A method for evaluating mapping accuracy of indoor laser radar scanning

Abstract

The application discloses a kind of methods for evaluating indoor laser radar scanning mapping precision, to solve the problem of dependence of existing evaluation technology on expensive true value system and difficulty in precision evaluation in large scene.The application does not need complex equipment, only needs to select several groups of well-parallel wall surfaces in indoor space in accordance with building specifications.By pasting reflective markers on the wall surfaces, the distance between these parallel wall surfaces can be measured using high-precision ranging equipment, and a series of spatial geometric constraints with known true values can be quickly obtained.The application is simple in process operation and extremely low in cost, and the mapping precision is accurately evaluated by quantifying key indicators.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of three-dimensional scanning, in particular to a method for evaluating the mapping accuracy of indoor laser radar scanning. BACKGROUND

[0002] With the maturity of laser radar (LiDAR) technology, three-dimensional scanning and reconstruction technology has been widely applied in the field of architecture, engineering and construction (AEC). Among them, the construction of indoor three-dimensional point cloud map provides a key spatial data basis for applications such as building digital management, BIM modeling, indoor navigation, and reconstruction design.

[0003] At present, indoor scanning mainly relies on two types of equipment: handheld laser radar and station-mounted laser radar. Handheld devices usually use SLAM (simultaneous localization and mapping) technology, which estimates its own pose (position and attitude) in real time while moving and splices consecutive laser frames to incrementally construct a map. However, the core challenge of the SLAM algorithm is that the pose estimation error will accumulate over time, causing the map to distort, overlap, or produce "ghost" phenomena. If you want to evaluate the mapping accuracy with high precision, you usually need to rely on an optical motion capture system as a pose truth reference. However, such a system is not only expensive, but also requires a large number of marker points to be pre-deployed, which is difficult to deploy in large-scale or complex indoor scenes, which greatly limits the precision authentication and wide application of handheld laser radar in large-scale indoor mapping scenarios.

[0004] On the other hand, although station-mounted laser radar has high measurement accuracy at a single station, when constructing a large scene map, the scanning results of multiple stations need to be spliced. The splicing and registration process between stations will introduce errors, which will also accumulate and spread with the increase in the number of stations, ultimately affecting the absolute accuracy and consistency of the overall map. At present, for the precision evaluation of station-mounted scanning results, there is also a lack of efficient and comprehensive data flow evaluation scheme, often relying on manual experience judgment or using total station equipment for spot checks, which is inefficient and difficult to fully reflect the overall map quality.

[0005] Therefore, whether it is for handheld SLAM devices or station-mounted scanning solutions, the industry urgently needs an indoor point cloud map precision calculation method that does not rely on expensive truth systems and can be implemented automatically to effectively evaluate and constrain the mapping error and ensure the reliability of three-dimensional spatial data. SUMMARY

[0006] The purpose of the present application is to provide a method for evaluating the mapping accuracy of indoor laser radar scanning, aiming to solve the problem of dependence on expensive truth systems (such as motion capture systems) and difficulty in precision evaluation in large scenes in existing evaluation techniques.

[0007] To achieve the above object, the technical scheme adopted by the present application is as follows:

[0008] A method for evaluating the mapping precision of indoor laser radar scanning, comprising the following steps:

[0009] S1, selecting a to-be-evaluated scene with parallel walls, pasting reflective markers on the walls, and measuring and recording the real distances between the parallel walls using a high-precision ranging device;

[0010] S2, scanning the to-be-evaluated scene using a laser radar scanning device to generate a point cloud map;

[0011] S3, based on the point cloud intensity value, identifying and extracting the point cloud clusters corresponding to the reflective markers from the point cloud map;

[0012] S4, taking each point cloud cluster as the core, gradually expanding outward, and extracting the complete point cloud data of the wall surface;

[0013] S5, performing plane fitting on the point cloud data of each wall surface after expansion to obtain the plane equation corresponding to each wall surface;

[0014] S6, based on the plane equation, calculating the measured distance between the parallel wall pairs in the point cloud map, comparing the measured distance with the real distance recorded in S1, and calculating the absolute distance error;

[0015] S7, analyzing the normal vector of the fitted plane equation to calculate the direction consistency error;

[0016] S8, based on the absolute distance error and the direction consistency error, quantitatively evaluating the mapping precision of the laser radar scanning.

