WO2022246826A1

WO2022246826A1 - Extrinsic calibration method and apparatus, movable platform, and storage medium

Info

Publication number: WO2022246826A1
Application number: PCT/CN2021/096900
Authority: WO
Inventors: 宫正; 李延召
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2021-05-28
Filing date: 2021-05-28
Publication date: 2022-12-01

Abstract

Embodiments of the present application disclose an extrinsic calibration method, comprising: acquiring point cloud data of a scene collected by a first lidar and a second lidar in the same time period; estimating a first motion trajectory of the first lidar according to the point cloud data collected by the first lidar, and estimating a second motion trajectory of the second lidar according to the point cloud data collected by the second lidar; establishing a constraint condition by using the first motion trajectory and the second motion trajectory, and, on the basis of the constraint condition, calculating an initial value of an external parameter; correcting the initial value of the external parameter, and obtaining a target external parameter, the target external parameter being used for a transformation between the world coordinate system of the first lidar and the world coordinate system of the second lidar. Embodiments of the present application disclose an extrinsic calibration method not requiring a given initial value of an external parameter, which can achieve rapid, convenient and robust high-precision extrinsic calibration.

Description

External parameter calibration method, device, movable platform and storage medium

technical field

The present application relates to the technical field of external reference calibration, and in particular to an external reference calibration method, device, movable platform and storage medium.

Background technique

LiDAR can collect the point cloud data of the scene and provide the mobile platform with the ability to perceive the scene. The movable platform is usually equipped with multiple lidars, and different lidars are installed at different positions of the movable platform. Since the point cloud data collected by each laser radar is based on its own coordinate system, these point cloud data based on different coordinate systems cannot be directly spliced or fused. It is necessary to unify the point cloud data collected by different laser radars into in the base coordinate system.

Contents of the invention

In view of this, one of the purposes of the embodiments of the present application is to provide an external parameter calibration method that does not need to give an initial value of the external parameter, and can realize high-precision external parameter calibration quickly, conveniently, and robustly.

The first aspect of the embodiment of the present application provides an external parameter calibration method, including:

Obtaining point cloud data of scenes collected by the first laser radar and the second laser radar within the same period of time, wherein the first laser radar and the second laser radar are fixed on the same movable platform;

Estimating a first motion track of the first laser radar according to the point cloud data collected by the first laser radar, and estimating a second motion of the second laser radar according to the point cloud data collected by the second laser radar track;

Establishing constraints by using the first trajectory and the second trajectory, and calculating an initial value of an external parameter based on the constraints;

The initial value of the external parameter is corrected to obtain a target external parameter, and the target external parameter is used for transformation between the world coordinate system of the first laser radar and the world coordinate system of the second laser radar.

The second aspect of the embodiment of the present application provides an external reference calibration device, including: a processor and a memory storing a computer program, and the processor implements the following steps when executing the computer program:

The third aspect of the embodiment of the present application provides a mobile platform, including:

body;

a driving device connected to the body for powering the movable platform;

The first laser radar and the second laser radar fixed at different positions of the body are used to collect point cloud data of the scene;

A processor and a memory storing a computer program, the processor, when executing the computer program, implements the following steps:

Obtaining point cloud data of scenes collected by the first laser radar and the second laser radar within the same period of time;

The fourth aspect of the embodiment of the present application provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the external parameter calibration method provided in the embodiment of the present application is implemented.

The external parameter calibration method provided by the embodiment of the present application can use the consistency between the first motion trajectory and the second motion trajectory to establish a constraint condition, and can calculate the initial value of the external parameter based on the constraint condition, so that the external parameter Accurate target extrinsic parameters are calculated on the basis of initial values. It can be seen that the method provided in the embodiment of the present application can automatically calculate the appropriate initial value of the external parameter by using the point cloud data collected by the laser radar, without manually setting the initial value of the external parameter, and realizes fast and robust automatic external parameter calibration .

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings that need to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

Fig. 1 is a flow chart of the external parameter calibration method provided by the embodiment of the present application.

Fig. 2 is a top view of the movable platform provided by the embodiment of the present application.

Fig. 3 is a scene diagram of the movement of the movable platform provided by the embodiment of the present application.

Fig. 4 is a schematic structural diagram of an external parameter calibration device provided in an embodiment of the present application.

Fig. 5 is a schematic structural diagram of a mobile platform according to an embodiment of the present application.

Detailed ways

The following will clearly and completely describe the technical solutions in the embodiments of the application with reference to the drawings in the embodiments of the application. Apparently, the described embodiments are only some of the embodiments of the application, not all of them. Based on the embodiments in this application, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the scope of protection of this application.

