An elevator shaft safety space data acquisition device and three-dimensional reconstruction measurement method

Abstract

The present invention discloses an elevator shaft safety space data acquisition device and a three-dimensional reconstruction measurement method. The operator only needs to place the device in the elevator shaft to automatically complete the three-dimensional data acquisition task of the elevator shaft. The device can fully cover all corners in the elevator shaft, ensuring the integrity and accuracy of data acquisition, and providing high-quality original data for space reconstruction and measurement. The data acquisition angle set by the method is based on empirical values, laying a foundation for subsequent registration. By combining the initial pose transformation and multi-view point cloud registration, the stability of multi-view point cloud registration is ensured. Finally, through the combination of TSDF volume reconstruction, the redundancy of the registered point cloud data is removed. There is no need for the operator to enter and perform real-time control, effectively realizing the three-dimensional data acquisition, reconstruction and measurement of the elevator shaft safety space, with convenient operation and high efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of elevator inspection, and particularly relates to a data acquisition device and a three-dimensional reconstruction measurement method for the safety space of an elevator shaft. Background Art

[0002] The safety spaces at the top and bottom pits of an elevator mainly serve to protect the car and its accessories, the equipment on the car top, and the operators working on the car top and in the bottom pit. Taking the top safety space as an example, the parameters to be measured mainly include the guiding distance of the car guide rail, the distance between two car guide rails, the distance between two counterweight guide rails, the safe vertical distance from the highest surface where people can stand on the car top to the lowest component at the top of the hoistway, the most unfavorable spacing between the components at the top of the hoistway and the equipment fixed on the car top, etc. Currently, the most common detection method requires the inspector to enter the safety space and hold a tape measure for measurement. In some cases, it is also necessary to work when the car is in motion, which poses risks of being squeezed, sheared, and trapped. Moreover, this method is highly dependent on the operator, and the sense of responsibility, experience, and mood of the operator will all affect the detection results. Existing elevator shaft detectors and detection systems need to place multiple laser ranging modules, and the measurement position cannot be changed during the measurement process; they have complex structures, cumbersome operations, and low detection efficiency. Additionally, there is a measurement system with a camera and a lidar that requires an experienced person to control the direction and select points in real time, which also has a dependence on the operator and low automation. Summary of the Invention

[0003] To solve the above technical problems, the present invention provides a data acquisition device and a three-dimensional reconstruction measurement method for the safety space of an elevator shaft, which do not require the operator to enter and do not require the operator to control in real time, effectively realizing the three-dimensional data acquisition, reconstruction, and measurement of the safety space of the elevator shaft, ensuring the absolute safety of the inspection personnel while being convenient to operate and having high efficiency.

[0004] The technical solution adopted by the present invention to solve its technical problems is:

[0005] A data acquisition device for the safety space of an elevator shaft includes a depth camera, a camera connector, a two-dimensional electric pan-tilt, a micro host, and a servo control board. The depth camera is electrically connected to the micro host, and the servo control board is electrically connected to the micro host. The two-dimensional electric pan-tilt includes a base, an L-shaped arm, an upper servo, and a lower servo. The upper servo is connected to the L-shaped arm. The two servos respectively control the horizontal rotation and pitching angle of the electric pan-tilt. The two servos are electrically connected through a servo control board based on STM32. The depth camera is installed at the end of the L-shaped arm through the camera connector and is electrically connected to the micro host.

[0006] The micro host uses a robot operating system for data management and control, sends control instructions to the servo control board through the serial port, and the servo control board controls the movement of two servos according to the received control instructions to control the rotation and pitching of the L-shaped arm respectively. When the servo moves to the specified angle, it sends a service request to the depth camera to control the depth camera to capture images at the set angle until all angles are collected.