[0017] Further, S1 comprises: S11, selecting multiple candidate parallel wall pairs in an indoor environment where the mapping precision needs to be evaluated, measuring the distance between the same wall pairs at multiple different positions using a high-precision ranging device and calculating the variance, and selecting the wall pairs with a variance less than a first preset threshold for evaluation, the first preset threshold being 3mm 2 ; S12, pasting reflective markers on the selected walls, and accurately measuring and recording the real distances between the parallel wall pairs.

[0018] Further, the reflective marker is a reflective strip, and the size of the reflective strip is not less than 30cm 30cm.

[0019] Further, step S3 includes: S31, filtering according to the intensity attributes of the point cloud, extracting points in the point cloud map whose intensity values ​​are higher than the second preset threshold, forming a set of high-brightness cloud, where the second preset threshold is the maximum value of the radar intensity range of the lidar; S32, performing Euclidean clustering analysis on the set of high-brightness cloud, dividing spatially adjacent high-brightness cloud into different point cloud clusters, with each point cloud cluster corresponding to a reflective marker.

[0020] Further, S4 includes: S41, calculating the directed bounding box of each point cloud cluster and using it as the initial search range; S42, based on the initial search range, iteratively expanding the region in the world coordinate system; after each iteration of expansion, searching for point clouds within the expanded region; S43, calculating the distance from the searched point cloud to the plane fitted to the corresponding initial point cloud cluster, retaining point clouds whose distance is less than a preset plane distance threshold and aggregating them into a single point cloud dataset.

[0021] Furthermore, in S42, each iteration is directed to the world coordinate system... X , Y , Z Expand by 0.3m in each direction, and iterate at least 10 times;

[0022] The iterative method is as follows: S421, Obtain a pair of diagonal vertices of the directed bounding box. and , , All are point cloud coordinates of vertices, with an extended region of r = 0.3 =(1,1,1); S422, let , Calculate the expanded region after the first iteration. If point clouds exist in the expanded region after the first iteration... satisfy Then dotted clouds The point cloud after the search, For parallel condition symbols; S423, each point cloud cluster must repeat operations S421 and S422 at least 10 times;

[0023] In S43, the preset plane distance threshold is 0.03m;

[0024] In S43, distance The calculation formula is: ,in These are the coordinates of the point cloud after the search. a , b , c These are the fitting planes corresponding to the initial point cloud clusters. of a i ,b i 、 c i 。

[0025] Further, in S5, the point cloud data of each wall surface after extension is subjected to plane fitting by using principal component analysis method.

[0026] Further, S6 comprises: S61, a pair of plane equations corresponding to the parallel wall surface pair measured in S1 is manually specified or automatically matched in the point cloud map; S62, the average distance is calculated for each pair of plane equations and taken as the measured distance; S63, the root mean square error between the measured distance of each parallel wall surface pair and the true distance recorded in S1 is calculated and taken as the absolute distance error.

[0027] Further, S7 comprises: S71, the fitted plane equations are divided into three groups perpendicular to the x axis, y axis and z axis, when the component on the x axis in the normal vector of the fitted plane equation is the largest, the plane equation is classified into the group perpendicular to the x axis, if the component on the y axis is the largest, the plane equation is classified into the group perpendicular to the y axis, if the component on the z axis is the largest, the plane equation is classified into the group perpendicular to the z axis; S72, the normal vectors of all plane equations in each group are subjected to normalization processing to obtain three normalized normal vectors; S73, the three normalized normal vectors are subjected to mean value, average vector length and error calculation in sequence to obtain the consistency error.

[0028] Further, in S8, the smaller the absolute distance error is, the higher the accuracy of the laser radar scanning mapping is, and vice versa; the closer the consistency error is to 0, the higher the accuracy of the laser radar scanning mapping is, and vice versa.

[0029] Compared with the prior art, the present application has the following beneficial effects:

[0030] The present application firstly proposes a standardized scene arrangement scheme with low cost and high convenience, without complex equipment, only a few groups of walls with good parallelism are selected in an indoor space (such as a corridor or a room) conforming to the building specification. By pasting reflective markers (such as low-cost reflective marker strips) on the walls and using high-precision distance measuring equipment to measure the distance between the parallel walls, a series of spatial geometric constraints (distance, parallelism) with known true values can be quickly obtained. This process is simple to operate and has extremely low cost, providing reliable true value references for accuracy evaluation.

