Unmanned vehicle multi-laser radar point cloud matching method, vehicle, device and medium

Abstract

The invention relates to an unmanned vehicle multi-laser radar point cloud matching method, a vehicle, a device and a medium. The method comprises the following steps: acquiring point cloud of each laser radar; fitting the feature points of the same vertical plane in each point cloud into a reference straight line, and judging whether all the reference straight lines coincide or not; if not, initial registration is carried out on the point clouds according to the features of the same marker in the point clouds, then accurate registration is carried out until all the reference straight lines coincide, and all the reference straight lines coincide, that is, all the point clouds are matched in a high-precision mode. According to the method, high-precision point cloud matching of the multiple laser radars of the unmanned vehicle can be simply and quickly realized, and the requirement of regular calibration of the device is met.

Description

Technical Field

[0001] The present invention relates to the field of automotive electronics, and in particular to a method, vehicle, device and medium for matching multi-laser radar point clouds of an unmanned vehicle. Background Art

[0002] Due to the continuous development and progress of laser technology, 3D laser scanning has become an emerging measurement method. Compared with traditional detection methods, 3D laser scanning can quickly obtain the 3D coordinates of the measurement target and can realize the measurement of narrow areas, dangerous areas or high-altitude areas, which makes the mine tunnel detection work under the mine more efficient and easy. The unmanned mining truck is equipped with laser radar. The front two laser radars are installed at a certain angle to each other, and the rear laser radar is installed in the center. They are used in a synchronous measurement mode to complete the full-section data measurement of the entire mine tunnel. The measurement data of the laser radar is based on the coordinate system of the sensor itself, while the full-section detection data of the tunnel line is observed and analyzed based on the coordinate system. In the case of using multiple laser radars to complete the full-section measurement, it is necessary to match the point clouds of multiple laser radars. Due to the complex and changeable road conditions under the mine, inaccurate point cloud matching will lead to inaccurate positioning and inaccurate obstacle distance detection. For unmanned mining trucks equipped with laser radars, how to simply and quickly achieve high-precision matching of point clouds of multiple laser radars and meet the needs of regular calibration of the device is a difficult problem faced in the application process of the device equipment. Summary of the invention

[0003] In order to solve the problems of low efficiency and inaccurate matching of point cloud of multiple laser radars of existing unmanned mining vehicles, the present invention provides a method, vehicle, device and medium for matching point cloud of multiple laser radars of an unmanned vehicle, which can simply and quickly realize high-precision matching of point clouds of multiple laser radars and meet the requirements of regular calibration of the device.

[0004] A first aspect of the present invention provides a method for matching multi-lidar point clouds of an unmanned vehicle, comprising:

[0005] Get the point cloud of each lidar;

[0006] Fit the feature points of the same vertical plane in each point cloud into a reference straight line, and determine whether all the reference straight lines coincide;

[0007] If not, the point clouds are initially aligned according to the feature points of the same landmark in each point cloud, and then accurately aligned until all reference lines coincide.

[0008] In conjunction with the first aspect of the present invention, in some embodiments, initially registering the point clouds according to feature points of the same marker in each point cloud includes:

[0009] The overlapping feature points of the same landmark in different point clouds are divided into a line feature point set and a surface feature point set;

[0010] Determine matching parameters of relevant point clouds according to the line feature point set or the surface feature point set;

[0011] The point cloud matching parameters are used to perform initial registration of related point clouds.

[0012] In conjunction with the first aspect of the present invention, in some implementations, determining point cloud matching parameters according to the line feature point set and the surface feature point set includes:

[0013] Determine the x-coordinate, y-coordinate and heading angle according to the line feature point set;

[0014] Determine the z coordinate and pitch angle based on the surface feature point set.

[0015] In combination with the first aspect of the present invention, in some embodiments, the unmanned vehicle includes a left front laser radar and a right front laser radar arranged in front of the vehicle, and a rear laser radar arranged in the middle of the rear of the vehicle; the point cloud matching parameters are determined according to the line feature point set and the surface feature point set, including

[0016] Determine the x-coordinate, y-coordinate and heading angle based on the line feature point set of the same marker that overlaps in the point cloud of the front laser radar and the right front laser radar;

[0017] The z coordinate and pitch angle are determined based on the overlapping surface feature point sets of the same marker in the left front laser radar point cloud, the right front laser radar point cloud, and the rear laser radar point cloud.

