Voxelization laser radar noise point filtering method based on signal intensity and distance
By using voxelization and dynamic weighting coefficients, noise in lidar is filtered out based on signal strength and distance, solving the problem of high noise misjudgment rate in existing technologies and achieving higher point cloud data quality and stability in subsequent processing.
Patent Information
- Application Number
- CN202511145230.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-15
- Publication Date
- 2025-11-18
AI Technical Summary
Existing noise removal methods based on local neighborhood statistics are prone to misclassifying valid points as noise in non-uniform point cloud distribution scenarios. The misclassification rate is particularly high when there is occlusion, long-distance observation, or significant changes in boundary structure, which affects the quality of point cloud data and the accuracy and stability of subsequent processing.
A voxelization filtering method based on signal strength and distance is adopted. By voxelizing the point cloud data, the mean and standard deviation of the distance value and signal strength value in each voxel grid are calculated. The weighting coefficient is dynamically adjusted in combination with the actual detection distance to construct a dynamic comprehensive value and set a threshold to remove noise.
It significantly reduces the false positive rate of noise, improves the accuracy of noise filtering, enhances the recognition stability and robustness of point cloud data in complex environments, and improves the accuracy and robustness of point cloud filtering.
Smart Images
Figure CN120976052A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of point cloud processing, and specifically to a voxelization method for filtering out noise in lidar based on signal strength and distance. Background Technology
[0002] The generation of noise in LiDAR point clouds involves multiple influencing factors, including the hardware characteristics of the laser and receiver, changes in external environmental conditions, the degree of electromagnetic interference, and the data processing mechanism during point cloud generation. Effective noise filtering is crucial in the point cloud preprocessing stage. It not only improves the overall quality of the raw data but also enhances the accuracy and stability of subsequent processing tasks such as target recognition and scene reconstruction. This significantly improves the interpretability and practical value of the data in complex environments, thereby enhancing the robustness and reliability of the entire system in real-world applications.
[0003] Statistical outlier removal is a noise filtering method based on the local neighborhood structure of point clouds. Its core idea is to construct a local neighborhood around each point, within a certain radius or based on a fixed number of neighboring points. By statistically analyzing the distance distribution characteristics from the center point in this neighborhood, such as calculating the average distance and standard deviation, it is determined whether the point significantly deviates from the overall local density distribution. If the average distance of a point's neighborhood is significantly greater than the statistical threshold of the average distance between adjacent points in the overall point cloud, the point is identified as an isolated point or outlier and removed. This effectively filters out invalid points introduced by local density anomalies or random noise, improving the overall quality and structural consistency of point cloud data.
[0004] Existing noise removal methods based on local neighborhood statistics rely on the stability of local point cloud density and the consistency of the distribution of distances between points. In scenarios with non-uniform point cloud distributions, effective points are easily misclassified as noise due to the naturally lower density of sparse local regions and edge structures, especially when there is occlusion, long-distance observation, or significant changes in boundary structures. Therefore, a method is needed to reduce the misclassification rate of noise in lidar point clouds, thereby improving the accuracy of noise filtering. Summary of the Invention
[0005] This invention provides a voxel-based method for filtering noise in lidar based on signal strength and distance, which can reduce the false positive rate of noise in lidar point clouds and thus improve the accuracy of noise filtering.
[0006] A first aspect of the present invention provides a voxel-based method for filtering noise in lidar based on signal strength and distance, the method comprising: Acquire point cloud data output by lidar; The point cloud data is voxelized, and the point cloud data points located in the same voxel grid are combined into a point cloud data set corresponding to the voxel. For each point cloud dataset, a differentiated result is calculated based on the distance and signal strength values of all point cloud data points in the point cloud dataset. Determine the weighting coefficients corresponding to the distance value and the signal strength value; The differential results are weighted and calculated according to the weighting coefficients to obtain the dynamic comprehensive value of each point cloud data point. If the dynamic composite value is determined to be greater than a preset threshold, the corresponding point cloud data point is identified as noise and removed.
[0007] Based on the above technical solutions, preferably, the voxelization processing of the point cloud data specifically includes: Based on the detection range of the lidar and the spatial distribution density of the point cloud data in the application scenario, a preset voxel size value matching the application scenario is determined, wherein the voxel size value is used to determine the spatial scale of each voxel grid in three-dimensional space.
[0008] Based on the above technical solutions, preferably, the voxelization processing of the point cloud data further includes: The average volume density is calculated based on the three-dimensional spatial range covered by the point cloud data and the number of point clouds. The voxel spatial volume is calculated based on the average volume density and the preset threshold for the number of point clouds within a single voxel grid. The voxel size value is determined based on the voxel space volume, where the voxel size value is the cube root of the voxel space volume.
[0009] Based on the above technical solutions, preferably, the step of forming a point cloud data set of corresponding voxels from point cloud data points located within the same voxel grid specifically includes: The three-dimensional space covered by the point cloud data is divided into a regular voxel grid based on the voxel size value, and each voxel grid has a unique spatial location identifier. Traverse all point cloud data points corresponding to the point cloud data points, determine the voxel grid to which the point cloud data points belong based on the three-dimensional coordinates of the point cloud data points, and classify all point cloud data points falling into the same voxel grid into the point cloud data set corresponding to the voxel grid. The point cloud data is organized into multiple point cloud data sets according to spatial location.
[0010] Based on the above technical solutions, preferably, the step of calculating the differentiated results for each point cloud data set based on the distance values and signal strength values of all point cloud data points in the point cloud data set specifically includes: For each point cloud data set corresponding to the voxel grid, the distance value and signal strength value of all point cloud data points in the point cloud data set are statistically analyzed. For each voxel grid, the mean distance value and the standard deviation of the corresponding distance value are calculated based on all distance values contained in the corresponding point cloud dataset. For each voxel grid, the mean signal strength value and the corresponding standard deviation of the signal strength value are calculated based on all distance values and signal strength values contained in the corresponding point cloud dataset. For any point cloud data point in the point cloud dataset, the distance difference result of the point cloud data point is calculated based on the distance value, the mean distance value, and the standard deviation of the distance value. The signal strength difference result of the point cloud data point is calculated based on the signal strength value, the mean signal strength value, and the standard deviation of the signal strength value.
