Method for improving precision of laser radar point cloud data

Abstract

The invention relates to a method for improving the precision of laser radar point cloud data. The method comprises the following steps: obtaining original low-precision laser radar point cloud data and original high-precision laser radar point cloud data; performing deletion processing on abnormal data in the original low-precision laser radar point cloud data; performing interpolation processing on the low-precision laser radar point cloud data after deletion processing so as to realize matching of the interpolated low-precision laser radar point cloud data and the original high-precision laser radar point cloud data; performing adaptive Fourier decomposition on the interpolated low-precision laser radar point cloud data until the similarity meets a preset condition; the similarity is calculated according to the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data.

Description

Technical Field

[0001] The present invention relates to the field of laser radar technology, and in particular to a method for improving the accuracy of laser radar point cloud data. Background Art

[0002] In the field of LiDAR technology, despite its multi-faceted application potential, high cost remains one of the main obstacles to its widespread adoption and application. Cost reduction strategies usually involve optimizing hardware costs and improving data processing efficiency. In terms of hardware, using lower-cost materials, improving manufacturing technology, and simplifying the design of LiDAR systems, such as optimizing laser transmitters and receivers, are all effective measures to reduce initial investment. In addition, improving the efficiency of data processing algorithms and reducing dependence on high-end computing resources are also key means to reduce overall costs.

[0003] In terms of point cloud data processing, it is particularly important to develop efficient processing algorithms in the face of problems such as large data volume, high noise, sparse data, and uneven distribution. However, these problems not only increase the complexity of data processing, but also place higher demands on computing resources. Occlusion and multipath effects in dynamic environments further increase the difficulty of real-time processing, resulting in huge technical challenges to overcome in extracting effective information and features. For example, the noise in point cloud data may come from errors in the lidar system itself, the material properties of the scanned object, or interference from external environmental factors.

[0004] Although the quality of low-precision LiDAR data can be improved by data interpolation, filtering and denoising, feature enhancement and other methods, these methods often introduce new problems such as over-smoothing and feature loss, and the processing efficiency and effect need to be improved. Therefore, developing a new method that can effectively improve the quality of low-precision LiDAR data while maintaining data features and improving processing efficiency has become an important research direction in this technical field. Summary of the invention

[0005] In order to solve the problem of poor processing efficiency and effect in the prior art when improving the quality of low-precision laser radar data, the purpose of the present invention is to provide a method for improving the accuracy of laser radar point cloud data. The technical solution adopted is as follows:

[0006] A method for improving the accuracy of laser radar point cloud data, the method comprising the following steps:

[0007] Obtain original low-precision LiDAR point cloud data and original high-precision LiDAR point cloud data;

[0008] Delete abnormal data in the original low-precision lidar point cloud data;

[0009] Performing interpolation processing on the deleted low-precision laser radar point cloud data to match the interpolated low-precision laser radar point cloud data with the original high-precision laser radar point cloud data;

[0010] Adaptive Fourier decomposition is performed on the interpolated low-precision lidar point cloud data until the similarity meets a preset condition; the similarity is calculated based on the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data.

[0011] Preferably, the abnormal data in the original low-precision laser radar point cloud data is deleted, including:

[0012] Get the nearest neighbor point of each data point in the original low-precision lidar point cloud data, and calculate the nearest neighbor distance from each data point to the nearest neighbor;

[0013] Obtain the average distance and standard deviation of all nearest neighbor distances by calculating;

[0014] Data points whose distance exceeds the mean distance plus twice the standard deviation are considered outliers and removed.

[0015] Preferably, cubic spline interpolation is used to interpolate the low-precision lidar point cloud data after the deletion process.

[0016] Preferably, calculating the similarity between the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data includes:

[0017] The dynamic time warping distance between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is calculated, and the similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the dynamic time warping distance.

[0018] Preferably, calculating the similarity between the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data includes:

[0019] Calculate the root mean square error between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data;

[0020] The similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the root mean square error.