[0018] In conjunction with the first aspect of the present invention, in some embodiments, performing precise registration includes:

[0019] Take any point cloud as the source point cloud for sampling to obtain the source point cloud subset;

[0020] Taking another point cloud as the target point cloud, an optimized subset is obtained from the source point cloud subset so that the statistical distance between the optimized subset and the target point cloud is not greater than a threshold.

[0021] In conjunction with the first aspect of the present invention, in some implementations, obtaining an optimized subset from a source point cloud subset includes:

[0022] Find the nearest point of each point in the source point cloud subset in the target point cloud, take each point in the source point cloud subset and its nearest point in the target point cloud as a point pair, and obtain the transformation matrix of the source point cloud and the target point cloud;

[0023] The erroneous point pairs are removed by using algorithms or limiting conditions, and then the transformation matrix is ​​optimized. The optimized transformation matrix is ​​applied to the source point cloud subset to obtain the optimized subset.

[0024] In conjunction with the first aspect of the present invention, in some implementations, making the statistical distance between the optimized subset and the target point cloud no greater than a threshold comprises:

[0025] Determine whether the statistical distance between the optimized subset and the target point cloud is not greater than a threshold;

[0026] When the statistical distance is not greater than the threshold, accurate registration is completed;

[0027] When the statistical distance is greater than the threshold, the optimized subset is obtained again from the source point cloud subset.

[0028] A second aspect of the present invention provides a multi-lidar point cloud matching device for an unmanned vehicle, the device comprising:

[0029] The acquisition module is used to obtain the point cloud of each laser radar;

[0030] A judgment module is used to fit the feature points of the same vertical plane in each point cloud into a reference straight line, and judge whether all the reference straight lines coincide;

[0031] The registration module is used to perform initial registration of the point clouds according to the feature points of the same marker in each point cloud when the judgment result of the judgment module is no, and then perform precise registration until all the reference straight lines coincide.

[0032] A third aspect of the present invention provides an unmanned vehicle, comprising a processor, a memory, and an unmanned vehicle multi-lidar point cloud matching program stored in the memory and executable by the processor, wherein when the unmanned vehicle multi-lidar point cloud matching program is executed by the processor, the steps of the unmanned vehicle multi-lidar point cloud matching method as described above are implemented.

[0033] A fourth aspect of the present invention provides a computer-readable storage medium, on which is stored a multi-lidar point cloud matching program for an unmanned vehicle. When the multi-lidar point cloud matching program for an unmanned vehicle is executed by a processor, the steps of the multi-lidar point cloud matching method for an unmanned vehicle as described above are implemented.

[0034] The beneficial effects brought by the technical solution provided by the embodiment of the present invention include:

[0035] The present invention fits the feature points of the same vertical plane in each point cloud into a reference straight line respectively, and determines whether all point clouds are matched with high precision by judging whether all the reference straight lines coincide. When the point clouds are not matched with high precision, the point clouds are initially aligned according to the feature points of the same marker in each point cloud, and then accurately aligned, which solves the problems of low matching accuracy and low matching efficiency of unmanned mining car laser radar point clouds in related technologies. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The present invention is a flowchart of an embodiment of a method for matching multi-lidar point clouds of an unmanned vehicle.

[0037] Figure 2 This is a schematic diagram of the functional modules of an embodiment of a multi-lidar point cloud matching device for an unmanned vehicle of the present invention.

[0038] Figure 3 The figure is a schematic diagram of the hardware structure of the unmanned vehicle involved in the embodiment of the present invention.

[0039] Figure 4 The figure is a schematic diagram of data transmission between the autonomous driving domain controller and the lidar involved in the embodiment of the present invention.

[0040] Figure 5 This is a laser radar obstacle point cloud matching scene built in the embodiment of the present invention. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] First, some technical terms in the present invention are explained to facilitate those skilled in the art to understand the present invention.

[0043] In the present invention, a "static obstacle group" refers to a plurality of stationary obstacles.

[0044] In the present invention, the "vertical plane" can be an independent vertical plane wall or a vertical plane wall of a house, but the parallelism of the house wall must be good enough within ±1°. When setting the "vertical plane", its length must ensure that the front and rear laser radars can scan the plane features. The length of the "vertical plane" is preferably more than 20 meters.