[0011] Based on the above technical solutions, preferably, determining the weighting coefficient corresponding to the distance value and the signal strength value specifically includes: For each point cloud data point, determine the actual detection distance corresponding to that point cloud data point; For each point cloud data point, the target distance interval to which it belongs is determined in the preset distance and weighting coefficient mapping table according to the actual detection distance. The determination of the weighting coefficients corresponding to the distance value and the signal strength value divides the detection distance range into multiple detection distance intervals. Each detection distance interval corresponds to a set of distance difference weighting coefficients and signal strength difference weighting coefficients. Extract the distance difference weighting coefficient and signal strength difference weighting coefficient corresponding to the target distance interval.
[0012] Based on the above technical solutions, preferably, determining the weighting coefficients corresponding to the distance value and the signal strength value further includes: For each point cloud data point, determine the actual detection distance corresponding to that point cloud data point; Based on the actual detection distance, the distance measurement error estimate and the signal strength attenuation error estimate are calculated respectively. The distance measurement error estimate is established based on the logarithmic function of the actual detection distance, and the signal strength attenuation error estimate is established based on the exponential attenuation function of the actual detection distance. Based on the inverse relationship between the distance measurement error estimate and the signal strength attenuation error estimate, a distance differentiation weighting coefficient and a signal strength differentiation weighting coefficient are calculated respectively, wherein the distance differentiation weighting coefficient is calculated based on the distance measurement error estimate, and the signal strength differentiation weighting coefficient is calculated based on the signal strength attenuation error estimate.
[0013] In a second aspect of the invention, a voxel-based noise reduction device for lidar based on signal strength and distance is provided. The device is used to perform a voxel-based noise reduction method for lidar based on signal strength and distance as described in any of the preceding embodiments. The device includes an acquisition module, a processing module, and an output module, wherein: The acquisition module is used to acquire point cloud data output by the lidar; The processing module is used to generate point cloud data sets corresponding to voxels; The processing module is used to calculate differentiated results for each point cloud data set based on the distance values and signal strength values of all point cloud data points in the point cloud data set. The processing module is used to determine the weighting coefficients corresponding to the distance value and the signal strength value; The processing module is used to perform a weighted calculation on the differential result and the weighting coefficient to obtain a dynamic comprehensive value for each point cloud data point; The output module is used to identify the corresponding point cloud data points as noise and remove them if it determines that the dynamic comprehensive value is greater than a preset threshold.
[0014] Based on the above technical solutions, preferably, the processing module is used to determine a preset voxel size value that matches the application scenario based on the detection range of the lidar and the spatial distribution density of the point cloud data in the application scenario, wherein the voxel size value is used to determine the spatial scale of each voxel grid in three-dimensional space.
[0015] Based on the above technical solutions, preferably, the processing module is used to calculate the average volume density according to the three-dimensional spatial range covered by the point cloud data and the number of point clouds. The processing module is used to calculate the voxel spatial volume based on the average volume density and a preset threshold for the number of point clouds within a single voxel grid. The processing module is used to determine the voxel size value based on the voxel space volume, wherein the voxel size value is the cube root of the voxel space volume.
[0016] Based on the above technical solutions, preferably, the processing module is used to divide the three-dimensional space covered by the point cloud data into a regular voxel grid according to the voxel size value, and each voxel grid has a unique spatial location identifier. The acquisition module is used to traverse all point cloud data points corresponding to the point cloud data, determine the voxel grid to which the point cloud data points belong based on the three-dimensional coordinates of the point cloud data points, and classify all point cloud data points falling into the same voxel grid into the point cloud data set corresponding to the voxel grid. The processing module is used to organize the point cloud data into multiple point cloud data sets according to their spatial location.
[0017] Based on the above technical solutions, preferably, the processing module is used to calculate the distance value and signal strength value of all point cloud data points in the point cloud data set for each point cloud data set corresponding to the voxel grid. The processing module is used to calculate the mean distance value and the standard deviation of the corresponding distance value for each voxel grid based on all distance values contained in the corresponding point cloud data set. The processing module is used to calculate the mean signal strength value and the corresponding standard deviation of the signal strength value for each voxel grid based on all distance signal strength values contained in the corresponding point cloud data set. The processing module is used to calculate the distance difference result of any point cloud data point in the point cloud data set based on the distance value, the mean distance value, and the standard deviation of the distance value of the point cloud data point, and to calculate the intensity difference result of the point cloud data point based on the signal strength value, the mean signal strength value, and the standard deviation of the signal strength value of the point cloud data point.
[0018] Based on the above technical solutions, preferably, the processing module is used to determine the actual detection distance corresponding to each point cloud data point; The processing module is used to determine the target distance interval to which each point cloud data point belongs based on the actual detection distance in a preset distance and weighting coefficient mapping table. The determination of the weighting coefficients corresponding to the distance value and the signal strength value divides the detection distance range into multiple detection distance intervals. Each detection distance interval corresponds to a set of distance difference weighting coefficients and signal strength difference weighting coefficients. The processing module is used to extract the distance difference weighting coefficient and the signal strength difference weighting coefficient corresponding to the target distance interval.
[0019] Based on the above technical solutions, preferably, the processing module is used to determine the actual detection distance corresponding to each point cloud data point; The processing module is used to calculate a distance measurement error estimate and a signal strength attenuation error estimate based on the actual detection distance. The distance measurement error estimate is established based on the logarithmic function of the actual detection distance, and the signal strength attenuation error estimate is established based on the exponential attenuation function of the actual detection distance. The processing module is used to calculate the distance difference weighting coefficient and the signal strength difference weighting coefficient respectively based on the inverse relationship between the distance measurement error estimate and the signal strength attenuation error estimate, wherein the distance difference weighting coefficient is calculated based on the distance measurement error estimate and the signal strength difference weighting coefficient is calculated based on the signal strength attenuation error estimate.
