A 3D simulation modeling system for construction safety training

Abstract

This invention relates to the field of simulation modeling, and more particularly to a 3D simulation modeling system for construction safety training, comprising: a scene analysis unit, used to determine the scene impact state based on the number of mobile devices and the distribution coefficient of mobile devices in the target scene, and to determine the region division method as mobile feature division or building feature division based on the scene impact state; a mobile division unit, used to acquire several device aggregation ranges and divide the aggregation ranges; an occlusion division unit, used to determine whether the dynamic occlusion range is a type of division region based on the dynamic occlusion frequency; an overlap analysis unit, used to determine the acquisition overlap rate based on the mobile characterization value, aggregation difference, or benchmark acquisition overlap rate; and an optimization adjustment unit, used to determine the optimization method based on the area of ​​the suspended region and the complexity of the surrounding area. This invention improves the usability of image acquisition data.

Description

Technical Field

[0001] This invention relates to the field of simulation modeling, and in particular to a three-dimensional simulation modeling system for construction safety training. Background Technology

[0002] With the advancement of digital twin technology, 3D interactive safety training has become widely used due to its advantages of more intuitively simulating accident scenarios and allowing trainees to learn response measures in a virtual environment, thereby improving training effectiveness. Among its key technologies, the construction of 3D scenes is crucial, and the accuracy of the construction is closely related to the realism and effectiveness of the training. However, existing safety training systems often fail to effectively reflect the realism of complex construction scenarios and suffer from insufficient accuracy when simulating them. Therefore, optimizing the scene realism of 3D modeling systems to ensure that training results closely resemble actual construction site environments is a critical technical problem that urgently needs to be solved by those skilled in the art.

[0003] Chinese Patent Publication No. CN117315160A discloses a method for 3D real-scene modeling of buildings, including: S1, acquiring original image samples through oblique shooting by a drone; calculating the offset between each sample and preset building data to obtain a k-neighborhood deviation value for each sample in the original image samples; using the k-neighborhood deviation value to adjust the point cloud data; constructing new samples together with the original image samples based on several selected samples; increasing candidate modeling data through flight path offset sampling; thereby enriching and enhancing the building's outline; S2, integrating the building's outline, optimizing the boundary modeling of the building's outline, correcting the point cloud data of the building's outline, and remodeling the outline. It is evident that the above solution has the following problems: the oblique shooting images acquired by the drone in this technical solution are taken horizontally at a constant speed along a preset flight path. When modeling large scenes with multiple buildings, it does not consider the distribution of buildings in different areas of the actual scene and the impact of different moving objects on the data acquisition effect, which easily leads to poor modeling accuracy. Summary of the Invention

[0004] To address this issue, the present invention provides a 3D simulation modeling system for construction safety training, which overcomes the problem in existing technologies that fail to consider the distribution of buildings in different areas of the actual scene and the impact of different moving objects on the data acquisition effect when modeling large scenes with multiple buildings, thus easily leading to poor modeling accuracy.

[0005] To achieve the above objectives, the present invention provides a three-dimensional simulation modeling system for construction safety training, comprising:

[0006] The scene analysis unit is used to determine the scene impact status based on the number of mobile devices and the distribution coefficient of mobile devices in the target scene, and to determine the area division method as mobile feature division or building feature division based on the scene impact status.

[0007] A mobile partitioning unit, which is connected to the scene analysis unit, is used to obtain several device aggregation ranges, and each device aggregation range is recorded as a first-class partitioned region, and the area of ​​the target scene other than the partitioned regions is recorded as a second-class partitioned region.

[0008] The occlusion segmentation unit, which is connected to the scene analysis unit, is used to execute the building segmentation method and determine whether the dynamic occlusion range is a type of segmentation area based on the dynamic occlusion frequency.

[0009] An overlap analysis unit is connected to the scene analysis unit, the motion segmentation unit, and the occlusion segmentation unit respectively, and is used to determine the acquisition overlap rate based on the motion characterization value, the clustering difference, or the baseline acquisition overlap rate.

[0010] The optimization and adjustment unit is connected to the scene analysis unit, the movement division unit, the occlusion division unit, and the overlap analysis unit, respectively, and is used to determine the optimization method based on the area of ​​the suspended region and the complexity of the surrounding area, which is to supplement the alternative model or adjust the flight altitude parameters.

[0011] Furthermore, the scene analysis unit determines the scene impact status based on the number of mobile devices in the target scene and the distribution coefficient of mobile devices, and determines the area division method based on the scene impact status;

[0012] If the scene impact state is that the number of mobile devices is greater than the preset number of mobile devices and the mobile device distribution coefficient is less than or equal to the preset mobile device distribution coefficient, then the region division method is mobile feature division.

[0013] If the scene's impact state is that the number of mobile devices is less than or equal to the preset number of mobile devices, or the mobile device distribution coefficient is greater than the preset mobile device distribution coefficient, then the area division method is building feature division.

