An elevator shaft real-time modeling monitoring method based on multiple planar array lidars

Abstract

The application provides an elevator shaft real-time modeling monitoring method based on multiple planar array laser radars, comprising: calibrating the pose relationship of multiple laser radars on the elevator car towards each face of the inner wall of the elevator shaft, and establishing current laser radar point cloud data; splicing the laser radar point cloud data of adjacent time points to obtain a pose change initial value; dividing the space in the elevator shaft into multiple grids, calculating the PCA result of each grid, and extracting all planes in the laser radar point cloud data as first plane parameters according to the eigenvalues of the PCA result; matching according to the pose change initial value and the first plane parameters, calculating the matching cost through a function, and obtaining the pose change initial value with the minimum matching cost; and merging the pose change initial values with the minimum matching cost at different time to obtain a scanning modeling result. The application solves the problems of the existing scanning modeling method of the elevator shaft through a drone, such as complex operation, low timeliness, and poor precision.

Description

TECHNICAL FIELD

[0001] The present application relates to the elevator technical field, and in particular, to an elevator shaft real-time modeling monitoring method based on multiple planar array laser radars. BACKGROUND

[0002] In the safety operation monitoring of the elevator, the real-time change of the shaft is of great significance, especially the geometric parameters of the shaft. At present, the 3D modeling and measurement method of the elevator shaft has begun to use laser radar as a sensor. Through high-precision guide rails or other methods with precise control sensors, the laser radar moves in the shaft, and then the laser point cloud is spliced to realize the 3D modeling and measurement of the shaft.

[0003] The method based on guide rails requires a very complicated layout process, high labor operation requirement and low safety factor. Other methods include a mechanical rotating laser radar with a 360-degree view carried under a drone to scan and reconstruct the shaft. A high-precision inertial navigation device and a range finder are carried on the drone to record the attitude and position of the flight in real time. Finally, the laser scanning data is spliced according to these information to complete the shaft modeling and measurement. The method based on the drone has a high degree of automation, but it also needs to configure high-performance auxiliary sensors, and the operation is relatively complex. Moreover, the drone cannot complete the measurement in real time during the operation of the elevator, and the timeliness is low. In addition, the number of lines of the rotating multi-line laser radar is generally small, and the angular resolution is low. The accuracy of the shaft reconstruction is difficult to guarantee. The cost of the surveying and mapping level high-precision laser scanner is too expensive, and it cannot be applied on a large scale. SUMMARY

[0004] The problem solved by the present application is that the existing method of scanning and modeling the elevator shaft by the drone is complex, low in timeliness and poor in accuracy.

[0005] To solve the above problems, the present application provides an elevator shaft real-time modeling monitoring method based on multiple planar array laser radars, which comprises: calibrating the pose relationship of multiple laser radars on the elevator car towards each face of the inner wall of the elevator shaft, establishing the current laser radar point cloud data; splicing the laser radar point cloud data of adjacent time to obtain the initial value of the pose change; dividing the space in the elevator shaft into multiple grids, calculating the PCA result of each grid, and extracting all planes in the laser radar point cloud data as the first plane parameter according to the eigenvalue of the PCA result; matching according to the initial value of the pose change and the first plane parameter, calculating the matching cost through a function, and obtaining the initial value of the pose change with the minimum matching cost; and merging the initial value of the pose change with the minimum matching cost at different times to obtain the scanning modeling result.

[0006] The technical effects achieved by the technical scheme are as follows: the laser radar can obtain better resolution and higher point cloud density, and the cost is lower than that of a surveying laser scanner; the automatic matching method of the point clouds of adjacent time points is used to realize full-automatic point cloud splicing and modeling, without the cooperation of additional precise mechanisms such as range finders and timers, and the operation is simple and time-efficient; in view of the special environment of the elevator shaft, the planar extraction is used as the base element of the point cloud matching, which can greatly improve the reliability of the full-automatic point cloud matching and splicing and improve the precision.