[0020] In a third aspect of the invention, an electronic device is provided, including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the preceding embodiments.
[0021] In a fourth aspect, the present invention provides a computer-readable storage medium storing instructions that, when executed, perform the method as described in any of the preceding claims.
[0022] In summary, one or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages: 1. This invention spatially reorganizes point cloud data by introducing a voxelized structure, enabling the anomaly assessment of each point cloud data point to be based on local statistical features rather than global distribution, effectively improving the representativeness of the difference expression; it constructs differentiated results by combining two key parameters, distance value and signal strength value, and dynamically adjusts the weighting coefficients according to the actual detection distance to achieve adaptive response to different dominant factors of measurement error; finally, it completes noise removal by comparing the weighted comprehensive value with a unified threshold, which not only suppresses misjudgment of edges and sparse regions, but also enhances the discrimination stability in complex environments, thereby significantly reducing the noise misjudgment rate and improving the accuracy of point cloud filtering.
[0023] 2. By jointly modeling the coverage area and number of point cloud data, calculating the average volume density and inversely calculating the volume of voxel space, the voxel size value is adaptively determined based on the threshold of the number of points within the target voxel, realizing the dynamic matching between voxel scale and point cloud distribution density, effectively improving the stability and filtering robustness of subsequent voxel statistics.
[0024] 3. Within each voxel grid, the mean and standard deviation of the distance and signal strength values are statistically analyzed, and two types of standardized deviation values are calculated for each point cloud data point. This achieves a two-dimensional measurement of point cloud spatial location anomalies and reflection characteristic anomalies, significantly improving the ability to identify local abnormal structures and avoiding misjudgments caused by relying on only a single feature.
[0025] 4. A continuous function is used to construct distance measurement error estimation and signal strength attenuation error estimation, and the weighting coefficient is calculated by inversely normalizing the two. This realizes an adaptive weighting strategy that combines high precision, continuity and device parameter adjustability, making the weighting mechanism have stronger physical interpretation and cross-device generalization ability, thereby improving the recognition accuracy and robustness in complex distance variation scenarios. Attached Figure Description
[0026] Figure 1 This is a schematic flowchart of a method for filtering out lidar noise based on signal strength and distance, as disclosed in an embodiment of the present invention. Figure 2 This is a schematic diagram of a voxelized lidar noise removal device based on signal strength and distance, as disclosed in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention.
[0027] Explanation of reference numerals in the attached drawings: 201, acquisition module; 202, processing module; 203, output module; 301, processor; 302, communication bus; 303, user interface; 304, network interface; 305, memory. Detailed Implementation
[0028] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.
[0029] In the description of the embodiments of the present invention, words such as "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "for example" or "for instance" in the embodiments of the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Rather, the use of words such as "for example" or "for instance" is intended to present the relevant concepts in a specific manner.
[0030] In the description of the embodiments of the present invention, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.
[0031] The generation of noise in lidar point clouds is influenced by multiple factors, including the hardware characteristics of the laser and receiver, environmental changes, electromagnetic interference, and data processing mechanisms. Effective noise removal is crucial for improving point cloud data quality and subsequent processing accuracy. Existing statistical outlier removal methods identify noise based on local neighborhood distance distribution characteristics. While these methods can effectively remove local outliers, they are prone to misjudgment in non-uniformly distributed or structural edge regions due to low local density, especially under conditions of occlusion, long-distance observation, or drastic boundary changes. Therefore, there is an urgent need to propose a point cloud noise identification and removal method that can reduce the misjudgment rate and improve filtering accuracy.
[0032] This embodiment discloses a voxel-based method for filtering noise in lidar based on signal strength and distance, referring to... Figure 1 This includes the following steps S110-S160: S110 acquires point cloud data output by the lidar.
[0033] This invention discloses a voxel-based noise reduction method for lidar based on signal strength and distance, which is applied to a server. The server includes, but is not limited to, electronic devices such as mobile phones, tablets, wearable devices, and PCs (Personal Computers), and can also be a backend server running the voxel-based noise reduction method for lidar based on signal strength and distance. The server can be implemented using a standalone server or a server cluster consisting of multiple servers.
[0034] After the lidar completes a full scanning cycle, it receives the output point cloud data stream in real time through a high-frequency communication channel. It analyzes the spatial location parameters, distance values, and signal strength values of each point cloud data point in the data stream and constructs a complete data frame structure based on the frame synchronization identifier. For continuously received data frames, it performs ordered caching and anomaly compensation mechanisms based on timestamps and echo order to remove duplicate and distorted frames, ensuring the continuity and spatial integrity of the resulting original point cloud dataset. The generated original point cloud dataset retains the original observation accuracy of the lidar and serves as the starting input for subsequent voxel mesh generation, point cloud dataset construction, and dynamic comprehensive value calculation. This data processing mainline runs through the entire noise removal process, ensuring that all operations in the processing chain are based on a unified accuracy standard and coordinate system.
[0035] S120 performs voxelization on the point cloud data, forming a set of point cloud data points located in the same voxel grid into corresponding voxel set.
[0036] In one possible implementation, the point cloud data is voxelized, specifically including: determining a preset voxel size value that matches the application scenario based on the detection range of the lidar in the application scenario and the spatial distribution density of the point cloud data, wherein the voxel size value is used to determine the spatial scale of each voxel grid in three-dimensional space.
[0037] Specifically, by first analyzing the lidar configuration file and real-time observation parameters, the maximum detection range parameters of the lidar under the current application scenario are obtained. The detection range is usually composed of the horizontal field of view, the vertical field of view, and the maximum distance to form a three-dimensional observation volume. Combined with the time span of point cloud data, a complete spatial coverage model can be constructed.