[0014] Furthermore, in response to the first division condition, the scene analysis unit executes a device clustering analysis strategy to obtain several device clustering ranges. The device clustering range is a rectangular area, and any mobile device within the rectangular area corresponds to at least one other mobile device at a distance less than a preset distance.

[0015] Each device aggregation range is categorized as a Class I region, and the target scene area excluding the categorized regions is categorized as a Class II region.

[0016] The first division condition is that the region division method is determined to be a mobile feature division.

[0017] Furthermore, the scene analysis unit responds to the second partitioning condition and determines the building partitioning method based on the distribution equilibrium coefficient of a type of building;

[0018] If the distribution balance coefficient of a certain type of building is greater than the preset distribution balance coefficient, then the building classification method is shading classification;

[0019] If the distribution balance coefficient of a type of building is less than or equal to the preset distribution balance coefficient, the building division method is to divide according to the degree of agglomeration, and the area with the degree of agglomeration greater than the preset degree of agglomeration is recorded as the first-class division area, and the other areas are recorded as the second-class division area.

[0020] The second division condition is that the area division method is determined by architectural features.

[0021] Furthermore, the occlusion segmentation unit responds to the occlusion segmentation conditions and performs occlusion analysis for each dynamic occlusion device. In the dynamic analysis of a single dynamic occlusion device, it determines whether the dynamic occlusion range is a Class I segmentation region based on the dynamic occlusion frequency.

[0022] If the dynamic occlusion frequency is greater than the preset dynamic occlusion frequency, the dynamic occlusion range is classified as a Class I region.

[0023] If the dynamic occlusion frequency is less than or equal to the preset dynamic occlusion frequency, the dynamic occlusion range is the second-class division area;

[0024] The occlusion division condition is that the scene analysis unit determines the building division method as occlusion division.

[0025] Furthermore, in response to the mobile category classification conditions, the overlap analysis unit collects the overlap rate based on the mobile characterization value of the mobile device corresponding to the region.

[0026] The acquisition overlap rate and the motion characterization value are positively correlated;

[0027] The condition for classifying the movement category is that the scene analysis unit determines the region division method as movement feature division and the region division is completed.

[0028] Furthermore, in response to the building category classification conditions, the overlap analysis unit determines the corresponding acquisition overlap rate based on the occlusion coefficient for the classified areas obtained through occlusion classification.

[0029] For the divided regions obtained by dividing based on the degree of clustering, the acquisition overlap rate corresponding to the first type of divided regions is determined based on the difference in the degree of clustering, and the acquisition overlap rate corresponding to the second type of divided regions is set as the benchmark acquisition overlap rate.

[0030] When determining the corresponding acquisition overlap rate based on the occlusion coefficient, the acquisition overlap rate and the occlusion coefficient are positively correlated.

[0031] When determining the acquisition overlap rate corresponding to a class of partitioned regions based on the aggregation difference, the acquisition overlap rate and the aggregation difference are positively correlated.

[0032] The building category classification condition is that the scene analysis unit determines the area division method as building feature division and the area division is completed.

[0033] Furthermore, the optimization and adjustment unit determines the optimization method based on the area of ​​the suspended region and the complexity of the surrounding area;

[0034] If the area of ​​the suspended region is greater than the preset area of ​​the suspended region and the complexity of the surrounding region is greater than the preset complexity of the surrounding region, the optimization method is to supplement with an alternative model.

[0035] If the area of ​​the suspended region is less than or equal to the preset area of ​​the suspended region, or the complexity of the surrounding area is less than or equal to the preset complexity of the surrounding area, the optimization method is to adjust the flight altitude parameters.

[0036] Furthermore, the first optimization condition of the unit response is optimized and adjusted. The model matching coefficient is determined based on the position accuracy influence coefficient and the appearance morphology influence coefficient, and the model with the largest model matching coefficient is recorded as the supplementary model.

[0037] The first optimization condition is that the optimization method is determined to be a supplementary alternative model.

[0038] Furthermore, the second optimization condition of the unit response is optimized and adjusted, and the flight altitude of a class of regions is determined based on the effective difference;

[0039] The flight altitude and the effective difference are positively correlated;

[0040] The second optimization condition is that the optimization method is determined to be the adjustment of flight altitude parameters.

[0041] Compared with the prior art, the beneficial effects of the present invention are as follows: the technical solution of the present invention determines the scene influence state based on the number of mobile devices and the distribution coefficient of mobile devices in the target scene, reflects the distribution of moving objects in the area through the number of mobile devices and the distribution coefficient of mobile devices, and determines the area division method as either mobile feature division or building feature division based on the scene influence state. The present invention takes into account the influence of moving objects in the actual scene and selects different area division methods accordingly to make the selection of area division methods more in line with the actual application scenario, avoiding the problems of low data collection efficiency and insufficient accuracy caused by the single area division method in the prior art, thereby improving the accuracy of subsequent analysis of the present invention and further ensuring the modeling quality.

