Road infrastructure bim model reverse construction method based on data fusion

Abstract

The application discloses a road infrastructure BIM model reverse construction method based on data fusion, and relates to the technical field of architectural design. The method comprises the following steps: acquiring first point cloud data of multiple road regions and their infrastructures; performing segmentation processing on the first point cloud data to obtain multiple first road sections; performing local optimal matching on point cloud data in the first road sections and standard point cloud data of the infrastructures based on a random sample consensus algorithm to obtain target point cloud data; screening the target point cloud data based on the recurrence degree of each target point cloud data at different road positions to perform denoising processing on the target point cloud data; and performing semantic segmentation and instance extraction processing on the denoised target point cloud data through a deep learning algorithm to construct a target road model containing different infrastructures. The application makes the road facilities in the constructed target road model more in line with actual road conditions.

Description

Technical Field

[0001] This application relates to the field of architectural design technology, and in particular to a method for reverse engineering a BIM model of road infrastructure based on data fusion. Background Technology

[0002] Road infrastructure refers to a series of fixed facilities that provide basic conditions for the passage of vehicles and pedestrians. Reverse modeling (Scan-to-BIM) technology digitally restores existing urban road facilities (such as roads, bridges, pipelines, etc.) to form a building information model (BIM) containing accurate geometric information and rich attribute information. It is of great significance for the operation and maintenance management of existing facilities, the planning and design of renovation and expansion, and the construction of smart cities.

[0003] Scan-to-BIM technology involves scanning existing urban road infrastructure into point clouds, performing initial denoising and registration on the point cloud data, and then using deep learning algorithms to segment the point cloud data into different road facilities, such as bridges, fences, and pipelines. However, the actual point cloud scanning process is affected by the non-smooth surfaces and structural errors of real-world infrastructure, such as infill particles on the road surface. This results in noise in the scanned 3D point cloud data, causing the road facilities segmented by the deep learning algorithm to differ from the actual infrastructure. Summary of the Invention

[0004] The main purpose of this application is to provide a method for reverse construction of BIM models of road infrastructure based on data fusion. This method aims to solve the technical problem in related technologies where, after scanning the point cloud of existing urban road infrastructure, the noise generated by the filling particles on the road surface cannot be removed by the preliminary denoising method, resulting in the road facilities divided by the deep learning algorithm not matching the actual situation.

[0005] To achieve the above objectives, embodiments of this application provide a method for reverse engineering a BIM model of road infrastructure based on data fusion, including:

[0006] Acquire first-point cloud data of multiple road areas and their infrastructure;

[0007] The first point cloud data is segmented to obtain multiple first road segments;

[0008] Based on the random sampling consensus algorithm, the point cloud data in the first road segment is locally optimally matched with the standard point cloud data of the infrastructure to obtain the target point cloud data.

[0009] Based on the reproducibility of each target point cloud data at different road locations, the target point cloud data is filtered to perform noise reduction processing.

[0010] The target road model containing different infrastructures is constructed by performing semantic segmentation and instance extraction on the denoised target point cloud data through a deep learning algorithm.

[0011] In a possible implementation of the present application, the first point cloud data is segmented to obtain a plurality of first road segments, including:

[0012] A vertical plane perpendicular to the road direction is taken as a road section of each road region along the corresponding road direction of the road region;

[0013] The number of point clouds of the first point cloud data corresponding to the data collection position in each road region is determined, and the number of point clouds is taken as the section density of each road section;

[0014] The section density is clustered to obtain a plurality of first road segments.

[0015] In a possible implementation of the present application, the section density is clustered to obtain a plurality of first road segments, including:

[0016] The density mean of the section density corresponding to all road regions is calculated, and the density mean is taken as a clustering threshold;

[0017] If the difference between the section density mean of any two clustering clusters is greater than the clustering threshold, then the two clustering clusters are determined to be divided into two first road segments, and until the clustering of all road regions is completed, a plurality of first road segments are obtained.

[0018] In a possible implementation of the present application, based on the random sample consensus algorithm, the point cloud data in the first road segment and the standard point cloud data of the infrastructure are locally optimally matched to obtain target point cloud data, including:

[0019] A preset number of second point cloud data is extracted from each first point cloud data through the random sample consensus algorithm;

[0020] The distance between the second point cloud data and the nearest point cloud data in the standard point cloud data corresponding to all infrastructures is set as the shortest matching distance of the second point cloud data;

[0021] The shortest matching distances corresponding to all second point cloud data are sorted in descending order to obtain a point cloud distance sorting sequence, and an optimal matching sequence segment in the point cloud distance sorting sequence is selected;

[0022] Based on the point cloud data and the shortest matching distance corresponding to the optimal matching sequence segment, the reliability of the point cloud data of the current iteration number is calculated;

[0023] When the reliability is greater than a preset threshold, the point cloud data of the current iteration number is taken as the target point cloud data.

[0024] In a possible implementation of the present application, the optimal matching sequence segment in the point cloud distance sorting sequence is selected, including:

[0025] The distance difference between the two shortest matching distances of the point cloud distance sorting sequence is calculated;

[0026] The point cloud distance sorting sequence is divided based on the position corresponding to the maximum value of the distance difference, to obtain a first sequence segment and a second sequence segment;

[0027] The sequence segment part corresponding to the minimum value of the distance average in the first sequence segment and the second sequence segment is taken as the optimal matching sequence segment.

