Engineering construction full life cycle management platform and method based on digital delivery

Abstract

The invention discloses an engineering construction full life cycle management platform and method based on digital delivery, and relates to the technical field of engineering construction management and informationization, and the method comprises the following steps: carrying out the high-frequency space-time granularity sampling of a construction site structure through a high-precision laser scanner, and obtaining the data information of a site structure; according to the invention, intelligent and adaptive dynamic control of laser point cloud sampling granularity is realized, and the problems of point cloud redundancy and noise pollution caused by blind high-frequency sampling are solved. Through regular facade detection and environment dynamic perception, the system can actively reduce the sampling density, reduce invalid data and reduce storage and calculation pressure for a structure stable region; in a complex or dynamic region, the system keeps high-precision sampling, and feature integrity is ensured. According to the method, the modeling precision and efficiency are considered, and the reliability and the application value of full-life-cycle dynamic modeling are improved.

Description

Technical Field

[0001] The present invention relates to the field of engineering construction management and information technology, and in particular to a full life cycle management platform and method for engineering construction based on digital delivery. Background Art

[0002] The full life cycle management of engineering construction based on digital delivery refers to the use of digital means (such as BIM, CIM, GIS, Internet of Things, cloud computing, artificial intelligence, etc.) to comprehensively collect, integrate, manage and share information at all stages of the life cycle of engineering construction projects, from planning, design, construction, operation to decommissioning, to form digital delivery results (such as digital models, digital archives, operational data, etc.) throughout the entire project process. On this basis, visual, collaborative and intelligent management of the entire life cycle of the project is achieved. Through digital delivery, data at all stages of the project can be efficiently circulated and reused, breaking the traditional "information silos", improving decision-making efficiency and management level, ensuring the comprehensive optimization of project quality, safety, progress and cost, and providing a reliable digital foundation for subsequent operation and maintenance and asset management, truly maximizing the value of the entire life cycle of the project.

[0003] The existing technology has the following deficiencies:

[0004] In the existing dynamic modeling process of engineering construction throughout its lifecycle based on laser point clouds, high-precision laser scanners are typically used to perform high-frequency spatiotemporal granularity acquisition of on-site structures. This is especially true for millimeter-level laser point cloud systems, which can capture tens of millions of point cloud data points per second. However, due to limitations in the default configuration of acquisition equipment and the influence of the dynamic on-site environment, direct real-time updates of massive point cloud data to engineering design models are often performed without effective sampling control and preprocessing, which can easily lead to the following problems:

[0005] The spatiotemporal granularity of point cloud acquisition far exceeds the requirements of engineering applications, resulting in a large amount of point data with duplicate spatial locations and redundant information in the model, which seriously increases the storage and computational burden of the model;

[0006] Excessive high-temporal and high-granularity noise point clouds will mask the true characteristics of the target structure after real-time updating, interfere with subsequent visualization, condition monitoring and deformation analysis, and may even cause misjudgment and lead to engineering management decision-making errors.

[0007] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0008] The purpose of the present invention is to provide a full life cycle management platform and method for engineering construction based on digital delivery, which realizes intelligent and adaptive dynamic control of the laser point cloud sampling granularity, and solves the point cloud redundancy and noise pollution problems caused by blind high-frequency sampling. Through regular facade detection and dynamic environmental perception, for structurally stable areas, the system can actively reduce the sampling density, reduce invalid data, and reduce storage and computing pressure; in complex or dynamic areas, the system maintains high-precision sampling to ensure feature integrity. This method takes into account both modeling accuracy and efficiency, improves the reliability and application value of full life cycle dynamic modeling, and solves the problems in the above-mentioned background technology.

[0009] To achieve the above objectives, the present invention provides the following technical solution: a method for managing the entire life cycle of engineering construction based on digital delivery, comprising the following steps:

[0010] Use high-precision laser scanners to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information;

[0011] Preprocess the acquired original point cloud data and uniformly store the normalized multi-time point cloud data into a data set;

[0012] Based on the established data set, feature engineering technology is used to extract characteristic dimensions reflecting that the current construction site is in a regular facade area from multiple dimensions. The extracted characteristic dimensions are comprehensively analyzed to quantify the degree of regularity of the target area.

[0013] The feature dimensions after comprehensive analysis are constructed into feature vectors and input into a pre-trained machine learning model to achieve intelligent evaluation of the regularity of the construction site structure and generate corresponding evaluation results;

[0014] When the target area is detected to be in a regular facade area, the spatiotemporal sampling granularity of the laser point cloud is dynamically adjusted according to the structural change characteristics and environmental dynamic characteristics of the target area, and the sampling granularity is controlled to reduce the point cloud sampling density while meeting the modeling accuracy requirements.

[0015] Preferably, the specific steps of performing high-frequency spatiotemporal granular sampling of the construction site structure by a high-precision laser scanner to obtain on-site structural data information are as follows:

[0016] First, based on the construction site layout, structure distribution, and monitoring requirements, the laser scanning equipment is rationally arranged to determine the scanning range, resolution, and sampling frequency.

[0017] Secondly, perform on-site point cloud acquisition tasks, and the laser scanner performs a full-scale scan of the target structure surface at high temporal frequency and high spatial resolution;

[0018] Subsequently, the coordinates of multiple frames of point clouds acquired at different times are unified and spliced to form a continuous set of structural point clouds with high temporal and spatial resolution.

[0019] Finally, the data set is organized, stored, and relevant metadata information is synchronized, laying the foundation for subsequent feature extraction, environmental assessment, and dynamic modeling.

[0020] Preferably, feature engineering technology is used to extract feature dimensions reflecting that the current construction site is in a regular facade area from multiple dimensions. The extracted feature dimensions include the time series variance of the feature change and the overlap of the forward and reverse scanning point clouds. The time series variance of the feature change and the overlap of the forward and reverse scanning point clouds are comprehensively analyzed under the detection window to generate a structural feature fluctuation reference value and a two-way consistency reference value respectively. The regularity of the target area is quantified through the structural feature fluctuation reference value and the two-way consistency reference value.