[0045] In the present invention, a "marker" is an object having plane and / or straight line features, preferably an object having a regular geometric shape, such as a cone, a cuboid, or a cube.

[0046] In the present invention, "point cloud" refers to a data set of laser points obtained by laser radar scanning static obstacles, including the position information of the laser points in three-dimensional space, such as: the world coordinates (x, y and z), heading angle (yaw) and pitch angle (pitch) of the laser points; a "point pair" is composed of any point in the source point cloud subset and its nearest point in the target point cloud, and a "point pair data set" is composed of multiple groups of "point pairs".

[0047] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0048] As described in the background technology, the measurement data of the lidar is based on the coordinate system of the sensor itself, while the full-section detection data of the tunnel line is observed and analyzed based on the coordinate system. In the case of using multiple lidars to cooperate to complete the full-section measurement, it is necessary to match the point clouds of multiple lidars. Due to the complex and changeable road conditions under the mine, inaccurate matching of the lidar point clouds will result in inaccurate positioning and inaccurate obstacle distance detection.

[0049] In order to solve the above problems, the present invention provides a method, vehicle, device and medium for matching point clouds of multiple laser radars of an unmanned vehicle, which can simply and quickly achieve high-precision matching of point clouds of multiple laser radars of an unmanned vehicle and meet the requirements of regular calibration of the device.

[0050] The technical solution of the present invention is described in detail below through specific embodiments.

[0051] In a first aspect, an embodiment of the present invention provides a method for matching multiple laser radar point clouds of an unmanned vehicle.

[0052] In one embodiment, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the method for matching multi-lidar point clouds for unmanned vehicles of the present invention. Figure 1 As shown in FIG. 1 , the multi-lidar point cloud matching method for an unmanned vehicle includes:

[0053] Step S10, obtaining the point cloud of each laser radar;

[0054] Step S20, fitting the feature points of the same vertical plane in each point cloud into a reference straight line, and determining whether all the reference straight lines coincide;

[0055] Step S30: If not, the point clouds are initially aligned according to the feature points of the same marker in each point cloud, and then accurately aligned until all reference lines coincide.

[0056] In the above method, if all the reference lines coincide, it proves that the multi-lidar point clouds match with high precision and no registration is required. The present invention determines whether all the reference lines coincide to determine whether the multi-lidar point clouds match with high precision. The method is simple and can ensure that the multi-lidar point clouds can match with high precision. The present invention performs an initial registration of the point clouds based on the feature points of the same marker in each point cloud and then performs a precise registration, which can improve the efficiency and accuracy of the precise registration. If the initial registration is not used and the algorithm is used directly for precise registration, the precise registration needs to be iterated multiple times, and the matching may be completely wrong.

[0057] Step S10, obtaining the point cloud of each laser radar; the unmanned mining vehicle involved in the embodiment of the present invention includes a left front laser radar and a right front laser radar arranged in front of the vehicle, and a rear laser radar arranged in the middle of the rear of the vehicle. This embodiment is only an example, and the radar setting method of the unmanned mining vehicle should not be used to limit the technical solution of the present invention. The point cloud obtained by scanning the static obstacle group by each laser radar of the present invention can be divided into a left front radar point cloud, a right front radar point cloud and a rear radar point cloud according to different scanning radars, and can also be divided into a point cloud of static obstacle A, a point cloud of static obstacle B, and a point cloud of static obstacle C according to different static obstacles. The static obstacle group includes several static obstacles, among which at least one of these static obstacles has a vertical plane, and at least one obstacle can be used as a marker. The marker is required to have geometric features that are easily distinguished from other objects. For example, a cone has a conical feature and a box has a rectangular feature, which is easy to distinguish from surrounding objects and is suitable for use as a marker.

[0058] In the initial state, the point clouds scanned by the two front laser radars and the rear laser radar do not overlap, and the direct use of algorithm registration is not effective. Figure 5 As shown, the present invention parks the unmanned vehicle next to a straight wall, and uses the front and rear laser radars to scan the wall as a straight line, that is, the angle is 180 degrees. If it is not 180 degrees, then adjust the point cloud alignment parameters until the wall point clouds scanned by the front and rear laser radars are in a straight line, indicating that the front and rear laser radar point clouds are aligned.