[0028] In a possible implementation of the present application, the reliability of the point cloud data of the current iteration number is calculated based on the point cloud data corresponding to the optimal matching sequence segment and the shortest matching distance, including:

[0029] The first proportion of the point cloud data corresponding to the optimal matching sequence segment and the first average value of each shortest matching distance are determined;

[0030] The reliability of the point cloud data of the current iteration number is calculated based on the ratio between the first proportion and the first average value.

[0031] In a possible implementation of the present application, the target point cloud data is filtered based on the recurrence degree of each target point cloud data at different road positions, including:

[0032] Along the road direction corresponding to the road area, a slice plane at a plurality of road positions in each road area is constructed;

[0033] Based on the distance between the target point cloud data in different slice planes, the suspected recurrent point cloud data of each target point cloud data is determined;

[0034] Based on the number of suspected recurrent point cloud data and the number of slice planes, the recurrence degree of each target point cloud data at different road positions is calculated;

[0035] The spatial distance between each target point cloud data and the fitting model constructed by the random sample consensus algorithm at the current iteration number is determined;

[0036] The spatial distance is weakened based on the recurrence degree, to obtain a weakened distance;

[0037] Based on the weakened distance, the target point cloud data is divided into inliers and outliers, and the divided inliers are retained.

[0038] In a possible implementation of the present application, the suspected recurring point cloud data of each target point cloud data is determined based on the distance between the target point cloud data in different slice planes, comprising:

[0039] For any slice plane, the target point cloud data between the current slice plane and the next adjacent slice plane is projected onto the current slice plane until there is point cloud data on all slice planes;

[0040] The point cloud data of each slice plane in each road region is projected onto the first slice plane with the road center as the origin;

[0041] For any target point cloud data in the first slice plane, the point cloud data in the preset adjacent region of the current target point cloud data is regarded as the suspected recurring point cloud data of the current target point cloud data.

[0042] In a possible implementation of the present application, the recurrence degree of each target point cloud data at different road positions is calculated based on the number of suspected recurring point cloud data, the number of slice planes, comprising:

[0043] All suspected recurring point cloud data of different target point cloud data are arranged in order according to the road direction, and the number of plane intervals between any two adjacent suspected recurring point cloud data of each target point cloud data is determined;

[0044] The average distance between each target point cloud data and its corresponding suspected recurring point cloud data is calculated;

[0045] Based on the ratio between the average distance and the number of plane intervals, and the ratio between the number of suspected recurring point cloud data and the number of slice planes, the recurrence degree of each target point cloud data at different road positions is calculated.

[0046] In a possible implementation of the present application, based on the weakened distance, the target point cloud data is divided into inner points and outer points, and the divided inner points are retained, comprising:

[0047] The weakened distance is compared with a preset distance threshold;

[0048] If the weakened distance is less than the preset distance threshold, the target point cloud data corresponding to the weakened distance is divided into an inner point, otherwise, it is divided into an outer point;

[0049] The number of inner points in the iteration number of the random sample consensus algorithm is determined, and the inner points in the iteration number are retained, and each outer point in the iteration number is screened out.

[0050] The application provides a road infrastructure BIM model reverse construction method based on data fusion. In the related art, after the built urban road infrastructure is scanned by point cloud, the noise caused by the filling particles on the road surface cannot be removed by the preliminary denoising method, resulting in that the road facilities obtained by the deep learning algorithm do not conform to the actual situation. In the present application, first point cloud data of multiple road regions and their infrastructures are obtained, the first point cloud data are processed by segmentation to obtain multiple first road segments, and the point cloud data in the first road segments and the standard point cloud data of the infrastructure are locally optimally matched by a random sample consensus algorithm to obtain target point cloud data. The target point cloud data are filtered according to the recurrence degree of the target point cloud data at different road positions, so as to denoise the target point cloud data. The infrastructure and the road are embedded and installed, and in this way, the point cloud data of the mutual embedding structure can be filtered to denoise the target point cloud data. Then, the target point cloud data after denoising are processed by semantic segmentation and instance extraction by a deep learning algorithm, so as to reconstruct a target road model containing different infrastructures, and the road facilities in the target road model obtained by division are more consistent with the actual road conditions. BRIEF DESCRIPTION OF DRAWINGS

[0051] Figure 1 A flowchart of a first embodiment of the road infrastructure BIM model reverse construction method based on data fusion of the present application is shown in the figure.

[0052] Figure 2 A flowchart of a second embodiment of the road infrastructure BIM model reverse construction method based on data fusion of the present application is shown in the figure.

[0053] Figure 3 A device structure diagram of a hardware running environment involved in the embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0054] It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0055] The embodiment of the present application provides a road infrastructure BIM model reverse construction method based on data fusion. Figure 1 The method comprises the following steps.

[0056] In step S10, first point cloud data of multiple road regions and their infrastructures are obtained.

[0057] In step S20, the first point cloud data are processed by segmentation to obtain multiple first road segments.