[0021] Preferably, the specific steps of comprehensively analyzing the time series variance of the feature variation in the detection window to generate the structural feature fluctuation reference value are as follows:

[0022] Assume that the characteristic value sequence of the structural features in the target area at continuous sampling time is f i , f i ={f1, f2, f3, ..., f n}, where f i It is the structural feature value extracted from the target area when the laser point cloud is sampled at time point i, n is the total number of time points, and the dynamic differential energy of the feature sequence is calculated. The calculation expression is as follows:

[0023]

[0024] , where Ef is the dynamic differential energy of the structural feature;

[0025] Based on the dynamic differential energy Ef of the structural feature, the structural feature fluctuation reference value is generated. The generation formula is as follows:

[0026]

[0027] , where SFFR is the structural characteristic fluctuation reference value.

[0028] Preferably, the specific steps of comprehensively analyzing the overlap of the forward and reverse scanning point clouds in the detection window to generate a bidirectional consistency reference value are as follows:

[0029] Get the forward scanning point cloud set P respectively f ={p fj}, and reverse scan point cloud set P b ={p bj}, where p fjis the jth point in the forward scan point cloud, p bj is the jth point in the reverse scan point cloud. The two sets of point clouds are aligned by point cloud registration to obtain the overlapping area. In the overlapping area, the corresponding point pair set is extracted, that is, the forward point p fj The reverse point p of its nearest neighbor bj To enhance the sensitivity to local in-plane changes, the pairing set is constructed to construct a relative position residual function, which is as follows:

[0030] R j =||(p fj -p bj )·n j ||

[0031] , where R j is the normal residual of a single point pair, n j It is p fj The local normal vector at the center;

[0032] After obtaining all normal residuals R j After that, a bidirectional consistency reference value is generated, and the generation formula is as follows:

[0033]

[0034] , where BCRV is the bidirectional consistency reference value, d j is the original Euclidean distance of the jth forward and reverse point pair, and N is the number of overlapping point pairs.

[0035] Preferably, the structural characteristic fluctuation reference value and the bidirectional consistency reference value after comprehensive analysis are constructed as a feature vector and input into a pre-trained machine learning model. The regularity variation coefficient is generated by the machine learning model, and the regularity of the construction site structure is intelligently evaluated by the regularity variation coefficient, and the corresponding evaluation results are generated.

[0036] Preferably, the regularity variation coefficient generated by the pre-trained machine learning model when performing intelligent evaluation of the regularity of the construction site structure is compared with a pre-set reference threshold value of the regularity variation coefficient to divide the construction site. The division steps are as follows:

[0037] If the regularity variation coefficient is less than a preset regularity variation coefficient reference threshold, the detection area is divided into a regular facade area;

[0038] If the regularity variation coefficient is greater than or equal to a preset regularity variation coefficient reference threshold, the detection area is divided into an irregular facade area.

[0039] Preferably, when the target area is detected to be in a regular facade area, the specific steps of controlling the sampling granularity to reduce the point cloud sampling density according to the structural change characteristics and environmental dynamic characteristics of the target area are as follows:

[0040] First, after detecting that the target area belongs to a regular facade area, the curvature entropy value is calculated based on the structural change characteristics in the area to measure the complexity of the surface geometry change in the area. The calculation expression is as follows:

[0041]

[0042] , where H c is the curvature entropy index, p a is the proportion of points in the ath curvature interval, and M is the number of curvature intervals;

[0043] Subsequently, in order to quantify the dynamic change characteristics of the environment in which the target area is located, the disturbance energy coefficient is calculated to reflect the disturbance intensity of the dynamic target in the environment on the point cloud. The calculation method is:

[0044]

[0045] , where T is the sampling period, H is the number of dynamic points detected in the target area, and v b (t) is the velocity vector of the bth dynamic point at time t, E d is the environmental dynamic disturbance energy coefficient, t0 is the starting time of the sampling period;

[0046] Comprehensive curvature entropy index H c and the environmental dynamic disturbance energy coefficient E d , the sampling granularity is adaptively adjusted through a dual-factor driven dynamic sampling adjustment strategy. The specific adjustment formula is as follows:

[0047]

[0048] , where ρ is the actual sampling granularity after dynamic adjustment, ρ0 is the initial sampling granularity, α is the curvature factor sensitivity coefficient, β is the curvature entropy exponential attenuation coefficient, γ is the dynamic disturbance factor sensitivity coefficient, λ is the dynamic disturbance exponential attenuation coefficient, and e is the natural base.

[0049] The preferred engineering construction lifecycle management platform based on digital delivery includes a high-frequency point cloud data acquisition module, a point cloud preprocessing and data set construction module, a regular facade feature extraction and quantification module, a regularity intelligent evaluation module, and a dynamic sampling granularity adjustment module:

[0050] The high-frequency point cloud data acquisition module uses a high-precision laser scanner to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information;

[0051] The point cloud preprocessing and data set construction module preprocesses the acquired raw point cloud data and uniformly stores the normalized multi-time point cloud data into a data set;

[0052] The regular facade feature extraction and quantification module, based on the established data set, uses feature engineering technology to extract feature dimensions from multiple dimensions that reflect the current construction site's location in the regular facade area. It then conducts a comprehensive analysis of the extracted feature dimensions to quantify the degree of regularity in the target area.

[0053] The regularity intelligent assessment module constructs the characteristic dimensions after comprehensive analysis into feature vectors and inputs them into a pre-trained machine learning model to realize intelligent assessment of the regularity of the construction site structure and generate corresponding assessment results;

[0054] The dynamic sampling granularity adjustment module dynamically adjusts the spatiotemporal sampling granularity of the laser point cloud according to the structural change characteristics and environmental dynamic characteristics of the target area when the target area is detected to be in a regular facade area. It also controls the sampling granularity to reduce the point cloud sampling density while meeting the modeling accuracy requirements.