[0059] The present invention determines that not all initial point clouds overlap, and thus performs initial registration on any two initial point clouds according to the line feature point set and the surface feature point set of the first marker in the initial point cloud, and then performs precise registration until all straight lines overlap, that is, all point clouds overlap.

[0060] Step S20, fitting the feature points of the same vertical plane in each point cloud into a reference straight line, and determining whether all the reference straight lines coincide;

[0061] In some implementations of the present invention, the feature points of the same vertical plane in each point cloud are fitted into a reference straight line to determine whether all the reference straight lines coincide. The purpose is to determine whether the point clouds coincide and whether they need to be registered. If any two straight lines do not coincide, the point clouds where the two straight lines are located do not coincide, and the point clouds where the two straight lines are located need to be registered; similarly, if two straight lines coincide, the two point clouds where they are located coincide, and the registration of these two point clouds can be omitted during registration; if all straight lines coincide, the point clouds obtained by each laser radar coincide, and no registration is required.

[0062] In some implementations of the present invention, the present invention fits the feature points of the same vertical plane in each point cloud into a straight line, and determines whether all the straight lines overlap; specifically, it includes: fitting the feature points of the first vertical plane in the left front radar point cloud into a straight line L, fitting the feature points of the first vertical plane in the right front radar point cloud into a straight line M, and fitting the feature points of the first vertical plane in the rear front radar point cloud into a straight line N, and determining whether the straight lines L, M, and N overlap. If any two straight lines do not overlap, it proves that the point clouds they are in do not overlap. The goal of the present invention is to make the left front radar point cloud, the right front radar point cloud, and the rear radar point cloud overlap, and the criterion for judging the overlap of the three is that the straight lines L, M, and N overlap.

[0063] Step S30, if not, then perform initial registration of the point clouds according to the feature points of the same marker in each point cloud, and then perform precise registration until all reference lines coincide; including:

[0064] Step S31, dividing the overlapping feature points of the same marker in different point clouds into a line feature point set and a surface feature point set;

[0065] Step S32, determining matching parameters of the relevant point cloud according to the line feature point set or the surface feature point set; and performing initial registration on the relevant point cloud using the point cloud matching parameters.

[0066] For example, Figure 5 As shown, in the laser radar obstacle point cloud matching scene constructed by the present invention, three cone barrels are placed in the point cloud overlapping area, four rectangular boxes are placed in the point cloud overlapping area, and the point cloud matching scene of the rear laser radar is constructed. The cone barrels are placed directly behind the rear radar, and the vehicle is parked in a place with a straight wall; Figure 4As shown in the figure, the point cloud scanned by three laser radars on the static obstacle is obtained through the autonomous driving domain controller (ADCU), and the point cloud rosbag data is recorded. The point cloud rosbag data includes the surface feature points and line feature points of the four rectangular boxes. These surface feature points and line feature points overlap in the point cloud of the two forward laser radars and can be divided into line feature point sets or surface feature point sets. The side edges and top edges of the four rectangular boxes can constrain the x-coordinate, y-coordinate and heading angle yaw. With the cooperation of the four rectangular boxes, the parameters of the x-coordinate, y-coordinate and heading angle yaw can be matched, that is, the matching parameters of the relevant point clouds are determined by the line feature point set. The four rectangular boxes and four cone barrels can be used to constrain the z-coordinate, pitch angle and roll angle. These parameters determine the relative position relationship of the point cloud.

[0067] In some embodiments, the driverless vehicle includes a left front laser radar and a right front laser radar arranged in front of the vehicle, and a rear laser radar arranged in the middle of the rear of the vehicle; point cloud matching parameters are determined according to the line feature point set and the surface feature point set, including: determining the x coordinate, y coordinate and heading angle according to the line feature point set of the same marker in the point cloud of the front laser radar and the right front laser radar; determining the z coordinate and pitch angle according to the surface feature point set of the same marker in the point cloud of the left front laser radar, the point cloud of the right front laser radar, and the point cloud of the rear laser radar. The two laser radars in front of the vehicle determine x, y and z as well as the pitch angle and heading angle, and the rear laser radar determines the registration parameters with the two front laser radars through a straight wall to determine the relative position relationship.

