Laser radar external parameter calibration method based on tunnel wall

Abstract

A laser radar external parameter calibration method based on a tunnel wall comprises the following steps: stopping a train in a long straight tunnel, obtaining point cloud data through a laser radar, naming the point cloud data as first point cloud data, filtering and clustering the first point cloud, extracting to obtain noisy-point-removed point cloud data, naming the point cloud data as second point cloud data, calculating a loss function for the second point cloud data, and calculating an external parameter of the laser radar according to the loss function. And optimizing the loss function by using an optimization theory to obtain an optimal parameter which is an external parameter between the laser radar needing to be calibrated and the vehicle coordinate system. The method is high in automation degree, does not need any calibration target plate, does not need GNSS / IMU / camera and other sensor signals, does not need carrier motion, does not need additional measurement equipment such as a total station / three-coordinate instrument and does not need a special calibration field, realizes one-key full-automatic calibration, and is extremely low in implementation complexity and high in robustness.

Description

Technical Field

[0001] This invention relates to a method for calibrating the extrinsic parameters of a lidar system based on tunnel walls. Background Technology

[0002] The operational safety of high-speed railways and passenger trains is a crucial link in the current railway safety control system, and the safety status of the perimeter environment has a significant impact on train operation safety. Railway perimeter safety incidents are typically sudden and sporadic, and traditional methods relying primarily on manual inspections and static control are no longer adequate to meet the real-time and comprehensive safety management requirements of modern efficient railway operations. Therefore, it is urgent to introduce advanced technologies to achieve automatic, rapid, and accurate identification of perimeter safety hazards, reducing the burden and complexity of manual inspections and comprehensively improving the safety assurance level of train operations. Existing LiDAR-based train collision avoidance and lookout systems are based on the principle of identifying perimeter intrusions by sensing the occlusion effect of intruding objects on the monitored area. The core target of detection is obstacles that may affect train operation safety, and whether they constitute a threat needs to be determined based on their relative positional relationship with the rails. The accurate identification of this positional relationship depends on the initial installation pose of the LiDAR sensor and the track, i.e., the extrinsic parameters between the LiDAR and the vehicle. The accuracy of these extrinsic parameters directly affects the accuracy of obstacle alarms.

[0003] Currently, there is a wealth of research on the external parameter calibration of lidar and vehicles, but each has its own application conditions, and most are not suitable for rail transit scenarios. A summary is provided below: I. Calibration methods based on calibration materials.

[0004] 1. Static calibration based on calibration objects is a traditional technical approach for lidar extrinsic parameter calibration. Its core principle is to establish a coordinate transformation relationship between the lidar coordinate system and the target coordinate system (such as the world coordinate system or vehicle coordinate system) by pre-setting a physical target with precise geometric features (such as a calibration plate, reflective cone, or planar diffuse reflector). The common technical process of this method can be summarized as follows: calibration object deployment → point cloud data acquisition → feature extraction and fitting → coordinate transformation parameter solution.

[0005] The key to this calibration method is obtaining the coordinates of the calibration object in the vehicle coordinate system beforehand. Generally, there are two methods: First, a total station can be used to accurately measure the positional relationship between the calibration object and the vehicle, thus obtaining the precise coordinates of the calibration object in the vehicle coordinate system. Second, a dedicated calibration field can be prepared in advance, where the positions of the vehicle and the calibration object are precisely determined and deployed, thus obtaining the coordinates of the calibration object in the vehicle coordinate system. However, the first method requires a total station, which is expensive and the calibration process is time-consuming and labor-intensive; the second method requires a dedicated calibration field, which is impractical for trains.

[0006] 2. Calibration method based on environmental characteristics.