[0042] Furthermore, in this invention, the building division method is determined based on the distribution balance coefficient of a type of building, either by the occlusion coefficient or by the clustering degree. The distribution balance coefficient reflects the distribution status of a type of building within the target area and corresponds to the two division methods. This avoids the problem that a single area division method is difficult to meet the actual situation and causes deviations between image acquisition and the actual situation, thereby improving the usability of image acquisition data.

[0043] Furthermore, this invention employs different methods for confirming the overlap rate of data collection for different regional divisions, avoiding the inability of a single method to meet actual conditions. The regional division makes the determination of the overlap rate more targeted, making the overlap rate more consistent with actual application scenarios, ensuring effective data collection in complex scenarios, and thus guaranteeing the accuracy of the 3D model.

[0044] Furthermore, in this invention, the optimization method is determined based on the area of ​​the suspended region and the complexity of the surrounding area. Since trees are prone to suspension due to factors such as leaf movement and lighting during the oblique photography process of the UAV, the accuracy and realism of image acquisition are affected, leading to a decrease in modeling accuracy and model quality. Therefore, the optimization method is determined based on the area of ​​the suspended region and the complexity of the surrounding area to improve the environmental adaptability of the model and avoid the problem of poor image acquisition effect and poor system accuracy caused by a single optimization method, thereby improving the accuracy of 3D modeling. Attached Figure Description

[0045] Figure 1 This is a unit connection diagram of the 3D simulation modeling system for construction safety training according to the present invention;

[0046] Figure 2 This is a flowchart illustrating the method for determining the region division based on the number of mobile devices and the distribution coefficient of mobile devices in the target scene according to the present invention.

[0047] Figure 3 This is a flowchart illustrating how the building classification method is determined based on the distribution equilibrium coefficient of a type of building, according to the present invention.

[0048] Figure 4 This is a flowchart illustrating the process of determining whether a dynamic occlusion range falls under a specific region based on the dynamic occlusion frequency, according to the present invention. Detailed Implementation

[0049] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0050] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0051] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0052] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0053] Please see Figures 1 to 4 As shown, the present invention provides a three-dimensional simulation modeling system for construction safety training, comprising:

[0054] The scene analysis unit is used to determine the scene impact status based on the number of mobile devices and the distribution coefficient of mobile devices in the target scene, and to determine the area division method as mobile feature division or building feature division based on the scene impact status.

[0055] A mobile partitioning unit, which is connected to the scene analysis unit, is used to obtain several device aggregation ranges, and each device aggregation range is recorded as a first-class partitioned region, and the area of ​​the target scene other than the partitioned regions is recorded as a second-class partitioned region.

[0056] The occlusion segmentation unit, which is connected to the scene analysis unit, is used to execute the building segmentation method and determine whether the dynamic occlusion range is a type of segmentation area based on the dynamic occlusion frequency.

[0057] An overlap analysis unit is connected to the scene analysis unit, the motion segmentation unit, and the occlusion segmentation unit respectively, and is used to determine the acquisition overlap rate based on the motion characterization value, the clustering difference, or the baseline acquisition overlap rate.

[0058] The optimization and adjustment unit is connected to the scene analysis unit, the movement division unit, the occlusion division unit, and the overlap analysis unit, respectively, and is used to determine the optimization method based on the area of ​​the suspended region and the complexity of the surrounding area, which is to supplement the alternative model or adjust the flight altitude parameters.

[0059] The application scenario of this invention is to conduct construction safety training using a 3D simulation modeling system, wherein the 3D simulation modeling process includes:

[0060] Images of the actual construction site were obtained using oblique photography by drones.

[0061] Filter out invalid images (blurred, occluded, etc.) and perform distortion correction and color equalization.

[0062] ContextCapture is used to perform joint adjustment of multi-view images, calculate the exterior orientation elements of each image, and generate sparse point clouds.

[0063] High-density point clouds containing the three-dimensional coordinates of ground features are generated through multi-view stereo matching (such as the SGM algorithm).

[0064] The point cloud is filtered, classified (ground, buildings, vegetation, etc.), and a TIN (triangular mesh irregularity) model is constructed.

[0065] The original image color information is mapped onto a mesh model to generate a 3D real-world model with realistic textures.

[0066] Images of the actual construction site were obtained through oblique photography by drones, and the image data of the corresponding construction site images were used for 3D simulation modeling. The scene captured by the drones was the construction site, and the construction site was recorded as the target scene.