[0007] Further, the calibration of the pose relationship of the plurality of laser radars on the elevator car towards each surface of the inner wall of the elevator shaft comprises: a plurality of laser radars are arranged circumferentially on the elevator car, and a calibration component is arranged in the scanning overlap area of adjacent laser radars, and the images of the calibration component are calibrated according to the adjacent laser radars to obtain a calibration initial value.

[0008] The technical effects achieved by the technical scheme are as follows: the combination of the plurality of planar laser radars, and the scanning overlap area of adjacent laser radars can realize 360° coverage and more comprehensive detection of the circumferential image of the elevator car; the images of adjacent laser radars can be obtained in the scanning overlap area, and the position relationship and the pose relationship of the adjacent laser radars are calibrated according to the images obtained by the two, so as to facilitate the splicing of the laser radar point cloud data.

[0009] Further, the calibration of the pose relationship of the plurality of laser radars on the elevator car towards each surface of the inner wall of the elevator shaft comprises: according to the calibration initial value, an iterative closest point algorithm is used to obtain a calibration accurate value.

[0010] The technical effects achieved by the technical scheme are as follows: the images of adjacent laser radars can be accurately spliced by the iterative closest point algorithm, and the accuracy of the laser radar point cloud data is improved.

[0011] Further, the splicing of the laser radar point cloud data of adjacent time points to obtain a pose change initial value comprises: the pose of the point cloud of adjacent time points is determined according to the running speed of the elevator to obtain the pose change initial value; wherein the up-down translation change amount during the running of the elevator is calculated by multiplying the running speed of the elevator by the running time, and the translation and rotation change amounts in other directions during the running of the elevator are set to 0.

[0012] The technical effects achieved by the technical scheme are as follows: during the running of the elevator, the up-down translation mainly occurs, and the translation and rotation in other directions are less, and the change of the pose can be quickly obtained according to the up-down translation change amount and the running time during the running of the elevator.

[0013] Further, the calculating PCA result of each grid, and extracting all planes in the laser radar point cloud data according to eigenvalues of the PCA result as the first plane parameter, comprises: calculating PCA result of each grid to obtain three eigenvalues of the PCA result, wherein the maximum value among the three eigenvalues is a, the intermediate value is b, and the minimum value is c; if any one of a-b≤k1 and a / b≤k2 is satisfied, and any one of b-c≥k3 and b / c≥k4 is satisfied, it is judged that the laser radar point cloud data in the grid is planar-like point cloud data, otherwise, it is non-planar point cloud data; wherein k1, k2, k3 and k4 are constants.

[0014] After the technical scheme is adopted, the following technical effects are achieved: if any one of a-b≤k1 and a / b≤k2 is satisfied, and any one of b-c≥k3 and b / c≥k4 is satisfied, that is, one of the three eigenvalues of the PCA result is much smaller than the other two, and the other two eigenvalues are close in size, at this time, the grid reflects that it is close to a plane, and in other cases, the grid reflects that it is an irregular surface.

[0015] Further, the calculating PCA result of each grid, and extracting all planes in the laser radar point cloud data according to eigenvalues of the PCA result as the first plane parameter, further comprises: searching the grid around the grid corresponding to the planar-like point cloud data, calculating the PCA result again, guiding the calculation of all grids, merging all planar-like point cloud data, ignoring all non-planar point cloud data, and obtaining the first plane parameter.

[0016] After the technical scheme is adopted, the following technical effects are achieved: detecting the grid around the plane can quickly identify the remaining plane, and is convenient for integration of adjacent planes, and accurate first plane parameters are obtained.

[0017] Further, the matching according to the pose change initial value and the first plane parameter, and calculating the matching cost through a function to obtain the pose change initial value with the minimum matching cost, comprises: calculating the Euclidean distance from a point P1 in the first plane parameter at T1 to a face A2 closest to P1 in the first plane parameter at T2, and taking the sum of the Euclidean distances of all points P1 to the face A2 as the matching cost; based on the pose change initial value, setting a search range, and searching for the pose change initial value with the minimum matching cost in the search range.