[0038] Simultaneously, statistical analysis is performed on the raw point cloud data received within the current period. The boundary range and total number of point clouds along the X, Y, and Z directions in three-dimensional space are extracted. Based on this, the point cloud density per unit volume is calculated. This point cloud density reflects the spatial density of the point cloud in the current acquisition scenario and is the core basis for judging the accuracy of subsequent voxel meshing. On this basis, according to the control target of point cloud density and the expected number of point clouds contained in the voxel, a preset voxel size value suitable for this application scenario is determined. The voxel size value is defined as the side length of the cube used to divide the three-dimensional space, which determines the spatial scale of each voxel mesh in the three-dimensional coordinate system and directly affects the scale and statistical representativeness of the point cloud data set in the voxel mesh. Taking autonomous driving as an example, if the original point cloud distribution density is 1500 points per cubic meter, and it is expected that each voxel grid contains about 10 to 20 points to support subsequent stable statistical analysis, then the calculated voxel side length is about 0.2 meters. The final generated voxel size value should meet the spatial resolution while ensuring that the number of points in each voxel grid is sufficient to support differential calculation and anomaly identification, avoiding the loss of target edge features due to excessively coarse voxel granularity, or the insufficient number of points in the voxel due to excessively fine granularity, which would affect the discrimination stability.
[0039] In one possible implementation, the point cloud data is voxelized, which further includes: calculating the average volume density based on the three-dimensional spatial range covered by the point cloud data and the number of points; calculating the voxel space volume based on the average volume density and a preset threshold for the number of points in a single voxel grid; and determining the voxel size value based on the voxel space volume, wherein the voxel size value is the cube root of the voxel space volume.
[0040] Specifically, first, all point cloud data points within the current period are traversed, and their X, Y, and Z 3D coordinate extreme values are extracted. The difference between the maximum and minimum values is calculated to obtain the coverage area of the point cloud data in each spatial dimension, which is used to construct the side length of the overall point cloud bounding box. Let the length in the X direction be denoted as . The length in the Y direction is The length in the Z direction is The volume of the three-dimensional space covered by the point cloud data is
[0041] in, Let be the overall spatial volume of the original point cloud. Further, count the number of point cloud data points in the original point cloud dataset, denoted as . Combine this with the total volume to calculate the average bulk density, denoted as:
[0042] in, It represents the number of point cloud data points per unit space, reflecting the average density of the point cloud within the observation area, and has statistical significance for physical spatial density.
[0043] Based on average volume density And the set threshold for the number of point clouds within the voxel mesh. Then, we can deduce the average spatial volume that each voxel mesh should possess. The formula for calculating this voxel spatial volume is:
[0044] in, This is the minimum number of points required within a single voxel to ensure sufficient sample support when performing statistical analysis at the voxel grid granularity. For example, if each voxel is expected to contain 20 points and the current average volume density is 1000 points per cubic meter, then the calculated volume of each voxel grid should be 0.02 cubic meters.
[0045] Finally, to construct a cubic voxel mesh structure with uniform side lengths, the aforementioned voxel space volume is converted into voxel size values, i.e.:
[0046] in, The voxel size value, i.e., the side length of the cubic mesh used for spatial partitioning, determines the resolution of subsequent voxel partitioning. The above formula realizes the function of dynamically and adaptively generating voxel scale based on statistical density and processing target, ensuring that the optimal granularity voxel structure can be constructed under the premise of ensuring computational stability, regardless of how the spatial distribution of point cloud data changes.
[0047] In one possible implementation, point cloud data points located within the same voxel grid are grouped into point cloud data sets corresponding to the voxels. Specifically, this includes: dividing the three-dimensional space covered by the point cloud data into regularly structured voxel grids based on the voxel size value, with each voxel grid having a unique spatial location identifier; traversing all point cloud data points corresponding to all point cloud data points, determining the voxel grid to which the point cloud data points belong based on their three-dimensional coordinates, and grouping all point cloud data points falling into the same voxel grid into the point cloud data set corresponding to the voxel grid; and organizing the point cloud data into multiple point cloud data sets according to their spatial location.
[0048] Specifically, based on the determined voxel size values, the point cloud data coverage space is divided into equal-interval segments along the X, Y, and Z coordinate directions, using each voxel size value as a partitioning unit, forming a set of voxel meshes with a regular structure. During the partitioning process, a three-dimensional integer coordinate index system is established, mapping each voxel mesh to a unique spatial location identifier represented by a combination of its index numbers along each axis direction, denoted as an index triplet (i, j, k), where i, j, and k represent the relative numbers of the current voxel in the X, Y, and Z directions, respectively. This spatial location identifier is used to identify the precise position of the voxel mesh within the overall partitioning structure and supports subsequent retrieval and positioning operations.
[0049] For each point in the original point cloud data within the current period, a coordinate assignment operation is performed. The 3D coordinate value of each point is extracted, divided by the voxel size and rounded down to obtain the index position (i, j, k) of the point in the 3D voxel mesh structure. Based on this, it is determined which voxel mesh the point should belong to. During the traversal, all point points with the same calculated index triplet are uniformly assigned to the corresponding voxel mesh, and a point cloud dataset is constructed for that voxel mesh as the object for subsequent processing.
[0050] After all point cloud data points are assigned to specific locations, multiple point cloud datasets corresponding to voxel grids are obtained. Point cloud data points within each dataset exhibit spatial proximity and structural consistency, forming statistically representative local spatial data units. This organization method preserves the spatial structure information of the point cloud while providing a structural foundation for subsequent differential analysis, weighted discrimination, and noise identification at the voxel granularity. For example, when the voxel size is 0.1 meters, and the coordinates of a data point in the original point cloud are (3.14, 1.87, 2.03), its voxel index position is (31, 18, 20). This point cloud data point is assigned to the point cloud dataset corresponding to the voxel grid with index (31, 18, 20).
[0051] S130: For each point cloud dataset, calculate the differential results based on the distance values and signal strength values of all point cloud data points in the dataset.