[0058] Step S30, based on the random sample consensus algorithm, locally optimally matching the point cloud data in the first section with the standard point cloud data of the infrastructure, obtaining target point cloud data;

[0059] Step S40, based on the recurrence degree of each target point cloud data at different road locations, screening the target point cloud data to perform denoising processing on the target point cloud data;

[0060] Step S50, performing semantic segmentation and instance extraction processing on the denoised target point cloud data through a deep learning algorithm, and constructing a target road model containing different infrastructures.

[0061] The embodiment aims to screen the target point cloud data based on the recurrence degree of the target point cloud data at different road locations, to perform denoising processing on the target point cloud data, and the infrastructure and the road are installed by embedding, so that the point cloud data of the mutual embedding structure can be screened, and the target point cloud data is denoised, so that the road facilities in the target road model obtained by division are more consistent with the actual road conditions.

[0062] The specific steps are as follows:

[0063] Step S10, obtaining first point cloud data of a plurality of road regions and infrastructures thereof.

[0064] As an example, the data fusion-based road infrastructure BIM model reverse construction method can be applied to a data fusion-based road infrastructure BIM model reverse construction device, and the data fusion-based road infrastructure BIM model reverse construction device belongs to a data fusion-based road infrastructure BIM model reverse construction system, and the data fusion-based road infrastructure BIM model reverse construction system belongs to a data fusion-based road infrastructure BIM model reverse construction apparatus.

[0065] As an example, the first point cloud data can be point cloud data of each road region and each infrastructure at each road region collected based on high-precision cameras, three-dimensional laser scanners and other data collection devices, and the infrastructure can be various road elements, such as lane lines, railings, signboards, official websites and street lamps.

[0066] As an example, the collection method of the first point cloud data can be:

[0067] By installing a laser radar on the top of the vehicle and a high-precision GPS device in the vehicle, the vehicle advances along the road direction, and the route of the road is acquired by the high-precision GPS device during the advancing process, and according to the road width where the vehicle is located, the GPS data collected by the laser radar in the local range with the vehicle as the center and the road width as the range is selected to avoid the influence of other interference outside the road infrastructure on the point cloud data.

[0068] Since the road infrastructure needs to be calibrated and semantically segmented, the types of existing road facilities are acquired in advance, and in the embodiments of the application, all categories of all known facilities on the current collected road and the standard point cloud data of each infrastructure are also read.

[0069] The above-mentioned all first point cloud data is represented as That is, the i-th point cloud data is represented as The three-dimensional coordinates thereof.

[0070] In step S20, the first point cloud data is segmented to obtain a plurality of first road sections.

[0071] As an example, since the facilities installed on different road sections are different, for example, the division between the two sides of the urban road and the non-road is the curb, the division between the two sides of the non-urban road and the non-road is the fence, the division between the two sides of the bridge road and the non-road is the cement fence, and the road also includes signs, traffic lights, etc., and the point cloud data collected along the direction of the road (i.e. the main direction of the point cloud data) is different in different road sections. Therefore, based on the main direction of each first point cloud data, the route of the road is confirmed, and each road is segmented according to the point cloud density of different road sections to obtain a plurality of first road sections, wherein the first road section represents a plurality of road sections obtained by segmenting the first point cloud data, so as to achieve the purpose of accurately dividing each road.

[0072] In step S30, the point cloud data in the first road section and the standard point cloud data of the infrastructure are locally optimally matched based on the random sample consensus algorithm to obtain target point cloud data.

[0073] As an example, the random sample consensus algorithm is used to randomly extract a small part from all data to fit a preliminary model, and then use this model to test all other data points to see how many points support (meet) this model, repeat this process many times, and finally select the model supported by the most data points as the best model.

[0074] In the related art, after a certain number of point cloud data is selected locally by the random sample consensus algorithm, the point cloud data is used for fitting. The normal road infrastructure is mainly in the form of lines and surfaces with obvious structures. Therefore, after random sampling, fitting can be performed. Points that do not conform to the fitting points are regarded as outliers, and point cloud data that conforms to the fitting model is regarded as inliers. The inliers are the points that conform to the fitting model. However, in this process, the road is not straight, and part of the structure may also appear randomly, such as signs. At the same time, the road infrastructure is usually connected to the road or other structures in an embedded manner, such as street lamps, signs, and fences, which are connected by embedding into the foundation. This results in different point cloud data at different structures. The random sample consensus algorithm uses multiple iterations to select the number of inliers to obtain the final fitting model. When encountering complex road structures, the number of outliers is large, which causes the outliers to participate in the fitting model during the fitting process, resulting in model errors and increasing the number of iterations to obtain the inliers.

[0075] In an embodiment, the application uses standard point cloud data of various known infrastructures, performs local optimal matching, and combines the recurrence of the infrastructure in the road direction to slice the collected first point cloud data in the road direction to obtain multi-layer point data. Thus, the multi-layer point cloud data and the standard point cloud data are combined to obtain data denoising in the reverse modeling process through data fusion.

[0076] As an example, due to the lane lines and rough structures on the road, there are a large number of interference points. If all the interference points are analyzed and calculated, the calculation amount will be huge, which will reduce the efficiency of the random sample consensus algorithm and Scan-to-BIM. Therefore, in this embodiment, when the local point is selected by the random sample consensus algorithm, the target point cloud data that needs to be analyzed and calculated is selected by judging the local optimal matching between the selected first point cloud data and the standard point cloud data.