[0021] Preferably, the interpolated low-precision lidar point cloud data is subjected to adaptive Fourier decomposition until the similarity meets a preset condition, including:

[0022] The low-precision lidar point cloud data after interpolation processing is converted into a linear curve, and decomposed by adaptive Fourier decomposition to extract M components;

[0023] The extracted M components are divided into N groups, each group has K components, one group of components is gradually eliminated starting from the high-frequency components, and the remaining components are merged to obtain the second low-precision lidar point cloud data;

[0024] Calculate the similarity between the second lowest precision LiDAR point cloud data at different elimination stages and the original high precision LiDAR point cloud data;

[0025] Compare the similarities corresponding to different elimination stages, and select the component combination with the greatest similarity as the optimal component combination;

[0026] Reconstruct low-precision lidar point cloud data using optimal component combination.

[0027] The low-precision lidar point cloud data after interpolation is further converted into a linear curve and decomposed through adaptive Fourier decomposition to extract M components, including:

[0028] Use Adaptive Basis Each basis function B n is called the modified Blaschke product and is defined as:

[0029]

[0030] where a n ∈D={z∈C:|z|<1}, C is the complex plane;

[0031] Initialization: Set the initial signal to f0(t);

[0032] Iterative decomposition: In each iteration, an optimal parameter a is selected n Make the energy converge as fast as possible, and calculate the next signal f through the recursive formula n+1 (t):

[0033] f n+1 (t) = f n (t)- <f n ,B n >B n

[0034] in, <f n ,B n > represents the signal f n (t) and basis function B n (t) Inner product.

[0035] A computer device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the method for improving the accuracy of laser radar point cloud data as described above when executing the computer program.

[0036] A computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to perform the method for improving the accuracy of laser radar point cloud data as described above.

[0037] The present invention has the following beneficial effects:

[0038] The present invention introduces adaptive Fourier decomposition (AFD) to solve the problems of data sparsity and insufficient precision in the prior art, and the loss of some important signal information while denoising, making it difficult to find a good balance between noise reduction and retaining data features. As a powerful signal processing tool, adaptive Fourier decomposition AFD can carefully decompose signals, more accurately identify and remove noise components, while retaining key signal features. Through this method, the present invention can effectively improve the quality of low-precision lidar point cloud data, solve the problems of data sparsity, large errors and noise, without the need for complex parameter adjustments, with relatively low computational cost and easy implementation. In addition, the AFD method does not rely on a large amount of training data, which makes the present invention have good applicability and generalization ability in different application scenarios.

[0039] The present invention uses an adaptive Fourier decomposition (AFD) processing method to improve the quality of point cloud data collected by low-cost, low-precision 2D LiDAR, making it closer to the performance of high-cost, high-precision LiDAR in terms of accuracy and data integrity. This technology is not only suitable for application scenarios in the fields of autonomous driving, robot navigation, environmental monitoring, etc. that have high requirements for LiDAR data accuracy, but can also provide support for LiDAR-based terrain mapping, three-dimensional reconstruction of buildings, and heritage protection. It especially demonstrates its important value and application potential in cost-sensitive projects. By improving the quality of low-precision LiDAR data, the present invention aims to broaden the scope of application of LiDAR technology and improve its practicality and economic benefits in multiple technical fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0041] Figure 1 A flowchart of a method for improving the accuracy of laser radar point cloud data provided by an embodiment of the present invention.

[0042] Figure 2This is a radar parameter comparison chart.

[0043] Figure 3 It is a schematic diagram of the experimental environment of the present invention.

[0044] Figure 4 This is a comparison chart of the outlier removal effect.

[0045] Figure 5 It is a schematic diagram of AFD decomposition.

[0046] Figure 6 This is a comparison chart after the first phase of the experiment.

[0047] Figure 7 This is a comparison chart after the second phase of the experiment.

[0048] Figure 8 This is a comparison chart after the third stage of the experiment.

[0049] Fig. 9 Schematic diagram for comparing different models.

[0050] Fig.10 A functional block diagram of a computer device provided in accordance with an embodiment of the present invention. DETAILED DESCRIPTION

[0051] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the method for improving the accuracy of laser radar point cloud data proposed by the present invention, its specific implementation method, structure, features and effects, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0052] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0053] With the rapid development of urbanization and economic society, LiDAR technology has been increasingly valued due to its wide application in key areas such as autonomous driving and environmental monitoring. The application of LiDAR technology is not limited to vehicle automatic navigation. Its high-precision and high-resolution measurement capabilities also make it play a vital role in environmental science and industrial applications. However, common 2D LiDAR faces many challenges in the processing process, especially the excessive noise and sparsity problems in the data, which seriously affect the availability and accuracy of the data, resulting in the generated point cloud contour image being blurred and full of noise.

