Optimization method based on non-uniform sampling of mechanical laser radar
By introducing local rate of change and weight factors, calculating curvature values and dynamically adjusting weight allocation, the problem of incomplete data sparsity and feature extraction of LiDAR point clouds is solved, and the accuracy and robustness of data processing are improved, and it is suitable for complex environments. autonomous driving and autonomous mobile robots.
Patent Information
- Application Number
- CN202410940909.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-15
- Publication Date
- 2024-11-15
AI Technical Summary
The existing LiDAR point cloud data is sparse at long distances, incomplete feature extraction, and unreasonable weight allocation, resulting in insufficient incompleteness and robustness of data processing, especially in complex environments to be sensitive to noise and interference.
By introducing local rate of change and weight factors, the curvature value of each point cloud point is calculated, the weight allocation of point cloud points is dynamically adjusted, the lidar position is optimized, the expression accuracy and comprehensive feature extraction of long-distance object point cloud data are improved, and the overall performance of long-distance object point cloud data is enhanced. The robustness of the algorithm.
It improves the expression accuracy of long-distance point cloud data, avoids feature extraction and missed detection, enhances the robustness and real-time processing capabilities of the algorithm in complex environments, and provides more reliable pose and map construction effects.
Smart Images

Figure CN118962641A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of automatic driving and autonomous mobile robots, and in particular to an optimization method based on uneven sampling of a mechanical laser radar. Background Art
[0002] Autonomous driving and autonomous mobile robot technologies have received widespread attention and developed rapidly in recent years. As a key perception device, LiDAR (laser radar) can provide high-precision three-dimensional environmental data, which is of great significance for autonomous driving and robot mapping. However, there are some problems in the application of existing LiDAR point cloud processing technology, which are mainly manifested in:
[0003] Sparsity of point cloud data: As the LiDAR scanning distance increases, the number of points in the same volume decreases, resulting in sparse representation of distant objects in the point cloud data, affecting the integrity and accuracy of the data.
[0004] Incomplete feature extraction: The existing technology uses a single-beam curvature calculation method, which is easy to miss the edge points of the step in a non-parallel state, resulting in incomplete feature extraction.
[0005] Unreasonable weight distribution: The existing LiDAR SLAM algorithm does not adjust the weight of each point. All points are given the same weight, ignoring the measurement uncertainty and signal-to-noise ratio difference, which reduces the robustness of the algorithm.
[0006] When attempting to solve the above problems, existing technologies have defects such as incomplete feature extraction and unreasonable weight distribution, which makes the optimization algorithm sensitive to noise and interference and fails to make full use of high-quality data. Summary of the invention
[0007] In view of the many problems existing in the above-mentioned prior art, the present invention provides an optimization method based on uneven sampling of mechanical lidar. The present invention improves the expression accuracy of point cloud data of distant objects by introducing local change rate and weight factor, and dynamically adjusts the weight of each point according to the distance from the point to the sensor, thereby enhancing the robustness and real-time processing capability of the algorithm in complex environments.
[0008] The optimization method based on uneven sampling of mechanical laser radar includes the following steps:
[0009] Point cloud data obtained by receiving laser radar scanning;
[0010] Calculate the local change rate of each point cloud point, where the local change rate is an indicator of the depth change of the neighboring area of the point cloud point;
[0011] According to the local change rate of the point cloud point, the curvature value of each point cloud point is calculated;
[0012] The point cloud point with the largest curvature value is defined as an edge point, and the point cloud point with the smallest curvature value is defined as a plane point; based on the edge point and the plane point, the residual from the point to the feature line or feature surface is calculated by selecting the corresponding feature point;
[0013] Design a weight factor based on adaptive residual weight to dynamically adjust the weight distribution of point cloud points;
[0014] Apply the adjusted weight distribution factor to optimize the lidar position.
[0015] Preferably, the receiving of point cloud data acquired by mechanical laser radar scanning includes preprocessing the original point cloud data to eliminate noise and abnormal points.