[0055] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0056] The present invention realizes intelligent, adaptive and dynamic control of the sampling granularity of laser point clouds, effectively solving the problems of point cloud data redundancy and noise pollution caused by blind high-frequency and high-density sampling in the existing technology. On the one hand, based on regular facade detection and dynamic environmental perception, the system can actively reduce the sampling resolution and sampling frequency for regular areas with stable structures and single features, reduce invalid or redundant point cloud data, and significantly reduce the storage and computing pressure of the model; on the other hand, in areas with complex structures or dynamic environments, the system can still maintain high-precision sampling to ensure the complete capture of important features. While ensuring modeling accuracy and structural feature expression capabilities, this solution achieves lightweight, precise and efficient point cloud data, significantly improving the modeling efficiency and reliability of dynamic modeling throughout the life cycle of engineering construction, and providing high-quality data support for subsequent visualization, status monitoring and intelligent decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction to the drawings required for use in the embodiments will be given below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0058] Figure 1This is a flow chart of the method for engineering construction full life cycle management based on digital delivery of the present invention.

[0059] Figure 2 This is a module diagram of the engineering construction full life cycle management platform based on digital delivery of the present invention. DETAILED DESCRIPTION

[0060] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art.

[0061] The present invention provides Figure 1 The engineering construction lifecycle management method based on digital delivery shown in the figure includes the following steps:

[0062] Use high-precision laser scanners to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information;

[0063] High-precision laser scanners are used to perform high-frequency, spatiotemporal sampling of construction site structures to acquire on-site structural data. This involves using laser scanners (such as terrestrial laser scanners, mobile lidar, or drone laser scanners) to continuously and densely collect 3D data of structures such as buildings, bridges, roads, and tunnels at a high temporal frequency (i.e., multiple samplings in a short period of time) and high spatial resolution (i.e., acquiring point clouds with millimeter or centimeter-level accuracy in 3D space). This sampling not only records static features such as the structural surface geometry, size, position, and boundaries in real time and detail, but also enables the capture of structural changes, deformation, or abnormal conditions during construction by comparing multiple moments. This high-frequency, spatiotemporal sampling provides authentic, complete, and highly accurate on-site structural data for subsequent modeling, deformation monitoring, safety analysis, and operations and maintenance management. Simply put, it involves "scanning" a building with lasers, continuously and meticulously capturing its true form and changes in time and space.

[0064] The specific steps for acquiring on-site structural data information by using high-precision laser scanners to perform high-frequency spatiotemporal granular sampling of construction site structures are as follows: First, the laser scanning equipment is rationally arranged based on the construction site layout, structure distribution, and monitoring requirements, and the scanning range, resolution, and sampling frequency are determined. Second, the on-site point cloud acquisition task is performed. The laser scanner scans the target structure surface in all directions at a high temporal frequency (e.g., multiple frames per second) and high spatial resolution (e.g., millimeter or centimeter-level point spacing), acquiring raw point cloud data such as its spatial coordinates, intensity, and reflectance information. Subsequently, the coordinates of the multiple point clouds acquired at different times are unified and spliced to form a continuous high-spatiotemporal resolution structural point cloud collection. Finally, this data collection is organized, stored, and synchronized with relevant metadata (e.g., acquisition time, acquisition equipment, and scanning parameters), laying the foundation for subsequent feature extraction, environmental assessment, and dynamic modeling. Through this process, the real, continuous three-dimensional spatial information and evolution characteristics of the structure during construction can be fully acquired.

[0065] Preprocess the acquired original point cloud data and uniformly store the normalized multi-time point cloud data into a data set;

[0066] The acquired raw point cloud data is pre-processed, and the standardized multi-time point cloud data is uniformly stored in the data set. This means that after completing the on-site point cloud data acquisition, the raw point cloud is first subjected to a series of data cleaning and standardization processes, including the removal of noise points (such as outliers, dynamic interference points), coordinate system calibration (ensuring that data acquired at different times or devices are aligned in the same spatial reference system), density balancing (avoiding overly dense or sparse local point cloud distribution), as well as format conversion and attribute supplementation, to ensure the accuracy, integrity and availability of the point cloud data. Subsequently, the standardized point cloud data collected at different times and in different areas are uniformly organized and archived to form a structured point cloud data set, providing a continuous, standardized and directly callable high-quality data foundation for subsequent feature extraction, scene change analysis, dynamic modeling and model updates. Simply put, it is to transform the "raw, messy and scattered" point cloud data into a "clean, unified and usable" data set through processing.

[0067] Based on the established data set, feature engineering technology is used to extract characteristic dimensions from multiple dimensions that reflect the current construction site being in a regular facade area. The extracted characteristic dimensions are then comprehensively analyzed to quantify the degree of regularity of the target area.

[0068] Comprehensive analysis of the extracted characteristic dimensions is commonly performed using methods such as statistical analysis, principal component analysis (PCA), weighted fusion, cluster analysis, and time series change analysis. Specifically, each characteristic dimension can first be standardized to eliminate interference from different dimensions. Then, principal component analysis (PCA) or factor analysis can be used to extract principal components or latent factors that comprehensively reflect the characteristics of regular facades. Furthermore, based on field experience or engineering weights, a weighted fusion model can be designed to perform a weighted summation or weighted scoring of multiple characteristic dimensions to form a unified "regular facade characteristic reference value." Furthermore, cluster analysis (such as K-means or DBSCAN) can be used to divide the structural surface into "regular regions" and "irregular regions." Combined with time series analysis, the fluctuation and stability of the characteristic dimension trends at successive moments can be examined to determine whether the structure maintains a long-term stable regular facade state. Through these various methods, characteristic dimensions can be fully integrated, de-redundant, and quantitatively analyzed, providing accurate decision-making for subsequent structural stability assessment and dynamic sampling adjustments.