[0031] In terms of data processing and precision calculation, the core of the application lies in the adoption of an innovative extended fitting algorithm. This algorithm first efficiently filters out the reflective marker point cloud based on the point cloud intensity value, and then accurately extracts and fits the plane equation of the entire wall surface. Compared with traditional methods, this algorithm can more effectively overcome problems such as sparse point cloud and noise interference, significantly improving the completeness of plane feature extraction and the accuracy of equation expression.

[0032] Finally, the application comprehensively quantifies and evaluates the mapping accuracy from two dimensions of plane consistency and distance error: on the one hand, by calculating the angle deviation between the fitted planes that should be parallel, the directional consistency accuracy of the scanning data is evaluated; on the other hand, by calculating the distance between the fitted planes and comparing it with the true value distance measured by the ranging device, the absolute distance error is accurately quantified. This method not only effectively evaluates the overall geometric accuracy of the point cloud map, but also provides intuitive and reliable basis for optimizing the SLAM algorithm and sensor calibration. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The method flowchart of the application. DETAILED DESCRIPTION

[0034] In order to make the purpose, technical scheme and advantages of the application clearer, the application will be further described in detail below with reference to the drawings. Obviously, the described embodiments are only part of the embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the application.

[0035] In the description of the application, it should be noted that the terms "first", "second", "third" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance.

[0036] As Figure 1 shown, the application provides a method for evaluating the scanning mapping accuracy of indoor laser radar, comprising the following steps:

[0037] S1, selecting a scene to be evaluated with parallel walls, pasting reflective markers on the walls, and using a high-precision ranging device to measure and record the true distance between the parallel walls;

[0038] S2, using a laser radar scanning device to scan the scene to be evaluated, generating a point cloud map;

[0039] S3, based on the point cloud intensity value, identifying and extracting the point cloud cluster corresponding to the reflective markers from the point cloud map;

[0040] S4, taking each point cloud cluster as the core, gradually expanding outward, and extracting complete point cloud data of the wall surface;

[0041] S5, performing plane fitting on the point cloud data of each expanded wall surface to obtain a plane equation corresponding to each wall surface;

[0042] S6, based on the plane equation, calculating the measured distance between parallel wall surfaces in the point cloud map, comparing the measured distance with the true distance recorded in S1, and calculating the absolute distance error;

[0043] S7, analyzing the normal vector of the fitted plane equation to calculate the direction consistency error;

[0044] S8, based on the absolute distance error and the direction consistency error, quantitatively evaluating the accuracy of the laser radar scanning mapping.

[0045] The present application does not require complex equipment, only needs to select several groups of well-parallel walls in indoor spaces (such as corridors, rooms) that meet the building specifications. By pasting low-cost reflective marker strips on the walls as feature markers and using a high-precision single-point laser range finder to measure the distance between these parallel walls, a series of spatial geometric constraints (distance, parallelism) with known true values can be quickly obtained.

[0046] Preferably, the high-precision ranging device of the present application is a single-point laser range finder, and the laser radar scanning device is a handheld SLAM laser radar scanner or a stand-mounted laser radar scanner.

[0047] In some embodiments, S1 includes: S11, selecting multiple candidate parallel wall pairs in an indoor environment where the mapping accuracy needs to be evaluated, measuring the distance between the same wall pairs at multiple different positions using a high-precision ranging device and calculating the variance, and selecting wall pairs with a variance less than a first preset threshold for evaluation, the first preset threshold being 3mm 2 ; S12, pasting reflective markers on the selected walls, accurately measuring and recording the true distance between each parallel wall pair. The selected parallel walls should meet the requirement of good parallelism, i.e. being flat and as parallel as possible.

[0048] The following is a specific embodiment, selecting multiple parallel wall pairs for the following tests: randomly selecting 10 different positions on one wall and placing a single-point range finder (accuracy error < 1mm) vertically to the wall, then measuring the distance to the opposite wall, recording the measurement result as , calculating the variance (unbiased estimate of the population variance) of multiple measurement results by the formula If the variance is < 3mm 2If the two walls are parallel, then they are considered parallel. Using this method, at least one pair of walls that meet the parallelism requirement must be found in the space where the mapping accuracy needs to be evaluated. Then, for each selected wall, a highly reflective reflective strip (at least 30cm in size) is used. Mark the location on the wall with a 30cm marker. Similarly, select 10 different locations and use a single-point distance meter (accuracy error < 1mm) placed perpendicular to the wall, then measure the distance to the opposite wall and record it as [the distance is not specified in the original text]. Calculate the mean As the distance between parallel walls, all measured and calculated distances between walls are recorded as follows. .

