Large-scale bridge three-dimensional laser scanning planning method in complex scene

Abstract

The invention relates to a large-scale bridge three-dimensional laser scanning planning method in a complex scene. The method comprises the following steps: 1) constructing a bridge simulation scene; 2) planning an observation station; 3) target planning; and 4) outputting a scanning scheme. According to the invention, an optimization method for minimizing the number of targets required by point cloud registration is provided, and the problem of low efficiency caused by a scanning scheme based on target registration is relieved; meanwhile, the scanning azimuth angle condition of a high-scanning-resolution observation station is considered, and the time cost can be reduced while the requirements for the detail degree and the integrity of the region of interest can be met; a general solution thought is provided for observation stations and target optimization arrangement in a complex bridge scanning scene, and the method can be applied to three-dimensional laser scanning planning in other bridge scenes.

Description

Technical Field

[0001] The present invention belongs to the technical field of bridge detection, and in particular relates to a three-dimensional laser scanning planning method for large-scale bridges in complex scenes. Background Art

[0002] As modern bridges develop towards large spans, high piers, high towers, and giant segments prefabricated in factories, higher and higher requirements are placed on the measurement of bridge geometry. The representative 3D laser scanner among advanced measurement equipment can quickly obtain geometric data related to the measured object with millimeter-level accuracy, providing a new technical means for accurate bridge geometry detection. In actual projects, the data acquisition work of 3D laser scanners usually relies on the subjective operation of inspectors, resulting in uncontrollable quality of collected data. Although some scholars have proposed scanning planning methods for bridge scenes, these methods are usually aimed at bridges with relatively simple geometric forms, such as beam bridges. The problem of 3D laser scanning planning for large-scale bridges in complex scenes has not yet been fully solved. There are two main differences between the scanning planning of large-scale bridges in complex scenes and the scanning planning of bridges with simple geometric forms.

[0003] First, there are limited locations for setting up 3D laser scanners in complex scenes. They are usually set up on the bridge deck to scan high-altitude components or under the bridge to scan the overall bridge shape. In order to meet the requirements for the level of detail and completeness of the area of ​​interest, a long-distance and high-resolution scanning mode must be adopted, and the use of panoramic scanning in high-resolution scanning mode will inevitably result in greater time costs. It is in line with actual needs to consider the scanning azimuth of the high-resolution scanning station when planning the scan to balance the scanning efficiency and data quality.

[0004] Secondly, the subsequent point cloud registration conditions need to be considered when formulating a plan. The general method of simplifying the registration conditions is to require that the overlap rate of the two registered point clouds is greater than 20%. However, the inherent noise and outliers in the point cloud data, as well as factors such as incomplete scanning, put forward strict requirements on the accuracy of the algorithm for extracting matching features, especially the lack of mature point cloud processing algorithms to achieve accurate registration of complex scenes. Another method is to introduce target-based point cloud registration, which is still a more common method. However, existing research lacks exploration of the scanning planning problem of target-based point cloud registration methods, mainly due to potential inefficiency. When formulating a plan, from the global perspective of station and target distribution, the number of targets required for scanning should be reduced as much as possible to improve cost-effectiveness.

[0005] Therefore, there is an urgent need to propose a large-scale bridge 3D laser scanning planning method in complex scenarios to formulate scanning plans and guide on-site scanning. Summary of the invention

[0006] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a large-scale bridge three-dimensional laser scanning planning method in complex scenes. The surface point cloud extracted from the bridge three-dimensional model is used as a simulation scene. Considering the user's data requirements and the multiple constraints of the scanner parameters used, the design is based on the weighted greedy algorithm, the ant colony algorithm and the density clustering algorithm.

[0007] The present invention solves the technical problem by the following technical solutions:

[0008] A large-scale bridge three-dimensional laser scanning planning method in a complex scene, the method comprising the following steps:

[0009] S1. Use drone photogrammetry to reconstruct the 3D model of the bridge, add scanning obstacles to the 3D model of the bridge, extract the center point and center point normal vector of the model triangle, and construct the bridge entity point cloud. Segment the area in the entity point cloud where the scanner can be set up and the target can be placed and discretize it into equidistant grid points as the candidate point set for the measuring station and target.