[0069] Feature engineering technology is used to extract feature dimensions reflecting that the current construction site is in a regular facade area from multiple dimensions. The extracted feature dimensions include the time series variance of the feature change and the overlap of the forward and reverse scanning point clouds. The time series variance of the feature change and the overlap of the forward and reverse scanning point clouds are comprehensively analyzed under the detection window to generate structural feature fluctuation reference values and bidirectional consistency reference values, respectively. The structural feature fluctuation reference values and bidirectional consistency reference values are used to quantify the degree of regularity of the target area.

[0070] Small fluctuations in the temporal variance of feature variation typically indicate that the structure on-site is a regular, flat, and featureless facade. This is because regular facades inherently possess a high degree of geometric flatness and stability; their surface structure does not undergo significant local deformation or feature changes over time, scanning angle, or environmental changes. Therefore, in a temporal point cloud constructed using high-frequency spatiotemporal sampling, if key features extracted (such as normal vectors, curvature, edge rates, and fitting residuals) exhibit minimal variation, stable variance, and limited fluctuations over multiple consecutive moments, it can be inferred that the surface state of that region remains consistent over time, reflecting the absence of complex structures, significant boundaries, significant curves, or detailed features, thus conforming to the typical characteristics of a regular facade. In contrast, complex structures or irregularly shaped facades typically exhibit greater volatility and uncertainty with changes in the spatiotemporal sampling scale or construction dynamics. Therefore, a small temporal variance in feature variation can serve as an important criterion for identifying regions with stable structures and regular morphology, effectively identifying areas with stable structures and regular morphology and providing a reliable basis for dynamic sampling and modeling strategies.

[0071] The specific steps for comprehensively analyzing the time series variance of feature changes within the detection window to generate a reference value for structural feature fluctuations are as follows:

[0072] Assume that the characteristic value sequence of the structural characteristics (such as strike angle, curvature, edge frequency, etc.) in the target area at continuous sampling time is f i , f i ={f1, f2, f3, ..., f n}, where f i It is the structural feature value extracted from the target area when the laser point cloud is sampled at time point i, n is the total number of time points, and the dynamic differential energy of the feature sequence is calculated. The calculation expression is as follows:

[0073]

[0074] , where Ef is the dynamic differential energy of the structural feature;

[0075] This step measures the cumulative energy intensity of adjacent fluctuations in the feature over time. If the structural feature changes slightly and is stable, the differential energy Ef tends to be small. Conversely, if the structural feature fluctuates violently (such as in deformed structures, crack development, or vibration), the Ef value increases significantly. This energy index can sensitively reflect the "volatility" of structural features at a microscale.

[0076] Based on the dynamic differential energy Ef of the structural feature, the structural feature fluctuation reference value is generated. The generation formula is as follows:

[0077]

[0078] , where SFFR is the structural characteristic fluctuation reference value.

[0079] By performing logarithmic suppression and normalization on the characteristic differential energy, a structural characteristic fluctuation reference value was constructed to reflect the strength of structural characteristic fluctuations. A smaller reference value indicates weaker temporal and spatial characteristic variations in the structure, lower volatility, and a region closer to a regular, flat, and characteristically stable facade. Conversely, a larger reference value indicates stronger structural variations or anomalies in the region.

[0080] The smaller the structural feature fluctuation reference value generated by post-analyzing the time-series variance of feature changes within the detection window, the closer the on-site structure is to a regular, flat, featureless facade. The principle is that regular facades in actual engineering exhibit good surface continuity, a single shape, and a lack of significant geometric features or local variations. Therefore, under high-frequency spatiotemporal sampling, the extracted geometric features (such as normal vectors, curvature, edge rates, fitting residuals, etc.) will remain highly stable in the temporal dimension, resulting in a smaller time-series variance of feature changes and, in turn, a lower structural feature fluctuation reference value. Conversely, if the monitored area has significant structural complexity, such as irregular curved surfaces, complex joints, local damage, or dynamic construction changes, the fluctuation amplitude of its features increases under continuous sampling, the time-series variance increases, and the structural feature fluctuation reference value increases accordingly.

[0081] A good degree of overlap between forward and reverse scan point clouds generally indicates that the structure on-site has a regular, flat, and featureless facade. This is because regular facades exhibit geometric symmetry, flatness, and continuity, and their geometric features are highly consistent at different angles and scanning directions. Specifically, when the structure's surface is laser scanned in both forward and reverse directions, if the point cloud data can achieve high-precision overlap in terms of spatial position, surface normal, and boundary contours, this indicates that the scanned surface lacks significant bumps, warping, irregularly shaped components, or complex local features, and instead exhibits flat, regular, and smooth surface characteristics. Conversely, if there are significant undulations, bends, complex details, or dynamic occlusion, the point clouds from the forward and reverse scans will often exhibit localized misalignment, normal deviation, or edge mismatches, resulting in reduced overlap. Therefore, the overlap between forward and reverse scan point clouds can indirectly reflect the flatness and geometric consistency of the structure's surface. The higher the overlap, the more likely it is that the area is a regular, flat, and featureless facade. This feature can be used as an important criterion for distinguishing regular surfaces from complex structures in point cloud feature engineering.

[0082] The specific steps for comprehensively analyzing the overlap of forward and reverse scan point clouds within the detection window to generate a bidirectional consistency reference value are as follows:

[0083] Get the forward scanning point cloud set P respectively f ={p fj}, and reverse scan point cloud set P b ={p bj}, where p fj is the jth point in the forward scan point cloud, p bj is the jth point in the reverse scan point cloud (with p fjThe two sets of point clouds are aligned by point cloud registration (such as ICP, NDT or other registration methods) to obtain the overlapping area. In the overlapping area, the corresponding point pair set is extracted, that is, the positive point p fj The reverse point p of its nearest neighbor bj To enhance the sensitivity to local in-plane changes, the pairing set is constructed to construct a relative position residual function, which is as follows:

[0084] R j =||(p fj -p bj )·n j ||

[0085] , where R j is the normal residual of a single point pair, n j It is p fj The local normal vector at the center;

[0086] By registering and matching forward and reverse point clouds, we extract a set of corresponding point pairs that truly reflect the local characteristics of the structure's surface. By constructing a residual function based on normal projection, we accurately characterize the slight deviation of the point cloud in the normal direction, providing a highly sensitive error basis for subsequent calculations.