[0054] To address these challenges, this embodiment proposes a method for improving the accuracy of laser radar point cloud data using adaptive Fourier decomposition, aiming to significantly improve the quality and accuracy of low-precision laser radar data.

[0055] A specific solution of a method for improving the accuracy of laser radar point cloud data provided by the present invention is described in detail below with reference to the accompanying drawings.

[0056] like Figure 1 As shown, a method for improving the accuracy of laser radar point cloud data includes the following steps:

[0057] Obtain original low-precision LiDAR point cloud data and original high-precision LiDAR point cloud data;

[0058] Delete abnormal data in the original low-precision lidar point cloud data;

[0059] Performing interpolation processing on the deleted low-precision laser radar point cloud data to match the interpolated low-precision laser radar point cloud data with the original high-precision laser radar point cloud data;

[0060] Adaptive Fourier decomposition is performed on the interpolated low-precision lidar point cloud data until the similarity meets a preset condition; the similarity is calculated based on the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data.

[0061] The present invention includes abnormal data removal processing, interpolation processing to enhance data continuity and the application of adaptive Fourier decomposition (AFD) technology. In particular, the introduction of AFD technology provides a novel solution for lidar data processing, which effectively reduces noise and improves data density and accuracy by gradually decomposing and reconstructing signals.

[0062] By comparing the optimized 2D low-precision N10 radar data with the high-precision M10 radar data, we not only verified the effectiveness of the proposed method, but also demonstrated the potential value of the technology in practical applications. For example, in the field of autonomous driving, by improving the accuracy and clarity of the data, the vehicle can more accurately identify and locate surrounding obstacles, thereby improving the safety of navigation and obstacle avoidance; in the field of environmental monitoring, accurate point cloud data can better perform terrain mapping and environmental quality assessment, providing strong technical support for environmental protection and resource management.

[0063] The present invention uses two 2D laser radars: low-precision laser radar N10 and high-precision laser radar M10 for data collection. These two laser radars have different parameters. For example, Figure 2 shown.

[0064] This embodiment takes into account that the performance of the laser radar may be affected by environmental factors and different dielectric materials, so the choice of the experimental environment becomes particularly critical. In order to ensure the accuracy and repeatability of the experimental results, this embodiment deliberately selected a semi-enclosed rectangular space with a size of 138 cm × 93 cm × 40 cm as the experimental site. Such an environmental design is intended to simulate various situations that may be encountered in actual applications, while minimizing the interference of external factors on the experimental data.

[0065] In the first stage of the experiment, Figure 3 As shown in the demonstration, 14 obstacles with a height of more than 10 cm were selected as test targets and distributed on the inner wall of the rectangular space. Then, the M10 and N10 LiDAR models were placed in the center of the box to detect these static obstacles. This set of experiments aims to evaluate the detection capability and accuracy of LiDAR in a relatively static and regularly distributed environment.

[0066] The purpose of preprocessing radar data is to reduce the impact of data errors on experimental results and ensure the accuracy and reliability of analysis. The preprocessing involves the elimination of abnormal data.

[0067] In a specific embodiment, the abnormal data in the original low-precision laser radar point cloud data is deleted, including:

[0068] Get the nearest neighbor point of each data point in the original low-precision lidar point cloud data, and calculate the nearest neighbor distance from each data point to the nearest neighbor;

[0069] Obtain the average distance and standard deviation of all nearest neighbor distances by calculating;

[0070] Data points whose distance exceeds the mean distance plus twice the standard deviation are considered outliers and removed.

[0071] In this example, it is critical to identify and handle outliers when processing LiDAR data. To achieve this, the functionality of the cKDTree class in the numpy and scipy libraries is used. Initially, the query functionality of cKDTree allows for efficient finding of the nearest neighbors of each data point and calculation of the nearest neighbor distance of each data point to the nearest neighbor. These nearest neighbor distances provide a quantitative expression of the relationship between data points and also reflect the spatial structure of the point cloud data.