[0018] After the technical scheme is adopted, the following technical effects are achieved: based on the point-to-plane cost function on the basis of the given pose change initial value, the possible pose range is searched, the stitching parameters are solved, and the most matched pose change initial value can be accurately obtained.

[0019] Further, the setting the search range based on the pose change initial value comprises: setting the search range in an X direction ±L1 range, a Y direction ±L2 range, a Z direction ±L3 range, an X axis rotation ±α angle range, a Y axis rotation ±β angle range, and a Z axis rotation ±γ angle range of the pose change initial value; wherein L1, L2, L3, α, β, and γ are constants.

[0020] The technical effects achieved after adopting the technical scheme are as follows: there is a certain translation or rotation error between the pose change initial value and the first plane parameter, translation search is performed based on the X axis, the Y axis, and the Z axis, and rotation search is performed based on the X axis, the Y axis, and the Z axis, which can cover matching to all possible first plane parameters, so that the matching cost is accurately calculated, and the best deflection angle and displacement amount estimation are obtained, and the class plane point cloud data is spliced into a whole.

[0021] Further, the elevator shaft real-time modeling monitoring method further comprises: setting a D1 time period, and eliminating the scanning modeling result before the D1 time.

[0022] The technical effects achieved after adopting the technical scheme are as follows: the continuous updating of the elevator shaft scanning modeling result is realized.

[0023] The application also provides an elevator shaft real-time modeling monitoring device based on multiple planar array laser radars, which is used to realize the elevator shaft real-time modeling monitoring method provided in any of the embodiments.

[0024] The technical effects achieved after adopting the technical scheme are as follows: any one or more of the technical effects of the above-mentioned embodiments can be achieved.

[0025] In summary, the above-mentioned technical solutions of the application can have one or more of the following advantages or beneficial effects: i) the laser radar can obtain better resolution and higher point cloud density, and the cost is lower than that of a surveying laser scanner; ii) the automatic matching method of the point clouds of adjacent time points is used to realize full-automatic point cloud splicing and modeling, without the cooperation of additional precise mechanisms such as range finders and timers, and the operation is simple and time-efficient; iii) for the special environment of the elevator shaft, the plane extraction is used as the primitive of the point cloud matching, which can greatly improve the reliability of the full-automatic point cloud matching and splicing and improve the precision. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 A flowchart of an elevator shaft real-time modeling monitoring method based on multiple planar array laser radars provided in an embodiment of the application;

[0027] Figure 2 A structural schematic diagram of an elevator shaft real-time modeling monitoring device based on multiple planar array laser radars provided in an embodiment of the application;

[0028] Figure 3 For Figure 2 Another perspective structural diagram.

[0029] Reference signs:

[0030] 100 - elevator shaft real-time modeling monitoring device; 110 - laser radar; 200 - elevator car; 300 - elevator shaft. DETAILED DESCRIPTION

[0031] The purpose of the present application is to provide an elevator shaft real-time modeling monitoring method based on multiple area array laser radars, which is used to achieve better resolution and higher point cloud density, and improve timeliness and accuracy.

[0032] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0033] Reference Figures 1-3 , the embodiment of the present application provides an elevator shaft real-time modeling monitoring method based on multiple area array laser radars, which comprises: calibrating the pose relationship of multiple laser radars on the elevator car towards each face of the inner wall of the elevator shaft, establishing the current laser radar point cloud data; splicing the laser radar point cloud data of adjacent time to obtain the initial value of pose change; dividing the space in the elevator shaft into multiple grids, calculating the PCA result of each grid, and extracting all planes in the laser radar point cloud data as the first plane parameter according to the eigenvalue of the PCA result; matching according to the initial value of pose change and the first plane parameter, calculating the matching cost through a function, and obtaining the initial value of pose change with the minimum matching cost; merging the initial value of pose change with the minimum matching cost at different times to obtain the scanning modeling result.