[0007] Preferably, it further includes a base bracket, which includes a clamp arm ball head, a clamp arm fixing plate, a bracket plate, a clamp arm, a clamp, a clamp fixing screw, a bracket leg, and a host storage box. The clamp arm ball head is installed on the clamp arm fixing plate, the clamp arm fixing plate is connected to the bracket plate, one end of the clamp arm is connected to the clamp arm fixing plate, and the other end is connected to the clamp. The clamp can adjust the fixing state through the clamp fixing screw. The bracket leg is connected to the bracket plate, and the host storage box is connected to the clamp arm fixing plate for fixing the micro host. The long arm of the L-shaped arm is longer than the radius of the clamp arm fixing plate.

[0008] Preferably, the rotation angle of the upper servo is 0° to 180°, and the rotation angle of the lower servo is 0° to 360°.

[0009] Preferably, the base bracket also includes a spring clamp. When the stability of the bracket cannot be guaranteed in the safe space of the elevator shaft, the spring clamp is clamped on the railing or other fixed positions in the elevator shaft to ensure the stability of the bracket and the equipment.

[0010] An elevator shaft safety space reconstruction and measurement method based on an elevator shaft safety space data acquisition device, the method includes the following steps:

[0011] S100: The micro host controls the servo according to the program to move the depth camera to multiple specified angles in sequence. Each time the depth camera moves to a position, it captures and saves a pair of depth images and color images.

[0012] S200: Generate a color point cloud based on each pair of depth images and color images obtained, and obtain N frames of color point clouds.

[0013] S300: Based on the angles of the upper and lower servos, the lengths of the long and short arms of the L-shaped arm, the distance from the starting point of the camera depth to the back of the camera, and the thickness of the camera connecting piece, construct a formula for calculating the pose of the depth camera at the end of the L-shaped arm, calculate the initial pose of each frame of color point cloud, and transform all point clouds to the world coordinate system for initial alignment.

[0014] S400: Use a multi-view point cloud registration method based on pose graph optimization. Based on the adjacency relationship between N frames of point clouds, construct a pose graph through ICP registration between each pair, and then perform global graph optimization on the pose graph to finally obtain an accurate pose graph.

[0015] S500: After the multi-view point cloud registration is completed, the initially aligned N-frame RGB-D data and the accurate pose map are input into the scalable truncated signed distance function (TSDF) volume for fusion reconstruction. Finally, the triangular mesh is extracted and output as a 3D reconstruction result in PLY format. Key measurement points are selected from the 3D reconstruction result to complete the parameter measurement of the safety space of the elevator shaft.

[0016] Preferably, S100 includes:

[0017] S110: Preset to pitch once every 30°, for a total of 6 angles; then set to rotate horizontally once every 30°, for a total of 12 angles. The two-dimensional pan-tilt head is controlled by the micro host to complete the pitching and rotating actions in sequence.

[0018] S120: Cooperate with the binocular structured light depth camera for shooting to obtain depth images and color images at a total of N angles. Among them, each depth image and color image correspond to the pitching angle and rotating angle of two servo motors.

[0019] Preferably, S200 includes:

[0020] S210: For each pair of depth images and color images, according to the focal length parameters fx, fy of the camera and the optical center parameters cx, cy of the camera, first convert the depth image into three-dimensional point coordinates.

[0021] S220: Then convert the color of each pixel in the color image into a normalized color value.

[0022] S230: Finally, combine the calculated three-dimensional point coordinates and color information to generate a colored point cloud, and finally obtain N colored point clouds.

[0023] Preferably, S300 includes:

[0024] S310: Read the rotation angle of the lower servo motor and the pitching angle of the upper servo motor corresponding to each camera pose, denoted as α, β.

[0025] S320: Define the world coordinate system a, the intermediate coordinate system b, and the end coordinate system c, and calculate the coordinate transformation matrix between the end coordinate system and the intermediate coordinate system and the coordinate transformation matrix between the intermediate coordinate system and the world coordinate system , specifically:

[0026] ;

[0027] ;

[0028] Among them, represents the rotation angle of the lower servo motor, represents the pitching angle of the upper servo motor, represents the length of the arm in the L-shaped arm that is close to the upper servo represents the sum of the length of the arm in the L-shaped arm that is far from the upper servo, the thickness of the camera connecting piece, and the distance from the back of the camera to the depth starting point;

[0029] Based on the above two coordinate transformation matrices, N original point clouds are transformed from the end coordinate system c to the world coordinate system a to complete the initial alignment.