[0049] In some embodiments, step S3 includes: S31, filtering based on the intensity attributes of the point cloud, extracting points in the point cloud map whose intensity values ​​are higher than a second preset threshold, forming a set of high-brightness cloud points, where the second preset threshold is the maximum value of the LiDAR's radar intensity range (e.g., if the LiDAR's own radar intensity value range is 0-255, then the second preset threshold is set to 255); S32, performing Euclidean clustering analysis on the set of high-brightness cloud points, dividing spatially adjacent high-brightness cloud points into different point cloud clusters, with each point cloud cluster corresponding to a reflective marker.

[0050] Since the target walls are marked with highly reflective strips, the intensity value attribute of the point cloud is used for filtering. Specifically, the complete point cloud in the point cloud map is denoted as F, and all points in F are filtered by intensity value (i.e., points with intensity values ​​higher than a second preset threshold), resulting in a set of high-intensity point clouds denoted as H. Then, the Euclidean clustering method is used to separate k wall marker point clouds, i.e., k point cloud clusters, from the high-intensity point cloud set H. To represent its set: .

[0051] In some embodiments, S4 includes: S41, calculating the directed bounding box of each point cloud cluster and using it as the initial search range; S42, based on the initial search range, iteratively expanding the region in the world coordinate system; after each iteration of expansion, searching for point clouds within the expanded region; S43, calculating the distance from the searched point cloud to the plane fitted to the corresponding initial point cloud cluster, retaining point clouds with a distance less than a preset plane distance threshold and aggregating them into a single point cloud dataset, where the preset plane distance threshold is 0.03m.

[0052] Specifically, the calculation of the directed bounding box (OBB) is as follows: First, perform PCA (Principal Component Analysis). PCA yields three principal directions: {PC1 (direction with the largest variance), PC2 (direction with the second largest variance), and PC3 (direction with the smallest variance)}. The coordinate system formed by these three principal directions is called the PCA coordinate system. The point cloud is first transformed to the PCA coordinate system, and the maximum and minimum coordinate points (max_point and min_point) are obtained. Then, the original coordinate system to which max_point and min_point are restored is denoted as […]. and This completes the calculation of the OBBs of all sub-point clouds in the point cloud cluster.

[0053] Because the local reflective point cloud does not cover the wall surface sufficiently, local irregularities may cause the feature vector calculated for the entire wall surface using PCA to be skewed by local noise. Therefore, increasing the coverage of the wall surface point cloud can effectively reduce the impact of local noise. Meanwhile, according to the Davis-Kahan theorem, the estimation error of the normal vector... It is readily apparent that increasing the sampling size n can effectively reduce the error in normal vector estimation. Therefore, it is necessary to use the point cloud set of highly reflective markers. Expand outwards to read as many point cloud data points as possible related to the wall surface to improve the final estimate of the plane on which the wall surface is located.

[0054] Then, based on the initial search range, the search range is iteratively expanded step by step in the world coordinate system. Each iteration expands the search range to the world coordinate system... X , Y , Z Extend by 0.3m in each direction, and iterate at least 10 times. The specific iteration method is as follows: S421, Obtain a pair of diagonal vertices of the directed bounding box. and , , All are point cloud coordinates of vertices, with an extended region of r = 0.3 =(1,1,1); S422, let , Calculate the expanded region after the first iteration. If point clouds exist in the expanded region after the first iteration... satisfy Then dotted clouds The point cloud after the search, These are parallel condition symbols, meaning they satisfy... and and dot clouds The point cloud after the search. Point cloud set. Any point in the cloud After the above operations, expand outward by 0.3m to obtain the point cloud. Repeat the operation more than 10 times to obtain the desired result. X , Y , Z Point cloud extending 3m in each direction , in turn Perform the above operations on all point clouds to obtain... After the above 10 expansions, the point cloud of the wall surface within a 3m range can finally be obtained.

[0055] Finally, the distance from the target point cloud after the search to the plane fitted to the corresponding initial point cloud cluster is calculated. Maintain distance Point clouds with distances less than a preset planar distance threshold are grouped together into a single point cloud dataset. Distance The calculation formula is: ,in These are the coordinates of the point cloud after the search. a , b , c These are the fitting planes corresponding to the initial point cloud clusters. of a i , b i , c i .