[0052] In one possible implementation, for each point cloud dataset, a differentiation result is calculated based on the distance and signal strength values of all point cloud data points in the dataset. Specifically, this includes: for each voxel grid corresponding to a point cloud dataset, calculating the distance and signal strength values of all point cloud data points in the dataset; for each voxel grid, calculating the mean distance value and the corresponding standard deviation of the distance value based on all distance values contained in the corresponding point cloud dataset; for each voxel grid, calculating the mean signal strength value and the corresponding standard deviation of the signal strength value based on all distance and signal strength values contained in the corresponding point cloud dataset; for any point cloud data point in the dataset, calculating the distance differentiation result of the point cloud data point based on its distance value, mean distance value, and standard deviation of the distance value, and calculating the intensity differentiation result of the point cloud data point based on its signal strength value, mean signal strength value, and standard deviation of the signal strength value.
[0053] Specifically, firstly, a statistical operation is performed on the point cloud data set corresponding to each voxel grid. The distance and signal strength values of all point cloud data points are extracted from the set, and distance value sequences and signal strength value sequences are constructed respectively, denoted as . and ,in This represents the number of point cloud data points within the voxel mesh. and They represent the first The distance and signal strength values of each point cloud data point. The distance value is the straight-line distance between the point cloud data point measured by the lidar and the laser emission source. The signal strength value is the intensity amplitude of the reflected echo from that point cloud data point, reflecting the surface material, angle, and reflection stability.
[0054] After constructing the above numerical sequence, the mean and standard deviation of the distance values for the voxel grid are calculated based on the distance value sequence. The mean distance value represents the average distribution of point cloud data points within the voxel at the detection depth, and the calculation formula is as follows:
[0055] in, The mean distance value is denoted as . The standard deviation represents the dispersion of each point relative to the mean, reflecting the consistency of the local spatial depth distribution. The calculation formula is:
[0056] in, This represents the standard deviation of the distance values.
[0057] Similarly, the mean and standard deviation of signal intensity values within the voxel grid are calculated based on the signal intensity value sequence. Mean signal intensity value The average level of the overall surface reflectivity of this local area is represented by the following formula:
[0058] Standard deviation of signal strength value The formula used to measure the stability of local reflection intensity is as follows:
[0059] After obtaining the four statistical parameters of the voxel grid, for any specific point cloud data point within it, based on the distance and signal strength values of that point, and combined with the corresponding mean and standard deviation, the degree of difference in the two feature dimensions is calculated. The formula for calculating the distance difference result is as follows:
[0060] The formula for calculating the signal strength difference result is:
[0061] in, and They represent the first The standardized deviation values of each point cloud data point in the distance and signal strength dimensions, i.e., the differential results, represent the degree of anomaly of the point relative to the central trend within its voxel grid. The above absolute value processing ensures that the deviation direction has no impact on subsequent weight calculations, and only the deviation magnitude is retained for comprehensive anomaly evaluation.
[0062] S140, determine the weighting coefficients corresponding to the distance value and the signal strength value.
[0063] In one possible implementation, determining the weighting coefficients corresponding to the distance value and the signal strength value specifically includes: for each point cloud data point, determining the actual detection distance corresponding to the point cloud data point; for each point cloud data point, determining the target distance interval to which it belongs based on the actual detection distance in a preset distance-weighting coefficient mapping table, determining the weighting coefficients corresponding to the distance value and the signal strength value to divide the detection distance range into multiple detection distance intervals, each detection distance interval corresponding to a set of distance-differentiated weighting coefficients and signal strength-differentiated weighting coefficients; and extracting the distance-differentiated weighting coefficients and signal strength-differentiated weighting coefficients corresponding to the target distance interval.
[0064] Specifically, for each point cloud data point, its distance value in the lidar coordinate system is extracted. This distance value is the spatial path length formed by the laser beam propagating from the source to the target surface and back to the receiver. It is usually accurately measured by the radar equipment based on the time of flight or phase difference principle. It is the core parameter characterizing the spatial position of the point cloud data point and the distance relationship between the source and the source. This actual detection distance not only determines the distribution of the point cloud in three-dimensional space, but also directly affects the intensity of the reflected signal, the stability of the echo and the measurement error. Therefore, it is the control variable for the subsequent weighting coefficient calculation.
[0065] A pre-defined distance-weighting coefficient mapping table is constructed. Based on the characteristics of lidar ranging error and the echo intensity attenuation model, the entire potential detection range is divided into several continuous target distance intervals. Each target distance interval is assigned a corresponding set of weighting coefficients, including distance-differential weighting coefficients and signal intensity-differential weighting coefficients. This set of weighting coefficients satisfies the condition that the sum of the two is 1, and exhibits a trend where the distance-differential weighting coefficient decreases while the signal intensity-differential weighting coefficient increases with increasing distance. This adapts to the physical law that ranging errors increase and echo intensity is relatively more stable at long distances. For example, the effective detection range from 0 to 300 meters can be divided into 6 intervals: [0,50], (50,100], (100,150], (150,200], (200,250], (250,+∞), and weighted pairs such as (0.9,0.1), (0.8,0.2), and (0.7,0.3) can be set to ensure that the weighting strategy has consistency in physical interpretation and continuity in function transition in different distance segments.
[0066] For any specific point cloud data point, after obtaining its actual detection distance, the target distance interval to which it belongs is located by referring to the aforementioned mapping table. Boundary comparison operations are used to determine which interval range its distance value falls within. Once the target distance interval is determined, the distance differentiation weighting coefficient and signal strength differentiation weighting coefficient can be extracted from the corresponding weighting coefficient pair. These are then used to weight and superimpose the distance differentiation result and signal strength differentiation result of the point cloud data point to generate a dynamic composite value. For example, if the detection distance of a point is 137 meters, it falls within the (100, 150] interval, corresponding to a distance differentiation weighting coefficient of 0.7 and a signal strength differentiation weighting coefficient of 0.3. This set of parameters is used in the weighted fusion calculation process for that point. The above processing implements a segmented continuous weighting adjustment strategy through a table-driven approach, making the weighting calculation highly adaptive and easily optimized systematically through empirical parameter tuning or equipment calibration.