[0087] After obtaining all normal residuals R j Finally, a bidirectional consistency reference value is generated. The core idea of the bidirectional consistency reference value is to measure the severity of the local in-plane normal projection deviation in the overlapping area. The generation formula is as follows:

[0088]

[0089] , where BCRV is the bidirectional consistency reference value, d j is the original Euclidean distance of the jth forward and reverse point pair, and N is the number of overlapping point pairs.

[0090] By calculating the residual energy of point pairs in the normal direction within the overlapping region, the local surface overlap error between the forward and reverse point clouds is accurately quantified, forming a consistency reference value that reflects the regularity of the structure. This effectively determines whether the on-site structure is flat and regular. The smaller the residual energy, the closer the area is to a regular facade.

[0091] The overlap of the forward and reverse scan point clouds is comprehensively analyzed within the detection window to generate a bidirectional consistency reference value. The smaller the bidirectional consistency reference value, the more regular, flat, and featureless the on-site structure is, indicating that it is a regular facade. The principle is that the bidirectional consistency reference value reflects the consistency of the structure surface under different scanning angles by quantifying the spatial deviation, overlap error, or fitting residual between the forward and reverse scan point clouds. For regular facades, due to the good flatness and geometric symmetry of their surfaces, no matter which direction they are scanned from, the collected point clouds can be highly consistent in terms of position, shape, edge contour, etc., with high overlap accuracy, resulting in a smaller bidirectional consistency reference value. However, if the on-site structure surface has obvious curves, special-shaped components, sudden edges, or complex details, the forward and reverse scan point clouds will be misaligned, partially missing, or have fitting errors when overlapped, resulting in a larger bidirectional consistency reference value.

[0092] The feature dimensions after comprehensive analysis are constructed into feature vectors and input into a pre-trained machine learning model to achieve intelligent evaluation of the regularity of the construction site structure and generate corresponding evaluation results;

[0093] The structural characteristic fluctuation reference value and the bidirectional consistency reference value after comprehensive analysis are constructed as a feature vector and input into a pre-trained machine learning model. The machine learning model generates a regularity variation coefficient, which is used to intelligently evaluate the regularity of the construction site structure and generate the corresponding evaluation results.

[0094] A pre-trained machine learning model refers to an intelligent model that has been trained to learn model parameters and pattern recognition capabilities through supervised or unsupervised learning methods based on historical construction site point cloud data or large amounts of simulation-generated data. In this solution, the model uses input feature vectors (such as structural feature fluctuation reference values and bidirectional consistency reference values) to automatically identify and determine whether the current construction site structure is a regular facade, and then quantify its regularity variation level. Before the model is officially put into use, it is usually trained on an existing sample dataset with annotated regularity status. By learning the distribution and variation patterns of "regular" and "irregular" structures in the feature space from the historical data, a nonlinear mapping relationship between features and regularity variation coefficients is established. Common machine learning models include support vector machines (SVMs), random forests (RFs), gradient boosted tree models (GBDTs), multi-layer perceptrons (MLPs), or convolutional neural networks (CNNs). After training, the model is capable of automatically identifying, scoring, and classifying new input features.

[0095] During the formal application of the system, the pre-trained machine learning model can quickly and accurately generate a regularity variation coefficient for the feature vectors of the on-site structural point cloud collected in real time or in stages. The coefficient value can be used to quantify the regularity of the current structure and reflect whether it meets the standards of a regular facade. At the same time, it combines historical change trends and on-site safety management needs to output intelligent structural assessment results. Compared with the traditional simple threshold-based judgment method, the machine learning model can automatically discover the complex patterns hidden behind the feature data in the high-dimensional feature space. It has higher stability, adaptability and judgment accuracy. It is especially capable of dealing with problems such as the changing construction site environment, large sampling uncertainty, and local structural complexity. It effectively improves the intelligence level of regularity detection and evaluation and avoids the risk of misjudgment caused by subjective setting of feature thresholds. By introducing this model, the system can realize the automation, data-driven and intelligent judgment of regularity, providing reliable support for the subsequent dynamic sampling strategy optimization, modeling accuracy control and safety risk identification.

[0096] The machine learning model is not limited here. Any machine learning model that can perform a comprehensive analysis of the structural feature fluctuation reference value SFFR and the bidirectional consistency reference value BCRV to generate the regularity variation coefficient RVC is acceptable. To implement the technical solution of the present invention, the present invention provides a specific implementation method.

[0097] The formula for generating the regularity variation coefficient RVC is as follows:

[0098]

[0099] , where b1 and b2 are the preset proportional coefficients of the structural feature fluctuation reference value SFFR and the bidirectional consistency reference value BCRV, respectively, and both b1 and b2 are greater than 0.

[0100] Preset scaling factors, denoted as b1 and b2, are fixed parameters used in the formula to balance and weight the influence of two different characteristic reference values (i.e., SFFR and BCRV) on the resulting regularity variation coefficient (RVC). Because SFFR (structural characteristic fluctuation reference value) and BCRV (bidirectional consistency reference value) may differ in their numerical distribution ranges, physical meanings, and importance in different engineering scenarios, direct addition or averaging may result in one characteristic overdoing or underdominating the RVC. Therefore, b1 and b2 are introduced as pre-set weighting factors. By properly setting the values of b1 and b2, the contribution ratio of structural fluctuation characteristics and consistency characteristics in the RVC calculation can be adjusted based on the varying levels of attention paid to them in actual applications. This ensures that the final regularity variation coefficient better reflects the actual regularity variation characteristics of the structure in the field and the requirements for engineering judgment. In short, the purpose of the preset scaling factors is to give appropriate "voice" to different characteristics, ensuring that the RVC calculation results are scientific, stable, and have engineering guidance significance.