[0072] The overall statistical characteristics of the lidar dataset are then obtained by calculating the mean and standard deviation of all nearest neighbor distances. The overall statistical characteristics include the mean and standard deviation. The mean reflects the general distance level between data points, while the standard deviation reveals the variability of these nearest neighbor distances. Based on these statistics, the threshold is set to "mean distance plus twice the standard deviation", and points outside this range are considered outliers. These outliers are far away from their nearest neighbors and may represent noise or anomalies.

[0073] In order to maintain the integrity of the data, the outliers are not discarded directly, but the coordinates of the outliers are replaced with the coordinates of the nearest neighbors, specifically the average coordinates of the two nearest neighboring data points before and after the data point. This method effectively reduces the impact of noise while preventing data loss. Figure 4 As shown, Figure 4 Part (a) shows the N10 model data before outliers are removed. Figure 4 Part (b) in Figure 3 describes the N10 model data after removing outliers.

[0074] In a specific embodiment, cubic spline interpolation is used to interpolate the low-precision lidar point cloud data after deletion.

[0075] Among them, cubic spline interpolation is a commonly used interpolation method in mathematics and computer science, which is used to construct a smooth function through a set of discrete data points. Among many interpolation techniques, cubic spline interpolation is particularly favored for its high flexibility and excellent approximation characteristics. It forms an overall smooth curve or surface by splicing multiple low-order polynomial functions to approximate or pass through all given data points. The following is the definition of spline function:

[0076] Suppose a partition is given on the interval [a,b]:

[0077] a=x0 <x1<…<x m-1 <x m =b

[0078] X m Refers to the last interpolation node, defining the end point of the interpolation interval.

[0079] T(x) is a real-valued function if it satisfies the following conditions:

[0080] T(x) in the subinterval [x k ,x k+1 ](k=0,1,…,m-1) is a polynomial of degree m;

[0081] T(x) has continuous derivatives up to order m-1 on [a,b], that is, T(x)∈C m-1[a, b],

[0082] Then \(T(x)\) is called a spline function of degree \(m\), and \(x_0, x_1, \ldots, x\) m are interpolation nodes.

[0083] The characteristics of cubic spline interpolation are as follows:

[0084] 1. The curve constructed by cubic spline interpolation is not only equal in value at all connection points, but also has continuous first and second derivatives, so the overall curve is very smooth.

[0085] 2. Local control: Each polynomial segment is determined only by the data points at its two ends. Therefore, when modifying or adding data points, only the adjacent curve segments are affected.

[0086] 3. Flexibility and adaptability: Cubic spline interpolation is applicable to various scenarios that require smooth curve approximation and has wide applications in both scientific research and engineering design.

[0087] Cubic spline interpolation: Suppose on the interpolation interval \([a, b]\), \(m + 1\) interpolation nodes \(x\) k (\(a = x_0 \lt x_1 \lt \ldots \lt x\) m-1 \(\lt X\) m = b\)) are given, and the function values \(y\) k at the nodes \(x\) k = f(x k )(k = 0, 1, \ldots, m). If the interpolation function \(s(x)\) satisfies the conditions:

[0088] 1. \(s(x k ) = y(x k ) = y k (k = 0, 1, \ldots, m);

[0089] 2. On each subinterval \([X k , x k+1 (k = 0, 1, \ldots, m - 1), \(s(x)\) is a cubic polynomial \(s k (x);

[0090] 3. \(s(x)\) is twice continuously differentiable on \([a, b]\), that is, \(s(x) \in C 2 [a + b]\), and \(C 2 represents the function of twice continuous differentiability.

[0091] Then \(s(x)\) is called the cubic spline interpolation function of \(f(x)\).

[0092] It should be noted that the low-precision lidar point cloud data after cubic spline interpolation is also a continuous signal.

[0093] In this embodiment, the root mean square error (RMSE) can be used as the similarity between the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data. The root mean square error is a common indicator for measuring the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data. Its calculation formula is as follows:

[0094]

[0095] Where: P i is the low-precision lidar point cloud data after the i-th interpolation; O i is the i-th original high-precision lidar point cloud data.