[0067] In one possible implementation, determining the weighting coefficients corresponding to the distance value and the signal strength value further includes: for each point cloud data point, determining the actual detection distance corresponding to the point cloud data point; calculating the distance measurement error estimate and the signal strength attenuation error estimate based on the actual detection distance, wherein the distance measurement error estimate is established based on the logarithmic function of the actual detection distance, and the signal strength attenuation error estimate is established based on the exponential attenuation function of the actual detection distance; and calculating the distance differentiation weighting coefficient and the signal strength differentiation weighting coefficient based on the inverse relationship between the distance measurement error estimate and the signal strength attenuation error estimate, wherein the distance differentiation weighting coefficient is calculated based on the distance measurement error estimate, and the signal strength differentiation weighting coefficient is calculated based on the signal strength attenuation error estimate.
[0068] Specifically, the distance value of each point cloud data point is extracted from the point cloud data as the actual detection distance for that point cloud data point. The actual detection distance reflects the propagation path length of the laser beam and directly affects the ranging accuracy and signal attenuation level, serving as the core variable for subsequent error modeling and weighted calculation. Two error estimation functions are constructed for this actual detection distance: On the one hand, a distance measurement error estimation function is constructed to characterize the ranging uncertainty caused by energy attenuation and decreased receiving time resolution of the lidar at long distances. A logarithmic function model is used to reflect the trend of the error slowly increasing with distance. This error estimate is denoted as:
[0069] in, The actual detection distance of the point cloud data points. This is the calibration coefficient, used to control the rate of error growth.
[0070] On the other hand, a signal strength attenuation error estimation function is constructed. Considering the influence of energy diffusion and medium absorption on the laser echo signal during propagation, its effective reflected signal will attenuate exponentially with distance. Therefore, a negative exponential function is used for modeling, and the error estimate is denoted as:
[0071] in, The initial attenuation factor of the signal. It is the exponential decay rate coefficient, used to characterize the rate at which signal strength decays with increasing distance.
[0072] Using the two error estimates above as relative weights, the distance difference weighting coefficient and the signal strength difference weighting coefficient are calculated according to the inverse proportionality principle. A normalization method is used to ensure that their sum is always 1. The calculation formula is as follows:
[0073] in, This is a distance-weighted coefficient for differences. These are the signal strength difference weighting coefficients, and both reflect the relative influence of the corresponding error-dominant factors on the comprehensive value at different detection distances.
[0074] S150 combines the differentiated results with a weighted calculation based on weighting coefficients to obtain the dynamic comprehensive value for each point cloud data point.
[0075] After calculating the distance and signal strength differential results for each point cloud data point and determining the corresponding distance and signal strength differential weighting coefficients, a weighted superposition operation is performed to generate the dynamic composite value for that point cloud data point. The dynamic composite value measures the overall anomaly degree of the point cloud data point within its voxel grid, integrating deviation information from both detection distance and reflection characteristics. It is a core evaluation metric for subsequent noise assessment.
[0076] Let the distance differential result of a certain point cloud data point be... The signal strength difference results are The corresponding distance difference weighting coefficient is The signal strength difference weighting coefficient is The formula for calculating its dynamic comprehensive value is:
[0077] in, This is a dynamic composite value, representing the weighted deviation of the point cloud data point under both distance and signal strength dimensions; and All deviations are normalized to standard deviation and are dimensionless. The form guarantees a positive value and reflects the degree of outlier. and Satisfy the normalization constraint, i.e. The values are derived from a preset mapping table or a dynamic error estimation function, ensuring that the comprehensive values have consistent dimensions and comparability.
[0078] S160, if it is determined that the dynamic comprehensive value is greater than the preset threshold, the corresponding point cloud data point is identified as noise and removed.
[0079] After calculating the dynamic composite value for each point cloud data point, a set of dynamic composite thresholds obtained through LiDAR equipment calibration is set as the decision boundary for noise identification. These preset thresholds are critical values determined based on the statistical characteristics of the distribution of valid points and noise points in the dynamic composite value from a large amount of experimental data. Their physical meaning lies in defining the upper limit of the severity of point cloud data points deviating from their normal distribution range under given detection accuracy and echo characteristics. Exceeding this threshold usually indicates a significant anomaly in the distance or signal strength dimension of that point.
[0080] The dynamic composite value of each point cloud data point is compared with a preset threshold. When the dynamic composite value is greater than the preset threshold, it indicates that the point deviates significantly from the overall distribution of its voxel grid. Its spatial position or signal reflection characteristics are inconsistent with the features of its neighborhood. It is very likely that it is a non-real reflection point caused by abnormal surface reflection, occlusion edges, environmental interference or system errors. Therefore, the point is judged as a noise point. Such point cloud data points are usually isolated, scattered and unstructured. Retaining them will lead to spatial modeling distortion or target recognition error. Therefore, they need to be removed from the point cloud data structure.
[0081] The removal operation can be implemented using logical masks or index markers. That is, a status marker is assigned to each point cloud data point; if its dynamic synthesis value exceeds a threshold, a removal marker is assigned, and it is excluded from the result set when generating the final processed point cloud. For example, if the dynamic synthesis threshold is set to 1.5, then all point cloud data points with a dynamic synthesis value greater than 1.5, such as a point cloud data point with a dynamic synthesis value of 1.82, are considered noise and removed from the point cloud data set. This processing mechanism ensures that local anomalies can be identified based on a unified standard in voxel meshes of different densities and reflection characteristics, thereby improving the coherence and geometric consistency of the overall point cloud structure.