[0077] In the step S30, the following steps are included:

[0078] In step S31, a preset number of second point cloud data is extracted from each first point cloud data by the random sample consensus algorithm.

[0079] As an example, the preset number can be several hundred, several thousand, etc., and is not limited in detail. The second point cloud data is a part of the first point cloud data.

[0080] In step S32, the distance between the second point cloud data and the nearest point cloud data in the standard point cloud data corresponding to all infrastructures is set as the shortest matching distance of the second point cloud data.

[0081] As an example, the random sample consensus algorithm has multiple iteration times in the calculation process, taking the current iteration time as the pth iteration, for the point cloud data of the pth iteration, the matching degree with all standard point cloud data of the infrastructure is calculated, and the matching degree is defined as: the set of distances between each second point cloud data in the iteration point cloud data and its nearest neighbor point in the standard point cloud data, and the distance set corresponding to each second point cloud data is recorded as the set of shortest matching distances of each point cloud data in the pth iteration.

[0082] In step S33, the shortest matching distances corresponding to all second point cloud data are sorted in descending order to obtain a point cloud distance sorting sequence, and an optimal matching sequence segment in the point cloud distance sorting sequence is selected.

[0083] As an example, after calculating the shortest matching distances, the shortest matching distances corresponding to all second point cloud data are sorted in descending order, that is, from large to small. After sorting, the point cloud distance sorting sequence is obtained, and the point cloud distance sorting sequence is divided according to the changes of each shortest matching distance in the point cloud distance sorting sequence, and a sequence segment with smaller shortest matching distance is selected as the optimal matching sequence segment.

[0084] In step S33, the shortest matching distances corresponding to all second point cloud data are sorted in descending order to obtain a point cloud distance sorting sequence, and an optimal matching sequence segment in the point cloud distance sorting sequence is selected.

[0085] The distance difference between the two adjacent shortest matching distances in the point cloud distance sorting sequence is calculated.

[0086] Based on the position corresponding to the maximum value of the distance difference, the point cloud distance sorting sequence is divided to obtain a first sequence segment and a second sequence segment.

[0087] As an example, the distance difference between each adjacent shortest matching distance in the point cloud distance sorting sequence is calculated, and the maximum value of each distance difference is determined to locate the position corresponding to the maximum value of the distance difference.

[0088] As an example, after locating the position corresponding to the maximum value of the distance difference, the point cloud distance sorting sequence is divided into two parts at this position to obtain a first sequence segment and a second sequence segment, and the position is the middle position of the two adjacent sequence values. After dividing the first sequence segment and the second sequence segment, the two sequence values are also divided into the first sequence segment and the second sequence segment, respectively.

[0089] The sequence segment part corresponding to the minimum value of the distance average value in the first sequence segment and the second sequence segment is taken as the optimal matching sequence segment.

[0090] As an example, the distance average value of each sequence value in the first sequence segment is calculated respectively, and the distance average value of each sequence value in the second sequence segment is calculated respectively, and the sequence segment corresponding to the minimum value of the two distance average values is selected as the optimal matching sequence segment, and the division mode is two-level division, and the purpose of two-level division is to avoid the influence of interference points, and on the other hand, since the facility and the road are embeddedly installed, there is a loss of point cloud data, which causes the process of calculating the distance of the standard point cloud data to have points that cannot be matched, resulting in an increase in the result, so only the point cloud data with smaller distance is selected.

[0091] In step S34, the reliability of the point cloud data of the current iteration number is calculated based on the point cloud data corresponding to the optimal matching sequence segment and the shortest matching distance.

[0092] As an example, the reliability of the point cloud data of the current iteration number is determined according to the proportion of the point cloud data corresponding to the optimal matching sequence segment in all second point cloud data and the shortest matching distance.

[0093] In step S34, the reliability of the point cloud data of the current iteration number is calculated based on the point cloud data corresponding to the optimal matching sequence segment and the shortest matching distance.

[0094] The first proportion of the point cloud data corresponding to the optimal matching sequence segment and the first average value of the shortest matching distances are determined.

[0095] The reliability of the point cloud data of the current iteration number is calculated based on the ratio between the first proportion and the first average value.

[0096] As an example, the first proportion is the proportion of the point cloud data corresponding to the optimal matching sequence segment in all point cloud data, and the first average value is the average value of the shortest matching distances in the optimal matching sequence segment.

[0097] As an example, the reliability of the point cloud data of the current iteration number may be calculated in the following manner:

[0098]

[0099] wherein, represents the first proportion, represents the first average value, and norm() represents normalization calculation. The reliability is used to represent the possible degree of the current batch of point cloud data for analysis and calculation. When the reliability meets certain conditions, it means that the current batch of point cloud data is accurate.

[0100] In step S35, when the reliability is greater than a preset threshold, the point cloud data of the current iteration number is taken as the target point cloud data.

[0101] As an example, the preset threshold value can be 0.68, 0.7, etc., and is not limited in particular.

[0102] As an example, when the calculated reliability degree is greater than the preset threshold value, the point cloud data of the current iteration number is selected as the target point cloud data, the recurrence degree of the target point cloud data is calculated, and the target point cloud data is screened, so as to achieve the purpose of data denoising.