[0016] Preferably, the local change rate is calculated by the following formula:
[0017]
[0018] Among them, V ij It's point p ij The local rate of change of ij and R mn Respectively represent point p ij and point p mn Depth measurement of It's point p ij A collection of neighboring points.
[0019] Preferably, the curvature value is calculated by the following formula:
[0020]
[0021] Among them, c ij It's point p ij The curvature value of W mn It's point p mn The weight factor of .
[0022] Preferably, the weight factor W mn Calculated by the following formula:
[0023] W mn =α·exp(-β·D mn -γ·ΔR mn )
[0024] Among them, α, β, γ are the adjustment of D mn and ΔR mn Weight coefficient; D mn It's point p mn With the center point p ij The spatial distance between.
[0025] Preferably, the residual from the calculation point to the characteristic line or characteristic surface comprises the following steps:
[0026] For edge points, the edge line is defined by selecting the two nearest points and calculating the residual from the point to the edge line;
[0027] For planar points, the incidence plane is defined by selecting three incidence points and the residuals from the points to the incidence plane are calculated.
[0028] Preferably, the residual calculation uses the following formula:
[0029] Edge point residual:
[0030]
[0031] Plane point residual:
[0032]
[0033] in, Represents the vector in the k+1th scanning cycle; and They respectively represent the corresponding vectors in the kth scanning cycle.
[0034] Preferably, the weight factor δ based on the adaptive residual weight is i Calculated by the following formula:
[0035]
[0036] Among them, T1 and T2 are set thresholds.
[0037] Preferably, the method for optimizing the laser radar position includes adjusting the posture of the laser radar to improve the matching accuracy of the point cloud data and the robustness of the algorithm.
[0038] Compared with the prior art, the advantages and beneficial effects of the present invention are:
[0039] The present invention improves the expression accuracy of point cloud data of distant objects by improving the curvature calculation method and solves the problem of sparsity of point cloud data.
[0040] The present invention achieves comprehensive feature extraction and avoids the problem of missing edge points by introducing local change rate and weight factor.
[0041] The present invention realizes dynamic adjustment of weights through an adaptive residual optimization weight distribution mechanism, thereby enhancing the robustness of the algorithm in complex environments.
[0042] The present invention improves the real-time processing capability by optimizing the algorithm structure and calculation method, thus meeting the needs of efficient data processing for automatic driving and autonomous mobile robots.
[0043] The present invention improves measurement accuracy and provides more reliable posture and map construction effects by refining feature point extraction and residual calculation. BRIEF DESCRIPTION OF THE DRAWINGS
[0044] Figure 1 Schematic diagram of dotted residuals in an embodiment of the present invention;
[0045] Figure 2 Schematic diagram of point-surface residual in an embodiment of the present invention. DETAILED DESCRIPTION
[0046] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.
[0047] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "comprise", "include", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.
[0048] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.
[0049] In the case of using expressions such as "at least one of A, B, and C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, and C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.). In the case of using expressions such as "at least one of A, B, or C, etc.", it should generally be interpreted in accordance with the meaning of the expression generally understood by those skilled in the art (for example, "a system having at least one of A, B, or C" should include but is not limited to a system having A alone, B alone, C alone, A and B, A and C, B and C, and / or A, B, C, etc.).
[0050] Whenever a new LiDAR scan data P is received k , first calculate the scan P k Every point p in k,i The degree of concave and convex on the harness is recorded as curvature c, and then sorted according to the size of c value to determine the point P k,i The feature type. Select the points with the maximum c value and the minimum c value as edge points E k and plane point H k .