[0034] In this embodiment, laser radar can obtain better resolution and higher point cloud density, and the cost is lower than that of surveying and mapping level laser scanner; the automatic matching method of adjacent time point cloud is used to realize full-automatic point cloud splicing and modeling, without the cooperation of additional range finders and timers and other precise mechanisms, which is simple to operate and has high timeliness; for the special environment of the elevator shaft, plane extraction is used as the primitive of point cloud matching, which can greatly improve the reliability of full-automatic point cloud matching and splicing, and improve the accuracy.

[0035] In one specific embodiment, the calibration of the pose relationship of multiple laser radars on the elevator car towards each face of the inner wall of the elevator shaft comprises: arranging multiple laser radars on the elevator car in a circumferential direction, and arranging calibration components in the scanning overlap area of adjacent laser radars, and calibrating the images of the calibration components according to adjacent laser radars to obtain the initial value of calibration.

[0036] It should be noted that the combination of multiple planar laser radars, the adjacent laser radars are provided with a scanning overlap area, which can realize the coverage of 360° view angle, and more comprehensively detect the image of the elevator car circumferentially; the adjacent laser radars can obtain images in the scanning overlap area, and the position relationship and attitude relationship of the adjacent laser radars are calibrated according to the images obtained by the two, so as to facilitate the splicing of the laser radar point cloud data.

[0037] Preferably, the number of laser radars is 4, for example, and the interval is 90°, and they are directed to four directions of the elevator shaft, wherein the horizontal field of view angle of each laser radar is 120°, and the vertical field of view angle is 22.5°.

[0038] Preferably, the calibration component is a three-dimensional component with concave and convex grids, and the three-dimensional component has a rectangular plane close to the inner wall of the elevator shaft. Among them, the adjacent laser radars align the concave grid and the convex grid of the three-dimensional component in a manual manner to obtain the initial calibration value.

[0039] In a specific embodiment, the calibration of the position and posture of the multiple laser radars on each surface of the elevator car towards the inner wall of the elevator shaft also includes: according to the initial calibration value, the iterative closest point algorithm is used to obtain the accurate calibration value.

[0040] It should be noted that the iterative closest point algorithm can accurately splice the images of adjacent laser radars and improve the accuracy of laser radar point cloud data.

[0041] In a specific embodiment, the laser radar point cloud data at adjacent time is spliced to obtain the initial value of the change of position and posture, which includes: judging the position and posture of the point cloud at adjacent time according to the running speed of the elevator to obtain the initial value of the change of position and posture; wherein the up-down translation change amount during the running of the elevator is calculated by the running speed of the elevator multiplied by the running time, and the translation and rotation change amount in other directions during the running of the elevator is set to 0.

[0042] It should be noted that during the running of the elevator, the up-down translation mainly occurs, and the translation and rotation in other directions are less, and the change of the position and posture can be quickly obtained according to the up-down translation change amount and the running time during the running of the elevator.

[0043] In a specific embodiment, the PCA result of each grid is calculated, and all planes in the laser radar point cloud data are extracted according to the eigenvalues of the PCA result as the first plane parameter, including: calculating the PCA result of each grid to obtain three eigenvalues of the PCA result, wherein the maximum value among the three eigenvalues is a, the intermediate value is b, and the minimum value is c; if any one of a-b≤k1 and a / b≤k2 is satisfied, and any one of b-c≥k3 and b / c≥k4 is satisfied, it is determined that the laser radar point cloud data in the grid is planar-like point cloud data, otherwise, it is non-planar point cloud data; wherein k1, k2, k3 and k4 are constants.

[0044] It should be noted that if any one of a-b≤k1 and a / b≤k2 is satisfied, and any one of b-c≥k3 and b / c≥k4 is satisfied, that is, one of the three eigenvalues of the PCA result is much smaller than the other two, and the sizes of the other two eigenvalues are close, at this time, the grid is close to a plane, and in other cases, the grid is an irregular surface.