[0007] Environmental feature-based methods calibrate extrinsic parameters by extracting naturally occurring geometric structures (such as planes and lines) or utilizing dynamic motion processes. Their core advantage lies in eliminating the need for manual calibration objects, enabling real-time or online calibration in actual operating scenarios. This significantly reduces reliance on specific locations and manual operations. Extrinsic parameter solutions are typically achieved through feature extraction, multi-frame matching, and optimization algorithms. While this method has become a mature technology in fields such as autonomous driving and robot navigation, it still faces challenges due to the unique environment of rail transit. This method generally requires plane extraction, but rail transit scenarios are complex (tunnels, ballast tracks, elevated lines, etc.), making it difficult to find flat planes, thus failing to meet requirements for plane fitting accuracy and feature robustness.

[0008] 3. Trajectory optimization method based on motion excitation.

[0009] Motion-driven trajectory optimization methods employ a calibration strategy that designs specific motion trajectories (such as 3D rotation, translation, or complex path driving) to enable sensors to acquire rich motion geometric constraint information, thereby achieving dynamic optimization of extrinsic parameters. Its core logic lies in leveraging the spatiotemporal consistency between point cloud data and external positioning data (such as GPS / GNSS, IMU) during device movement, and minimizing trajectory residuals to solve for the transformation relationship between the LiDAR and other sensors or the vehicle's coordinate system. A typical scheme, such as LiDAR_IMU_Init, requires the device to remain stationary for 5 seconds to build an initial map, followed by 15 seconds of 3D rotation and translation to excite extrinsic parameter errors. The system automatically detects the excitation level and optimizes online.

[0010] The most significant characteristic of this approach is that it requires the vehicle to perform sufficiently varied motions and GNSS or IMU data. In the rail transit field, trains must follow fixed track patterns, making it impossible to perform serpentine or rotating excitations, which greatly reduces the calibration effectiveness. Furthermore, tunnels are GNSS-denied environments, making it impossible to provide GNSS data. Adding an IMU would also introduce additional costs and complexity. Motion excitation methods rely on testing conditions of "non-planar motion + low-speed travel + long-distance complex routes + rich environmental features," but the inherent attributes of rail transit—"fixed trajectory + high-speed operation + closed lines + structured environment"—make it impossible to meet these requirements, becoming the main bottleneck for the application of this type of method in rail transit scenarios.

[0011] II. The unique characteristics of rail transit scenarios impose limitations on lidar extrinsic parameter calibration.

[0012] 1. Venue and equipment limitations.

[0013] Rail transit lines have significant unique characteristics and spatial enclosed features, and their special site constraints pose fundamental constraints on the calibration of lidar external parameters. For example, there are spatial constraints in underground tunnels and time windows for daily maintenance.

[0014] 2. Complex environmental characteristics.

[0015] The complexity of the environment in rail transit scenarios is reflected in multiple dimensions such as structural diversity, variability of interference factors, and differences in target characteristics, which directly affect the stability and accuracy of lidar extrinsic parameter calibration. Ballast tracks have a large number of uneven sleepers and gravel, which are highly repetitive and prone to misjudgment; tunnel walls are smooth and featureless, and have curved surfaces; elevated and other open scenarios lack vertical features, and the overall environment lacks flat planes, which poses a great challenge to the calibration of traditional lidar extrinsic parameters.

[0016] 3. Motion constraints.

[0017] Existing kinematic methods impose multiple stringent requirements on the carrier motion for extrinsic parameter calibration, which significantly conflict with the rail transit scenario. Rail transit trains can only run along a pre-set rail path, mostly in near-straight lines, with a minimum turning radius of only 300 meters, making large-angle yaw impossible, and roll and pitch are even more out of the question. These constraints collectively constitute the unique challenge of lidar extrinsic parameter calibration in rail transit scenarios. Traditional methods relying on motion excitation are unable to adapt to its operational characteristics and cannot meet the stringent positioning accuracy requirements for safe operation.

[0018] 4. Signal rejection.

[0019] Most urban rail and subway lines are underground, making it impossible to obtain high-precision GNSS positioning signals. The signal limitations and cost constraints (high-precision sensors, edge computing, maintenance costs) in rail transit scenarios are intertwined, making existing methods that rely on GNSS / INS fusion, high-precision maps, or complex calibration tools unusable. There is an urgent need for new calibration solutions that are low-cost and resistant to interference.