[0051] In order to adapt to point cloud distributions with different sparsity and measurement distances, a new curvature calculation method is defined. First, in the initial stage of point cloud data processing, the local change rate of each point is calculated. The local change rate V ij is to quantify each point p ij An indicator of the variation in depth within its immediate vicinity. The meaning of this term is defined by the formula:
[0052]
[0053] In the formula, R ij and R mn Respectively represent point p ij and point p mn Depth measurement. is a two-dimensional matrix representing point p ij The depth information distribution of the neighboring point set, The set of elements in the neighboring points is counted and defined as:
[0054]
[0055] Where m and n represent the row and column indices of the neighboring points, respectively, and Δi and Δj define the vertical and horizontal selection ranges of the neighboring points, respectively.
[0056] The present invention proposes a weight factor W based on the relative position and geometric attributes between points. mn Calculation method to recalculate the curvature c value:
[0057]
[0058] In the formula, c ij Represents point p ij Numerical results of the curvature of W. mn is a weight factor that takes into account the difference in spatial distance and range values, and is calculated using the formula:
[0059] W mn =α·exp(-β·D mn -γ·ΔR mn )
[0060] Where D mn Represents point p mn With the center point p ij The spatial distance between mn It's point p mn With point p ij The depth value difference between α, β, and γ is used to adjust D mn and ΔR mn The weight coefficient is adjusted according to the characteristics of the point cloud data. In the present invention, the values are 0.70, 0.50, and 0.70 respectively.
[0061] On this basis, the threshold θ is set v , to determine the point p ij Is it in a complex and noisy area or a simple and stable area? ij The local rate of change V ij When the threshold is exceeded, the point p is considered ij The geometric structure of the area is complex or contains a high noise level. In this case, the point p is calculated by the following formula ij Therefore, the present invention defines a decision function F(i,j), based on the point p k,i Local rate of change V ij Select the appropriate curvature calculation method:
[0062]
[0063] Where c represents the curvature of each point in the scan, represents the linear distance from the kth time point to the LiDAR, Indicates the current scan data P k Returns a set of continuous points, threshold θ v It is set according to the severity of the scene depth change. It is set to 0.80 in outdoor environment and 0.65 in indoor environment.
[0064] Finally, according to point p k,i The curvature calculation result of the LiDAR point cloud scanning data P k Each point in is divided into edge points E k (edge point) and plane point H k (plane point).
[0065] 2. Feature matching and residual calculation
[0066] For edge feature point set By selecting the kth scan, with point P k The two closest points j and i define the corresponding edge line. Point p k,i The residual to the edge line L Calculated by the formula:
[0067]
[0068] In the formula, Indicates that in the k+1th scanning cycle, the LiDAR center is used as the starting point and the measurement point p is used as the k,i is the vector of the end point; similarly, and They represent the corresponding vectors in the kth scanning cycle, and the schematic diagram of the dotted residual is as follows Figure 1 shown.
[0069] Point i is the corner feature from the k+1th frame, and points j and i are the corner features from the kth frame. The dotted line is the dotted line residual
[0070] For a plane feature point set The correlation plane P is defined by selecting 3 points j, i and m in the kth scan. Point p k,i Residual of the plane Calculated by the formula:
[0071]
[0072] In the formula, Represents the vector in the k+1th scanning cycle; and They represent the corresponding vectors in the kth scanning cycle. The point-surface residual needs to find three related points in the k+1th scanning. The schematic diagram of the point-surface residual is as follows Figure 2 shown.
[0073] Point i is the surface feature point from the k+1th frame, and points j, i, and m are the surface feature points from the kth frame. The dotted line is the point-surface residual.
[0074] Next, for each point p k,i The residual of the calculation of an adaptive adjustment coefficient based on distance classification, whose value is affected by the point p k,i The distance r to the sensor affects the dynamic residual weight coefficient δ i The calculation method is:
[0075]
[0076] Where ZT1 and ZT2 are the set thresholds, which are set to 8.0 and 9.0 respectively. For r<30, For r>60, And when ratio_dis>3, limit it to 3.