[0030] Preferably, S400 includes:

[0031] S410: Since the camera captures images at a preset angle and the adjacent relationship between all frames is known, the N point clouds obtained from the initial alignment are directly registered according to the known adjacency relationship. For all adjacent point clouds, multi-scale color ICP registration is used for precise alignment to obtain the transformation matrix between adjacent point clouds, and the confidence information matrix of each edge is calculated. Using the initial alignment pose of the point cloud as the node and the registered transformation matrix and the calculated information matrix as the edge, the construction of the pose graph is completed accordingly;

[0032] S420: Read the pose graph and call the GlobalOptimizationLevenbergMarquardt() function in the Open3D library to perform global optimization on it. The convergence criterion for optimization, the maximum corresponding point distance, the edge pruning threshold, and the preference for loop closure registration need to be set. Finally, the optimized pose graph is output and saved.

[0033] Preferably, S500 includes:

[0034] S510: After completing the multi-view point cloud registration, the N frames of RGB-D data after the initial alignment and the precise pose graph are jointly input into the scalable truncated signed distance function TSDF volume for fusion reconstruction. The initialization parameters of the TSDF are set based on experience, and the RGB8 color type is used to store color information; during the fusion process, each frame of RGB-D image is read in sequence, and combined with the optimized pose corresponding in the pose graph, the RGB-D data is integrated into the TSDF volume through inverse transformation;

[0035] S520: After completing the fusion of all frame data, extract the triangular mesh from the TSDF, output it as a three-dimensional reconstruction result in PLY format, extract the complete point cloud data of the elevator shaft safety space, and select key measurement points and measure the elevator shaft space parameters in the point cloud visualization software.

[0036] The above elevator shaft safety space data acquisition device and three-dimensional reconstruction measurement method execute a data acquisition program through a micro host, and can set a delay according to requirements. The operator only needs to place the device in the elevator shaft to automatically complete the three-dimensional data acquisition task of the elevator shaft. The device can fully cover all corners in the elevator shaft to ensure the integrity and accuracy of data acquisition, and provide high-quality original data for subsequent space reconstruction and measurement. The angles set by the method are based on empirical values, which can not only avoid wasting computing resources due to too many perspectives, but also achieve a reasonable overlap rate between point clouds, laying a foundation for subsequent registration. The scheme of initial alignment of point clouds based on the servo angle, color ICP point cloud registration between adjacent frames and graph optimization ensures the stability of multi-view point cloud registration and is applicable to various types of elevator shaft structures. Finally, through the combination of TSDF volume reconstruction, the redundancy of the registered point cloud data is removed. There is no need for the operator to enter or manipulate in real time, effectively realizing the three-dimensional data acquisition, reconstruction and measurement of the elevator shaft safety space, ensuring the absolute safety of inspectors, and being convenient to operate and highly efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a module diagram of the elevator shaft safety space data acquisition device according to an embodiment of the present invention;

[0038] Figure 2 It is a schematic structural diagram of the overall elevator shaft safety space data acquisition device according to an embodiment of the present invention;

[0039] Figure 3 is Figure 2 a schematic structural diagram of the shown two-dimensional electric pan-tilt and measurement unit;

[0040] Figure 4 is Figure 2 a schematic structural diagram of the shown bracket;

[0041] Figure 5 It is a data acquisition flow chart according to an embodiment of the present invention;

[0042] Figure 6 It is a flow chart of the elevator shaft safety space reconstruction measurement method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to enable those skilled in the art of the present technology to better understand the technical solutions of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] In one embodiment, as Figure 1 , Figure 2 and Figure 3As shown in the figure, an elevator shaft safety space data acquisition device 100 includes a depth camera 111, a camera connection member 112, a two-dimensional electric pan-tilt head 114, a micro host 129, and a servo control board 117. The depth camera 111 is electrically connected to the micro host 129, and the servo control board 117 is electrically connected to the micro host 129. The two-dimensional electric pan-tilt head 114 includes a base, an L-shaped arm 113, an upper servo 115, and a lower servo 116. The upper servo 115 is connected to the L-shaped arm 113. The two servos respectively control the horizontal rotation and pitch angle of the electric pan-tilt head 114. The two servos are electrically connected through a servo control board 117 based on STM32. The depth camera 111 is installed at the end of the L-shaped arm 113 through the camera connection member 112 and is electrically connected to the micro host 129.