[0068] If the point clouds of the first two laser radars are not initially matched, the matching efficiency will be low and the matching accuracy will be insufficient. The present invention uses the obstacle feature method to perform initial registration of the source point cloud and the target point cloud, which can greatly improve the efficiency of the algorithm registration. If the initial registration is not used and the algorithm is directly used for registration, then there may even be a situation where the matching is completely wrong.

[0069] After the initial registration, the effect and error of the point cloud registration cannot meet the required accuracy. The present invention reduces the error by using precise registration, thereby achieving the required accuracy.

[0070] In step S30, accurate registration is performed, including:

[0071] Step S33, taking any point cloud as the source point cloud for sampling to obtain a source point cloud subset;

[0072] Step S34, taking another point cloud as the target point cloud, obtaining an optimized subset from the source point cloud subset, so that the statistical distance between the optimized subset and the target point cloud is less than a threshold.

[0073] In some embodiments, step S34, taking another point cloud as a target point cloud and obtaining an optimized subset from a source point cloud subset, includes:

[0074] Step S341, searching the target point cloud for the nearest point of each point in the source point cloud subset, taking each point in the source point cloud subset and its nearest point in the target point cloud as a point pair, and obtaining the transformation matrix of the source point cloud and the target point cloud;

[0075] Step S342, using an algorithm or limiting conditions to remove erroneous point pairs, then optimizing the transformation matrix, and applying the optimized transformation matrix to the source point cloud subset to obtain an optimized subset.

[0076] In step S34, making the statistical distance between the optimized subset and the target point cloud smaller than a threshold includes:

[0077] Step S343, determining whether the statistical distance between the optimized subset and the target point cloud is less than a threshold;

[0078] Step S344, when the statistical distance is less than the threshold, the precise registration is completed; when the statistical distance is greater than the threshold, the optimized subset is re-obtained from the source point cloud subset.

[0079] The initial registration algorithm provides a good initial position for the accurate registration of the two point clouds, and the accurate registration further reduces the registration error.

[0080] Given two point clouds P and Q to be registered, with P as the source point cloud and Q as the target point cloud, the source point cloud P and the target point cloud Q need to coincide, and the position of the source point cloud P needs to be transformed. The position transformation of the source point cloud P is actually the coordinates of the source point cloud P multiplied by the optimized transformation matrix. The optimized transformation matrix is ​​divided into a rotation matrix and a translation matrix.

[0081] The source point cloud P can be expressed as P = {p i ∈R 3 , i=1,2,L,N p}, R 3 represents the full set of three-dimensional coordinates, Np represents the limit boundary of the points in the source point cloud set, and the target point cloud Q can be expressed as {q i ∈R 3 , i=1,2,L,N q}, Nq represents the limit boundary of the points in the target point cloud set. i For any point in the point cloud P, find the distance p from the target point cloud Q i The point q with the shortest Euclidean distance i , and p i and q i The transformation matrix is ​​obtained as a point pair, and the ICP algorithm is continuously iterated. Formula (1) is used as the condition for stopping the iteration, and finally the optimized transformation matrix is ​​obtained so that the source point cloud P and the target point cloud Q coincide.

[0082]

[0083] Where R represents the rotation transformation matrix, T represents the translation transformation matrix, and pi represents any point in the source point cloud P; q i Represents a point in the target point cloud Q; f(R, T) represents the function of the statistical distance d between the source point cloud P and the target point cloud Q; min represents the minimum value of the objective function, and k represents the number of iterations.

[0084] The essence of the ICP algorithm is to continuously rotate and translate by matching points, using the least squares method as a criterion until the distance between points reaches a preset threshold. The present invention uses the ICP algorithm to calculate the optimized transformation matrix of the source point cloud P and the target point cloud Q, and minimizes the registration error of the source point cloud P and the target point cloud Q by rotation and translation, thereby achieving the best registration effect.