[0020] 5. Calibration process complexity.

[0021] Because rail transit needs to serve public transportation, its maintenance is generally carried out at night with a very short window of only about 3 hours. This places high demands on the complexity and automation of the calibration process. The shortcomings of existing methods in terms of automation, process efficiency, and environmental robustness highlight the urgency of developing automated calibration technologies specifically for rail transit. This requires reducing manual intervention, simplifying multi-sensor collaborative processes, and optimizing environmentally adaptable algorithms to meet the demands of high-frequency, high-precision calibration.

[0022] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art. Summary of the Invention

[0023] The purpose of this invention is to provide a lidar extrinsic parameter calibration method based on tunnel walls. This method is highly automated, requires no calibration target plate, no other sensor signals such as GNSS / IMU / camera, no carrier movement, no additional measuring equipment such as total station / coordinate measuring machine, and no special calibration field. It achieves one-click fully automated calibration with extremely low implementation complexity and high robustness.

[0024] To achieve the above objectives, the present invention provides a method for calibrating the extrinsic parameters of a lidar system based on tunnel walls, comprising: The train was stopped in the long, straight tunnel; Point cloud data acquired using lidar is called the first point cloud data. By filtering and clustering the first point cloud, noise-removed point cloud data is extracted, which is called the second point cloud data. The loss function is calculated for the second point cloud data. The loss function is then optimized using optimization theory. The optimal parameters obtained are the extrinsic parameters between the LiDAR and the vehicle coordinate system to be calibrated.

[0025] The length of the tunnel is greater than or equal to 30m.

[0026] Optionally, the loss function is: f = D y + D z ,or f = D y × D z ; in, xyz Using the vehicle coordinate system, draw the bounding box of the entire tunnel point cloud. y The side length along the axis is D y enclosed box z The side length along the axis is D z , f It is a function of heading angle and pitch angle.

[0027] Optionally, the loss function is: g = d ; Among these methods, a circular equation is fitted to the point cloud of the tunnel wall, and the thickness of the point cloud of the tunnel wall is... d , gIt is a function of heading angle and pitch angle.

[0028] Optionally, the loss function is the projected area.

[0029] The optimization method includes at least one of the following: direct search method, numerical optimization method based on calculus, and metaheuristic method.

[0030] The present invention has the following beneficial effects: 1. No calibration target plate is required, truly achieving targetless and fully automatic external parameter calibration.

[0031] 2. It does not require GNSS, IMU or other sensor signals, and is still applicable in environments where signals are blocked, such as underground tunnels and outdoor areas, making it suitable for rail transit applications.

[0032] 3. No carrier movement is required, and no motion constraints need to be met. External parameter calibration can be completed while the entire process is stationary, which reduces the complexity of the calibration process and improves stability and robustness, making it user-friendly for rail transit applications.

[0033] 4. No additional measuring equipment is required, which reduces costs, simplifies the calibration process, and improves stability and robustness, making it user-friendly for rail transit applications.

[0034] 5. No special customized calibration field is required; only a long, straight tunnel is needed. This is very easy for rail transit scenarios, reducing the complexity of the calibration process, improving stability and robustness, making it user-friendly for rail transit applications, highly reusable, and easy to port.

[0035] 6. The algorithm used is simple and does not involve complex black-box algorithms such as neural networks. The entire process is controllable and transparent, with strong interpretability and reliability. The security principle analysis is user-friendly and can achieve SIL4 security certification through 2-out-of-2 and heterogeneous architecture design. Attached Figure Description

[0036] Figure 1 This is a flowchart of a lidar extrinsic parameter calibration method based on a tunnel wall provided in an embodiment of the present invention.

[0037] Figure 2 This is a point cloud data map in one embodiment.