[0010] S2. Under unconstrained conditions, the entity point cloud is virtually pre-scanned to obtain the surface point cloud, downsample and segment special areas as the bridge simulation scene;

[0011] S3. According to user needs and scanner parameters, set the constraints of global area point cloud density, special area point cloud density, maximum incident angle, maximum scanning distance and scanning angle resolution range, carry out ray tracing simulation analysis on the bridge simulation scene under the constraints, and use the weighted greedy algorithm to optimize the two-stage station layout, select the optimal station position, scanning angle resolution and scanning azimuth of the high scanning angle resolution station, and plan the path of the optimal station through the ant colony algorithm based on the station layout optimization results, with the shortest scanner moving path and the maximum number of overlapping points of adjacent stations.

[0012] S4. According to the station path planning results, cluster the neighboring stations into station groups. Based on the target candidate point set and the distribution of the station groups, optimize the target layout with the principle of minimizing the station-target mutual visibility and target coordinate covariance condition number, and find the optimal target position for each station group.

[0013] S5. Summarize the optimization results of the station and target arrangement and output a scanning plan. The scanning plan includes the optimal station position and station group, the optimal target position of each station group, the scanning angle resolution, the scanning azimuth angle of the high scanning angle resolution station, and the scanning path.

[0014] Moreover, the specific process of the virtual pre-scan in S2 is as follows:

[0015] First, according to the scanning planning experience of measurement professionals, a group of virtual pre-scanning station locations are selected;

[0016] Secondly, based on the station positions of the virtual pre-scan, ray tracing simulation analysis is performed on the physical point cloud under unconstrained conditions, and all visible points are saved as the bridge surface point cloud.

[0017] Moreover, the specific process of the two-stage station arrangement optimization in S3 is as follows:

[0018] First, a panoramic scanning mode with low scanning angle resolution is used to count the locations of areas that meet the constraints in the scene and determine whether special areas meet the constraints.

[0019] Secondly, if the constraint conditions are not met, a high scanning angle resolution limited azimuth scanning mode simulation is first used for the special area, and then a low scanning resolution panoramic scanning mode simulation is performed;

[0020] Thirdly, according to the number of scene points that meet the constraints, the weight coefficient of each candidate station point is calculated, and the station with the largest weight is selected iteratively until all scanning points meet the completeness requirements, or the difference in coverage between two adjacent iterations is less than the specified threshold. Then the iteration is stopped to obtain a set of optimal station positions and scanning angle resolutions.

[0021] Moreover, the specific process of path planning in S3 is as follows:

[0022] Firstly, the ant colony algorithm is used to assign a number of ants to each station. According to the distance between stations and the scanning overlap rate, the heuristic function calculates the transfer probability of each ant to the next station. The next station is selected by the roulette method. The order of stations that the ants have passed and the length of the path are recorded until all stations are visited. The elite ants are selected by sorting the total length of each ant's path in ascending order.

[0023] Repeat the above steps until the preset number of iterations is met. At this time, the order of the stations passed by the elite ant and the path length are the global path planning.

[0024] Moreover, the specific process of station grouping in S4 is as follows:

[0025] First, the median of the distance between all two adjacent stations is counted, and a neighborhood circle is drawn with a station as the center and the median as the radius to perform neighborhood detection. If the number of stations in the neighborhood is equal to or exceeds the threshold, then this station is a core station, otherwise it is a noise station. The threshold is set to 2;

[0026] Secondly, all the stations in the neighborhood are tested until the noise station is detected. All the detected core stations are considered as stations in the same group, and the noise station is a separate station group.

[0027] Finally, repeat the above steps for the remaining stations until all stations are divided into multiple station groups.