[0103] In step S40, the target point cloud data is screened based on the recurrence degree of the target point cloud data at different road positions, so as to perform denoising processing on the target point cloud data.

[0104] As an example, the recurrence degree is used to represent the regularity of the point cloud data appearing at the same relative position at equal distance intervals on the road, and a plurality of different road positions can be included on a road. The distance between each road position can be 1 meter, 2 meters, etc., and is not limited in particular.

[0105] As an example, the screening method of the target point cloud data can be that the recurrence degree of each target point cloud data at different road positions is calculated, and the spatial distance between each point cloud data and the fitting model constructed in the pth iteration number of the random sample consensus algorithm is weakened through the recurrence degree, and then the distances after weakening are used to divide the inliers and the outliers, so as to achieve the purpose of data screening.

[0106] In step S40, the target point cloud data is screened based on the recurrence degree of the target point cloud data at different road positions, so as to perform denoising processing on the target point cloud data.

[0107] In step S41, a slice plane at a plurality of road positions in each road region is constructed along the road direction corresponding to the road region.

[0108] As an example, before this, the road positions have been divided, and the slice planes at the road positions are constructed along the road direction (road center line) corresponding to each road region. The construction method of the slice plane is that a plane perpendicular to the road direction is taken as the slice plane at the road position. In the present application, the road positions are preset intervals, for example, 1 meter intervals.

[0109] In step S42, the suspected recurrence point cloud data of each target point cloud data is determined based on the distance between the target point cloud data in different slice planes.

[0110] As an example, in different slice planes, the suspected recurrence point cloud data of the target point cloud data in the preset adjacent region is selected according to the distance between each target point cloud data. The suspected recurrence point cloud data indicates that the point cloud data can be the same point cloud data repeatedly appearing at the next road position, for example, a sign belongs to road infrastructure, which will also repeatedly appear after passing an interval distance.

[0111] In step S42, the following steps are performed:

[0112] For each slice plane, the target point cloud data between the current slice plane and the next adjacent slice plane is projected onto the current slice plane until there is point cloud data on all slice planes.

[0113] As an example, after determining the slice planes at each road location, the target point cloud data between two adjacent slice planes (the current slice plane and the next slice plane) is projected onto the current slice plane until there is point cloud data on all slice planes, and the last slice plane is not calculated.

[0114] The point cloud data of each slice plane in each road region is projected onto the first slice plane with the road center as the origin.

[0115] As an example, the road centerline is the skeleton of the road, and if the target point cloud data belongs to the road infrastructure, the road infrastructure is repeatedly present, and the corresponding target point cloud data is also repeatedly present. Therefore, after slicing, the positions of the planes where the target point cloud data of the repeated road facilities is located are the same, and the intervals of the target point cloud data of the same feature points between the planes are similar. Therefore, according to the average degree of the interval of the point cloud data in the approximate position in the same plane after projecting the target point cloud data onto a plane, the degree of recurrence of the target point cloud data is obtained.

[0116] As an example, the point cloud data of each slice plane in each road region is projected onto the same slice plane with the road center as the origin, that is, the first slice plane. At this time, the point cloud data of all slice planes is projected onto a slice plane, so as to determine the suspected recurrent point cloud data of each target point cloud data on different slice planes on the same slice plane.

[0117] For any target point cloud data in the first slice plane, the point cloud data in the preset adjacent region of the current target point cloud data is taken as the suspected recurrent point cloud data of the current target point cloud data.

[0118] As an example, the preset adjacent region can be a circular region expanded by a preset radius, and the preset radius can be 0.1 meters. For any target point cloud data in the first slice plane, the point cloud data in the preset adjacent region of the current target point cloud data is taken as the suspected recurrent point cloud data of the current target point cloud data.

[0119] It should be noted that, since there can be multiple point cloud data in the preset adjacent area, multiple point cloud data in the same slice plane can be selected as suspected recurring point cloud data, therefore, the suspected recurring point cloud data needs to be screened, and the screening method is: if there are multiple point cloud data in the same slice plane as suspected recurring point cloud data, only the point cloud data closest to the current target point cloud data in the same slice plane is taken as the suspected recurring point cloud data, so as to ensure that there is at most one suspected recurring point cloud data in each slice plane for the same target point cloud data.

[0120] In step S43, the recurrence degree of each target point cloud data at different road positions is calculated based on the number of suspected recurring point cloud data and the number of slice planes.

[0121] As an example, for each target point cloud data, there are multiple suspected recurring point cloud data, and the recurrence degree of each target point cloud data at different road positions is calculated according to the number of suspected recurring point cloud data, the number of slice planes, and the regularity of slice plane distribution between two adjacent suspected recurring point cloud data.

[0122] In step S43, the recurrence degree of each target point cloud data at different road positions is calculated based on the number of suspected recurring point cloud data and the number of slice planes.

[0123] All suspected recurring point cloud data of different target point cloud data are arranged in order of road direction, and the number of plane intervals between any two adjacent suspected recurring point cloud data of each target point cloud data is determined.

[0124] As an example, all suspected recurring point cloud data of each target point cloud data on different slice planes are arranged in order of road direction, and the number of plane intervals between any two adjacent suspected recurring point cloud data on the first slice plane is determined, wherein the number of plane intervals is the number of slice planes between the two adjacent suspected recurring point cloud data.