[0067] This invention establishes target judgment values ​​and relevant reference values. The correspondence between the target judgment values ​​and relevant reference values ​​is expressed through a calculation formula that includes an adjustment coefficient and a baseline judgment value. The calculation formula is: Target Judgment Value = Baseline Judgment Value + Relevant Reference Value × Adjustment Coefficient. Specifically, the target judgment values ​​in this invention include flight altitude and acquisition overlap rate, and the relevant reference values ​​include effective difference, occlusion coefficient, clustering difference, and motion characterization value. It is understood that each target judgment value has one or more corresponding relevant reference values. For example, when determining the corresponding acquisition overlap rate based on the occlusion coefficient, the acquisition overlap rate and the occlusion coefficient are positively correlated; or, when determining the acquisition overlap rate corresponding to a class of segmented regions based on the clustering difference, the acquisition overlap rate and the clustering difference are positively correlated. Furthermore, each target judgment value corresponds to... There should be a baseline judgment value, which is set by the user. For example, the baseline judgment value corresponding to the acquisition overlap rate can be obtained by extracting the average value of the acquisition overlap rate corresponding to the historical records that meet the user's needs. The baseline judgment value corresponding to the flight altitude is similar and will not be elaborated here. Each relevant reference value has a separate adjustment coefficient, which is set by the user. For example, the value of the adjustment coefficient corresponding to the occlusion coefficient can be set according to the user's needs. The user can determine the acquisition overlap rate corresponding to different occlusion coefficients based on historical records, thereby determining the degree of influence of the occlusion coefficient on the acquisition overlap rate, and thus determining the adjustment coefficient. Different adjustment coefficients are selected accordingly. The greater the degree of influence, the larger the adjustment coefficient. The adjustment coefficient is positive when there is a positive correlation and negative when there is a negative correlation.

[0068] This invention includes several historical records, each of which includes at least the number of mobile devices, the distribution coefficient of mobile devices, the distance between mobile devices within the clustered area, the reflective surface area, the regularity of the shape, the building characterization coefficient, the distribution balance coefficient, the clustering degree, the dynamic occlusion rate, the area of ​​the floating region, the complexity of the surrounding area, the weighting coefficient, and the acquisition overlap rate. Each historical record is also marked with a pass / fail flag, indicating whether the historical record meets the user's requirements. Whether a historical record meets the user's requirements can be determined, but is not limited to, based on the accuracy of the final constructed model. Determining whether a historical record meets the user's requirements based on the user-defined model accuracy index is a concept already understood by those skilled in the art and is not limited here.

[0069] Specifically, the scene analysis unit determines the scene impact status based on the number of mobile devices in the target scene and the distribution coefficient of mobile devices, and determines the area division method based on the scene impact status;

[0070] If the scene impact state is that the number of mobile devices is greater than the preset number of mobile devices and the mobile device distribution coefficient is less than or equal to the preset mobile device distribution coefficient, then the region division method is mobile feature division.

[0071] If the scene's impact state is that the number of mobile devices is less than or equal to the preset number of mobile devices, or the mobile device distribution coefficient is greater than the preset mobile device distribution coefficient, then the area division method is building feature division.

[0072] Mobile devices are work devices that can move within the target scene. Work devices include, but are not limited to, tower cranes and dump trucks. The distribution coefficient of mobile devices is the average of the reference distances corresponding to all mobile devices. For a single mobile device, the mobile device is recorded as the target mobile device, and other mobile devices other than the target mobile device are recorded as reference mobile devices. The shortest distance from the target mobile device to each reference mobile device is detected, and the average of each shortest distance is recorded as the reference distance.

[0073] The user can determine the preset number of mobile devices and the preset mobile device distribution coefficient based on the actual application scenario. It is understood that the size, speed, and reflective materials on the surface of mobile devices can easily affect the model construction. The lower the user's tolerance for the influence of mobile devices, the smaller the preset number of mobile devices and the larger the preset mobile device distribution coefficient. This paper provides a preset number of mobile devices and the mobile device distribution coefficient. The corresponding number of mobile devices and the mobile device distribution coefficient in the historical records that meet the user's needs are extracted. Outliers are filtered out, and the average values ​​of the number of mobile devices and the mobile device distribution coefficient after removing outliers are recorded as the preset number of mobile devices and the preset mobile device distribution coefficient, respectively. The outlier filtering method can be, but is not limited to, the 3σ criterion method or the IQR method.

[0074] Specifically, the scene analysis unit responds to the first division condition and executes the device clustering analysis strategy to obtain several device clustering ranges. The device clustering range is a rectangular area, and any mobile device within the rectangular area corresponds to at least one other mobile device at a distance less than a preset distance.

[0075] Each device aggregation range is categorized as a Class I region, and the target scene area excluding the categorized regions is categorized as a Class II region.

[0076] The first division condition is that the region division method is determined to be a mobile feature division.

[0077] The device clustering analysis strategy involves performing several clustering analyses to obtain several device clustering ranges. A single clustering analysis is performed by taking the location of any mobile device not included in the device clustering range as the starting detection point and performing distance detection. The distance detection involves detecting the distance between the starting detection point and other mobile devices. Other mobile devices with a distance less than a preset distance are then marked as starting detection points again, and the distance detection continues until there are no other mobile devices with a distance less than the preset distance corresponding to the current starting detection point. The single clustering analysis ends, and the smallest rectangle that can include the mobile devices corresponding to each distance detection in this clustering analysis is marked as a device clustering range.

