan

[0079] Table 3. Construction costs of the three options

[0080] 5. Optimization of low-carbon construction solutions: 1) Device configuration optimization One electric loader, consuming 36,000 kWh (the proportion of electric equipment increased by 16.8%); 4 electric dump trucks, consuming 9,000 kWh (the proportion of electric equipment increased by 4.2%); The number of diesel equipment decreased by 60%, while the proportion of electric equipment increased by 60%; There are 15 electric drilling rigs, with electricity consumption distributed among 214,200 kWh (the proportion of electric equipment increased by 70.5%).

[0081] 2) Optimization of energy structure Photovoltaic panels cover an area of 800m 2 , coverage rate is 40% (photovoltaic system coverage rate optimization); Equipped with a 200kWh energy storage system, it effectively supports construction electricity.

[0082] 3) Transport route planning optimization Steel transportation uses 30-ton trucks with an energy consumption of 1.8 liters of diesel per kilometer (a 15% reduction in diesel consumption); Concrete transportation adopts 12m 3 Tanker trucks, consuming 20 liters of diesel per kilometer (a 12% reduction in diesel consumption); People commute via electric buses, which consume 1,680 kWh of energy (a 100% reduction in gasoline consumption).

[0083] 4) Construction sequence: Combined with the 3D model to predict the excavation progress, the equipment scheduling strategy is dynamically adjusted, and the idle rate is significantly reduced.

[0084] 6. Implementation effect verification: 1) Improved accuracy: The error in earthwork calculation has been reduced from 10.2% to 1.8% using traditional methods, and the error in vegetation destruction area measurement is less than 2%. 2) Efficiency optimization: The carbon emission assessment cycle for a single project was shortened from 72 hours to 3 hours; 3) Emission reduction benefits: The optimized solution reduced total carbon emissions by 34.3%, and the five-year operation and maintenance costs decreased by 31.6% (saving approximately RMB 800,000); 4) Decision support: Through Pareto frontier analysis, a multi-objective trade-off solution of carbon emissions, costs and construction period is provided to support the construction party in selecting the optimal solution based on the weights.

[0085] The above embodiments are intended to illustrate the present invention, not to limit the present invention. Therefore, changes in illustrative values or substitutions of equivalent components should still fall within the scope of the present invention.

[0086] From the above detailed description, it will be clear to those skilled in the art that the present invention can indeed achieve the aforementioned objectives and is in compliance with the provisions of the Patent Law.

[0087] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as covering the preferred embodiments and all changes and modifications that fall within the scope of the invention. The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

[0088] It should be noted that the above description of the relevant processes is for illustration and purpose only and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the processes under the guidance of this specification. However, such modifications and changes are still within the scope of this specification.

[0089] The basic concepts have been described above. It will be apparent to those skilled in the art after reading this application that the above disclosures are merely illustrative and do not constitute limitations on this application. Although not explicitly stated herein, those skilled in the art may make various modifications, improvements, and amendments to this application. Such modifications, improvements, and amendments are suggested in this application and remain within the spirit and scope of the exemplary embodiments of this application.

[0090] At the same time, this application uses specific terms to describe the embodiments of this application. For example, "one embodiment," "an embodiment," and / or "some embodiments" refer to a certain feature, structure, or characteristic related to at least one embodiment of this application. Therefore, it should be emphasized and noted that "one embodiment," "an embodiment," or "an alternative embodiment" mentioned twice or more in different places in this specification does not necessarily refer to the same embodiment. In addition, certain features, structures, or characteristics in one or more embodiments of this application may be appropriately combined.

[0091] Furthermore, those skilled in the art will appreciate that various aspects of the present application may be illustrated and described in terms of a number of patentable categories or situations, including any new and useful process, machine, product, or combination of substances, or any new and useful improvement thereof. Thus, various aspects of the present application may be implemented entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software. Each of the above hardware and software may be referred to as a "unit," "module," or "system." Furthermore, various aspects of the present application may take the form of a computer program product embodied in one or more computer-readable media, with computer-readable program code embodied therein.

[0092] The computer program code required for the operation of each part of this application can be written in any one or more programming languages, including object-oriented programming languages such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages such as C programming language, Visual Basic, Fortran2103, Perl, COBOL2102, PHP, ABAP, dynamic programming languages such as Python, Ruby and Groovy, or other programming languages. The program code can be run entirely on the user's computer, or as a standalone software package on the user's computer, or partly on the user's computer and partly on a remote computer, or entirely on a remote computer or server. In the latter case, the remote computer can be connected to the user's computer via any network form, such as a local area network (LAN) or a wide area network (WAN), or connected to an external computer (e.g., via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0093] In addition, unless expressly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some embodiments of the invention that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installation on an existing server or mobile device.

[0094] Similarly, it should be noted that in order to simplify the presentation of this disclosure and thereby facilitate understanding of one or more of the invention's embodiments, the foregoing descriptions of the embodiments of this disclosure sometimes combine multiple features into a single embodiment, figure, or description thereof. However, this approach should not be interpreted as reflecting an intention that the claimed subject matter requires more features than expressly recited in each claim. Rather, the subject matter of the invention may possess fewer features than the single embodiment described above.