[0028] Moreover, the specific process of the target arrangement optimization in S4 is as follows:

[0029] Firstly, for the same station group, three target layout points are selected from the visible target candidate points based on the principle of mutual visibility of the targets within the group and the minimum condition number of the coordinate covariance matrix;

[0030] Secondly, for the adjacent station groups, the two nearest stations are determined according to the path planning results. Based on the principle of mutual visibility of the targets of the two adjacent stations and the minimum condition number of the coordinate covariance matrix, three target positions are selected from the visible target candidate points. The above steps are repeated until all station groups are traversed.

[0031] Moreover, the maximum incident angle is no greater than 70 degrees.

[0032] The advantages and beneficial effects of the present invention are:

[0033] 1. The present invention proposes a scanning planning method for scenes on and under bridges, providing guidance for formulating scanning plans for typical bridge scenes;

[0034] 2. The present invention takes into account the scanning azimuth conditions of the high scanning resolution measuring station, which can meet the requirements of detail and completeness of the area of ​​interest while reducing the time cost;

[0035] 3. The present invention proposes an optimization method to minimize the number of targets required for point cloud registration, thereby alleviating the inefficiency problem caused by the scanning scheme based on target registration;

[0036] 4. The scanning planning method proposed in the present invention provides a general solution for the optimal arrangement of measuring stations and targets in complex bridge scanning scenes, and can be applied to three-dimensional laser scanning planning problems in other bridge scenes. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flow chart of the present invention;

[0038] Figure 2 It is a schematic diagram of the three-dimensional model processing process of the present invention;

[0039] Figure 3 A process diagram for establishing a candidate point set for a station and a target of the present invention;

[0040] Figure 4 Establishing a simulation scenario process diagram for the present invention;

[0041] Figure 5 This is a schematic diagram of the scanning azimuth angle of the present invention;

[0042] Figure 6 This is a schematic diagram of target layout optimization of the present invention;

[0043] Figure 7 Scanning planning scheme diagram for the present invention;

[0044] Figure 8 This is a diagram of the bridge scanning result of the present invention. DETAILED DESCRIPTION

[0045] The present invention is further described in detail below through specific examples. The following examples are only illustrative and not restrictive, and the protection scope of the present invention cannot be limited thereto.

[0046] like Figure 1 As shown, a large-scale bridge 3D laser scanning planning method in a complex scene is innovative in that the method comprises the following steps:

[0047] Step 1: For an existing 3D bridge model or a 3D bridge model reconstructed using drone photogrammetry, use CloudCompare software to add scanning obstacles to the 3D model bridge deck, extract the center point and center point normal vector of the model triangle, and construct a bridge entity point cloud, such as Figure 2 As shown in , the area where the scanner can be set up and the target can be placed in the segmented entity point cloud is discretized into equidistant grid points as the candidate point set for the measuring station and target, as shown in Figure 3 As shown;

[0048] In step 1, the candidate point set of the measuring station and the target is constructed, and the specific process is as follows:

[0049] The bridge deck area of ​​the entity point cloud is segmented and discretized into sparse point clouds at intervals of 1m as the candidate point set for the bridge station. The top surface area of ​​the occluded 3D model is segmented and discretized into sparse point clouds at intervals of 1m, and then the candidate point set for the bridge station is fused as the candidate point set for the bridge target. The location of the bridge is searched through Google Maps, and the bridge and the surrounding scene area in the map are intercepted. The scene image is converted into a grayscale image, and then the white edge contour is discretized into a sparse point cloud at intervals of 1m as the candidate point set for the bridge station.

[0050] Step 2: According to user requirements and instrument parameters, set the global area point cloud density, special area point cloud density, maximum incident angle, maximum scanning distance and scanning angle resolution range constraints. Under unconstrained conditions, perform a virtual pre-scan on the entity point cloud to obtain the surface point cloud, downsample and segment the special area as the bridge simulation scene. The specific process of this step is as follows: Figure 4 As shown;

[0051] In step 2, the maximum incident angle set by the constraint condition is no more than 70 degrees, the global LOD requirement is set to 0.025m, special LOD requirements are set to 0.008m and 0.005m for the arch end and the main beam notch section, the scanning distance on the bridge is 120m, and the scanning distance under the bridge is 270m. According to the user interface of the scanner Leica P50, the maximum scanning distance of 120m corresponds to the optional options of 0.005 degrees, 0.009 degrees, 0.018 degrees, and 0.036 degrees, and the maximum scanning distance of 270m corresponds to the optional options of 0.005 degrees, 0.009 degrees, 0.018 degrees, and 0.036 degrees.