[0096] RMSE squares the difference between each interpolated low-precision LiDAR point cloud data and the corresponding original high-precision LiDAR point cloud data, then takes the average, and finally takes the square root of the average. The square root ensures that the RMSE maintains the same units as the original data, which is easy to understand and interpret. A smaller RMSE value indicates a smaller prediction error, which means that the interpolated low-precision LiDAR point cloud data is more consistent with the original high-precision LiDAR point cloud data.

[0097] In a specific embodiment, preferably, calculating the similarity between the original high-precision LiDAR point cloud data and the interpolated low-precision LiDAR point cloud data includes:

[0098] The dynamic time warping distance between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is calculated, and the similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the dynamic time warping distance.

[0099] Preferably, calculating the similarity between the original high-precision laser radar point cloud data and the interpolated low-precision laser radar point cloud data includes:

[0100] Calculate the root mean square error between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data;

[0101] The similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the root mean square error.

[0102] Traditional sequence similarity comparison methods based on norm distance, such as Euclidean distance and Manhattan distance, have two significant limitations when processing sequence data: they cannot effectively compare sequences of inconsistent lengths and are easily affected by local time offsets.

[0103] In view of the difference in the number of radar data points between M10 and N10, one of the goals of this embodiment is to measure the similarity between two unequal length curves. Among the methods for measuring the similarity of unequal length curves, Hausdorff distance, Fréchet distance and dynamic time warping (DTW) distance are all commonly used measurement standards.

[0104] Considering the unique advantages of DTW distance, that is, it not only allows comparison between sequences of unequal lengths, but also effectively handles the problem of local time offset in time series, the present invention prefers DTW distance as the main evaluation tool. DTW minimizes the difference between the overall time series by finding the best matching path between the two sequences, thus providing a flexible and accurate way to evaluate the similarity of M10 and N10 radar data after optimization. This allows us to more accurately measure the effect of the optimization strategy, especially when dealing with radar data with different lengths and potential time offsets.

[0105] The DTW (Dynamic Time Warping) distance is an algorithm used to measure the similarity of unequal time series. It can find the best matching path between two time series and minimize their cumulative distance. The following are the detailed steps of DTW distance calculation:

[0106] 1. Construct a distance matrix: Two unequal length curves can be recorded as a finite set of points:

[0107]

[0108] where x i ,y i They correspond to the different grid data points on their respective curves, and n and m are the lengths of curves A and B respectively.

[0109] Calculate the distance between each pair of points in the two sequences and construct an n×m matrix D. The elements in the matrix d i,j It represents the distance between the i-th point in sequence A and the j-th point in sequence B, usually using the Euclidean distance.

[0110] d i,j =||a i -b j || 2

[0111] 2. Calculate the cumulative distance matrix: Use dynamic programming techniques to calculate the best path from the upper left corner to the lower right corner of D. Define a cumulative distance matrix D', where the element D'(n,m) represents the minimum cumulative distance from the upper left corner of D to the element d(i,j).

[0112] D′(i,j)=d(i,j)+min{D′(i-1,j),D′(i,j-1),D′(i-1,j-1)}

[0113] The boundary conditions are:

[0114] D′(1,1)=d(1,1)

[0115] D′(i,1)=d(i,1)+D′(i-1,1)(2≤i≤n)

[0116] D′(1,j)=d(1,j)+D′(1,j-1)(2≤j≤m)

[0117] 3. Find the best path: Start from the lower right corner of the matrix D′ and trace back to the upper left corner to find the path with the smallest cumulative distance. This path represents the best correspondence between the two time series.

[0118] The steps of the backtracking path are as follows:

[0119] (i,j)←argmin{D′(i-1,j),D′(i,j-1),D′(i-1,j-1)}

[0120] Until you reach the upper left corner of the matrix.

[0121] 4. Calculate the DTW distance: The final DTW distance is the value in the lower right corner of the cumulative distance matrix D′, that is, D′(n,m), which reflects the total distance between sequence Q and sequence C.

[0122] DTW(Q,C)=D′(n,m)

[0123] Through the above steps, DTW can effectively handle the inconsistency of time series in speed and time, and provide a similarity measurement method for sequences of different lengths or different speed changes.