[0045] In a specific embodiment, the PCA result of each grid is calculated, and all planes in the laser radar point cloud data are extracted according to the eigenvalues of the PCA result as the first plane parameter, further including: searching the grid around the grid corresponding to the planar-like point cloud data, calculating the PCA result again, guiding the calculation of all grids, merging all planar-like point cloud data, ignoring all non-planar point cloud data, and obtaining the first plane parameter.

[0046] It should be noted that detecting the grid around the plane can quickly identify the remaining plane, and facilitate the integration of adjacent planes to obtain accurate first plane parameters.

[0047] In a specific embodiment, the matching is performed according to the pose change initial value and the first plane parameter, the matching cost is calculated by a function, and the pose change initial value with the minimum matching cost is obtained, including: calculating the Euclidean distance from the point P1 in the first plane parameter at T1 to the face A2 closest to P1 in the first plane parameter at T2, and taking the sum of the Euclidean distances from all points P1 to the face A2 as the matching cost; based on the pose change initial value, a search range is set, and the pose change initial value with the minimum matching cost is searched in the search range.

[0048] It should be noted that based on the point-to-plane cost function under the given pose change initial value, the possible pose range is searched, and the splicing parameter is solved, so that the most matched pose change initial value can be accurately obtained.

[0049] In a specific embodiment, based on the pose change initial value, the search range is set to include: in the X direction ± L1 range, in the Y direction ± L2 range, in the Z direction ± L3 range, and in the X axis rotation ± α angle range, in the Y axis rotation ± β angle range, and in the Z axis rotation ± γ angle range.

[0050] It should be noted that there is a certain translation or rotation error between the pose change initial value and the first plane parameter. Based on the X axis, Y axis and Z axis, the translation search is performed, and then based on the X axis, Y axis and Z axis, the rotation search is performed. All possible first plane parameters can be included to accurately calculate the matching cost, obtain the best deflection angle and displacement estimation, and splice the planar point cloud data into a whole.

[0051] Preferably, L1, L2, L3, α, β, and γ are constants. For example, L1, L2, and L3 are all 10 cm, and α, β, and γ are all 5°.

[0052] Preferably, when performing translation search, the translation search interval is, for example, 1 cm, and when performing rotation search, the angle search interval is, for example, 0.05°, to obtain accurate search results, which are not limited here.

[0053] In a specific embodiment, the elevator shaft real-time modeling monitoring method further includes: setting a D1 time period, and excluding the scanning modeling results before D1 time to realize continuous updating of the elevator shaft scanning modeling results. For example, D1 is, for example, 10 minutes.

[0054] Referring to Figures 2-3 The embodiments of the present application also provide an elevator shaft 300 real-time modeling monitoring device 100 based on a plurality of planar array laser radars 110, which is used to realize the elevator shaft 300 real-time modeling monitoring method provided by any of the above embodiments. Any one or more of the above technical effects can be achieved.

[0055] Preferably, the elevator shaft 300 real-time modeling monitoring device 100 based on multiple planar array laser radars 110 includes 4 laser radars 110 located on the elevator car 200 and facing each face of the inner wall of the elevator shaft 300, and the adjacent laser radars 110 are at an angle of 90°, for establishing the current laser radar 110 point cloud data. The elevator shaft 300 real-time modeling monitoring device 100 further includes a calculation module, for splicing the laser radar 110 point cloud data at adjacent time points to obtain a pose change initial value, dividing the space in the elevator shaft 300 into multiple grids, calculating the PCA result of each grid, extracting all planes in the laser radar 110 point cloud data according to the eigenvalues of the PCA result as first plane parameters, matching according to the pose change initial value and the first plane parameters, calculating the matching cost through a function, obtaining the pose change initial value with the minimum matching cost, and merging the pose change initial values with the minimum matching cost at different time points to obtain a scanning modeling result.

[0056] Although the present application has been disclosed as above, it is not limited to the above. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application, and the protection scope of the present application should be subject to the scope defined by the claims.