[0085] The following is a detailed description of the process of registering the source point cloud P and the target point cloud Q using the ICP algorithm:

[0086] (1) Sample the source point cloud P to obtain the source point cloud subset P 0 , P 0 ∈P, represents a subset of the source point cloud P;

[0087] (2) Search in the target point cloud Q to find the subset P of the source point cloud 0 For each point p i The nearest point q i ,Right now p i and q i Form a point pair to obtain the initial correspondence between the source point cloud P and the target point cloud Q; use the ICP algorithm or limit conditions to remove the wrong point pairs; the limit condition is: after the initial registration, the statistical distance d between the source point cloud P and the target point cloud Q should not exceed 10 times the threshold t. If the statistical distance d is greater than 10t, then p is considered i and q i It is an incorrect point pair and needs to be deleted; statistical distance P i is any point in the source point cloud P, Q i is any point in the target point cloud Q.

[0088] (3) Calculate the transformation matrix of the source point cloud P and the target point cloud Q according to the initial correspondence between the source point cloud P and the target point cloud Q in step (2) to minimize the value of formula (1) and obtain the optimized transformation matrix; apply the optimized transformation matrix to the source point cloud subset P 0 , after transformation, we get the optimized subset P 0 '.

[0089] (4) Determine whether the iteration is terminated based on the size of the statistical distance d. If the statistical distance d is greater than the preset threshold τ, return to process (2) to continue iterating. If the statistical distance d is less than the preset threshold τ, or reaches the previously set number of iterations k, the algorithm converges and the iteration stops.

[0090] In a second aspect, an embodiment of the present invention also provides a multi-lidar point cloud matching device for an unmanned vehicle. Figure 2 Schematic diagram of functional modules of an embodiment of a multi-lidar point cloud matching device for an unmanned vehicle according to the present invention. Figure 2 As shown, the multi-lidar point cloud matching device for an unmanned vehicle includes:

[0091] An acquisition module 10 is used to acquire the point cloud of each laser radar;

[0092] A judgment module 20 is used to fit the feature points of the same vertical plane in each point cloud into a reference straight line, and judge whether all the reference straight lines coincide with each other;

[0093] The registration module 30 is used to perform initial registration of the point clouds according to the feature points of the same marker in each point cloud when the judgment result of the judgment module is no, and then perform precise registration until all the reference straight lines coincide.

[0094] Furthermore, in one embodiment, the registration module 30 includes an initial registration module and a precise registration module.

[0095] Furthermore, in one embodiment, the initial registration module is used to: divide the overlapping feature points of the same marker in different point clouds into a line feature point set and a surface feature point set; determine the matching parameters of the relevant point clouds based on the line feature point set or the surface feature point set; and use the point cloud matching parameters to perform initial registration on the relevant point clouds.

[0096] Furthermore, in one embodiment, the precise registration module is used to: sample any point cloud as a source point cloud to obtain a source point cloud subset; take another point cloud as a target point cloud, obtain an optimized subset from the source point cloud subset, so that the statistical distance between the optimized subset and the target point cloud is less than a threshold.

[0097] Among them, the functional implementation of each module in the above-mentioned unmanned vehicle multi-lidar point cloud matching device corresponds to the various steps in the above-mentioned unmanned vehicle multi-lidar point cloud matching method embodiment, and its functions and implementation processes will not be repeated here one by one.

[0098] In a third aspect, an embodiment of the present invention further provides an unmanned vehicle, such as Figure 3As shown, the unmanned vehicle includes a processor, a memory, and an unmanned vehicle multi-lidar point cloud matching program stored in the memory and executable by the processor, wherein when the unmanned vehicle multi-lidar point cloud matching program is executed by the processor, the steps of the unmanned vehicle multi-lidar point cloud matching method as described above are implemented.

[0099] Furthermore, in one embodiment, it includes a left front laser radar and a right front laser radar arranged in front of the vehicle, and a rear laser radar arranged in the middle of the rear of the vehicle.

[0100] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium.

[0101] The computer-readable storage medium of the present invention stores an unmanned vehicle multi-lidar point cloud matching program, wherein when the unmanned vehicle multi-lidar point cloud matching program is executed by a processor, the steps of the unmanned vehicle multi-lidar point cloud matching method as described above are implemented.

[0102] Among them, the method implemented when the unmanned vehicle multi-lidar point cloud matching program is executed can refer to the various embodiments of the unmanned vehicle multi-lidar point cloud matching method of the present invention, and will not be repeated here.

[0103] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0104] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present invention.

[0105] The terms "including" and "having" and any variations thereof in the specification and claims of the present invention and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or equipment. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.

[0106] In the description of the embodiments of the present invention, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present invention should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.