[0077] Furthermore, based on the adaptive residual weight δ i The calculated result is applied to the adjustment point p k,i The feature weight calculation, weight redistribution result δ i The calculation formula is:
[0078]
[0079] In the formula, the weight s i It adapts to different measurement conditions and data quality in the form of piecewise functions, which increases its dominant position in the residual optimization process and directly affects the contribution of each point and the optimization effect.
[0080] With the point p k,i The closer to the LiDAR sensor, the greater the k i The value of increases accordingly, thereby increasing the weight of these high-quality data points. This means that as point p k,i The farther away from the LiDAR sensor, the i The value of will be reduced accordingly to reduce the impact of these low-quality data points. In order to make the weight difference concrete, the weight s is adjusted in the form of second-order weighting. i Further refinement to highlight its status:
[0081]
[0082] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0083] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0084] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0085] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0086] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0087] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.
[0088] Computer readable media include permanent and non-permanent, removable and non-removable media that can be used to store information by any method or technology. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media 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 disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in the present invention, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.
[0089] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.
[0090] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims
1. An optimization method based on uneven sampling of mechanical laser radar, characterized in that: The following steps are involved: Point cloud data obtained by receiving laser radar scanning; Calculate the local change rate of each point cloud point, where the local change rate is an indicator of the depth change of the neighboring area of the point cloud point; According to the local change rate of the point cloud point, the curvature value of each point cloud point is calculated; The point cloud point with the largest curvature value is defined as an edge point, and the point cloud point with the smallest curvature value is defined as a plane point; based on the edge point and the plane point, the residual from the point to the feature line or feature surface is calculated by selecting the corresponding feature point; Design a weight factor based on adaptive residual weight to dynamically adjust the weight distribution of point cloud points; Apply the adjusted weight distribution factor to optimize the lidar position.
2. The optimization method based on uneven sampling of mechanical laser radar according to claim 1, characterized in that: The receiving of point cloud data acquired by mechanical laser radar scanning includes preprocessing the original point cloud data to eliminate noise and abnormal points.
3. The optimization method based on uneven sampling of mechanical laser radar according to claim 1, characterized in that: The local rate of change is calculated by the following formula: Among them, V ij It's point p ij The local rate of change of ij and R mn Respectively represent point p ij and point p mn Depth measurement of It's point p ij A collection of neighboring points.
4. The optimization method based on uneven sampling of mechanical laser radar according to claim 1, characterized in that: The curvature value is calculated by the following formula: Among them, c ij It's point p ij The curvature value of W mn It's point p mn The weight factor of .
5. The optimization method based on non-uniform sampling of mechanical laser radar according to claim 4 is characterized in that: The weight factor W mm Calculated by the following formula: W mn =α·exp(-β·D mn -γ·△R mn ) Among them, α, β, and γ are the adjustment of D mn and ΔR mn Weight coefficient; D mn It's point p mn With the center point p ij The spatial distance between.
6. The optimization method based on non-uniform sampling of mechanical laser radar according to claim 1, characterized in that: The calculation of the residual from the point to the characteristic line or the characteristic surface comprises the following steps: For edge points, the edge line is defined by selecting the two nearest points and calculating the residual from the point to the edge line; For planar points, the incidence plane is defined by selecting three incidence points and the residuals from the points to the incidence plane are calculated.
7. The optimization method based on non-uniform sampling of mechanical laser radar according to claim 6 is characterized in that: The residual is calculated using the following formula: Edge point residual: Plane point residual: in, Represents the vector in the k+1th scanning cycle; and They respectively represent the corresponding vectors in the kth scanning cycle.
8. The optimization method based on non-uniform sampling of mechanical laser radar according to claim 1, characterized in that: The weight factor δ based on the adaptive residual weight i Calculated by the following formula: Among them, T1 and T2 are set thresholds.
9. The optimization method based on non-uniform sampling of mechanical laser radar according to claim 8, characterized in that: The method for optimizing the laser radar position includes adjusting the posture of the laser radar to improve the matching accuracy of the point cloud data and the robustness of the algorithm.