[0056] In some embodiments, principal component analysis is used to analyze the point cloud data of each wall surface after expansion in S5. A plane fitting is performed on all point clouds in the dataset, and the result is denoted as... .

[0057] In some embodiments, S6 includes: S61, manually specifying or automatically matching a pair of plane equations corresponding to the parallel wall surfaces measured in S1 in the point cloud map; S62, calculating the average distance for each pair of plane equations and using it as the measured distance; S63, calculating the root mean square error between the measured distance of each parallel wall surface and the actual distance recorded in S1 and using it as the absolute distance error.

[0058] Specifically, display All point clouds are used to manually specify or automatically match the wall surfaces that require subsequent operations through interactive operations, and recorded as follows: , ,Will The specified wall faces correspond one-to-one with the wall faces filtered by S1. Use The distance between the planes represented by each pair of point clouds is calculated. For example, the calculation results in S5 The middle is saved The plane fitting information of all point clouds in the middle is set The corresponding plane fitting information is , traverse All points in, calculate the distance of each point to , select the smallest one as , and select from , calculate the distance D1 of to and the distance D2 of to , take the average of D1 and D2 as , calculate the distance of each point cloud pair in , record as . Calculate the root mean square error between the measured distance of each parallel wall surface and the true distance recorded in S1 .

[0059] In some embodiments, S7 comprises: S71, divide the fitted plane equation into three groups perpendicular to x axis, y axis and z axis, when the component of the normal vector of the fitted plane equation in the x coordinate axis is the largest, classify the plane equation to the group perpendicular to x axis, if the component in the y coordinate axis is the largest, classify to the group perpendicular to y axis, if the component in the z coordinate axis is the largest, classify to the group perpendicular to z axis; S72, normalize the normal vector of all plane equations in each group to obtain three normalized normal vectors; S73, calculate the direction consistency error of the three normalized normal vectors in turn by mean, average vector length and error.

[0060] Specifically, suppose the normal vector of any fitted plane equation is (a', b', c'), and then classify it. If its |a'| (x component) is the largest, it means that the normal vector is mainly along the x axis direction, then the plane is most perpendicular to the x axis, so it is parallel to the YZ plane, if its |b'| (y component) is the largest, it means that the normal vector is mainly along the y axis direction, then the plane is most perpendicular to the y axis, so it is parallel to the XZ plane, if its |c'| (z component) is the largest, it means that the normal vector is mainly along the z axis direction, then the plane is most perpendicular to the z axis, so it is parallel to the XY plane. Classify all planes in according to the above method, and The three groups are respectively recorded as 、 、 . The normal vectors of all plane equations in each group are normalized , and three normalized normal vectors are obtained 、 、 . Then, the three normalized normal vectors are sequentially subjected to mean value, average vector length and error calculation: first, mean value calculation , , then average vector length 、 、 , and finally, consistency error calculation, to obtain the directional consistency error 、 、 .

[0061] After the above calculation, the absolute distance error and the directional consistency error are used to quantitatively evaluate the accuracy of the laser radar scanning mapping. The smaller the absolute distance error, the higher the accuracy of the laser radar scanning mapping, and vice versa. The closer the consistency error is to 0, the higher the accuracy of the laser radar scanning mapping, and vice versa.

[0062] Finally, it should be noted that: the above embodiments are only the preferred embodiments of the present application for explaining the technical solutions of the present application, but not limiting them, and of course, not limiting the patent scope of the present application; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application; that is, any modification or polishing without substantial significance in the main design idea and spirit of the present application, and the technical problems solved are still consistent with the present application, which should be included in the protection scope of the present application; in addition, the technical solutions of the present application are directly or indirectly applied to other related technical fields, which are also included in the patent protection scope of the present application.