[0107] In the description of the embodiments of the present invention, unless otherwise specified, “ / ” means or. For example, A / B can mean A or B. The “and / or” in the text is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of the present invention, “multiple” refers to two or more than two.

[0108] In some processes described in the embodiments of the present invention, multiple operations or steps appearing in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present invention or may be executed in parallel, and the sequence numbers of the operations are only used to distinguish different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0109] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for matching multi-lidar point clouds of an unmanned vehicle, characterized in that: include: Get the point cloud of each lidar; Fit the feature points of the same vertical plane in each point cloud into a reference straight line, and determine whether all the reference straight lines coincide; If not, the point clouds are initially aligned according to the feature points of the same landmark in each point cloud, and then accurately aligned until all reference lines coincide.

2. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 1, characterized in that: The point clouds are initially registered based on the feature points of the same landmark in each point cloud, including: The overlapping feature points of the same landmark in different point clouds are divided into a line feature point set and a surface feature point set; Determine matching parameters of relevant point clouds according to the line feature point set or the surface feature point set; The point cloud matching parameters are used to perform initial registration of related point clouds.

3. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 2, characterized in that: Determining point cloud matching parameters based on line feature point sets and surface feature point sets includes: Determine the x-coordinate, y-coordinate and heading angle according to the line feature point set; Determine the z coordinate and pitch angle based on the surface feature point set.

4. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 3, characterized in that: The driverless car includes a left front laser radar and a right front laser radar arranged in front of the vehicle, and a rear laser radar arranged in the middle of the rear of the vehicle; the point cloud matching parameters are determined according to the line feature point set and the surface feature point set, including Determine the x-coordinate, y-coordinate and heading angle based on the line feature point set of the same marker that overlaps in the point cloud of the front laser radar and the right front laser radar; The z coordinate and pitch angle are determined based on the overlapping surface feature point sets of the same marker in the left front laser radar point cloud, the right front laser radar point cloud, and the rear laser radar point cloud.

5. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 1, characterized in that: Perform precise registration, including: Take any point cloud as the source point cloud for sampling to obtain the source point cloud subset; Taking another point cloud as the target point cloud, an optimized subset is obtained from the source point cloud subset so that the statistical distance between the optimized subset and the target point cloud is not greater than a threshold.

6. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 5, characterized in that: Get the optimized subset from the source point cloud subset, including: Find the nearest point of each point in the source point cloud subset in the target point cloud, take each point in the source point cloud subset and its nearest point in the target point cloud as a point pair, and obtain the transformation matrix of the source point cloud and the target point cloud; The erroneous point pairs are removed by using algorithms or limiting conditions, and then the transformation matrix is ​​optimized. The optimized transformation matrix is ​​applied to the source point cloud subset to obtain the optimized subset.

7. The method for matching multi-lidar point clouds of an unmanned vehicle according to claim 6, characterized in that: Make the statistical distance between the optimized subset and the target point cloud no greater than the threshold, including: Determine whether the statistical distance between the optimized subset and the target point cloud is not greater than a threshold; When the statistical distance is not greater than the threshold, accurate registration is completed; When the statistical distance is greater than the threshold, the optimized subset is obtained again from the source point cloud subset.

8. A multi-lidar point cloud matching device for an unmanned vehicle, characterized in that: The device comprises: The acquisition module is used to obtain the point cloud of each laser radar; A judgment module is used to fit the feature points of the same vertical plane in each point cloud into a reference straight line, and judge whether all the reference straight lines coincide; The registration module is used to perform initial registration of the point clouds according to the feature points of the same marker in each point cloud when the judgment result of the judgment module is no, and then perform precise registration until all the reference straight lines coincide.

9. An unmanned vehicle, characterized in that: The unmanned vehicle includes a processor, a memory, and an unmanned vehicle multi-lidar point cloud matching program stored in the memory and executable by the processor, wherein when the unmanned vehicle multi-lidar point cloud matching program is executed by the processor, the steps of the unmanned vehicle multi-lidar point cloud matching method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an unmanned vehicle multi-lidar point cloud matching program, wherein when the unmanned vehicle multi-lidar point cloud matching program is executed by the processor, the steps of the unmanned vehicle multi-lidar point cloud matching method as described in any one of claims 1 to 7 are implemented.

Images