[0101] It can be seen from the regularity variation coefficient that the smaller the structural feature fluctuation reference value generated by post-analyzing the time-series variance of the feature variation under the detection window, and the smaller the bidirectional consistency reference value generated by comprehensively analyzing the overlap of the forward and reverse scanning point clouds under the detection window, the smaller the regularity variation coefficient generated when the regularity of the construction site structure is intelligently evaluated by the pre-trained machine learning model, indicating that the probability that the on-site structure is in a regular, flat, and regular facade with no obvious features is greater. Conversely, the probability that the on-site structure is in a regular, flat, and regular facade with no obvious features is smaller.

[0102] The regularity variation coefficient generated by the pre-trained machine learning model during the intelligent evaluation of the regularity of the construction site structure is compared with the pre-set reference threshold of the regularity variation coefficient to divide the construction site. The division steps are as follows:

[0103] If the regularity variation coefficient is less than a preset regularity variation coefficient reference threshold, the detection area is divided into a regular facade area;

[0104] If the regularity variation coefficient is greater than or equal to a preset regularity variation coefficient reference threshold, the detection area is divided into an irregular facade area.

[0105] When the target area is detected to be in a regular vertical area, the spatiotemporal sampling granularity of the laser point cloud is dynamically adjusted according to the structural change characteristics of the target area and the dynamic characteristics of the environment. Under the condition of meeting the modeling accuracy requirements, the sampling granularity is controlled to reduce the point cloud sampling density.

[0106] When the target area is detected as a regular facade, the spatiotemporal sampling granularity of the laser point cloud is dynamically adjusted based on the structural changes and environmental dynamics of the target area. While meeting modeling accuracy requirements, the sampling granularity is controlled to reduce the point cloud sampling density. Its core function is to adaptively adjust the sampling strategy to effectively reduce the amount of redundant data, improve modeling efficiency, and ensure model accuracy and dynamic responsiveness. On construction sites, regular facade areas typically exhibit smooth surfaces, minimal curvature changes, few edge features, and high structural stability. Maintaining high spatiotemporal granularity sampling in such areas often results in the collection of a large amount of duplicate and redundant point cloud data, which not only consumes storage and computing resources but also can lead to a surge in computational complexity for subsequent modeling and rendering, impacting engineering modeling efficiency and system real-time performance. Through this dynamic adjustment strategy, the system can proactively reduce the temporal sampling frequency and spatial sampling density based on the stability and structural characteristics of the target area. This reduces unnecessary high-frequency scanning and excessively high-resolution sampling, keeping the point cloud quantity within a reasonable range. On the premise of meeting the accuracy requirements of the model for this area, retaining the necessary structural feature points and removing redundant points can significantly reduce the scale of point cloud data, reduce system load, and improve the response speed of data processing, modeling, and visualization. At the same time, this mechanism can also be combined with the dynamic monitoring mechanism. Once dynamic changes or potential anomalies (such as structural displacement, external force interference) occur in the regular facade area, the system can increase the sampling granularity in real time and restore high-density monitoring capabilities, thereby ensuring comprehensive perception and high-precision modeling of key engineering processes, and achieving the unity of lightweight modeling and accuracy assurance. On the basis of ensuring the accuracy of engineering modeling, this mechanism effectively solves the problems of serious redundancy of high-precision point clouds, high computing pressure, and slow update response in existing technologies, and significantly improves the level of intelligence in modeling throughout the entire life cycle of the project.

[0107] When the target area is detected to be in a regular facade area, the specific steps for controlling the sampling granularity to reduce the point cloud sampling density are as follows based on the structural change characteristics and environmental dynamic characteristics of the target area:

[0108] First, after detecting that the target area belongs to a regular facade area, the curvature entropy value is calculated based on the structural change characteristics in the area to measure the complexity of the surface geometry change in the area. The calculation expression is as follows:

[0109]

[0110] , where H c is the curvature entropy index, p a is the proportion of points in the ath curvature interval, M is the number of curvature intervals, which means that the curvature distribution in the target area is divided into M intervals;

[0111] Through curvature entropy, we can comprehensively measure whether there are still small-scale undulations, seams, detailed structures and other features inside the regular facade, so as to avoid missing details due to large-scale regional regularity.

[0112] Subsequently, in order to quantify the dynamic change characteristics of the environment in which the target area is located, the disturbance energy coefficient is calculated to reflect the disturbance intensity of the dynamic target in the environment on the point cloud. The calculation method is:

[0113]

[0114] , where T is the sampling period, H is the number of dynamic points detected in the target area, and v b (t) is the velocity vector of the bth dynamic point at time t, E d is the environmental dynamic disturbance energy coefficient, t0 is the starting time of the sampling period;

[0115] By using the energy coefficient in integral form, we can avoid misjudgment based on instantaneous speed or position alone, and can fully reflect the long-term impact of the dynamic environment of the construction site on the stability of the point cloud, providing a reliable basis for subsequent sampling and control.

[0116] Comprehensive curvature entropy index H c and the environmental dynamic disturbance energy coefficient E d , the sampling granularity is adaptively adjusted through a dual-factor driven dynamic sampling adjustment strategy. The specific adjustment formula is as follows:

[0117]

[0118] , where ρ is the actual sampling granularity after dynamic adjustment, ρ0 is the initial sampling granularity, α is the curvature factor sensitivity coefficient, and the control curvature entropy index H c The intensity of the influence on the sampling granularity adjustment, β is the curvature entropy exponential decay coefficient, which controls the decay rate of curvature entropy on granularity adjustment, β>0, γ is the dynamic disturbance factor sensitivity coefficient, which controls the disturbance energy coefficient E d The intensity of the influence on the particle size adjustment, λ is the exponential attenuation coefficient of the dynamic disturbance, and the energy coefficient of the dynamic disturbance of the control environment E d For the decay rate of granularity adjustment, λ>0, e is the natural base.