[0078] The preset distance can be set adaptively by the user according to the actual application needs. It can be understood that the present invention divides the area with dense mobile devices in the target scene through the device clustering analysis strategy. The greater the user's tolerance for the density of mobile devices, the smaller the preset distance value. The present invention provides a method for setting the preset distance, which extracts the corresponding distance in the historical records that meet the user's needs, filters out the outliers through machine learning methods, and records the average value of the distance after removing outliers as the preset distance.

[0079] Specifically, the scene analysis unit responds to the second division condition and determines the building division method based on the distribution balance coefficient of a type of building;

[0080] If the distribution balance coefficient of a certain type of building is greater than the preset distribution balance coefficient, then the building classification method is shading classification;

[0081] If the distribution balance coefficient of a type of building is less than or equal to the preset distribution balance coefficient, the building division method is to divide according to the degree of agglomeration, and the area with the degree of agglomeration greater than the preset degree of agglomeration is recorded as the first-class division area, and the other areas are recorded as the second-class division area.

[0082] The second division condition is that the area division method is determined by architectural features.

[0083] Buildings with a building characterization coefficient greater than the preset building characterization coefficient are classified as a type of building, where the building characterization coefficient is the structural characteristic degree of the building;

[0084] Structural characteristic degree of a building = Reflective surface area / Preset reflective surface area + Shape regularity / Preset shape regularity;

[0085] The reflective material areas are marked on the 3D model of the building surface, and the reflective surfaces are identified and their areas are calculated using Rhino+Grasshopper. The Rhino+Grasshopper is a concept that is easily understood by those skilled in the art and will not be elaborated upon here.

[0086] The method for determining the regularity of the shape is to convert the building outline into a frequency domain signal using Fourier descriptors. The regularity of the building's shape = low-frequency signal / total frequency domain signal. The higher the proportion of low-frequency signal, the more regular the shape.

[0087] Users can adaptively set the preset reflective surface area and preset shape regularity values ​​according to actual application scenarios. It is understood that the greater the influence of the reflective surface area and shape regularity of a building on the structural characteristics of the building, the smaller the preset reflective surface area value and the preset shape regularity value will be. This invention provides a method for setting the preset reflective surface area and preset shape regularity values, which extracts the corresponding reflective surface area and shape regularity values ​​from the historical records that meet the user's needs, filters out outliers, and records the average values ​​of the reflective surface area and shape regularity after removing outliers as the preset reflective surface area and preset shape regularity, respectively.

[0088] The preset building representation coefficient can be set by the user according to actual needs. It can be understood that the larger the building representation coefficient, the more critical the structural features of the building are to the modeling effect. Therefore, the greater the user's need for analysis of the structural features of the building, the smaller the preset building representation coefficient should be. This invention provides a method for setting the preset building representation coefficient by extracting the corresponding building representation coefficients from the historical records that meet the user's needs, filtering out outliers, and recording the building representation coefficients after removing outliers as the preset building representation coefficients.

[0089] The distribution balance coefficient is the actual average nearest neighbor distance. The actual average nearest neighbor distance is determined by performing distance analysis on each type of building to obtain the nearest neighbor distance for each type of building. The average of the nearest neighbor distances of all types of buildings is recorded as the actual average nearest neighbor distance. When performing distance analysis on a single type of building, the type of building is recorded as the target type of building, and all other types of buildings are recorded as reference types of buildings. The distance between the target type of building and each reference type of building is detected, and the minimum distance is recorded as the nearest neighbor distance.

[0090] The user can adaptively set the value of the preset distribution balance coefficient according to the actual application scenario. It can be understood that the higher the user's requirement for the accuracy of the building division method, the smaller the preset distribution balance coefficient will be. This invention provides a method for setting the value of the preset distribution balance coefficient, which extracts the corresponding distribution balance coefficient in the historical records that meet the user's needs, filters out the outliers, and records the average value of the distribution balance coefficient after removing the outliers as the preset distribution balance coefficient.

[0091] When dividing based on clustering degree, the first step is to obtain a cluster of buildings of a certain type. The method for identifying a cluster of buildings of a certain type is to divide the target scene evenly to obtain several rectangular areas of the same area. The number of buildings of a certain type corresponding to any rectangular area is the clustering degree of that rectangular area. The value of the preset clustering degree can be adaptively set by the user according to the actual application needs. It can be understood that the higher the user's tolerance for the impact of clustering degree on model accuracy, the smaller the value of the preset clustering degree will be. This invention provides a method for setting the value of the preset clustering degree by extracting the corresponding clustering degree in the historical records that meet the user's needs, filtering out outliers, and recording the average value of the clustering degree after removing outliers as the preset clustering degree.