LiDAR can collect the point cloud data of the scene and provide the mobile platform with the ability to perceive the scene. The movable platform is usually equipped with multiple lidars, and different lidars are installed at different positions of the movable platform. Since the point cloud data collected by each laser radar is based on its own coordinate system, these point cloud data based on different coordinate systems cannot be directly spliced or fused. It is necessary to unify the point cloud data collected by different laser radars into in the base coordinate system. The process of transforming the point cloud data collected by a laser radar into the reference coordinate system is the external parameter calibration of the laser radar, in other words, to determine an external parameter so that the point cloud data collected by the laser radar After the transformation of the parameters, it can be transformed into the reference coordinate system.

There are many ways to calibrate the external parameters of multi-lidar. In one embodiment, calibration can be performed using a calibration wall and a fixed turntable. This method has high requirements on the site and working environment, and it also needs to manually match the point cloud, which is troublesome to operate and high in labor costs. In one embodiment, the point cloud maps created by the two laser radars can be matched to obtain the external parameters between the two laser radars, but this method requires a relatively good initial value of the external parameters to converge. Accurate results, it is difficult to perform effective calibration when the initial value of the external reference is not ideal.

The embodiment of the present application provides a kind of external parameter calibration method, can refer to Fig. 1, and Fig. 1 is the flowchart of the external parameter calibration method that the present application embodiment provides, and this method comprises the following steps:

S102. Acquire point cloud data of the scene collected by the first lidar and the second lidar.

S104. Estimate the first movement track of the first laser radar according to the point cloud data collected by the first laser radar, and estimate the first movement trajectory of the second laser radar according to the point cloud data collected by the second laser radar. Two motion tracks.

S106. Establish constraint conditions by using the first motion trajectory and the second motion trajectory, and calculate an initial value of an external parameter based on the constraint conditions.

S107. Correct the initial value of the external parameter to obtain the target external parameter.

The first laser radar and the second laser radar may be two radars mounted on the same movable platform. It can be understood that the movable platform may be equipped with multiple laser radars, the first laser radar may be any one of the multiple radars, and the second laser radar may be any one of the multiple laser radars that is different from the first laser radar. The first laser radar and the second laser radar may be installed at different positions on the movable platform, and the field of view of the first laser radar may or may not overlap with the field of view of the second laser radar.

The mobile platform can be a variety of platforms with mobile capabilities such as vehicles, aircraft, ships, and robots. In one example, refer to FIG. 2 , the movable platform can be a vehicle, and the vehicle can be equipped with multiple laser radars, two of which are shown in the figure (lidar A and laser radar B), and two laser radars There is no overlap between the fields of view.

The point cloud data collected by the first laser radar and the point cloud data collected by the second laser radar can be collected in the same period. While moving, the first laser radar and the second laser radar on the movable platform can collect point cloud data respectively, so that the point cloud data collected by the first laser radar and the second laser radar are collected in the same period of time, and the two The point cloud data collected by the operator is the point cloud data corresponding to the same scene.

In one embodiment, the movable platform may be triggered by a user to start moving. For example, after obtaining the start calibration instruction input by the user, the movable platform can automatically start moving according to the specified trajectory, and during the movement process, the lidar carried by itself can collect point cloud data. In one example, the movable platform can also move along a specified trajectory under the control of the user. Specifically, after obtaining the start calibration instruction input by the user, the user can be prompted to control the movable platform. point cloud data. Here, there are many ways to prompt the user, for example, the specified track can be displayed on the screen, or the operation matching the specified track can be played through voice.

The trajectory of the movable platform can be a preset trajectory, or a trajectory automatically planned according to a scene or an electronic map. In one embodiment, by designing the specified trajectory, the lidar mounted on the movable platform can collect relatively complete point cloud data of the scene after the movable platform completes the movement along the specified trajectory. In one embodiment, the designated trajectory may include a U-shaped trajectory or an 8-shaped trajectory. Here, the U-shaped locus, that is, the shape of the locus is approximately U-shaped, and the figure-eight locus, that is, the shape of the locus is approximately an 8-shape. Reference can be made to FIG. 3 , which shows the movement trajectories of the laser radar A and the laser radar B carried on the vehicle after the vehicle moves along the 8-shaped trajectory.

Understandably, even if the first lidar and the second lidar on the movable platform do not overlap in the field of view, after the movable platform moves sufficiently, the first lidar and the second lidar can collect The point cloud data of the same area, so there is an overlap in the detection area on the point cloud data, therefore, in an embodiment of the following, the map established by the point cloud data collected by the two can be used to map the two corresponding The map is matched to calibrate the extrinsic parameters.