[0125] The average distance between each target point cloud data and its corresponding suspected recurring point cloud data is calculated.

[0126] Based on the ratio between the average distance and the number of plane intervals, and the ratio between the number of suspected recurring point cloud data and the number of slice planes, the recurrence degree of each target point cloud data at different road positions is calculated.

[0127] As an example, there is a spatial distance between each target point cloud data and its corresponding suspected recurring point cloud data, and the average of these spatial distances is calculated to obtain the average distance.

[0128] As an example, taking the mth target point cloud data of the first slice plane as an example, the recurrence degree of the mth target point cloud data of the first slice plane is The calculation method is as follows:

[0129]

[0130] in, This represents the number of suspected reproducing point cloud data points for the m-th target point cloud data in the first slice plane. This represents the total number of slice planes within this road segment. A larger value indicates higher reproducibility; The acquisition method is as follows: Arrange all suspected reproducible point cloud data in order of road direction, calculate the number of planar intervals between two adjacent suspected reproducible point cloud data, and denot the kurtosis of all planar intervals as . This indicates the clustering of the number of planar intervals; the larger the value, the more obvious the clustering, indicating a regular arrangement. This is the average distance between the suspected reproducible point cloud data and the m-th point cloud data. The smaller the value, the stronger the reproducibility. +1 is used to prevent the denominator from being zero.

[0131] Step S44: Determine the spatial distance between each target point cloud data and the fitted model constructed by the random sampling consensus algorithm in the current iteration number.

[0132] Step S45: Based on the degree of reproducibility, the spatial distance is weakened to obtain the weakened distance.

[0133] As an example, the spatial distance between each target point cloud data and the fitted model constructed by the random sampling consensus algorithm in the current iteration is determined, and this spatial distance can be directly obtained in the current iteration.

[0134] As an example, by calculating the reproducibility of each target point cloud data, the corresponding spatial distance is negatively weighted and calculated to weaken the spatial distance, resulting in a weakened distance. The negatively weighted calculation method can be: weakened distance = spatial distance (1 - reproducibility).

[0135] Step S46: Based on the weakened distance, divide the target point cloud data into interior points and exterior points, and retain the interior points.

[0136] Step S46 includes:

[0137] The weakened distance is compared with a preset distance threshold.

[0138] If the reduced distance is less than the preset distance threshold, the target point cloud data corresponding to the reduced distance is classified as an interior point; otherwise, it is classified as an exterior point.

[0139] As an example, after obtaining the weakened distance, the target point cloud data is divided into an inner point or an outer point by comparing the weakened distance corresponding to each target point cloud data with a preset distance threshold, and the target point cloud data is divided into an inner point when the weakened distance is less than the preset distance threshold, and is divided into an outer point otherwise.

[0140] The number of inner points in each iteration of the random sample consensus algorithm is determined, and the inner points in the iteration with the largest number of inner points are retained, and each outer point in the iteration is screened out.

[0141] As an example, in the random sample consensus algorithm, there are multiple iterations, and the fitting model of the iteration with the largest number of inner points is selected, the inner points are retained, and the outer points are screened out.

[0142] In step S50, the denoised target point cloud data is processed by a deep learning algorithm for semantic segmentation and instance extraction, and a target road model containing different infrastructures is constructed.

[0143] As an example, after selecting the final denoised target point cloud data, the point cloud data is processed by a deep learning algorithm (such as a point cloud panoramic segmentation model) for semantic segmentation and instance extraction, and road elements are automatically identified, thereby constructing a target road model containing different infrastructures, wherein the processing method of the deep learning algorithm is a prior art and will not be described here.

[0144] All point cloud data is denoised in the above manner to screen out the outer points, a deep learning algorithm is trained based on the standard point cloud data as a training set, and the denoised point cloud data is processed for semantic segmentation to obtain various infrastructures in the road, thereby realizing reverse modeling of Scan-to-BIM using point cloud data.

[0145] The application provides a road infrastructure BIM model reverse construction method based on data fusion. In the related art, after the built urban road infrastructure is scanned by point cloud, the noise generated by the filling particles on the road surface cannot be removed by the preliminary denoising method, resulting in that the road facilities obtained by the deep learning algorithm do not match the actual situation. In the application, first point cloud data of multiple road regions and their infrastructures are obtained, the first point cloud data is segmented to obtain multiple first road sections, and the point cloud data in the first road section and the standard point cloud data of the infrastructure are locally optimally matched by the random sample consensus algorithm to obtain target point cloud data. The target point cloud data is filtered according to the recurrence degree of the target point cloud data at different road positions, so as to denoise the target point cloud data. The infrastructure and the road are installed by embedding. In this way, the point cloud data of the mutual embedding structure can be filtered, so as to denoise the target point cloud data. Then, the semantic segmentation and instance extraction of the denoised target point cloud data are performed by the deep learning algorithm, so as to reconstruct the target road model containing different infrastructures, and the road facilities in the target road model obtained by division are more consistent with the actual road conditions.

[0146] Further, referring to Figure 2 , based on the first embodiment in the application, another embodiment of the application is provided, in which the step S20 of segmenting the first point cloud data to obtain multiple first road sections includes:

[0147] Step S21, along the road direction corresponding to the road region, taking the vertical plane perpendicular to the road direction as the road section of each road region.