[0092] Specifically, the occlusion segmentation unit responds to the occlusion segmentation conditions and performs occlusion analysis for each dynamic occlusion device. In the dynamic analysis of a single dynamic occlusion device, it determines whether the dynamic occlusion range is a class of segmented regions based on the dynamic occlusion frequency.

[0093] If the dynamic occlusion frequency is greater than the preset dynamic occlusion frequency, the dynamic occlusion range is classified as a Class I region.

[0094] If the dynamic occlusion frequency is less than or equal to the preset dynamic occlusion frequency, the dynamic occlusion range is the second-class division area;

[0095] The occlusion division condition is that the scene analysis unit determines the building division method as occlusion division.

[0096] The dynamic occlusion device is a mobile device that is in motion and obstructs buildings. This invention applies a cyclical detection cycle. The duration of a single detection cycle can be set by the user according to actual needs. It can be understood that the higher the user's requirement for the speed of dynamic occlusion frequency determination, the shorter the detection cycle duration. In this invention, the duration of a single detection cycle is 5 minutes.

[0097] The method for confirming the dynamic occlusion frequency of any dynamic occlusion device is to detect the total duration of the dynamic occlusion device occluding each building within the most recent complete detection cycle, which is recorded as the occlusion duration. Dynamic occlusion frequency = occlusion duration / duration of a single detection cycle.

[0098] The preset dynamic occlusion rate can be adaptively set by the user according to the actual application scenario. It is understood that the dynamic occlusion rate affects the accuracy of the 3D model. The higher the dynamic occlusion rate, the lower the model accuracy. Therefore, the smaller the user's tolerance for the impact of the dynamic occlusion rate, the larger the preset dynamic occlusion rate will be. This invention provides a method for setting the preset dynamic occlusion rate by extracting the corresponding dynamic occlusion rate from the historical records that meet the user's needs, filtering out outliers, and recording the average value of the dynamic occlusion rate after removing outliers as the preset dynamic occlusion rate.

[0099] Specifically, the overlap analysis unit responds to the mobile category classification conditions and collects the overlap rate based on the mobile characterization value of the mobile device corresponding to the region.

[0100] The acquisition overlap rate and the motion characterization value are positively correlated;

[0101] The condition for classifying the movement category is that the scene analysis unit determines the region division method as movement feature division and the region division is completed.

[0102] Mobility characterization value = number of mobile devices × speed of mobile devices. The mobility characterization value has no unit. Among them, the number of mobile devices is the total number of mobile devices in the area, and the speed of mobile devices is the average speed of each type of mobile device in the area.

[0103] Specifically, in response to the building category classification conditions, the overlap analysis unit determines the corresponding acquisition overlap rate based on the occlusion coefficient for the classified areas obtained through occlusion classification.

[0104] For the divided regions obtained by dividing based on the degree of clustering, the acquisition overlap rate corresponding to the first type of divided regions is determined based on the difference in the degree of clustering, and the acquisition overlap rate corresponding to the second type of divided regions is set as the benchmark acquisition overlap rate.

[0105] When determining the corresponding acquisition overlap rate based on the occlusion coefficient, the acquisition overlap rate and the occlusion coefficient are positively correlated.

[0106] When determining the acquisition overlap rate corresponding to a class of partitioned regions based on the aggregation difference, the acquisition overlap rate and the aggregation difference are positively correlated.

[0107] The building category classification condition is that the scene analysis unit determines the area division method as building feature division and the area division is completed.

[0108] In actual construction sites, during drone photography, surrounding buildings, trees, and mobile equipment (such as cranes) can easily obstruct buildings. Therefore, the occlusion coefficient = the sum of the obscured surface areas of each building / the sum of the complete surface areas of each building. It can be understood that 0 < occlusion coefficient < 1, and the aggregation difference = aggregation degree - preset aggregation degree.

[0109] The user can adaptively set the baseline acquisition overlap rate according to the actual application scenario. It is understood that terrain complexity, modeling accuracy requirements, and flight altitude will all affect the acquisition overlap rate. Among them, the higher the terrain complexity, the greater the acquisition overlap rate. This invention provides a method for determining the baseline acquisition overlap rate by extracting the corresponding acquisition overlap rate from the historical records that meet the user's needs, filtering out outliers, and recording the average acquisition overlap rate after removing outliers as the baseline acquisition overlap rate. This invention provides a value for the baseline acquisition overlap rate, with a heading baseline acquisition overlap rate of 80% and a lateral baseline acquisition overlap rate of 70%.

[0110] Specifically, the optimization adjustment unit response model construction conditions are determined based on the area of ​​the suspended region and the complexity of the surrounding area;

[0111] If the area of ​​the suspended region is greater than the preset area of ​​the suspended region and the complexity of the surrounding region is greater than the preset complexity of the surrounding region, the optimization method is to supplement with an alternative model.