[0124] In a specific embodiment, the interpolated low-precision LiDAR point cloud data is adaptively decomposed until the similarity meets a preset condition, including:

[0125] The low-precision lidar point cloud data after interpolation processing is converted into a linear curve, and decomposed by adaptive Fourier decomposition to extract M components;

[0126] The extracted M components are divided into N groups, each group has K components, one group of components is gradually eliminated starting from the high-frequency components, and the remaining components are merged to obtain the second low-precision lidar point cloud data;

[0127] Calculate the similarity between the second lowest precision LiDAR point cloud data at different elimination stages and the original high precision LiDAR point cloud data;

[0128] Compare the similarities corresponding to different elimination stages, and select the component combination with the greatest similarity as the optimal component combination;

[0129] Reconstruct low-precision lidar point cloud data using optimal component combination.

[0130] In this embodiment, adaptive Fourier decomposition (AFD) is a signal processing method evolved from traditional Fourier decomposition, which achieves fast energy convergence by using an adaptive basis. AFD is mainly used to decompose signals so that they can analyze and process local characteristics of signals in more detail. Adaptive Fourier decomposition (AFD) is a complex signal decomposition technology that plays an important role in digital signal processing (DSP). Unlike traditional Fourier transform, AFD can adapt to the local characteristics of signals and is particularly suitable for non-stationary signals whose frequency components change over time. This method decomposes continuous signals into linear combinations of sine and cosine components of different frequencies in an adaptive manner, thereby enhancing its application in various fields such as lidar technology.

[0131] The low-precision lidar point cloud data after interpolation is further converted into a linear curve and decomposed by adaptive Fourier decomposition to extract M components, including:

[0132] Use Adaptive Basis Each basis function B n is called the modified Blaschke product and is defined as:

[0133]

[0134] where a n ∈D={z∈C:|z|<1}, C is the complex plane;

[0135] D represents the unit disk on the complex plane, usually defined as D = {z∈C:|z|<1}; Z represents a complex variable; e jt Represents the complex exponential function, representing a complex number on the unit circle, often used to represent rotational or oscillatory signals, which can be viewed as a function of time t, often associated with Fourier transforms or frequency domain analysis.

[0136] Initialization: Set the initial signal to f0(t);

[0137] Iterative decomposition: In each iteration, an optimal parameter a is selected n Make the energy converge as fast as possible, and calculate the next step signal f through the recursive formula n+1 (t):

[0138] f n+1 (t) = f n (t)- <fn ,B n >B n

[0139] in, <f n ,B n > represents the signal f n (t) and basis function B n (t) Inner product.

[0140] The selection of adaptive basis is the key to AFD. The optimal basis function is selected through the matching pursuit algorithm. In each iteration, the basis function with the fastest energy convergence is selected according to the characteristics of the current signal, so that the signal can be quickly decomposed in a fewer number of iterations. By selecting an adaptive basis, AFD achieves faster energy convergence than traditional Fourier decomposition. This means that the main features of the signal can be captured in fewer iterations, thereby improving the efficiency and accuracy of signal processing. The decomposition results are shown in Figure 1. Figure 5 shown.

[0141] This example demonstrates the effectiveness of the present invention through three groups of experiments, (a) is M10; (b) is before N10 treatment; (c) is after N10 treatment.

[0142] In the first stage of the experiment, Figure 6 As shown, 14 obstacles with a height of more than 10cm were selected as test targets and were placed along the inner wall of the rectangular space. Then, the M10 and N10 LiDAR devices were placed in the center of the box to detect these static obstacles. This set of experiments aims to evaluate the detection capability and accuracy of LiDAR in a relatively static and regularly distributed environment.

[0143] In the second phase of the experiment, the number of obstacles was reduced from 14 to 12, while keeping their distribution along the inner walls of the rectangle, in order to test the performance of the lidar in a more complex environment, especially the resolution capability and accuracy when facing more targets, such as Figure 7 Both the M10 and N10 were located in the exact center of the box to ensure consistency across the experiments.

[0144] In the third phase of the experiment, while the number of obstacles remained at 12, their distances relative to the LiDAR were changed by bringing the obstacles closer to the device, e.g. Figure 8 shown.