[0148] As an example, along the road direction corresponding to the road region, taking the vertical plane corresponding to the road center line as the road section of the corresponding road region.

[0149] Step S22, determining the point cloud quantity of the first point cloud data corresponding to the data acquisition position in each road region, and taking the point cloud quantity as the cross-sectional density of each road section.

[0150] As an example, the way to obtain the cross-sectional density can be:

[0151] The position of the vehicle collecting data on the road is obtained, and the position is taken as the center and the road as the width to calculate the point cloud quantity of the first point cloud data in the region as the cross-sectional density of the road section corresponding to each road position.

[0152] Step S23, clustering the cross-sectional density to obtain multiple first road sections.

[0153] As an example, a clustering space is constructed with road positions as the horizontal axis and cross-sectional densities as the vertical axis, the cross-sectional densities are clustered by using a hierarchical clustering algorithm, and a plurality of first road segments are obtained.

[0154] The step S23 comprises:

[0155] The density average of the cross-sectional densities corresponding to all road regions is calculated, and the density average is taken as a clustering threshold.

[0156] If the difference between the cross-sectional density averages of any two clustering clusters is greater than the clustering threshold, it is determined that the two clustering clusters are divided into two first road segments, and until the clustering of all road regions is completed, a plurality of first road segments are obtained.

[0157] As an example, in the clustering process, since a threshold is needed as the end limit of clustering, the average of the cross-sectional densities of all road positions in each road region is calculated, and 5% of the average is taken as the clustering threshold. When the difference between the cross-sectional density averages of two clustering clusters in the clustering process is greater than 5%, it is considered that two first road segments are divided. When the clustering of all road regions is completed, a plurality of first road segments are obtained.

[0158] In this embodiment, the point cloud data contained by the point cloud data under different road segments is different along the direction of the road, that is, the main direction of the point cloud data. By dividing each road, a plurality of first road segments are obtained. Based on the point cloud data of the first road segment, subsequent road division and data denoising are facilitated.

[0159] Referring to Figure 3 , Figure 3 is a device structure schematic diagram of a hardware running environment involved in the embodiment scheme of the application.

[0160] As Figure 3 shown, the data fusion-based road infrastructure BIM model reverse construction device can include a processor 1001, a memory 1003, and a communication bus 1002. The communication bus 1002 is used to realize the connection communication between the processor 1001 and the memory 1003.

[0161] Optionally, the data fusion-based road infrastructure BIM model reverse construction device can further include a user interface, a network interface, a camera, an RF (Radio Frequency, radio frequency) circuit, a sensor, a WiFi module, and the like. The user interface can include a display screen (Display), an input sub-module such as a keyboard (Keyboard), and the optional user interface can further include a standard wired interface, a wireless interface. The network interface can include a standard wired interface, a wireless interface (such as a WI-FI interface).

[0162] Those skilled in the art can understand that,Figure 3 The data fusion based road infrastructure BIM model reverse construction device structure shown in the figure does not constitute a limitation of the data fusion based road infrastructure BIM model reverse construction device, and can include more or fewer components than the figure, or combine certain components, or different component arrangements.

[0163] As shown in the figure, the memory 1003 as a storage medium can include an operating system, a network communication module, and a data fusion based road infrastructure BIM model reverse construction program. The operating system is a program that manages and controls the hardware and software resources of the data fusion based road infrastructure BIM model reverse construction device, supports the running of the data fusion based road infrastructure BIM model reverse construction program and other software and / or programs. The network communication module is used to realize the communication between the components inside the memory 1003, and the communication with other hardware and software in the data fusion based road infrastructure BIM model reverse construction system. Figure 3 In the data fusion based road infrastructure BIM model reverse construction device shown in the figure, the processor 1001 is used to execute the data fusion based road infrastructure BIM model reverse construction program stored in the memory 1003, and realize the steps of any one of the above data fusion based road infrastructure BIM model reverse construction methods.

[0164] Figure 3 The data fusion based road infrastructure BIM model reverse construction device embodiment of the present application is basically the same as the above data fusion based road infrastructure BIM model reverse construction method embodiments, and will not be repeated here.

[0165] It should be noted that in this paper, the term "includes", "contains" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or system including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or system. Without more limitations, the element defined by the statement "includes a" does not exclude the presence of another identical element in the process, method, article or system including the element.

[0166] It should be noted that the above sequence number of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments.

[0167] The above sequence number of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments.

[0168] ​Those skilled in the art can clearly understand the above-mentioned embodiment method can be realized by means of software and the necessary general hardware platform, of course, it can also be realized by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disc) and includes a number of instructions for causing a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the methods of various embodiments of the present application.

[0169] The above are only preferred embodiments of the present application, and do not limit the scope of the application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, is also included in the scope of protection of the present application.