[0112] If the area of ​​the suspended region is less than or equal to the preset area of ​​the suspended region, or the complexity of the surrounding area is less than or equal to the preset complexity of the surrounding area, the optimization method is to adjust the flight altitude parameters.

[0113] The model is built upon the completion of the 3D reality model.

[0114] Calculate the point cloud density within the range of the bottom height of each tree in the 3D reality model that is less than 0.5m. If it is less than the density of the edge ground, the smallest 3D rectangle that can contain the tree is recorded as the floating region. If there is no floating region, there is no need to determine the optimization method.

[0115] For any suspended region, its corresponding enclosure is a circular area that can include the suspended region under top-down conditions, and the area of ​​the circular area is larger than the area of ​​the suspended region. The user can adaptively set the value of the area of ​​the circular area according to the actual application scenario. It can be understood that the larger the area of ​​the circular area, the larger the corresponding enclosure range. The present invention provides a value for the area of ​​the circular area: Circular area = 1.2 × Suspended region area.

[0116] Enclosure complexity = number of sub-regions / n + reflectivity factor; divide the enclosure into n equal-area sector regions, and denote the sector regions with relevant points as sub-regions. The relevant point is any point selected on the building within the sector region. The reflectivity factor is the sum of the reflective areas of the corresponding buildings within the sub-region / the sum of the surface areas of the corresponding buildings within the enclosure region.

[0117] The method for setting the preset floating area area and preset surrounding complexity allows users to adapt them to the actual scenario. It is understood that the larger the floating area area and the greater the surrounding complexity, the greater the difficulty of drone data collection and the lower the drone data collection efficiency. Therefore, the greater the user's demand for drone data collection efficiency, the smaller the preset floating area area and the smaller the preset surrounding complexity. This invention provides a method for setting the preset floating area area and preset surrounding complexity by extracting the corresponding floating area area and surrounding complexity from the historical records that meet the user's needs, filtering out outliers using machine learning methods, and recording the average values ​​of the floating area area and surrounding complexity after removing outliers as the preset floating area area and preset surrounding complexity, respectively.

[0118] Specifically, the first optimization condition of the unit response is optimized and adjusted. The model matching coefficient is determined based on the position accuracy influence coefficient and the appearance morphology influence coefficient, and the model with the largest model matching coefficient is recorded as the supplementary model.

[0119] The first optimization condition is that the optimization method is determined to be a supplementary alternative model.

[0120] Model matching coefficient = α1 × positional accuracy influence coefficient + α2 × appearance morphology influence coefficient; where α1 is the first weight coefficient and α2 is the second weight coefficient. The values ​​of α1 and α2 can be set directly by the user based on domain experience, or by using statistical methods such as regression analysis or principal component analysis (PCA) to determine the contribution of the positional accuracy influence coefficient and the appearance morphology influence coefficient to the model matching coefficient, thereby determining the corresponding weight coefficient values. The greater the contribution, the larger the weight coefficient value. The weights can also be adjusted through historical training (such as machine learning).

[0121] The positional accuracy influence coefficient is determined by detecting the minimum distance between the relevant points on the buildings in each sub-region and the center point of the suspended area. The positional accuracy influence coefficient = 1 / minimum distance. The apparent morphology influence coefficient is determined by detecting the vertical projection area of ​​the tree canopy on the ground and the vertical distance between the highest point of the tree canopy and the ground. The vertical projection area is recorded as the tree canopy area.

[0122] Specifically, the unit response is optimized and adjusted according to the second optimization condition, and the flight altitude of a class of regions is determined based on the effective difference.

[0123] The flight altitude and the effective difference are positively correlated;

[0124] The second optimization condition is that the optimization method is determined to be the adjustment of flight altitude parameters.

[0125] The effective difference ΔH = max(H1, H2), where H is the effective difference, H1 = the area of ​​the preset floating region - the area of ​​the floating region, and H2 = the complexity of the preset domain - the complexity of the domain.