[0052] In step 2, the virtual pre-scanning steps are as follows:

[0053] First, according to the scanning planning experience of measurement professionals, a group of virtual pre-scanning station locations are selected;

[0054] Secondly, based on the station positions of the virtual pre-scan, ray tracing simulation analysis is performed on the physical point cloud without constraints, and all visible points are saved as the bridge surface point cloud.

[0055] Step 3: Under the constraints, ray tracing simulation analysis is carried out on the bridge simulation scene, and a two-stage station layout optimization is performed through the weighted greedy algorithm to select a set of optimal station locations and scanning angle resolutions, such as Figure 5 As shown in the figure, according to the optimization results of the station layout, the path planning of the optimal station is carried out by using the ant colony algorithm with the goal of shortest scanning path and maximum number of overlapping points of adjacent stations;

[0056] In step 3, the specific process of two-stage station layout optimization is as follows:

[0057] First, a panoramic scanning mode with low scanning angle resolution (0.036 degrees, 0.018 degrees, and 0.009 degrees) is used to count the locations of areas that meet the constraints in the scene and determine whether special areas meet the constraints.

[0058] Secondly, if the constraint conditions are not met, a high scanning angle resolution (0.005 degrees) is used to limit the azimuth scanning mode simulation for the special area, and then a low scanning resolution panoramic scanning mode simulation is performed. The smaller the scanning angle resolution mentioned above, the higher the resolution;

[0059] Thirdly, according to the number of scene points that meet the constraints, the weight coefficient of each candidate station point is calculated, and the station with the largest weight is selected iteratively until all scanning points meet the completeness requirements, or the difference in coverage between two adjacent iterations is less than 2%. Then the iteration is stopped to obtain a set of optimal station positions and scanning angle resolutions.

[0060] In step 3, the weight coefficient is calculated as follows:

[0061] First, according to the visibility analysis, the candidate measuring station S is determined. i For observation point T j Visibility index V ij :

[0062] V ij ={0,1}

[0063] Where: V ij The value is 0, indicating that the scan object T j Candidate scanning site S i Visible; V ij The value is 1, indicating that the scan object T j Cannot be scanned by candidate site S i Visible; S i is any element in the candidate scanning site set {S}; T j is any element in the candidate scan object set {T}.

[0064] Secondly, according to the visibility index V ij Calculate the weight coefficient n for each observation point i , the calculation formula is as follows:

[0065]

[0066] Again, according to the visibility index V ij Calculate the weighted weight coefficient w for each observation point i , calculation formula

[0067] as follows:

[0068]

[0069] Finally, according to the weight coefficient n of the observation point i and w i Calculate the normalized weight coefficient W for each candidate station i , the calculation formula is as follows:

[0070]

[0071] In step 3, the specific process of path planning is as follows:

[0072] First, a number of ants are assigned to each station. According to the distance between stations and the scanning overlap rate, the heuristic function calculates the transfer probability of each ant to the next station. The next station is selected by the roulette method. The order of the stations the ants have passed and the length of the path are recorded until all stations have been visited. The elite ants are selected by sorting the total length of each ant's path in ascending order.

[0073] Repeat the above steps until the preset number of iterations is met. At this time, the order of the stations passed by the elite ant and the path length are the global path planning.