[0170] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0171] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. A method for reverse construction of a road infrastructure BIM model based on data fusion, characterized in that, The method comprises: obtaining a plurality of first point cloud data of road regions and their infrastructures; segmenting the first point cloud data to obtain a plurality of first road segments; based on a random sample consensus algorithm, locally optimally matching the point cloud data in the first road segments with standard point cloud data of the infrastructures to obtain target point cloud data; the random sample consensus algorithm, locally optimally matching the point cloud data in the first road segments with standard point cloud data of the infrastructures to obtain target point cloud data, comprises: extracting a preset number of second point cloud data from each of the first point cloud data by a random sample consensus algorithm; setting the distance between the second point cloud data and the nearest point cloud data in the standard point cloud data corresponding to all infrastructures as the shortest matching distance of the second point cloud data; sorting the shortest matching distances corresponding to all second point cloud data in descending order to obtain a point cloud distance sorting sequence, and selecting an optimal matching sequence segment in the point cloud distance sorting sequence; based on the point cloud data corresponding to the optimal matching sequence segment and the shortest matching distance, calculating the reliability of the point cloud data of the current iteration number; when the reliability is greater than a preset threshold, the point cloud data of the current iteration number is taken as the target point cloud data; based on the recurrence degree of each target point cloud data at different road locations, filtering the target point cloud data to denoise the target point cloud data; based on the recurrence degree of each target point cloud data at different road locations, filtering the target point cloud data, comprises: constructing a slice plane at a plurality of road locations in each road region along the road direction corresponding to the road region; determining the suspected recurrent point cloud data of each target point cloud data based on the distance between the target point cloud data in different slice planes; based on the number of suspected recurrent point cloud data and the number of slice planes, calculating the recurrence degree of each target point cloud data at different road locations; determining the spatial distance between each target point cloud data and the fitting model constructed by the random sample consensus algorithm at the current iteration number; based on the recurrence degree, weakening the spatial distance to obtain a weakened distance; based on the weakened distance, dividing the target point cloud data into inliers and outliers, and retaining the divided inliers, which are the denoised target point cloud data; performing semantic segmentation and instance extraction processing on the denoised target point cloud data by a deep learning algorithm to construct a target road model containing different infrastructures.

2. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The segmentation of the first point cloud data to obtain a plurality of first road segments comprises: taking a vertical plane perpendicular to the road direction as the road section of each road region along the road direction corresponding to the road region; determining the point cloud number of the first point cloud data corresponding to the data collection position in each road region, and taking the point cloud number as the cross-sectional density of each road section; clustering the cross-sectional density to obtain a plurality of first road segments.

3. The data fusion based road infrastructure BIM model reverse construction method of claim 2, wherein, The clustering of the cross-sectional density to obtain a plurality of first road segments comprises: Calculate the density average of the cross-sectional density corresponding to all road regions, and take the density average as the clustering threshold; If the difference between the cross-sectional density averages of any two clustering clusters is greater than the clustering threshold, it is determined that the two clustering clusters are divided into two first road segments, and after all road region clustering is completed, a plurality of first road segments are obtained.

4. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The selecting the optimal matching sequence segment in the point cloud distance sorting sequence comprises: Calculating the distance difference between the two shortest matching distances of the adjacent point cloud distance sorting sequence; Based on the position corresponding to the maximum value of the distance difference, the point cloud distance sorting sequence is divided to obtain a first sequence segment and a second sequence segment; The sequence segment part corresponding to the minimum value of the distance average in the first sequence segment and the second sequence segment is taken as the optimal matching sequence segment.

5. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The calculation of the reliability of the point cloud data of the current iteration number based on the point cloud data corresponding to the optimal matching sequence segment and the shortest matching distance comprises: Determine the first proportion of the point cloud data corresponding to the optimal matching sequence segment, and the first average value of each shortest matching distance; Based on the ratio between the first proportion and the first average value, the reliability of the point cloud data of the current iteration number is calculated.

6. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The determination of the suspected recurring point cloud data of each target point cloud data based on the distance between the target point cloud data in different slice planes comprises: For any slice plane, the target point cloud data between the current slice plane and the next adjacent slice plane is projected onto the current slice plane until there is point cloud data on all slice planes; Project the point cloud data of each slice plane in each road region to a first slice plane with the road center as the origin; For any target point cloud data in the first slice plane, the point cloud data in the preset adjacent region of the current target point cloud data is taken as the suspected recurring point cloud data of the current target point cloud data.

7. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The calculation of the recurrence degree of each target point cloud data at different road positions based on the number of suspected recurring point cloud data and the number of slice planes comprises: Arrange all suspected recurring point cloud data of different target point cloud data in order of road direction, and determine the plane interval number between any two adjacent suspected recurring point cloud data of each target point cloud data; Calculate the distance average between each target point cloud data and its corresponding suspected recurring point cloud data; Based on the ratio between the distance average and the plane interval number, and the ratio between the number of suspected recurring point cloud data and the number of slice planes, the recurrence degree of each target point cloud data at different road positions is calculated.

8. The data fusion based road infrastructure BIM model reverse construction method of claim 1, wherein, The division of the target point cloud data into inliers and outliers based on the weakened distance, and the retention of the divided inliers, comprises: Comparing the weakened distance with a preset distance threshold; If the weakened distance is less than the preset distance threshold, the target point cloud data corresponding to the weakened distance is divided into inliers, otherwise, it is divided into outliers; Determine the iteration number with the most number of interior points in the multiple iterations of the random sampling consistent algorithm, and retain the interior points at the iteration number, and screen out each exterior point at the iteration number.

Images