Claims

1. A real-time modeling and monitoring method for elevator shafts based on multiple array laser radars, characterized in that: The elevator shaft real-time modeling and monitoring method comprises: Calibrate the pose relationships of multiple LiDAR sensors on the elevator car facing the inner walls of the elevator shaft to create the current LiDAR point cloud data; splicing the lidar point cloud data at adjacent moments to obtain an initial value of the pose change, including: determining the poses of the point clouds at adjacent moments according to the elevator running speed to obtain the initial value of the pose change; wherein the vertical translation change during the elevator running is calculated by multiplying the elevator running speed by the running time, and the translation and rotation changes in other directions during the elevator running are set to 0; Divide the space in the elevator shaft into multiple grids, calculate the PCA result of each grid, and extract all planes in the laser radar point cloud data according to the eigenvalues ​​of the PCA result as the first plane parameters; including: calculating the PCA result of each grid to obtain three eigenvalues ​​of the PCA result, wherein the maximum value of the three eigenvalues ​​is a, the middle value is b, and the minimum value is c. If any one of ab≤k1 and a / b≤k2 is satisfied, and any one of bc≥k3 and b / c≥k4 is satisfied, then the laser radar point cloud data in the grid is judged to be planar point cloud data, otherwise it is non-planar point cloud data; wherein k1, k2, k3, and k4 are constants; Matching the initial value of the pose change and the first plane parameter, calculating the matching cost through a function, and obtaining the initial value of the pose change with the minimum matching cost; including: calculating the Euclidean distance from point P1 in the first plane parameter at time T1 to the surface A2 in the first plane parameter closest to P1 at time T2, and taking the sum of the Euclidean distances from all points P1 to surface A2 as the matching cost; setting a search range based on the initial value of the pose change, and searching for the initial value of the pose change with the minimum matching cost within the search range; The initial values ​​of the posture changes with the minimum matching costs at different times are combined to obtain a scanning modeling result.

2. The elevator shaft real-time modeling and monitoring method according to claim 1 is characterized in that: The calibration pose relationship of multiple laser radars on the elevator car facing each surface of the inner wall of the elevator shaft includes: A plurality of laser radars are circumferentially arranged on the elevator car, and a calibration component is arranged in the scanning overlap area of ​​adjacent laser radars. The calibration component is calibrated according to the image of the adjacent laser radar to obtain an initial calibration value.

3. The elevator shaft real-time modeling and monitoring method according to claim 2 is characterized in that: The calibration posture relationship of multiple laser radars on the elevator car facing each surface of the inner wall of the elevator shaft also includes: According to the initial calibration value, an iterative closest point algorithm is used to obtain an accurate calibration value.

4. The elevator shaft real-time modeling and monitoring method according to claim 1, characterized in that: The step of calculating the PCA result of each grid and extracting all planes in the laser radar point cloud data as first plane parameters according to the eigenvalues ​​of the PCA result further includes: Search for grids around the grid corresponding to the plane-like point cloud data, calculate the PCA result again, guide the calculation of all grids, merge all plane-like point cloud data, ignore all non-planar point cloud data, and obtain the first plane parameters.

5. The elevator shaft real-time modeling and monitoring method according to claim 1, characterized in that: The setting of the search range based on the initial value of the posture change includes: The search range is set within the range of ±L1 in the X direction, ±L2 in the Y direction, ±L3 in the Z direction, and within the range of ±α angle rotation along the X axis, ±β angle rotation along the Y axis, and ±γ angle rotation along the Z axis of the initial value of the posture change; Among them, L1, L2, L3, α, β, and γ are constants.

6. The elevator shaft real-time modeling and monitoring method according to claim 1, characterized in that: The elevator shaft real-time modeling and monitoring method further includes: A D1 time period is set, and the scanning modeling results before D1 are discarded.

7. A real-time modeling and monitoring device for elevator shafts based on multiple array laser radars, characterized in that: Used to implement The elevator shaft real-time modeling and monitoring method according to any one of claims 1 to 6.

Images