[0038] Figure 3 This is a point cloud data map in another embodiment. Detailed Implementation

[0039] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become clearer from the following description. It should be noted that the drawings are in a very simplified form and use non-precise proportions, used only to facilitate and clearly illustrate the embodiments of the present invention. Please refer to the drawings to make the objectives, features, and advantages of the present invention more apparent and understandable. It should be understood that the structures, proportions, sizes, etc., depicted in the accompanying drawings are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the implementation conditions of the present invention. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in proportions, or adjustments to the size, without affecting the effects and objectives achieved by the present invention, should still fall within the scope of the technical content disclosed in the present invention.

[0040] This invention provides a method for calibrating the extrinsic parameters of a lidar system based on tunnel walls, comprising the following steps: Step 101: Stop the train in the long, straight tunnel; This is also the only limiting condition of the present invention. For rail transit scenarios, long and straight tunnels exist in large numbers in the lines, and for subways, they are even the main operating scenario. Therefore, this limiting condition is easy to meet and is very applicable to rail transit scenarios. The tunnel length is related to the required calibration accuracy. The longer the tunnel, the higher the calibration accuracy. It is generally recommended that the tunnel length be greater than or equal to 30m. Step 102: Acquire point cloud data using lidar, referred to as the first point cloud data; Data is acquired using the radar that needs to be calibrated; no data is needed for radars that do not require calibration. There are no requirements for the radar installation location. The first point cloud data is a frame of point cloud scanned by radar, which contains all the information in the scene and is the most original data directly output by radar; Step 103: By filtering and clustering the first point cloud, relatively clean point cloud data after removing noise is extracted, which is called the second point cloud data; The main function of this step is to remove noise from the point cloud and improve the accuracy and stability of the calibration results; Step 104: Calculate the loss function for the second point cloud data, and optimize the loss function using optimization theory; The loss function here is designed according to certain rules / logic and can take many forms; There are many methods for optimizing the loss function, and this invention does not impose any limitations on them. As an example, common methods for optimizing the loss function include, but are not limited to: direct search methods (grid search, pattern search, etc.), numerical optimization methods based on calculus (gradient descent, Newton's method, BFGS, etc.), and metaheuristic methods (simulated annealing, genetic algorithm, particle swarm optimization, etc.). Step 105: The optimal parameters obtained through optimization in step 104 are the extrinsic parameters between the LiDAR and the vehicle coordinate system to be calibrated.

[0041] In one embodiment of the present invention, Figure 2 (a) shows the second point cloud data obtained after steps 101, 102, and 103, with the lower right corner showing... xyz The coordinate system is the vehicle coordinate system. x The axis is perpendicular to the plane and faces inward. y The axis is to the left. z (Axis facing upwards). It can be seen that due to the radar installation angle deviation, the point cloud of the tunnel wall is tilted in the vehicle coordinate system. At this point, the bounding box of the entire tunnel point cloud (i.e., bounded by the smallest...) can be drawn. x , minimum y , minimum z ,maximum x ,maximum y ,maximum z (a cube formed by vertices), the bounding box is in y The side length along the axis is D y , z The side length along the axis is D z .like Figure 2 As shown in (b), when the second point cloud is rotated so that its depth direction is aligned with the train coordinate system... x When the positive direction of the axis is parallel, it is easy to obtain the following according to Euclidean geometry: D y and D z It should be the minimum. Therefore, based on the principle of minimizing the bounding box side length, a loss function can be written for step 104, namely... f = D y + D z ,or f = D y × D z , f It is a function of the heading and pitch angles. Then, using the heading and pitch angles as parameters to be optimized, an optimization problem argmin is established. fThe loss function is obtained. f Take the heading and pitch angles corresponding to the minimum values ​​to complete the calibration of the lidar extrinsic parameters.