[0119] The adjustment formula can be based on the structural regularity of the region (given by H c characterization) and environmental dynamics (by E d Characterization) Jointly control sampling granularity:

[0120] When H c and E d are all small, indicating that the target area is a flat, stable and regular facade, which can effectively reduce the sampling granularity;

[0121] When H c or E d The larger the size, the higher the sampling granularity is, ensuring that potential dynamic changes in the structure or environment are captured.

[0122] The present invention realizes intelligent, adaptive and dynamic control of the sampling granularity of laser point clouds, effectively solving the problems of point cloud data redundancy and noise pollution caused by blind high-frequency and high-density sampling in the existing technology. On the one hand, based on regular facade detection and dynamic environmental perception, the system can actively reduce the sampling resolution and sampling frequency for regular areas with stable structures and single features, reduce invalid or redundant point cloud data, and significantly reduce the storage and computing pressure of the model; on the other hand, in areas with complex structures or dynamic environments, the system can still maintain high-precision sampling to ensure the complete capture of important features. While ensuring modeling accuracy and structural feature expression capabilities, this solution achieves lightweight, precise and efficient point cloud data, significantly improving the modeling efficiency and reliability of dynamic modeling throughout the life cycle of engineering construction, and providing high-quality data support for subsequent visualization, status monitoring and intelligent decision-making.

[0123] The present invention provides Figure 2 The engineering construction full lifecycle management platform based on digital delivery shown in the figure includes a high-frequency point cloud data acquisition module, a point cloud preprocessing and data collection construction module, a regular facade feature extraction and quantification module, a regularity intelligent evaluation module, and a dynamic sampling granularity adjustment module:

[0124] The high-frequency point cloud data acquisition module uses a high-precision laser scanner to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information;

[0125] The point cloud preprocessing and data set construction module preprocesses the acquired raw point cloud data and uniformly stores the normalized multi-time point cloud data into a data set;

[0126] The regular facade feature extraction and quantification module, based on the established data set, uses feature engineering technology to extract feature dimensions from multiple dimensions that reflect the current construction site's location in the regular facade area. It then conducts a comprehensive analysis of the extracted feature dimensions to quantify the degree of regularity in the target area.

[0127] The regularity intelligent assessment module constructs the characteristic dimensions after comprehensive analysis into feature vectors and inputs them into a pre-trained machine learning model to realize intelligent assessment of the regularity of the construction site structure and generate corresponding assessment results;

[0128] The dynamic sampling granularity adjustment module dynamically adjusts the spatiotemporal sampling granularity of the laser point cloud according to the structural change characteristics and environmental dynamic characteristics of the target area when the target area is detected to be in a regular facade area. It also controls the sampling granularity to reduce the point cloud sampling density while meeting the modeling accuracy requirements.

[0129] The engineering construction full life cycle management method based on digital delivery provided in the embodiment of the present invention is realized through the above-mentioned engineering construction full life cycle management platform based on digital delivery. The specific methods and processes of the engineering construction full life cycle management platform based on digital delivery are detailed in the embodiment of the above-mentioned engineering construction full life cycle management method based on digital delivery, which will not be repeated here.

[0130] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0131] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

[0132] It should be noted that, in this document, if there are relational terms such as first and second, etc., they are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "comprising" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprising a ..." does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0133] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0134] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0135] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0136] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0137] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0138] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

[0139] The above description is merely illustrative of certain exemplary embodiments of the present invention. It goes without saying that those skilled in the art will be able to modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims.

Claims

1. A full lifecycle management method for engineering construction based on digital delivery, characterized by: The following steps are involved: Use high-precision laser scanners to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information; Preprocess the acquired original point cloud data and uniformly store the normalized multi-time point cloud data into a data set; Based on the established data set, feature engineering technology is used to extract characteristic dimensions reflecting that the current construction site is in a regular facade area from multiple dimensions. The extracted characteristic dimensions are comprehensively analyzed to quantify the degree of regularity of the target area. The feature dimensions after comprehensive analysis are constructed into feature vectors and input into a pre-trained machine learning model to achieve intelligent evaluation of the regularity of the construction site structure and generate corresponding evaluation results; When the target area is detected to be in a regular facade area, the spatiotemporal sampling granularity of the laser point cloud is dynamically adjusted according to the structural change characteristics and environmental dynamic characteristics of the target area, and the sampling granularity is controlled to reduce the point cloud sampling density while meeting the modeling accuracy requirements.

2. The engineering construction full life cycle management method based on digital delivery according to claim 1 is characterized in that: The specific steps for obtaining on-site structural data information by using a high-precision laser scanner to perform high-frequency spatiotemporal granular sampling of the construction site structure are as follows: First, based on the construction site layout, structure distribution, and monitoring requirements, the laser scanning equipment is rationally arranged to determine the scanning range, resolution, and sampling frequency. Secondly, perform on-site point cloud acquisition tasks, and the laser scanner performs a full-scale scan of the target structure surface at high temporal frequency and high spatial resolution; Subsequently, the coordinates of multiple frames of point clouds acquired at different times are unified and spliced to form a continuous set of structural point clouds with high temporal and spatial resolution. Finally, the data set is organized, stored, and relevant metadata information is synchronized, laying the foundation for subsequent feature extraction, environmental assessment, and dynamic modeling.

3. The engineering construction full life cycle management method based on digital delivery according to claim 1 is characterized in that: Feature engineering technology is used to extract feature dimensions reflecting that the current construction site is in a regular facade area from multiple dimensions. The extracted feature dimensions include the time series variance of the feature change and the overlap of the forward and reverse scanning point clouds. The time series variance of the feature change and the overlap of the forward and reverse scanning point clouds are comprehensively analyzed under the detection window to generate structural feature fluctuation reference values and bidirectional consistency reference values, respectively. The structural feature fluctuation reference values and bidirectional consistency reference values are used to quantify the degree of regularity of the target area.