[0074] Step 4: According to the station path planning results, cluster the neighboring stations into station groups. Based on the target candidate point set and the distribution of the station group, the target layout is optimized with the principle of minimizing the station-target mutual visibility and target coordinate covariance condition number to find the optimal target position for each group of stations, such as Figure 6 As shown;

[0075] In step 4, the specific process of station grouping is as follows:

[0076] First, the median of the distance between all two adjacent stations is counted, and a neighborhood circle is drawn with a station as the center and the median as the radius to perform neighborhood detection. If the number of stations in the neighborhood is equal to or exceeds the number threshold, then this station is a core station, otherwise it is a noise station. The number threshold is set to 2;

[0077] Secondly, all the stations in the neighborhood are tested until the noise station is detected. All the detected core stations are considered as the same group of stations, and the noise station is a separate station group.

[0078] Finally, repeat the above steps for the remaining stations until all stations are divided into multiple station groups.

[0079] In step 4, the specific process of target arrangement optimization is as follows:

[0080] Firstly, for the stations in the same group, three target placement points are selected from the visible target candidate points based on the principle of mutual visibility of the targets in the group and the minimum condition number of the coordinate covariance matrix;

[0081] Secondly, for the adjacent station group, two adjacent stations are determined based on the path planning results, and three target layout points are selected from the visible target candidate points based on the principle of mutual visibility of the targets of the two adjacent stations and the minimum condition number of the coordinate covariance matrix. Repeat the above steps until all station groups are traversed.

[0082] In step 4, the covariance matrix condition number formula is as follows:

[0083] k(S t )=Cand(Cov(S t )) / S St ,(S t ∈Ω)

[0084] Among them, S t are the three selected target sphere coordinate matrices, Cov represents the covariance of the matrix, Cand represents the condition number of the matrix, S Strepresents the area enclosed by the three selected targets, and Ω is the intersection of the visible areas.

[0085] Step 5: Summarize the station and target layout optimization results and visualize the output scanning plan, such as Figure 7 As shown, a scanning plan is used to guide on-site scanning.

[0086] Among them, the station is a red triangle, and the station group number and scanning resolution are marked next to the station. There are five groups of stations in total, sorted from 0 to 5. The scanning resolution is represented by 1 / 1 (angular resolution is 0.005°), 1 / 2 (angular resolution is 0.009°) and 1 / 3 (angular resolution is 0.018°). The same group of targets is represented by the same color, and the station group number that can be used for splicing stations is marked. For example, target 0-1 means that station group 0 and station group 1 scan this group of targets, and the green straight line represents the scanning path. The P4S scheme has a total of 20 stations, including 16 stations on the bridge and 4 stations under the bridge.

[0087] When scanning on the bridge, it takes about 5 minutes to move the station, set the scanning parameters and scan the targets, and about 3 minutes to place the targets. There are 16 stations and 9 groups of targets, and the scanning time is 270 minutes. When scanning under the bridge, it takes about 10 minutes to move the station and set the scanning parameters. There are 4 stations and the scanning time is 148 minutes. Therefore, the total scanning time is 418 minutes.

[0088] The solution sets a scanning angle of 15 degrees for the two high-scanning resolution measuring stations on the bridge (scanning resolution is 1 / 1), reducing the number of targets required for the alignment of the measuring stations on the bridge from 15 groups (45) to 9 groups (27).

[0089] Compared with the solution without setting the scanning angle and simplifying the number of targets, the time for placing and scanning a group of targets is calculated as 6 minutes. The solution of the present invention shortens the total scanning time from 554 minutes to 418 minutes, which is about 25% shorter.

[0090] Although the embodiments and drawings of the present invention are disclosed for illustrative purposes, those skilled in the art will appreciate that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the scope of the present invention is not limited to the contents disclosed in the embodiments and drawings.