[0082] This embodiment also discloses a voxel-based noise filtering device for lidar based on signal strength and distance, referring to... Figure 2 The device includes an acquisition module 201, a processing module 202, and an output module 203. It is used to execute any of the above-described voxel-based noise reduction methods for lidar based on signal strength and distance, wherein: The acquisition module 201 is used to acquire point cloud data output by the lidar.
[0083] Processing module 202 is used to generate point cloud data sets corresponding to voxels.
[0084] The processing module 202 is used to calculate differentiated results for each point cloud dataset based on the distance values and signal strength values of all point cloud data points in the point cloud dataset.
[0085] Processing module 202 is used to determine the weighting coefficients corresponding to the distance value and the signal strength value.
[0086] The processing module 202 is used to perform a weighted calculation on the differential results according to the weighting coefficients to obtain the dynamic comprehensive value of each point cloud data point.
[0087] The output module 203 is used to identify the corresponding point cloud data points as noise and remove them if the dynamic comprehensive value is determined to be greater than a preset threshold.
[0088] In one possible implementation, the processing module 202 is used to determine a preset voxel size value that matches the application scenario based on the detection range of the lidar and the spatial distribution density of the point cloud data in the application scenario, wherein the voxel size value is used to determine the spatial scale of each voxel grid in three-dimensional space.
[0089] In one possible implementation, the processing module 202 is used to calculate the average volume density based on the three-dimensional spatial range covered by the point cloud data and the number of point clouds.
[0090] The processing module 202 is used to calculate the voxel spatial volume based on the average volume density and a preset threshold for the number of point clouds in a single voxel grid.
[0091] Processing module 202 is used to determine the voxel size value based on the voxel space volume, where the voxel size value is the cube root of the voxel space volume.
[0092] In one possible implementation, the processing module 202 is used to divide the three-dimensional space covered by the point cloud data into a regular voxel grid according to the voxel size value, and each voxel grid has a unique spatial location identifier.
[0093] The acquisition module 201 is used to traverse all point cloud data points corresponding to the point cloud data, determine the voxel grid to which the point cloud data points belong based on the three-dimensional coordinates of the point cloud data points, and classify all point cloud data points falling into the same voxel grid into the point cloud data set corresponding to the voxel grid.
[0094] The processing module 202 is used to organize point cloud data into multiple point cloud data sets according to spatial location.
[0095] In one possible implementation, the processing module 202 is used to calculate the distance value and signal strength value of all point cloud data points in the point cloud data set for each voxel grid.
[0096] The processing module 202 is used to calculate the mean distance value and the standard deviation of the corresponding distance value for each voxel grid based on all distance values contained in the corresponding point cloud data set.
[0097] The processing module 202 is used to calculate the mean signal strength value and the corresponding standard deviation of the signal strength value for each voxel grid based on all distance values and signal strength values contained in the corresponding point cloud data set.
[0098] The processing module 202 is used to calculate the distance difference result of any point cloud data point in the point cloud data set based on the distance value, the mean distance value, and the standard deviation of the distance value, and to calculate the intensity difference result of the point cloud data point based on the signal strength value, the mean signal strength value, and the standard deviation of the signal strength value.
[0099] In one possible implementation, the processing module 202 is used to determine the actual detection distance corresponding to each point cloud data point.
[0100] The processing module 202 is used to determine the target distance interval to which each point cloud data point belongs based on the actual detection distance in a preset distance and weighting coefficient mapping table, determine the weighting coefficients corresponding to the distance value and the signal strength value, divide the detection distance range into multiple detection distance intervals, and each detection distance interval corresponds to a set of distance difference weighting coefficients and signal strength difference weighting coefficients.
[0101] The processing module 202 is used to extract the distance difference weighting coefficient and signal strength difference weighting coefficient corresponding to the target distance interval.
[0102] In one possible implementation, the processing module 202 is used to determine the actual detection distance corresponding to each point cloud data point.
[0103] The processing module 202 is used to calculate the distance measurement error estimate and the signal strength attenuation error estimate based on the actual detection distance. The distance measurement error estimate is established based on the logarithmic function of the actual detection distance, and the signal strength attenuation error estimate is established based on the exponential attenuation function of the actual detection distance.
[0104] The processing module 202 is used to calculate the distance difference weighting coefficient and the signal strength difference weighting coefficient respectively based on the inverse relationship between the distance measurement error estimate and the signal strength attenuation error estimate. The distance difference weighting coefficient is calculated based on the distance measurement error estimate, and the signal strength difference weighting coefficient is calculated based on the signal strength attenuation error estimate.
[0105] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.
[0106] This embodiment also discloses an electronic device, as shown in the reference. Figure 3 The electronic device may include: at least one processor 301, at least one communication bus 302, user interface 303, network interface 304, and at least one memory 305.
[0107] The communication bus 302 is used to enable communication between these components.
[0108] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.
[0109] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).
[0110] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications. The GPU is responsible for rendering and drawing the content required for display. The modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.
[0111] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory may include a non-transitory computer-readable storage medium. The memory 305 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the various method embodiments described above, etc. The data storage area may store data involved in the various method embodiments described above. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. As a computer storage medium, the memory 305 may include an operating system, a network communication module, a user interface 303 module, and an application program for a voxelization noise filtering method for lidar based on signal strength and distance.
[0112] exist Figure 3 In the illustrated electronic device, the user interface 303 is primarily used to provide an input interface for the user and to acquire user input data. The processor 301 can be used to call an application program stored in the memory 305 that uses a voxelization method for filtering lidar noise based on signal strength and distance. When executed by one or more processors 301, this causes the electronic device to perform one or more methods as described in the above embodiments.
[0113] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, as some steps can be performed in other orders or simultaneously according to the present invention. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the present invention.
[0114] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0115] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.
[0116] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0117] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.
[0118] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 305 and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned memory 305 includes various media capable of storing program code, such as a USB flash drive, external hard drive, magnetic disk, or optical disk.
[0119] The present invention also discloses a computer-readable storage medium storing instructions. When executed by one or more processors 301, these instructions cause an electronic device to perform one or more methods as described in the above embodiments.