[0145] This adjustment aims to explore the impact of changes in obstacle distance on the LiDAR ranging accuracy. With this configuration, not only can the performance of LiDAR under different distance settings be analyzed, but also its ability to detect close-range targets can be evaluated, further verifying the potential application of LiDAR technology in complex environments. These two sets of experiments jointly verified the adaptability and accuracy of LiDAR in the face of changes in obstacle density and distance, providing strong data support for the optimization of LiDAR in practical applications.

[0146] Through these three sets of experiments, the goal is to comprehensively evaluate the performance of the M10 and N10 models in various environmental settings. The first set of experiments was designed as an experimental group to evaluate the detection capability and accuracy of the lidar in a relatively static and regularly distributed environment. Subsequently, the second and third sets of experiments served as validation groups to test the repeatability and reliability of the first set of results to ensure that its performance was not accidental. By adjusting the number and position of obstacles in the validation group, the performance of the lidar in more complex environments was further tested, especially the impact of changes in obstacle distance on the lidar ranging accuracy was studied. This experimental design aims to gain an in-depth understanding of the operating characteristics of the lidar and its potential limitations in practical applications through comparative analysis under different settings, so as to ensure the accuracy and practical application value of the research results.

[0147] In order to deepen the analysis and optimization of LiDAR data, AFD technology was applied, and EMD and VMD were introduced as comparison benchmarks. Fig. 9 Comparing the three sets of data shown, it can be observed that under similar processing conditions, the noise reduction performance of AFD exceeds that of EMD and VMD methods.

[0148] observe Fig. 9 Comparison of the three sets of data shown in Figure 1 clearly shows that under similar processing conditions, AFD significantly outperforms EMD and VMD in noise reduction performance. Specifically, in the first set of data (experimental group), AFD's performance is better than the other two methods in terms of improvement percentage and error index. In particular, the noise reduction advantage of AFD technology is further verified in the second and third sets of data. This finding not only highlights the efficiency of AFD technology in processing radar signals.

[0149] The present invention solves the key problems faced by existing lidar data processing technology through a series of innovative processing strategies and technical optimizations, including noise sensitivity, data sparsity, processing capacity limitations, lack of flexibility and adaptability, as well as complexity and resource consumption. The following is a detailed introduction to the optimization processing strategy of the present invention and its improvement over the shortcomings of the existing technology:

[0150] 1. Improvements on noise sensitivity:

[0151] In the prior art, LiDAR systems are sensitive to environmental noise such as rain, fog, and dust, which significantly affect data quality. The present invention uses adaptive Fourier decomposition (AFD) technology to effectively separate useful signals from noise by accurately analyzing frequency components, greatly improving the robustness of data in complex environments.

[0152] 2. Improvements to data sparsity:

[0153] The present invention improves the sparse data point problem in low-precision LiDAR data through cubic spline interpolation. These technologies can effectively fill in data gaps without significantly increasing the computational burden, thereby improving the clarity of object contours and the accuracy of target recognition.

[0154] 3. Improvements to address processing power limitations:

[0155] By optimizing the data processing algorithm, the present invention reduces the reliance on high-performance processing hardware. The application of adaptive Fourier decomposition (AFD) technology reduces the computational intensity of the algorithm, making data processing more efficient, reducing the demand for storage space, and thus reducing the overall system cost, which is particularly suitable for real-time or near real-time data processing applications.

[0156] 4. Improvements to address lack of flexibility and adaptability:

[0157] Traditional data processing methods have limited adaptability in different environments or for different types of target objects. The present invention introduces adaptive Fourier decomposition to enable the system to dynamically adjust the processing strategy according to the characteristics of the data itself, thereby maintaining high performance in various environments.

[0158] 5. Improvements on complexity and resource consumption:

[0159] The present invention simplifies the data processing flow, reduces the complex algorithms and large amount of computing resources involved in the traditional method, effectively reduces the computing burden and improves the system response speed and efficiency through the removal of outliers and the fine adjustment of cubic spline interpolation.

[0160] In summary, the present invention not only solves the core challenges of the existing technology in lidar data processing through a series of innovative optimization processing strategies, but also improves the robustness of data processing, reduces the dependence on high-performance hardware, and greatly improves the overall quality of data and the practicality of the system. These improvements give the present invention significant advantages in a variety of practical application scenarios, especially in situations where fast, accurate and efficient processing of lidar data is required.