[0126] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0127] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A three-dimensional simulation modeling system for construction safety training, characterized by, The application comprises: a scene analysis unit, configured to determine a scene influence state according to the number of mobile devices in a target scene and a mobile device distribution coefficient, and determine a region division manner as mobile feature division or building feature division according to the scene influence state; a mobile division unit, connected to the scene analysis unit, configured to obtain a plurality of device aggregation ranges, correspondingly record each device aggregation range as a first division region, and record a region of the target scene other than the division region as a second division region; a shelter division unit, connected to the scene analysis unit, configured to execute a building division manner, and determine whether a dynamic shelter range is a first division region according to a dynamic shelter frequency; an overlap analysis unit, connected to the scene analysis unit, the mobile division unit and the shelter division unit, configured to determine a collection overlap rate according to a mobile representation value, an aggregation degree difference value, a reference collection overlap rate or a shelter coefficient; an optimization adjustment unit, connected to the scene analysis unit, the mobile division unit, the shelter division unit and the overlap analysis unit, configured to determine an optimization manner as a replacement model supplement or a flight height parameter adjustment according to a suspension region area and a surrounding area complexity; the scene analysis unit is configured to determine a building division manner according to a distribution balance coefficient of a first building in response to a second division condition; if the distribution balance coefficient of the first building is greater than a preset distribution balance coefficient, the building division manner is shelter division; if the distribution balance coefficient of the first building is less than or equal to the preset distribution balance coefficient, the building division manner is division according to an aggregation degree, and a region with an aggregation degree greater than a preset aggregation degree is recorded as a first division region, and other regions other than the first division region are recorded as second division regions; the second division condition is that the region division manner is building feature division; a building with a building representation coefficient greater than a preset building representation coefficient is recorded as a first building; the overlap analysis unit is configured to determine a collection overlap rate according to a mobile representation value of a mobile device corresponding to a region in response to a mobile category division condition; the collection overlap rate and the mobile representation value are in a positive correlation relationship; the mobile category division condition is that the scene analysis unit determines that the region division manner is mobile feature division and the region division is completed; the overlap analysis unit is configured to determine a corresponding collection overlap rate based on a shelter coefficient for a divided region obtained through shelter division in response to a building category division condition; for a divided region obtained through division according to an aggregation degree, a collection overlap rate corresponding to a first division region is determined based on an aggregation degree difference value, and a collection overlap rate corresponding to a second division region is set as a reference collection overlap rate; when the corresponding collection overlap rate is determined based on the shelter coefficient, the collection overlap rate and the shelter coefficient are in a positive correlation relationship; when the corresponding collection overlap rate of the first division region is determined based on the aggregation degree difference value, the collection overlap rate and the aggregation degree difference value are in a positive correlation relationship; the building category division condition is that the scene analysis unit determines that the region division manner is building feature division and the region division is completed.

2. The three-dimensional simulation modeling system for construction safety training of claim 1, wherein, The scene analysis unit determines a scene influence state according to the number of mobile devices in the target scene and a mobile device distribution coefficient, and determines a region division manner according to the scene influence state; if the scene influence state is that the number of mobile devices is greater than a preset number of mobile devices and the mobile device distribution coefficient is less than or equal to a preset mobile device distribution coefficient, the region division manner is mobile feature division; if the scene influence state is that the number of mobile devices is less than or equal to a preset number of mobile devices or the mobile device distribution coefficient is greater than a preset mobile device distribution coefficient, the region division manner is building feature division.

3. The three-dimensional simulation modeling system for construction safety training of claim 2, wherein, The mobile division unit executes a device aggregation analysis strategy to obtain a plurality of device aggregation ranges in response to a first division condition, the device aggregation range is a rectangular region, and any mobile device in the rectangular region corresponds to at least one other mobile device with a distance less than a preset distance; each device aggregation range is recorded as a division region of a first type, and a region of the target scene other than the division region is recorded as a division region of a second type; wherein the first division condition is that the region division manner is determined to be mobile feature division.

4. The three-dimensional simulation modeling system for construction safety training of claim 1, wherein, The occlusion division unit performs occlusion analysis on each dynamic occlusion device in response to an occlusion division condition, wherein the dynamic analysis for a single dynamic occlusion device determines whether the dynamic occlusion range is a division region of a first type according to a dynamic occlusion frequency; if the dynamic occlusion frequency is greater than a preset dynamic occlusion frequency, the dynamic occlusion range is a division region of a first type; if the dynamic occlusion frequency is less than or equal to a preset dynamic occlusion frequency, the dynamic occlusion range is a division region of a second type; wherein the occlusion division condition is that the scene analysis unit determines the building division manner to be occlusion division.

5. The three-dimensional simulation modeling system for construction safety training of claim 1, wherein, The optimization adjustment unit determines an optimization manner according to a floating region area and a surrounding area complexity; if the floating region area is greater than a preset floating region area and the surrounding area complexity is greater than a preset surrounding area complexity, the optimization manner is replacement model supplement; if the floating region area is less than or equal to a preset floating region area or the surrounding area complexity is less than or equal to a preset surrounding area complexity, the optimization manner is flight height parameter adjustment.

6. The three-dimensional simulation modeling system for construction safety training of claim 5, wherein, The optimization adjustment unit determines a model matching coefficient according to a position accuracy influence coefficient and an apparent form influence coefficient in response to a first optimization condition, and records the model with the largest model matching coefficient as a supplement model; wherein the first optimization condition is that the optimization manner is determined to be replacement model supplement.

7. The three-dimensional simulation modeling system for construction safety training of claim 5, wherein, The optimization adjustment unit determines a flight height of a division region of a first type according to an effective difference in response to a second optimization condition; the flight height and the effective difference are in a positive correlation; wherein the second optimization condition is that the optimization manner is determined to be flight height parameter adjustment.

Images