[0045] The micro host 129 uses a robot operating system for data management and control, and sends control commands to the servo control board 117 through a serial port. The servo control board 117 controls the movement of the two servos according to the received control commands, respectively controlling the rotation and pitch actions of the L-shaped arm 113. When the servo moves to the specified angle, it sends a service request to the depth camera 111 to control the depth camera 111 to capture images at the set angle until all angles are collected.

[0046] Specifically, the data acquisition device 100 further includes a power supply for supplying electrical energy to the micro host and the servo control board. This device executes a data acquisition program through the micro host 129 and can set a delay according to requirements. The operator only needs to place the device in the elevator shaft to automatically complete the three-dimensional data acquisition task of the elevator shaft without personnel entering the elevator shaft, greatly reducing potential safety hazards during the operation and ensuring the safety of the operator. Secondly, the two-dimensional electric pan-tilt head 114 in the device controls the rotation and pitch actions of the L-shaped arm 113 through two servos respectively, and cooperates with the viewing angle range of the binocular structured light depth camera 111 to realize the data acquisition of full-angle coverage of the elevator shaft safety space. Through this design, the device can comprehensively cover all corners in the elevator shaft, ensuring the integrity and accuracy of data acquisition, and providing high-quality raw data for subsequent space reconstruction and measurement. In addition, the device of this system has a simple structure, reasonable design, and convenient operation, which not only improves the operation efficiency, but also ensures the safety and accuracy of data acquisition, reducing the possibility of human intervention and misjudgment.

[0047] In one embodiment, as Figure 4As shown in the figure, it further includes a base bracket 120, which comprises a clamping arm ball head 121, a clamping arm fixing plate 122, a bracket plate 123, clamping arms 124, clips 125, clip fixing screws 126, bracket legs 127 and a main unit storage box 128. The clamping arm ball head 121 is installed on the clamping arm fixing plate 122, the clamping arm fixing plate 122 is connected to the bracket plate 123, one end of the clamping arm 124 is connected to the clamping arm fixing plate 122, and the other end is connected to the clip 125. The clip 125 can adjust the fixing state through the clip fixing screw 126. The bracket legs 127 are connected to the bracket plate 123, and the main unit storage box 128 is connected to the clamping arm fixing plate 122 for fixing the micro main unit 129. The long arm of the L-shaped arm 113 is longer than the radius of the clamping arm fixing plate 122.

[0048] Specifically, the base bracket 120 is used to stabilize the position of the data acquisition device to ensure the stability and accuracy of the device during data acquisition; the long arm of the L-shaped arm 113 is longer than the radius of the clamping arm fixing plate 122, so that the shooting of the binocular structured light camera 111 is not blocked.

[0049] In one embodiment, the rotation angle of the upper servo 115 is 0° to 180°, and the rotation angle of the lower servo 116 is 0° to 360°.

[0050] Specifically, the depth camera 111 itself has a certain field of view, so as to cover the data acquisition of the full-angle scene of the elevator shaft safety space.

[0051] In one embodiment, the base bracket 120 further includes a spring clip. When it is impossible to ensure the stability of the bracket in the elevator shaft safety space, the spring clip is clamped on the railing or other fixed positions in the elevator shaft to ensure the stability of the bracket and the device.