During the movement of the movable platform, the first laser radar and the second laser radar mounted on the movable platform are also moving, therefore, the first laser radar has its own motion trajectory, and the second laser radar also has its own motion track. Here, the first motion track of the first laser radar can be estimated according to the point cloud data collected by the first laser radar, and the second motion track of the second laser radar can be estimated according to the point cloud data collected by the second laser radar. After estimating the first trajectory and the second trajectory, the initial value of the external parameters can be solved using the trajectory consistency constraint. Specifically, the estimated first trajectory and the second trajectory can be used to establish constraints, and the The constraint condition calculates the initial value of the external parameter.

There are many ways to estimate motion trajectory from point cloud data. In one embodiment, SLAM (Simultaneous Localization and Mapping) technology can be used to estimate the motion trajectory. Taking the first laser radar as an example here, it can be understood that the first motion track of the first laser radar includes the pose (position and attitude) of the first laser radar at each moment, and the pose here is in the position of the first laser radar. The pose in the world coordinate system. During the movement of the first laser radar, the local coordinate system of the first laser radar changes in real time (the local coordinate system takes the real-time position of the first laser radar as the origin), so in an example, the first laser radar can be The local coordinate system of the first frame where the lidar starts to move is the world coordinate system, then the pose of the first lidar at the first moment can correspond to the origin of the world coordinate system, and the point cloud of the first frame is completed in the first lidar After the data is collected, the point cloud data of the first frame can be used to establish the point cloud map corresponding to the first frame. After the first lidar completes the collection of the point cloud data of the second frame, the point cloud map of the second frame can be used The cloud data establishes a point cloud map corresponding to the second frame, and can match the point cloud map of the second frame with the point cloud map of the first frame to calculate the pose of the first lidar at the second moment. Afterwards, the point cloud map of the third frame can be matched with the point cloud map of the second frame to calculate the pose of the first lidar at the third moment... Repeat the above iterative calculation process, and finally the first lidar can be calculated. The pose of the lidar at each moment is estimated to obtain the first motion track of the first lidar. Regarding the second motion track of the second lidar, its calculation process can refer to the calculation process of the above-mentioned first motion track, which will not be repeated here.

As mentioned above, the point cloud data collected by each laser radar needs to be unified into the reference coordinate system. In one embodiment, the world coordinate system of the first laser radar can be used as the reference coordinate system, and the target of external parameter calibration It is to determine an external parameter, so that the point cloud data collected by the second laser radar can be transformed from the world coordinate system of the second laser radar to the world coordinate system of the first laser radar through the transformation of the external parameter. The extrinsic parameter that you want to finally obtain can be called the target extrinsic parameter. Assuming that the target extrinsic parameter is T, T=[R,t], where R is the rotation parameter and t is the translation parameter, then based on the constraint of trajectory consistency, The following derivation can be obtained:

in,

is the attitude of the second lidar at the nth moment,

is the position of the second lidar at the nth moment,

is the attitude of the first lidar at the nth moment,

is the attitude of the first lidar at the nth moment.

In the above derivation, the constraints established by using the first motion trajectory and the second motion trajectory may include: the sum of the pose differences at each moment is the smallest, here, the position gap at a moment is the pose of the second lidar at that moment The difference between the pose of the first lidar and the first lidar at this moment after the transformation of the initial value of the external parameter. With the goal of minimizing the sum of the pose differences at each moment, the initial value of the external parameter can be optimized with the help of various optimization algorithms, and the gap between the initial value of the external parameter and the final target external parameter will not be too large, which can be used for subsequent corrections Get accurate target external parameters to provide guarantee.

It can be understood that the extrinsic parameters generally include rotation parameters and/or translation parameters, so the transformation performed by the extrinsic parameters includes rotation transformation and/or translation transformation.

Considering that there are inevitable errors in the estimated first and second motion trajectories, the optimized initial value of the extrinsic parameter still needs further correction. By correcting the initial value of the external parameter, the final target extrinsic parameter can be obtained. .

There are many ways to modify the initial value of the external parameter. In one embodiment, after acquiring the point cloud data collected by the first laser radar and the second laser radar respectively, in addition to estimating the motion trajectory based on the point cloud data, a map can also be established based on the point cloud data. Specifically, the first map of the scene can be established according to the point cloud data collected by the first laser radar, and the second map of the scene can be established according to the point cloud data collected by the first laser radar, and the first map and the second map can be used Correct the initial value of the external parameter to obtain the target external parameter. Here, the established map can be a point cloud map or a feature point map. There are many ways to create maps, including but not limited to SLAM mapping, feature point registration, and iterative closest point algorithm ICP (Iterative Closest Point), G-ICP and other registration algorithms for mapping.