4. The engineering construction full life cycle management method based on digital delivery according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the time series variance of feature changes within the detection window to generate a reference value for structural feature fluctuations are as follows: Assume that the characteristic value sequence of the structural features in the target area at continuous sampling time is f i , f i ={f1, f2, f3, ..., f n }, where f i It is the structural feature value extracted from the target area when the laser point cloud is sampled at time point i, n is the total number of time points, and the dynamic differential energy of the feature sequence is calculated. The calculation expression is as follows: , where Ef is the dynamic differential energy of the structural feature; Based on the dynamic differential energy Ef of the structural feature, the structural feature fluctuation reference value is generated. The generation formula is as follows: , where SFFR is the structural characteristic fluctuation reference value.

5. The engineering construction full life cycle management method based on digital delivery according to claim 3 is characterized in that: The specific steps for comprehensively analyzing the overlap of forward and reverse scan point clouds within the detection window to generate a bidirectional consistency reference value are as follows: Get the forward scanning point cloud set P respectively f ={p fj }, and reverse scan point cloud set P b ={p bj }, where p fj is the jth point in the forward scan point cloud, p bj is the jth point in the reverse scan point cloud. The two sets of point clouds are aligned by point cloud registration to obtain the overlapping area. In the overlapping area, the corresponding point pair set is extracted, that is, the forward point p fj The reverse point p of its nearest neighbor bj To enhance the sensitivity to local in-plane changes, the pairing set is constructed to construct a relative position residual function, which is as follows: R j =∥(p fj -p bj )·n j ∥, where R j is the normal residual of a single point pair, n j It is p fj The local normal vector at the center; After obtaining all normal residuals R j After that, a bidirectional consistency reference value is generated, and the generation formula is as follows: , where BCRV is the bidirectional consistency reference value, d j is the original Euclidean distance of the jth forward and reverse point pair, and N is the number of overlapping point pairs.

6. The engineering construction full life cycle management method based on digital delivery according to claim 3 is characterized in that: The structural characteristic fluctuation reference value and bidirectional consistency reference value after comprehensive analysis are constructed as feature vectors and input into a pre-trained machine learning model. The regularity variation coefficient is generated by the machine learning model, and the regularity of the construction site structure is intelligently evaluated by the regularity variation coefficient, and the corresponding evaluation results are generated.

7. The engineering construction full life cycle management method based on digital delivery according to claim 6 is characterized in that: The regularity variation coefficient generated by the pre-trained machine learning model during the intelligent evaluation of the regularity of the construction site structure is compared with the pre-set reference threshold of the regularity variation coefficient to divide the construction site. The division steps are as follows: If the regularity variation coefficient is less than a preset regularity variation coefficient reference threshold, the detection area is divided into a regular facade area; If the regularity variation coefficient is greater than or equal to a preset regularity variation coefficient reference threshold, the detection area is divided into an irregular facade area.

8. The engineering construction full life cycle management method based on digital delivery according to claim 1 is characterized in that: When the target area is detected to be in a regular facade area, the specific steps for controlling the sampling granularity to reduce the point cloud sampling density are as follows based on the structural change characteristics and environmental dynamic characteristics of the target area: First, after detecting that the target area belongs to a regular facade area, the curvature entropy value is calculated based on the structural change characteristics in the area to measure the complexity of the surface geometry change in the area. The calculation expression is as follows: , where H c is the curvature entropy index, p a is the proportion of points in the ath curvature interval, and M is the number of curvature intervals; Subsequently, in order to quantify the dynamic change characteristics of the environment in which the target area is located, the disturbance energy coefficient is calculated to reflect the disturbance intensity of the dynamic target in the environment on the point cloud. The calculation method is: , where T is the sampling period, H is the number of dynamic points detected in the target area, and v b (t) is the velocity vector of the bth dynamic point at time t, E d is the environmental dynamic disturbance energy coefficient, t0 is the starting time of the sampling period; Comprehensive curvature entropy index H c and the environmental dynamic disturbance energy coefficient E d , the sampling granularity is adaptively adjusted through a dual-factor driven dynamic sampling adjustment strategy. The specific adjustment formula is as follows: , where ρ is the actual sampling granularity after dynamic adjustment, ρ0 is the initial sampling granularity, α is the curvature factor sensitivity coefficient, β is the curvature entropy exponential attenuation coefficient, γ is the dynamic disturbance factor sensitivity coefficient, λ is the dynamic disturbance exponential attenuation coefficient, and e is the natural base.

9. A digitally delivered engineering construction full life cycle management platform, used to implement the digitally delivered engineering construction full life cycle management method described in any one of claims 1 to 8, characterized in that: It includes high-frequency point cloud data acquisition module, point cloud preprocessing and data set construction module, regular facade feature extraction and quantification module, regularity intelligent evaluation module and dynamic sampling granularity adjustment module: The high-frequency point cloud data acquisition module uses a high-precision laser scanner to perform high-frequency spatiotemporal granular sampling of construction site structures to obtain on-site structural data information; The point cloud preprocessing and data set construction module preprocesses the acquired raw point cloud data and uniformly stores the normalized multi-time point cloud data into a data set; The regular facade feature extraction and quantification module, based on the established data set, uses feature engineering technology to extract feature dimensions from multiple dimensions that reflect the current construction site's location in the regular facade area. It then conducts a comprehensive analysis of the extracted feature dimensions to quantify the degree of regularity in the target area. The regularity intelligent assessment module constructs the characteristic dimensions after comprehensive analysis into feature vectors and inputs them into a pre-trained machine learning model to realize intelligent assessment of the regularity of the construction site structure and generate corresponding assessment results; The dynamic sampling granularity adjustment module dynamically adjusts the spatiotemporal sampling granularity of the laser point cloud according to the structural change characteristics and environmental dynamic characteristics of the target area when the target area is detected to be in a regular facade area. It also controls the sampling granularity to reduce the point cloud sampling density while meeting the modeling accuracy requirements.

Images