Claims

1. A method for evaluating the accuracy of indoor lidar scanning mapping, characterized in that, Includes the following steps: S1. Select the scene to be evaluated with parallel walls, paste reflective markers on the walls, and use a high-precision ranging device to measure and record the actual distance between each parallel wall. S2. Use a lidar scanning device to scan the scene to be evaluated and generate a point cloud map; S3. Based on the point cloud intensity value, identify and extract the point cloud clusters corresponding to the reflection markers from the point cloud map; S4. Using each point cloud cluster as the core, gradually expand outward to extract the complete point cloud data of the wall surface. S5. Perform plane fitting on the point cloud data of each expanded wall surface to obtain the plane equation corresponding to each wall surface. S6. Based on the plane equation, calculate the measured distance between parallel wall surfaces in the point cloud map, compare the measured distance with the actual distance recorded in S1, and calculate the absolute distance error. S7. Analyze the normal vector of the fitted plane equation and calculate the direction consistency error; S8. The accuracy of lidar scanning mapping is quantitatively evaluated based on absolute distance error and direction consistency error. S4 includes: S41, calculating the directed bounding box of each point cloud cluster and using it as the initial search range; S42. Based on the initial search range, iteratively expand the region in the world coordinate system; after each iteration, search for point clouds within the expanded region; S43. Calculate the distance from the searched point cloud to the plane fitted by the corresponding initial point cloud cluster, retain point clouds whose distance is less than the preset plane distance threshold, and aggregate them into a single point cloud data set. In S42, each iteration is directed to the world coordinate system. X , Y , Z Expand by 0.3m in each direction, and iterate at least 10 times; The iterative method is as follows: S421, Obtain a pair of diagonal vertices of the directed bounding box. and , , All are point cloud coordinates of vertices, with an extended region of r = 0.3 =(1,1,1); S422, let , Calculate the expanded region after the first iteration. If point clouds exist in the expanded region after the first iteration... satisfy Then dotted clouds The point cloud after the search, For parallel condition symbols; S423, each point cloud cluster must repeat operations S421 and S422 at least 10 times; In S43, the preset plane distance threshold is 0.03m; In S43, distance The calculation formula is: ,in These are the coordinates of the point cloud after the search. a , b , c These are the fitting planes corresponding to the initial point cloud clusters. of a i , b i , c i .

2. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, S1 includes: S11. In an indoor environment where the mapping accuracy needs to be evaluated, select multiple sets of candidate parallel wall faces. Use a high-precision distance measuring device to measure the distance between the same wall faces at multiple different locations and calculate the variance. Select wall faces with a variance less than a first preset threshold for evaluation. The first preset threshold is 3mm. 2 S12. Affix reflective markers to each of the selected wall surfaces, and accurately measure and record the actual distance between each parallel wall surface.

3. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, The reflective markers are reflective strips, and the reflective strips are at least 30cm in size. 30cm.

4. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, Step S3 includes: S31, filtering based on the intensity attributes of the point cloud, extracting points in the point cloud map whose intensity values ​​are higher than a second preset threshold to form a set of high-brightness cloud points, where the second preset threshold is the maximum value of the radar intensity range of the lidar; S32, performing Euclidean clustering analysis on the set of high-brightness cloud points, dividing spatially adjacent high-brightness cloud points into different point cloud clusters, with each point cloud cluster corresponding to a reflective marker.

5. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, In S5, principal component analysis is used to perform planar fitting on the point cloud data of each expanded wall surface.

6. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, S6 include: S61. In the point cloud map, manually specify or automatically match a pair of plane equations corresponding to the parallel wall surfaces measured in S1; S62. Calculate the average distance for each pair of plane equations and use it as the measured distance; S63. Calculate the root mean square error between the measured distance and the actual distance recorded in S1 for each parallel wall surface and use it as the absolute distance error.

7. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, S7 includes: S71, dividing the fitted plane equation into those perpendicular to... x axis, y shaft and z The three sets of axes, when the normal vector of the fitted plane equation is in x When the components on the coordinate axes are at their maximum, the plane equation is classified as perpendicular to the coordinate axes. x A set of axes, if in y When the component on the coordinate axis is at its maximum, classify it to be perpendicular to the axis. y A set of axes, if in z When the component on the coordinate axis is at its maximum, classify it to be perpendicular to the axis. z S72. Normalize the normal vectors of all plane equations in each group to obtain three normalized normal vectors; S73. Calculate the mean, length of the average vector, and error of the three normalized normal vectors in turn to obtain the direction consistency error.

8. The method for evaluating the accuracy of indoor lidar scanning mapping according to claim 1, characterized in that, In S8, the smaller the absolute distance error, the higher the accuracy of the LiDAR scanning and mapping, and vice versa; the closer the consistency error is to 0, the higher the accuracy of the LiDAR scanning and mapping, and vice versa.

Images