In one embodiment, the first map of the scene can be a point cloud map obtained by fusion of each frame of point clouds collected by the first laser radar, and the second map of the scene can be a point cloud map of each frame collected by the second laser radar The fused point cloud maps, that is, the first map and the second map are maps covering the entire scene. Then, when using the first map and the second map to correct the initial value of the external reference, since the point cloud data in the second map is based on the world coordinate system of the second lidar, the initial value of the external reference can be used to correct the The second map is transformed, and the second map is transformed to the world coordinate system of the first lidar, so that the second map and the first map are in the same coordinate system, and then the transformed second map and the first map can be Matching is performed to determine an extrinsic correction value that enables the second map to coincide with the first map after transformation. Using the external parameter correction value to correct the external parameter initial value, the target external parameter can be obtained. Here, in an example, the target extrinsic parameter may be obtained by multiplying the extrinsic parameter correction and the extrinsic initial value.

In an implementation manner, the second map of the scene may include N sub-maps, and each sub-map may correspond to a partial area in the scene. Specifically, each submap may be obtained by fusing several frames of point clouds collected by the second lidar. For example, in an example, each submap may be obtained by fusing five frames of point clouds collected by the second lidar. Then, when using the first map and the second map to correct the initial value of the external parameter, the first motion track and the initial value of the external parameter calculated based on the trajectory constraints can be used to transform the first sub-map, and the transformed first sub-map The map can be matched with the first map to obtain the correction value of the external parameter. After using the correction value of the external parameter to correct the initial value of the external parameter, the correction result can be obtained, and the correction result can be used as a new initial value of the external parameter to participate in the correction of the first external parameter. During the transformation of the 2 submaps. The above iterations are repeated until the matching of the last submap and the first map is completed, and the final correction result can be determined as the target extrinsic reference.

For the nth submap, since the nth submap is obtained by fusion of several frames of point clouds collected by the second lidar, the point cloud data in the nth submap is based on the local coordinates corresponding to the several frames of point clouds system, the point cloud data in the nth sub-map needs to be transformed from the local coordinate system to the global coordinate system (that is, the world coordinate system of the second lidar). The nth submap is obtained by fusion of several frames of point clouds, so the moment corresponding to a certain frame in the several frames can be used as the moment corresponding to the nth submap, here, the moment corresponding to the nth submap can be recorded as for t. When transforming the nth submap, the pose corresponding to the first lidar at time t can be determined according to the first motion trajectory, and the pose corresponding to the time t can be used to transform the nth submap, so that the The point cloud data in the n submaps can be transformed from the local coordinate system to the global coordinate system, and quickly positioned near the target area of the first map. The target area is the local area of the scene corresponding to the nth submap. Afterwards, the nth submap can be transformed by the current initial value of the external parameter, and the nth submap can be transformed from the world coordinate system of the second lidar to the world coordinate system of the first lidar, so that the transformed submap can be The map is matched with the first map to obtain an extrinsic correction value.

When matching the transformed submap with the first map, the nearest neighbor point corresponding to each point in the submap can be searched in the first map using the nearest neighbor search algorithm, and the points formed by the searched nearest neighbor points The set can be called the nearest neighbor point set, and the transformed submap can be matched with the nearest neighbor point set to obtain the extrinsic correction value.

In one case, if the FOV of the lidar mounted on the movable platform is small, there may not be enough feature points between the front and rear frames collected by the lidar, resulting in incorrect matching results. For example, if the nth submap corresponds to a part of the white wall in the scene, when the nth submap is matched with the first map, it is difficult to accurately find the part points corresponding to the nth submap in the first map Therefore, it is very likely to match the wrong external parameter correction value. If the wrong external parameter correction value is used to correct the external parameter initial value, it will cause the external parameter initial value to deviate in the wrong direction in this round of iterations, and eventually lead to The obtained target extrinsics are not accurate.

Considering the above problems, in one embodiment, when the nth sub-map is matched with the first map, the matching score of the two can be calculated, and if the matching score of the two is higher than the preset threshold, the matching The obtained external parameter correction value is used to correct the external parameter initial value, and the correction result enters the next iteration as a new external parameter initial value. Conversely, if the matching score is lower than the preset threshold, the external parameter correction value obtained by the matching is not used to correct the external parameter initial value, and can directly enter the next iteration, and the external parameter initial value used in the next iteration is still the original The initial value of the external parameter. Filtering out unreliable matching results by matching scores can greatly improve the accuracy of external parameters and the robustness of external parameter calibration.