[0120] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the specification and the disclosure of practical truths. This invention is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.
Claims
1. A voxel-based method for filtering noise in lidar based on signal strength and distance, characterized in that, The method includes: Acquire point cloud data output by lidar; The point cloud data is voxelized, and the point cloud data points located in the same voxel grid are combined into a point cloud data set corresponding to the voxel. For each point cloud dataset, a differentiated result is calculated based on the distance and signal strength values of all point cloud data points in the point cloud dataset. Determine the weighting coefficients corresponding to the distance value and the signal strength value; The differential results are weighted and calculated according to the weighting coefficients to obtain the dynamic comprehensive value of each point cloud data point. If the dynamic composite value is determined to be greater than a preset threshold, the corresponding point cloud data point is identified as noise and removed.
2. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 1, characterized in that, The voxelization process for the point cloud data specifically includes: Based on the detection range of the lidar and the spatial distribution density of the point cloud data in the application scenario, a preset voxel size value matching the application scenario is determined, wherein the voxel size value is used to determine the spatial scale of each voxel grid in three-dimensional space.
3. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 1, characterized in that, The voxelization process for the point cloud data further includes: The average volume density is calculated based on the three-dimensional spatial range covered by the point cloud data and the number of point clouds. The voxel space volume is calculated based on the average volume density and the preset threshold for the number of point clouds within a single voxel grid. The voxel size value is determined based on the voxel space volume, where the voxel size value is the cube root of the voxel space volume.
4. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 2, characterized in that, The step of forming a point cloud data set of corresponding voxels from point cloud data points located within the same voxel grid specifically includes: The three-dimensional space covered by the point cloud data is divided into a regular voxel grid based on the voxel size value, and each voxel grid has a unique spatial location identifier. Traverse all point cloud data points corresponding to the point cloud data points, determine the voxel grid to which the point cloud data points belong based on the three-dimensional coordinates of the point cloud data points, and classify all point cloud data points falling into the same voxel grid into the point cloud data set corresponding to the voxel grid. The point cloud data is organized into multiple point cloud data sets according to spatial location.
5. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 1, characterized in that, For each point cloud data set, the differential result is calculated based on the distance values and signal strength values of all point cloud data points in the point cloud data set, specifically including: For each point cloud data set corresponding to the voxel grid, the distance value and signal strength value of all point cloud data points in the point cloud data set are statistically analyzed. For each voxel grid, the mean distance value and the standard deviation of the corresponding distance value are calculated based on all distance values contained in the corresponding point cloud dataset. For each voxel grid, the mean signal strength value and the corresponding standard deviation of the signal strength value are calculated based on all distance values and signal strength values contained in the corresponding point cloud dataset. For any point cloud data point in the point cloud dataset, the distance difference result of the point cloud data point is calculated based on the distance value, the mean distance value, and the standard deviation of the distance value. The signal strength difference result of the point cloud data point is calculated based on the signal strength value, the mean signal strength value, and the standard deviation of the signal strength value.
6. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 1, characterized in that, The determination of the weighting coefficients corresponding to the distance value and the signal strength value specifically includes: For each point cloud data point, determine the actual detection distance corresponding to that point cloud data point; For each point cloud data point, the target distance interval to which it belongs is determined in the preset distance and weighting coefficient mapping table according to the actual detection distance. The determination of the weighting coefficients corresponding to the distance value and the signal strength value divides the detection distance range into multiple detection distance intervals. Each detection distance interval corresponds to a set of distance difference weighting coefficients and signal strength difference weighting coefficients. Extract the distance difference weighting coefficient and signal strength difference weighting coefficient corresponding to the target distance interval.
7. The method for voxel-based noise removal of lidar based on signal strength and distance according to claim 1, characterized in that, The determination of the weighting coefficients corresponding to the distance value and the signal strength value further includes: For each point cloud data point, determine the actual detection distance corresponding to that point cloud data point; Based on the actual detection distance, the distance measurement error estimate and the signal strength attenuation error estimate are calculated respectively. The distance measurement error estimate is established based on the logarithmic function of the actual detection distance, and the signal strength attenuation error estimate is established based on the exponential attenuation function of the actual detection distance. Based on the inverse relationship between the distance measurement error estimate and the signal strength attenuation error estimate, a distance differentiation weighting coefficient and a signal strength differentiation weighting coefficient are calculated respectively, wherein the distance differentiation weighting coefficient is calculated based on the distance measurement error estimate, and the signal strength differentiation weighting coefficient is calculated based on the signal strength attenuation error estimate.
8. A voxel-based noise filtering device for lidar based on signal strength and distance, characterized in that, The device is used to perform a method for filtering out lidar noise based on signal strength and distance as described in any one of claims 1-7. The device includes an acquisition module (201), a processing module (202), and an output module (203), wherein: The acquisition module (201) is used to acquire point cloud data output by the lidar; The processing module (202) is used to perform voxelization processing on the point cloud data, and to form a point cloud data set of corresponding voxels by point cloud data points located in the same voxel grid. The processing module (202) is used to calculate a differentiated result for each point cloud data set based on the distance value and signal strength value of all point cloud data points in the point cloud data set. The processing module (202) is used to determine the weighting coefficients corresponding to the distance value and the signal strength value; The processing module (202) is used to perform a weighted calculation on the differential result and according to the weighting coefficient to obtain the dynamic comprehensive value of each point cloud data point; The output module (203) is used to determine the corresponding point cloud data points as noise and remove them if it is determined that the dynamic comprehensive value is greater than a preset threshold.
9. An electronic device, characterized in that, The device includes a processor (301), a communication bus (302), a user interface (303), a network interface (304), and a memory (305). The memory (305) is used to store instructions. The user interface (303) and the network interface (304) are both used to communicate with other devices. The communication bus (302) is used to realize the connection and communication between the components within the electronic device. The processor (301) is used to execute the instructions stored in the memory (305) so that the electronic device performs the method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.