[0052] Furthermore, the micro main unit 129 installs the Ubuntu system, and the camera 111 comes with a ROS2 package. It can take pictures through the ROS2 service. Write the ROS2 node according to the Figure 5 shown process. On the one hand, send serial commands to the servo control board 117 to control the rotation angles of the two servos. On the other hand, when the servo moves to the specified angle each time, the client sends a shooting request, and the camera 111 takes pictures. The angles of the upper and lower servos 116 corresponding to all N viewing angles have been preset in advance, and the acquisition process ends after all pictures are taken.

[0053] As Figure 6 shown, an elevator shaft safety space reconstruction and measurement method based on an elevator shaft safety space data acquisition device, the method includes the following steps:

[0054] S100: The micro main unit controls the servo according to the program to move the depth camera to multiple specified angles in sequence. Each time the depth camera moves to a position, it takes and saves a pair of depth images and color images;

[0055] S200: Generate a colored point cloud based on each pair of depth images and color images obtained, and obtain N frames of colored point clouds;

[0056] S300: Based on the angles of the upper and lower servos, the lengths of the long and short arms of the L-shaped arm, the distance from the starting point of the camera depth to the back of the camera, and the thickness of the camera connector, construct a formula for calculating the pose of the depth camera at the end of the L-shaped arm, calculate the initial pose of each frame of colored point cloud, and transform all point clouds to the world coordinate system for initial alignment;

[0057] S400: Use a multi-view point cloud registration method based on pose graph optimization. Based on the adjacency relationship between N frames of point clouds, construct a pose graph through ICP registration between pairs, and then perform global graph optimization on the pose graph to finally obtain an accurate pose graph;

[0058] S500: After the multi-view point cloud registration is completed, input the N frames of RGB-D data after initial alignment and the accurate pose graph into an extensible truncated signed distance function (TSDF) volume for fusion reconstruction. Finally, extract the triangular mesh and output it as a 3D reconstruction result in PLY format. Select key measurement points from the 3D reconstruction result to complete the parameter measurement of the safety space of the elevator shaft.

[0059] Specifically, before S100, the staff adjusts the position of the car, arranges the device in the safety space and starts it. The steps of device arrangement and startup include: taking the safety space at the top of the elevator shaft as an example, first raise the standing surface on the top of the car to be flush with the highest floor and open the elevator hall door (outer door); the staff judges the flatness of the standing surface in the elevator shaft, opens the tripod at the bottom of the data acquisition device or clips the spring clip on the fence of the top space to ensure that the device is stably placed and then starts the device; close the hall door, control the elevator to rise to the specified height required for measurement, and the data acquisition device starts to work after the program delay ends.

[0060] In one embodiment, S100 includes:

[0061] S110: Preset to pitch once every 30°, a total of 6 angles (excluding the vertically upward angle); then set to rotate horizontally once every 30°, a total of 12 angles, and control the two-dimensional pan-tilt head to complete the pitching and rotation actions in sequence through the micro host;

[0062] S120: Cooperate with the binocular structured light depth camera for shooting to obtain depth images and color images at a total of N angles. Among them, each depth image and color image corresponds to the pitching angle and rotation angle of the two servos.

[0063] In one embodiment, S200 includes:

[0064] S210: For each pair of depth image and color image, according to the focal length parameters fx, fy of the camera and the optical center parameters cx, cy of the camera, first convert the depth image into three-dimensional point coordinates;

[0065] S220: Then convert the color of each pixel in the color image into a normalized color value;

[0066] S230: Finally, combine the calculated three-dimensional point coordinates and color information to generate a colored point cloud, and finally obtain N colored point clouds.

[0067] In one embodiment, S300 includes:

[0068] S310: Read the rotation angle of the lower servo and the pitch angle of the upper servo corresponding to each camera pose, denoted as α, β;

[0069] S320: Define a world coordinate system a, an intermediate coordinate system b, and an end coordinate system c, and calculate the coordinate transformation matrix between the end coordinate system and the intermediate coordinate system and the coordinate transformation matrix between the intermediate coordinate system and the world coordinate system , specifically:

[0070] ;

[0071] ;

[0072] Among them, represents the rotation angle of the lower servo, represents the pitch angle of the upper servo, represents the length of the arm in the L-shaped arm close to the upper servo, represents the sum of the length of the arm in the L-shaped arm far from the upper servo, the thickness of the camera connecting piece, and the distance from the back of the camera to the depth starting point;

[0073] Based on the above two coordinate transformation matrices, convert the N original point clouds from the end coordinate system c to the world coordinate system a to complete the initial alignment.