In one embodiment, when building the first map and/or the second map of the scene, planar feature points in the point cloud data may be obtained to build the map, and non-planar feature points in the point cloud data may be removed. On the one hand, this can improve the matching speed, on the one hand, it can avoid the introduction of non-planar noise, and improve the matching accuracy.

In an implementation manner, when establishing the first map and/or the second map of the scene, down-sampling may also be performed on point clouds belonging to the ground in the map. Due to the high density of point clouds on the ground in most scenes, the constraints on the matching process of ground points are too strong, and the constraints of other features are submerged. Therefore, the sparse processing of ground points by downsampling can reduce the constraints of the ground. Improve the accuracy of matching results.

Reference can be made to FIG. 4 below. FIG. 4 is a schematic structural diagram of an external parameter calibration device provided by an embodiment of the present application, which includes: a processor 410 and a memory 420 storing a computer program, and when the processor executes the computer program Implement the following steps:

Optionally, the point cloud data is collected by the first laser radar and the second laser radar when the movable platform moves along a specified track.

Optionally, the specified trajectory includes: a U-shaped trajectory or an 8-shaped trajectory.

Optionally, the movable platform starts to move under the trigger of the user.

Optionally, the first motion trajectory includes the pose of the first laser radar at each moment, and the second motion trajectory includes the pose of the second laser radar at each moment;

The constraint conditions include: the sum of the pose differences corresponding to each moment is the smallest, and the pose difference is the pose of the second laser radar after the initial value transformation of the external parameters and the pose of the first laser radar gap between.

Optionally, the processor corrects the initial value of the external parameter, and when obtaining the target external parameter, it is used for:

Using the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter, wherein the first map is a point cloud collected according to the first lidar The second map is established based on the point cloud data collected by the second lidar.

Optionally, when the processor uses the first map and the second map to correct the initial value of the external parameter, it is used to:

transforming the second map by using the initial value of the external parameter, and matching the transformed second map with the first map to obtain a correction value of the external parameter;

The initial value of the external parameter is corrected by using the correction value of the external parameter.

Optionally, the second map includes N submaps, each submap corresponds to a part of the scene, and N is an integer greater than 1.

Optionally, when the processor uses the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter, it is used to:

After entering the iteration of the nth submap, transform the nth submap according to the first motion track and the current initial value of the external parameter, and match the transformed submap with the first map to obtain the external parameter correction value;

Using the external parameter correction value to correct the current external parameter initial value, and using the correction result as a new external parameter initial value, enter the iteration of the n+1th sub-map;

After the iteration of the last submap is completed, the final correction result is used as the target external parameter, wherein the n is an integer greater than 0 and less than N.

Optionally, when the processor uses the external parameter correction value to correct the current external parameter initial value, it is used to:

If the matching score between the transformed sub-map and the first map is higher than a preset threshold, the current initial value of the extrinsic parameter is corrected using the extrinsic correction value.

Optionally, the processor is also used for:

If the matching score between the transformed submap and the first map is lower than the preset threshold, directly enter into iteration of the n+1th submap.

Optionally, when the processor matches the transformed submap with the first map:

A nearest neighbor point set corresponding to the transformed submap is determined in the first map, and the transformed submap is matched with the nearest neighbor point set.

Optionally, the first map and/or the second map have been processed to remove non-planar feature points.

Optionally, the point clouds belonging to the ground in the first map and/or the second map have undergone down-sampling processing.

Optionally, the transformation includes: rotation transformation and/or translation transformation.

Optionally, there is no overlap between the field of view of the first laser radar and the field of view of the second laser radar.

Various implementations of the external reference calibration device have been provided above. For specific implementation, reference may be made to relevant descriptions in the foregoing methods, and details are not repeated here.

The external parameter calibration device provided by the embodiment of the present application can use the consistency between the first motion trajectory and the second motion trajectory to establish a constraint condition, and can calculate the initial value of the external parameter based on the constraint condition, so that the external parameter Accurate target extrinsic parameters are calculated on the basis of initial values. It can be seen that the device provided in the embodiment of the present application can automatically calculate the appropriate initial value of the external parameter by using the point cloud data collected by the laser radar, without manually setting the initial value of the external parameter, and realizes fast and robust automatic external parameter calibration .