Claims

1. A large-scale bridge 3D laser scanning planning method in complex scenes, characterized by: The steps of the method are: S1. Use drone photogrammetry to reconstruct the 3D model of the bridge, add scanning obstacles to the 3D model of the bridge, extract the center point and center point normal vector of the model triangle, and construct the bridge entity point cloud. Segment the area in the entity point cloud where the scanner can be set up and the target can be placed and discretize it into equidistant grid points as the candidate point set for the measuring station and target. S2. Under unconstrained conditions, the entity point cloud is virtually pre-scanned to obtain the surface point cloud, downsample and segment special areas as the bridge simulation scene; S3. According to user needs and scanner parameters, set the constraints of global area point cloud density, special area point cloud density, maximum incident angle, maximum scanning distance and scanning angle resolution range, carry out ray tracing simulation analysis on the bridge simulation scene under the constraints, and use the weighted greedy algorithm to optimize the two-stage station layout, select the optimal station position, scanning angle resolution and scanning azimuth of the high scanning angle resolution station, and plan the path of the optimal station through the ant colony algorithm based on the station layout optimization results, with the shortest scanner moving path and the maximum number of overlapping points of adjacent stations. S4. According to the station path planning results, cluster the neighboring stations into station groups. Based on the target candidate point set and the distribution of the station groups, optimize the target layout with the principle of minimizing the station-target mutual visibility and target coordinate covariance condition number, and find the optimal target position for each station group. S5. Summarize the optimization results of the station and target arrangement and output a scanning plan. The scanning plan includes the optimal station position and station group, the optimal target position of each station group, the scanning angle resolution, the scanning azimuth angle of the high scanning angle resolution station, and the scanning path.

2. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The specific process of the virtual pre-scan in S2 is as follows: First, according to the scanning planning experience of measurement professionals, a group of virtual pre-scanning station locations are selected; Secondly, based on the station positions of the virtual pre-scan, ray tracing simulation analysis is performed on the physical point cloud under unconstrained conditions, and all visible points are saved as the bridge surface point cloud.

3. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The specific process of the two-stage station layout optimization in S3 is as follows: First, a panoramic scanning mode with low scanning angle resolution is used to count the locations of areas that meet the constraints in the scene and determine whether special areas meet the constraints. Secondly, if the constraint conditions are not met, a high scanning angle resolution limited azimuth scanning mode simulation is first used for the special area, and then a low scanning resolution panoramic scanning mode simulation is performed; Thirdly, according to the number of scene points that meet the constraints, the weight coefficient of each candidate station point is calculated, and the station with the largest weight is selected iteratively until all scanning points meet the completeness requirements, or the difference in coverage between two adjacent iterations is less than the specified threshold. Then the iteration is stopped to obtain a set of optimal station positions and scanning angle resolutions.

4. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The specific process of path planning in S3 is as follows: Firstly, the ant colony algorithm is used to assign a number of ants to each station. According to the distance between stations and the scanning overlap rate, the heuristic function calculates the transfer probability of each ant to the next station. The next station is selected by the roulette method. The order of stations that the ants have passed and the length of the path are recorded until all stations are visited. The elite ants are selected by sorting the total length of each ant's path in ascending order. Repeat the above steps until the preset number of iterations is met. At this time, the order of the stations passed by the elite ant and the path length are the global path planning.

5. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The specific process of station grouping in S4 is as follows: First, the median of the distance between all two adjacent stations is counted, and a neighborhood circle is drawn with a station as the center and the median as the radius to perform neighborhood detection. If the number of stations in the neighborhood is equal to or exceeds the threshold, then this station is a core station, otherwise it is a noise station. The threshold is set to 2; Secondly, all the stations in the neighborhood are tested until the noise station is detected. All the detected core stations are considered as stations in the same group, and the noise station is a separate station group. Finally, repeat the above steps for the remaining stations until all stations are divided into multiple station groups.

6. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The specific process of target arrangement optimization in S4 is as follows: Firstly, for the same station group, three target layout points are selected from the visible target candidate points based on the principle of mutual visibility of the targets within the group and the minimum condition number of the coordinate covariance matrix; Secondly, for the adjacent station groups, the two nearest stations are determined according to the path planning results. Based on the principle of mutual visibility of the targets of the two adjacent stations and the minimum condition number of the coordinate covariance matrix, three target positions are selected from the visible target candidate points. The above steps are repeated until all station groups are traversed.

7. The large-scale bridge 3D laser scanning planning method in complex scenes according to claim 1 is characterized by: The maximum incident angle is no greater than 70 degrees.

Images