[0074] Specifically, if the starting point and the ending point of the angle ranges of the two servos are changed, that is, and the definition of the angle, then the transformation matrix needs to be appropriately changed. After the above is completed, perform the corresponding coordinate transformation on each point cloud, and save the updated point cloud information after converting it to the world coordinate system.

[0075] In one embodiment, S400 includes:

[0076] S410: Since the camera captures images at a preset angle and the adjacent relationships between all frames are known, the N point clouds obtained from the initial alignment are directly registered according to the known adjacency relationships. For all adjacent point clouds, multi-scale color ICP registration is used for precise alignment to obtain the transformation matrix between adjacent point clouds, and the confidence information matrix of each edge is calculated. Using the initial alignment pose of the point cloud as the node and the registered transformation matrix and the calculated information matrix as the edges, the pose graph is constructed accordingly. For example, the relative transformation matrix after registration from point cloud Pi’ to Pj’ is Tij. Taking point cloud Pi’ to Pj’ as the node and Tij as the corresponding edge, it is added to the pose graph.

[0077] S420: Read the pose graph and call the GlobalOptimizationLevenbergMarquardt() function in the Open3D library to perform global optimization on it. The convergence criterion for optimization, the maximum corresponding point distance, the edge pruning threshold, and the preference for loop closure registration need to be set. Finally, the optimized pose graph is output and saved.

[0078] Specifically, according to the mesh distribution relationship of each frame in the elevator shaft, the adjacent relationship between frames is confirmed. Generally speaking, in the graph optimization method of multi-view registration, the adjacent relationship of the pose graph is only limited to the front and back frames. However, since the data acquisition in the elevator shaft is controllable, the frames around a frame can be defined as adjacent frames.

[0079] Furthermore, in the process of global pose optimization, using the poses of each frame as the optimization variables and the transformation matrix of the edges as the constraint conditions, by minimizing the error function, the poses of all frames reach global consistency. The information matrix is used as the optimization weight to ensure that the edges with high confidence have a greater impact on the result. The optimized poses are used to realign each frame of point cloud, and finally a 3D point cloud model of the elevator shaft with smaller error and higher precision is generated.

[0080] In one embodiment, S500 includes:

[0081] S510: After completing the multi-view point cloud registration, input the N frames of RGB-D data after initial alignment and the precise pose graph into the scalable truncated signed distance function (TSDF) volume for fusion reconstruction. Based on experience, set the initialization parameters of the TSDF, such as voxel length (voxel_length), truncation distance (sdf_trunc), etc., and use the RGB8 color type to store color information. During the fusion process, read each frame of RGB-D image in sequence and combine the optimized pose (the transformation matrix from the camera to the world coordinate system) corresponding in the pose graph. Through inverse transformation, the RGB-D data is integrated into the TSDF volume.

[0082] S520: After fusing all frame data, extract the triangular mesh from the TSDF and output it as a 3D reconstruction result in PLY format. Extract the complete point cloud data of the elevator shaft safety space, and select key measurement points and measure the space parameters of the elevator shaft in a point cloud visualization software.

[0083] Furthermore, the measurement steps include: You can choose meshlab software, open the saved complete point cloud, and confirm the object to be measured. Taking the guiding distance of the car guide rail as an example, first find any one of the main car guide rails, click the measuring tool in the menu bar above meshlab, first click the highest point of the guide rail in space, and then click the point flush with the highest point of the pulley (drag the view if necessary to facilitate the second selection). At this time, meshlab will display the connection line and the distance between the two points, which is the measurement result. By analogy, measure other parameters to be measured, and then the measurement of the key parameters of the elevator shaft safety space can be completed. Comparing with the reference data, it can be judged whether the elevator shaft safety space is qualified.