Claims

1. A construction scheme optimization method based on carbon emission assessment based on deep learning, characterized by: include: S1. Acquire construction site image data using drones, perform temporal consistency analysis, illumination normalization, and hybrid denoising on the image data, extract image features, perform image registration, and generate a 3D model of the real scene using deep learning. S2. Segment the point cloud data of the real-world 3D model to extract different construction areas. Compare the models before and after reconstruction to determine the amount of earthwork excavation and vegetation destruction. Combined with multi-dimensional carbon emission factor data and adaptive dynamic factors, calculate the carbon emissions during the construction process. S3. Use a multi-objective optimization approach to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

2. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: S1 includes: S11. Perform temporal consistency analysis and illumination normalization on the construction site image data, and remove noise using a hybrid denoising strategy that combines median filtering and edge-preserving bilateral filtering. S12. Image matching is performed using a dynamic matching threshold function based on local image density. Feature points are extracted using a feature extraction algorithm guided by local geometric consistency. Registration and global stitching are performed using a partitioned parallel incremental reconstruction method. S13. In the dense point cloud reconstruction stage, a depth-guided multi-view stereo method is used to reconstruct the hole areas of the point cloud, and a lightweight 3D model is generated through a model compression framework.

3. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The point cloud data segmentation steps in S2 include: S21. Use an airborne visual camera to acquire multi-view 3D images and video data to construct an original spatial scene dataset; S22. Perform local neighborhood voxelization on the original point cloud of the original spatial scene dataset, and perform feature encoding through the coordinate attention mechanism and high-order semantic fusion module to obtain local features at each scale; S23. Use the multi-scale feature aggregation module to fuse local features at each scale and send them to the semantic classification layer to predict regional labels for functional area division, including the main structure area, excavation area, processing area, and living area.

4. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The calculation steps for earthwork excavation and vegetation destruction in S2 include: S24. Obtain a high-density point cloud model before and after construction, and perform model registration using multi-scale rigid registration and Gaussian error field weight adjustment strategy; S25. Perform grid projection difference analysis on the registered model, calculate grid cell height difference, and obtain earth excavation volume; S26. Extract the area and type of vegetation damage based on the before and after comparison of the 3D model.

5. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The carbon emissions accounting steps in S2 include: S27. Based on regional geometric characteristics and emission factor method, an adaptive dynamic factor is introduced to calculate the carbon emissions of excavation projects; S28. Calculate carbon emissions from vegetation destruction by combining the area of vegetation destruction, biomass per unit area, carbon content, and respiration rate; S29. Summarize the carbon emissions from excavation and vegetation destruction to arrive at the total carbon emissions.

6. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The steps for optimizing device configuration in S3 include: S31. With the goal of minimizing carbon emissions, determine the optimal number of electric excavators, loaders, and dump trucks to meet the total workload requirements without exceeding the maximum number of equipment; S32. Optimize the ratio of electric equipment to diesel equipment through genetic algorithms to reduce the use of diesel equipment.

7. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The energy structure optimization steps in S3 include: S33. Design the photovoltaic system coverage area and energy storage capacity, calculate the daily power generation of the photovoltaic system, and achieve low-carbon substitution for construction electricity demand; S34. Determine the optimal photovoltaic coverage and energy storage configuration based on sunshine hours, photovoltaic efficiency, and construction electricity demand.

8. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The construction process optimization steps in S3 include: S35. Combine the 3D model to predict excavation progress and dynamically adjust equipment scheduling strategies to reduce equipment no-load energy consumption; S36. Optimize transportation route planning to reduce carbon emissions during material transportation.

9. The construction scheme optimization method based on carbon emission assessment based on deep learning according to claim 1 is characterized in that: The S1 also includes: S14. When preprocessing the data collected by the drone, use high-pass filters and median filters to suppress noise, and use edge detection and contour extraction to retain important features; S15. Use principal component analysis to achieve data dimensionality reduction and improve the efficiency of feature extraction and registration.

10. A construction plan optimization system based on carbon emission assessment based on deep learning, characterized by: The construction scheme optimization method based on carbon emission assessment of deep learning according to any one of claims 1 to 9 is applied, and the construction scheme optimization system based on carbon emission assessment of deep learning comprises: The 3D model reconstruction module is used to acquire construction site image data through drones, perform time-series consistency analysis, illumination normalization, and hybrid denoising on the construction site image data, extract image features and perform registration, and use deep learning to generate a real-scene 3D model; The carbon emission calculation module is used to segment the point cloud data of the real-life 3D model, extract different construction areas, compare the models before and after reconstruction to determine the amount of earth excavation and vegetation destruction, and combine multi-dimensional carbon emission factor data and adaptive dynamic factors to calculate the carbon emissions during the construction process; The construction plan optimization module uses a multi-objective optimization method to comprehensively evaluate and adjust equipment configuration, energy structure, and construction processes, generating a low-carbon construction plan that includes increasing the proportion of electric equipment, optimizing photovoltaic system coverage, and planning transportation routes.

Images