Reference can be made to FIG. 5 below. FIG. 5 is a schematic structural diagram of a mobile platform provided by an embodiment of the present application. The mobile platform includes:

Body 510;

A drive device connected to the body 510, used to provide power to the movable platform;

The first laser radar 520 and the second laser radar 530 fixed at different positions of the body 510 are used to collect point cloud data of the scene;

A processor 540 and a memory 550 storing a computer program, the processor implements the following steps when executing the computer program:

Optionally, the movable platform starts to move under the trigger of the user.

Optionally, the processor is also used for:

Various implementations of the mobile platform are provided above, and for the specific implementation, reference may be made to relevant descriptions in the foregoing methods, and details will not be repeated here.

The mobile platform provided by the embodiment of the present application can use the consistency between the first motion trajectory and the second motion trajectory to establish a constraint condition, and can calculate the initial value of the external parameter based on the constraint condition, so that the initial value of the external parameter can be obtained. Accurate target extrinsic parameters are calculated based on the values. It can be seen that the mobile platform provided by the embodiment of the present application can automatically calculate the appropriate initial value of the external parameter by using the point cloud data collected by the laser radar, without manually setting the initial value of the external parameter, and realizes a fast and robust fully automatic external parameter. Reference calibration.

The embodiment of the present application also provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any external parameter calibration method provided in the embodiment of the present application.

A variety of implementations are provided above for each protection subject. On the basis of no conflict or contradiction, those skilled in the art can freely combine various implementations according to actual conditions, thus forming various technical solutions. . However, the document of the present application is limited in length, and it is not possible to explain all the combined technical solutions, but it can be understood that these technical solutions that cannot be developed also belong to the scope of the disclosure of the embodiments of the present application.

Embodiments of the present application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having program code embodied therein. Computer usable storage media includes both volatile and non-permanent, removable and non-removable media, and may be implemented by any method or technology for information storage. Information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for computers include, but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Flash memory or other memory technology, Compact Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, Magnetic tape cartridge, tape magnetic disk storage or other magnetic storage device or any other non-transmission medium that can be used to store information that can be accessed by a computing device.

It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is a relationship between these entities or operations. There is no such actual relationship or order between them. The term "comprises", "comprises" or any other variation thereof is intended to cover a non-exclusive inclusion such that a process, method, article or apparatus comprising a set of elements includes not only those elements but also other elements not expressly listed elements, or also elements inherent in such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

The methods and devices provided by the embodiments of the present invention have been described in detail above. The principles and implementation methods of the present invention have been explained by using specific examples in this paper. The descriptions of the above embodiments are only used to help understand the methods and methods of the present invention. core idea; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation and scope of application. In summary, the content of this specification should not be construed as limiting the present invention .

Claims

An external parameter calibration method is characterized in that it comprises:

Obtaining point cloud data of scenes collected by the first laser radar and the second laser radar within the same period of time, wherein the first laser radar and the second laser radar are fixed on the same movable platform;

Estimating a first motion track of the first laser radar according to the point cloud data collected by the first laser radar, and estimating a second motion of the second laser radar according to the point cloud data collected by the second laser radar track;

Establishing constraints by using the first trajectory and the second trajectory, and calculating an initial value of an external parameter based on the constraints;

The initial value of the external parameter is corrected to obtain a target external parameter, and the target external parameter is used for transformation between the world coordinate system of the first laser radar and the world coordinate system of the second laser radar.
The method according to claim 1, wherein the point cloud data is collected by the first laser radar and the second laser radar when the movable platform moves along a specified track.
The method according to claim 2, characterized in that the specified trajectory comprises: a U-shaped trajectory or an 8-shaped trajectory.
The method according to claim 2, wherein the movable platform starts to move under the trigger of a user.
The method according to claim 1, wherein the first motion trajectory includes the pose of the first laser radar at each moment, and the second motion trajectory includes the pose of the second laser radar at each moment. pose;

The constraint conditions include: the sum of the pose differences corresponding to each moment is the smallest, and the pose difference is the pose of the second laser radar after the initial value transformation of the external parameters and the pose of the first laser radar gap between.
The method according to claim 1, wherein said modifying the initial value of said external parameter to obtain a target external parameter comprises:

Using the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter, wherein the first map is a point cloud collected according to the first lidar The second map is established based on the point cloud data collected by the second lidar.
The method according to claim 6, wherein the modifying the initial value of the external reference by using the first map and the second map includes:

transforming the second map by using the initial value of the external parameter, and matching the transformed second map with the first map to obtain a correction value of the external parameter;

The initial value of the external parameter is corrected by using the correction value of the external parameter.
The method according to claim 6, wherein the second map includes N submaps, each submap corresponds to a partial area in the scene, and the N is an integer greater than 1.
The method according to claim 8, wherein said using the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter comprises:

After entering the iteration of the nth submap, transform the nth submap according to the first motion track and the current initial value of the external parameter, and match the transformed submap with the first map to obtain the external parameter correction value;

Using the external parameter correction value to correct the current external parameter initial value, and using the correction result as a new external parameter initial value, enter the iteration of the n+1th sub-map;

After the iteration of the last submap is completed, the final correction result is used as the target external parameter, wherein the n is an integer greater than 0 and less than N.
The method according to claim 9, characterized in that, using the external parameter correction value to correct the current external parameter initial value comprises:

If the matching score between the transformed sub-map and the first map is higher than a preset threshold, the current initial value of the extrinsic parameter is corrected using the extrinsic correction value.
The method according to claim 10, characterized in that the method further comprises:

If the matching score between the transformed submap and the first map is lower than the preset threshold, directly enter into iteration of the n+1th submap.
The method according to claim 9, wherein said matching the transformed submap with the first map comprises:

A nearest neighbor point set corresponding to the transformed submap is determined in the first map, and the transformed submap is matched with the nearest neighbor point set.
The method according to claim 6, characterized in that, the first map and/or the second map are processed to remove non-planar feature points.
The method according to claim 6, characterized in that, the point cloud belonging to the ground in the first map and/or the second map has undergone down-sampling processing.
The method according to claim 1, wherein the transformation comprises: rotation transformation and/or translation transformation.
The method according to claim 1, wherein the field of view of the first laser radar does not overlap with the field of view of the second laser radar.
An external reference calibration device is characterized in that it comprises: a processor and a memory that stores a computer program, and the processor implements the following steps when executing the computer program:

Obtaining point cloud data of scenes collected by the first laser radar and the second laser radar within the same period of time, wherein the first laser radar and the second laser radar are fixed on the same movable platform;

Estimating a first motion track of the first laser radar according to the point cloud data collected by the first laser radar, and estimating a second motion of the second laser radar according to the point cloud data collected by the second laser radar track;

Establishing constraints by using the first trajectory and the second trajectory, and calculating an initial value of an external parameter based on the constraints;

The initial value of the external parameter is corrected to obtain a target external parameter, and the target external parameter is used for transformation between the world coordinate system of the first laser radar and the world coordinate system of the second laser radar.
The device according to claim 17, wherein the point cloud data is collected by the first laser radar and the second laser radar when the movable platform moves along a specified track.
The device according to claim 18, characterized in that, the specified trajectory comprises: a U-shaped trajectory or an 8-shaped trajectory.
The apparatus according to claim 18, wherein the movable platform is initiated to move by a user.
The device according to claim 17, wherein the first motion trajectory includes the pose of the first laser radar at each moment, and the second motion trajectory includes the pose of the second laser radar at each moment. pose;

The constraint conditions include: the sum of the pose differences corresponding to each moment is the smallest, and the pose difference is the pose of the second laser radar after the initial value transformation of the external parameters and the pose of the first laser radar gap between.
The device according to claim 17, wherein the processor corrects the initial value of the external parameter, and when obtaining the target external parameter is used for:

Using the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter, wherein the first map is a point cloud collected according to the first lidar The second map is established based on the point cloud data collected by the second lidar.
The device according to claim 22, wherein the processor is configured to: when using the first map and the second map to correct the initial value of the external parameter:

transforming the second map by using the initial value of the external parameter, and matching the transformed second map with the first map to obtain a correction value of the external parameter;

The initial value of the external parameter is corrected by using the correction value of the external parameter.
The device according to claim 22, wherein the second map includes N submaps, each submap corresponds to a part of the scene, and N is an integer greater than 1.
The device according to claim 24, wherein the processor uses the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter. At:

After entering the iteration of the nth submap, transform the nth submap according to the first motion track and the current initial value of the external parameter, and match the transformed submap with the first map to obtain the external parameter correction value;

Using the external parameter correction value to correct the current external parameter initial value, and using the correction result as a new external parameter initial value, enter the iteration of the n+1th sub-map;

After the iteration of the last submap is completed, the final correction result is used as the target external parameter, wherein the n is an integer greater than 0 and less than N.
The device according to claim 25, wherein the processor is used for: when using the external parameter correction value to correct the current external parameter initial value:

If the matching score between the transformed sub-map and the first map is higher than a preset threshold, the current initial value of the extrinsic parameter is corrected using the extrinsic correction value.
The device according to claim 26, wherein the processor is further configured to:

If the matching score between the transformed submap and the first map is lower than the preset threshold, directly enter into iteration of the n+1th submap.
The device according to claim 25, wherein the processor matches the transformed submap with the first map for:

A nearest neighbor point set corresponding to the transformed submap is determined in the first map, and the transformed submap is matched with the nearest neighbor point set.
The device according to claim 22, characterized in that, the first map and/or the second map have been processed to remove non-planar feature points.
The device according to claim 22, characterized in that, the point clouds belonging to the ground in the first map and/or the second map have undergone down-sampling processing.
The device according to claim 17, wherein the transformation comprises: rotation transformation and/or translation transformation.
The device according to claim 17, wherein the field of view of the first laser radar does not overlap with the field of view of the second laser radar.
A mobile platform, characterized in that it comprises:

body;

a driving device connected to the body for powering the movable platform;

The first laser radar and the second laser radar fixed at different positions of the body are used to collect point cloud data of the scene;

A processor and a memory storing a computer program, the processor, when executing the computer program, implements the following steps:

Obtaining point cloud data of scenes collected by the first laser radar and the second laser radar within the same period of time;

Estimating a first motion track of the first laser radar according to the point cloud data collected by the first laser radar, and estimating a second motion of the second laser radar according to the point cloud data collected by the second laser radar track;

Establishing constraints by using the first trajectory and the second trajectory, and calculating an initial value of an external parameter based on the constraints;

The initial value of the external parameter is corrected to obtain a target external parameter, and the target external parameter is used for transformation between the world coordinate system of the first laser radar and the world coordinate system of the second laser radar.
The movable platform according to claim 33, wherein the point cloud data is collected by the first laser radar and the second laser radar when the movable platform moves along a specified track.
The movable platform according to claim 34, wherein the specified trajectory comprises: a U-shaped trajectory or an 8-shaped trajectory.
The movable platform according to claim 34, wherein the movable platform starts to move when triggered by a user.
The movable platform according to claim 33, wherein the first motion trajectory includes the pose of the first laser radar at each moment, and the second motion trajectory includes the poses of the second laser radar at each moment. pose at the moment;

The constraint conditions include: the sum of the pose differences corresponding to each moment is the smallest, and the pose difference is the pose of the second laser radar after the initial value transformation of the external parameters and the pose of the first laser radar gap between.
The mobile platform according to claim 33, wherein the processor corrects the initial value of the external parameter, and when obtaining the target external parameter is used for:

Using the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter, wherein the first map is a point cloud collected according to the first lidar The second map is established based on the point cloud data collected by the second lidar.
The movable platform according to claim 38, wherein the processor is configured to: when using the first map and the second map to correct the initial value of the external reference:

transforming the second map by using the initial value of the external parameter, and matching the transformed second map with the first map to obtain a correction value of the external parameter;

The initial value of the external parameter is corrected by using the correction value of the external parameter.
The mobile platform according to claim 38, wherein the second map includes N submaps, each submap corresponds to a part of the scene, and N is an integer greater than 1.
The movable platform according to claim 40, wherein the processor uses the first map of the scene and the second map of the scene to correct the initial value of the external parameter to obtain the target external parameter when used in:

After entering the iteration of the nth submap, transform the nth submap according to the first motion track and the current initial value of the external parameter, and match the transformed submap with the first map to obtain the external parameter correction value;

Using the external parameter correction value to correct the current external parameter initial value, and using the correction result as a new external parameter initial value, enter the iteration of the n+1th sub-map;

After the iteration of the last submap is completed, the final correction result is used as the target external parameter, wherein the n is an integer greater than 0 and less than N.
The movable platform according to claim 41, wherein the processor is used for: when using the external parameter correction value to correct the current external parameter initial value:

If the matching score between the transformed sub-map and the first map is higher than a preset threshold, the current initial value of the extrinsic parameter is corrected using the extrinsic correction value.
The mobile platform according to claim 42, wherein the processor is further used for:

If the matching score between the transformed submap and the first map is lower than the preset threshold, directly enter into iteration of the n+1th submap.
The mobile platform according to claim 41, wherein the processor matches the transformed sub-map with the first map for:

A nearest neighbor point set corresponding to the transformed submap is determined in the first map, and the transformed submap is matched with the nearest neighbor point set.
The mobile platform according to claim 38, characterized in that, the first map and/or the second map have been processed to remove non-planar feature points.
The mobile platform according to claim 38, characterized in that, the point clouds belonging to the ground in the first map and/or the second map have undergone down-sampling processing.
The movable platform according to claim 33, wherein the transformation comprises: rotation transformation and/or translation transformation.
The movable platform according to claim 33, wherein the field of view of the first laser radar does not overlap with the field of view of the second laser radar.
A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1-16 is implemented.