[0084] The above method for reconstructing and measuring the elevator shaft safety space obtains spatial data of the elevator shaft from multiple angles by presetting fixed angles for the pitch and rotation of the pan-tilt. The setting of this angle is based on empirical values, which can not only avoid wasting computing resources due to too many perspectives, but also achieve a reasonable overlap rate between point clouds, laying a foundation for subsequent registration; The scheme of initial alignment of point clouds based on the servo angle, color ICP point cloud registration between adjacent frames and graph optimization ensures the stability of multi-view point cloud registration and is applicable to various types of elevator shaft structures; Through the combination of TSDF volume reconstruction, the redundancy of the point cloud data after registration is removed.

[0085] The above has introduced in detail the elevator shaft safety space data acquisition device and the 3D reconstruction and measurement method provided by the present invention. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the core idea of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and modifications can be made to the present invention, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. An elevator shaft safety space data acquisition device, characterized in that, It includes a depth camera, a camera connecting piece, a two-dimensional electric pan-tilt, a micro host and a servo control board. The depth camera is electrically connected to the micro host, and the servo control board is electrically connected to the micro host. The two-dimensional electric pan-tilt includes a base, an L-shaped arm, an upper servo and a lower servo. The upper servo is connected to the L-shaped arm. The two servos respectively control the horizontal rotation and pitching angle of the electric pan-tilt. The two servos are electrically connected through a servo control board based on STM32. The depth camera is installed at the end of the L-shaped arm through the camera connecting piece and is electrically connected to the micro host. The device also includes a power supply, which is used to supply electrical energy to the micro host and the servo control board; The micro host uses the robot operating system for data management and control, and sends control instructions to the servo control board through the serial port. The servo control board controls the movement of the two servos according to the received control instructions, respectively controlling the rotation and pitching actions of the L-shaped arm. When the servo moves to the specified angle, it sends a service request to the depth camera to control the depth camera to capture images at the set angle until all angles are collected.

2. The device according to claim 1, characterized in that, It also includes a base bracket, which includes a clamp arm ball head, a clamp arm fixing plate, a bracket plate, a clamp arm, a clamp, a clamp fixing screw, a bracket leg and a host storage box. The clamp arm ball head is installed on the clamp arm fixing plate, and the clamp arm fixing plate is connected to the bracket plate. One end of the clamp arm is connected to the clamp arm fixing plate, and the other end is connected to the clamp. The clamp can adjust the fixing state through the clamp fixing screw. The bracket leg is connected to the bracket plate, and the host storage box is connected to the clamp arm fixing plate for fixing the micro host. The long arm of the L-shaped arm is longer than the radius of the clamp arm fixing plate.

3. The device according to claim 2, characterized in that, The rotation angle of the upper servo is 0° to 180°, and the rotation angle of the lower servo is 0° to 360°.

4. The device according to claim 3, characterized in that, The base bracket also includes a spring clamp. When the stability of the bracket cannot be guaranteed in the safe space of the elevator shaft, the spring clamp is clamped on the railing or other fixed positions in the elevator shaft to ensure the stability of the bracket and the equipment.