[0042] In another embodiment of the invention, the tunnel wall is a standard circle due to construction requirements. For example... Figure 3 As shown, a circular equation can be fitted to the point cloud of the tunnel wall, at which point the thickness of the point cloud of the tunnel wall is... d Similarly, it is easy to obtain from Euclidean geometric relations that when the second point cloud is rotated, it becomes: d When the minimum value is reached, the extrinsic parameter calibration of the heading and pitch angles is complete. Therefore, based on this principle of minimizing tunnel wall thickness, another loss function in step 104 can be written, namely... g = d , g It is a function of the heading and pitch angles. Then, using the heading and pitch angles as parameters to be optimized, an optimization problem argmin is established. g The loss function is obtained. g Take the heading and pitch angles corresponding to the minimum values ​​to complete the calibration of the lidar extrinsic parameters.

[0043] In addition to the two loss functions listed in the examples, the projected area can also be used. For example, the projected area of ​​the tunnel onto the yz plane can be used as the loss function. Furthermore, the loss function can be modified, such as by wrapping it with a kernel function to smooth it out, or by multiplying Dx and Dy by different numbers to adjust their weights.

[0044] Generally, for optimization problems, the loss function should be designed to be smooth, differentiable, and preferably convex, which is beneficial for later optimization searches. If the designed loss function is not differentiable or inconvenient to differentiate, then numerical optimization methods based on calculus are not applicable. This invention utilizes the special properties of rail transit scenarios to propose a lidar extrinsic parameter calibration method suitable for rail transit. By using the point cloud of the tunnel wall in the rail transit tunnel environment, the extrinsic parameter calibration process is transformed into an optimization problem of the loss function. Extrinsic parameter calibration can be achieved through optimization search. It has a high degree of automation, requires no calibration target, no other sensor signals such as GNSS / IMU / camera, no carrier movement, no additional measuring equipment such as total station / coordinate measuring machine, and no special calibration field. It achieves one-click fully automated calibration with extremely low implementation complexity and high robustness.

[0045] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0046] In the description of this invention, it should be understood that the terms "center," "height," "thickness," "upper," "lower," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0047] In the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0048] In this invention, unless otherwise explicitly specified and limited, "above" or "below" the second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of the second feature includes the first feature directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" the second feature includes the first feature directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature.

[0049] Although the present invention has been described in detail through the preferred embodiments above, it should be understood that the above description should not be considered as a limitation of the present invention. Various modifications and substitutions to the present invention will be apparent to those skilled in the art after reading the above description. Therefore, the scope of protection of the present invention should be defined by the appended claims.

Claims

1. A method for calibrating extrinsic parameters of a tunnel-wall-based lidar, characterized in that, Comprise: Stopping the train in a long straight tunnel; Obtaining point cloud data by laser radar, referred to as first point cloud data; Through filtering and clustering of the first point cloud, point cloud data after removing noise is extracted, referred to as second point cloud data; Calculating a loss function for the second point cloud data, and optimizing the loss function by optimization theory to obtain the best parameters, which are the extrinsic parameters between the laser radar and the vehicle coordinate system required for calibration.

2. The tunnel wall based laser radar extrinsic calibration method of claim 1, wherein, The length of the tunnel is greater than or equal to 30m.

3. The tunnel wall based laser radar extrinsic calibration method of claim 1, wherein, The loss function is: f = D y + D z , or f = D y × D z ; wherein, xyz The coordinate system is a vehicle coordinate system, and a bounding box of the entire tunnel point cloud is drawn, the bounding box being in y The edge length in the axis direction is D y , the bounding box being in z The edge length in the axis direction is D z , f is a function of the heading angle and the pitch angle.

4. The tunnel wall based laser radar extrinsic calibration method of claim 1, wherein, The loss function is: g = d ； wherein a circle equation is fitted to the point cloud of the tunnel wall, the thickness of the tunnel wall point cloud is d , g is a function of the heading angle and the pitch angle.

5. The tunnel wall based laser radar extrinsic calibration method of claim 1, wherein, The loss function is the projected area.

6. The tunnel wall based laser radar extrinsic calibration method of claim 1, wherein, The optimization method at least contains any one of the following: direct search method, numerical optimization method based on calculus, meta-heuristic method.

Images