5. An elevator shaft safety space reconstruction and measurement method based on the elevator shaft safety space data acquisition device described in any one of claims 1 to 4, characterized in that, The method includes the following steps: S100: The micro host controls the servo according to the program to move the depth camera to multiple specified angles in sequence. Each time the depth camera moves to a position, it captures and saves a pair of depth images and color images; S110: It is preset to pitch once every 30°, for a total of 6 angles; and then it is set to rotate horizontally once every 30°, for a total of 12 angles. The two-dimensional pan-tilt is controlled by the micro host to complete the pitching and rotation actions in sequence; S120: Cooperate with the binocular structured light depth camera for shooting to obtain depth images and color images at a total of N angles. Among them, each depth image and color image corresponds to the pitching angle and rotation angle of the two servos; S200: Generate color point clouds according to each pair of depth images and color images obtained, and obtain N frames of color point clouds; S300: Based on the angles of the upper and lower servos, the lengths of the long and short arms of the L-shaped arm, the distance from the camera depth starting point to the back of the camera, and the thickness of the camera connecting piece, construct a formula for calculating the pose of the depth camera at the end of the L-shaped arm, calculate the initial pose of each frame of color point cloud, and transform all point clouds to the world coordinate system for initial alignment; S400: Use the multi-view point cloud registration method optimized based on the pose graph. Based on the adjacency relationship between N frames of point clouds, construct a pose graph through pairwise ICP registration, then perform global graph optimization on the pose graph, and finally obtain an accurate pose graph; S500: After multi-view point cloud registration is completed, input the initially aligned N frames of RGB-D data and the accurate pose graph into the scalable truncated signed distance function (TSDF) volume for fusion reconstruction. Finally, extract the triangular mesh and output it as a 3D reconstruction result in PLY format. Select key measurement points from the 3D reconstruction result to complete the parameter measurement of the safety space of the elevator shaft.

6. The method according to claim 5, characterized in that S200 includes: S210: For each pair of depth images and color images, according to the focal length parameters fx, fy of the camera and the optical center parameters cx, cy of the camera, first convert the depth image into three-dimensional point coordinates; S220: Then convert the color of each pixel in the color image into a normalized color value; S230: Finally, combine the calculated three-dimensional point coordinates and color information to generate a colored point cloud, and finally obtain N colored point clouds.

7. The method according to claim 6, wherein S300 includes: S310: Read the rotation angle of the lower servo and the pitch angle of the upper servo corresponding to each camera pose, denoted as α, β; S320: Define the world coordinate system a, the intermediate coordinate system b, and the end coordinate system c, and calculate the coordinate transformation matrix between the end coordinate system and the intermediate coordinate system and the coordinate transformation matrix between the intermediate coordinate system and the world coordinate system respectively respectively Specifically, ； Among them, represents the rotation angle of the lower servo, represents the pitch angle of the upper servo, represents the length of the arm in the L-shaped arm close to the upper servo, represents the sum of the length of the arm in the L-shaped arm far from the upper servo, the thickness of the camera connecting piece, and the distance from the back of the camera to the depth starting point; Based on the above two coordinate transformation matrices, transform the N original point clouds from the end coordinate system c to the world coordinate system a to complete the initial alignment.

8. The method according to claim 7, characterized in that S400 includes: S410: Since the camera takes pictures at a preset angle and the adjacency relationship between all frames is known, directly register the N point clouds obtained by the initial alignment according to the known adjacency relationship. For all adjacent point clouds, use multi-scale colored ICP registration for accurate alignment to obtain the transformation matrix between adjacent point clouds, and calculate the confidence information matrix of each edge. Use the initial alignment pose of the point cloud as the node, and the registered transformation matrix and the calculated information matrix as the edge to complete the construction of the pose graph accordingly; S420: Read the pose graph and call the GlobalOptimizationLevenbergMarquardt() function in the Open3D library to perform global optimization on it. It is necessary to set the convergence criterion for optimization, the maximum corresponding point distance, the edge pruning threshold, and the preference for loop closure registration. Finally, output and save the optimized pose graph.

9. The method according to claim 8, characterized in that, S500 includes: S510: After completing multi-view point cloud registration, input the initially aligned N frames of RGB-D data and the accurate pose graph into the scalable truncated signed distance function (TSDF) volume for fusion reconstruction. Set the initialization parameters of the TSDF based on experience and use the RGB8 color type to store color information; during the fusion process, read each frame of RGB-D image in sequence, and combine the optimized pose corresponding to it in the pose graph, and integrate the RGB-D data into the TSDF volume through inverse transformation; S520: After completing the fusion of all frame data, extract the triangular mesh from the TSDF and output the three-dimensional reconstruction result in PLY format. Extract the complete point cloud data of the elevator shaft safety space, and select key measurement points and measure the spatial parameters of the elevator shaft in the point cloud visualization software.

Images