[0161] See also Fig.10 , Fig.105 is a schematic block diagram of a computer device provided by an embodiment of the present invention. The computer device 500 is a server, which can be an independent server or a server cluster composed of multiple servers.

[0162] See also Fig.10 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a non-volatile storage medium 503 and an internal memory 504 .

[0163] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 can execute a multi-audio data intelligent processing method.

[0164] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0165] The internal memory 504 provides an environment for the operation of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute the multi-audio data intelligent processing method.

[0166] The network interface 505 is used for network communication, such as providing data information transmission, etc. Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 500 to which the solution of the present invention is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0167] The processor 502 is used to run a computer program 5032 stored in the memory to implement the multi-audio data intelligent processing method disclosed in the embodiment of the present invention.

[0168] Those skilled in the art will understand that Fig.10 The embodiments of the computer device shown in the figure do not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such embodiments, the structure and function of the memory and the processor are the same as those of the embodiment of the present invention. Fig.10 The embodiments shown are consistent and will not be described again here.

[0169] It should be understood that in the embodiment of the present invention, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0170] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a non-volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the multi-audio data intelligent processing method disclosed in the embodiment of the present invention is implemented.

[0171] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described equipment, devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0172] In the several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. Units with the same function may also be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0173] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0174] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0175] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially 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. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), disk or optical disk and other media that can store program codes.

[0176] It should be noted that the sequence of the above embodiments of the present invention is only for description and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0177] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

Claims

1. A method for improving the accuracy of laser radar point cloud data, characterized in that: The method comprises the following steps: Obtain original low-precision LiDAR point cloud data and original high-precision LiDAR point cloud data; Delete abnormal data in the original low-precision lidar point cloud data; Performing interpolation processing on the deleted low-precision laser radar point cloud data to match the interpolated low-precision laser radar point cloud data with the original high-precision laser radar point cloud data; Adaptive Fourier decomposition is performed on the interpolated low-precision lidar point cloud data until the similarity meets a preset condition; the similarity is calculated based on the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data.

2. The method for improving the accuracy of laser radar point cloud data according to claim 1, characterized in that: Delete abnormal data in the original low-precision lidar point cloud data, including: Get the nearest neighbor point of each data point in the original low-precision lidar point cloud data, and calculate the nearest neighbor distance from each data point to the nearest neighbor; Obtain the average distance and standard deviation of all nearest neighbor distances by calculating; Data points whose distance exceeds the mean distance plus twice the standard deviation are considered outliers and removed.

3. The method for improving the accuracy of laser radar point cloud data according to claim 1, characterized in that: Cubic spline interpolation is used to interpolate the low-precision lidar point cloud data after deletion.

4. The method for improving the accuracy of laser radar point cloud data according to claim 1, characterized in that: Calculate the similarity between the original high-precision LiDAR point cloud data and the interpolated low-precision LiDAR point cloud data, including: The dynamic time warping distance between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is calculated, and the similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the dynamic time warping distance.

5. The method for improving the accuracy of laser radar point cloud data according to claim 1, characterized in that: Calculate the similarity between the original high-precision LiDAR point cloud data and the interpolated low-precision LiDAR point cloud data, including: Calculate the root mean square error between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data; The similarity between the original high-precision lidar point cloud data and the interpolated low-precision lidar point cloud data is evaluated based on the root mean square error.

6. The method for improving the accuracy of laser radar point cloud data according to claim 1, characterized in that: Adaptive Fourier decomposition is performed on the interpolated low-precision lidar point cloud data until the similarity meets the preset conditions, including: The low-precision lidar point cloud data after interpolation processing is converted into a linear curve, and decomposed by adaptive Fourier decomposition to extract M components; The extracted M components are divided into N groups, each group has K components, one group of components is gradually eliminated starting from the high-frequency components, and the remaining components are merged to obtain the second low-precision lidar point cloud data; Calculate the similarity between the second lowest precision LiDAR point cloud data at different elimination stages and the original high precision LiDAR point cloud data; Compare the similarities corresponding to different elimination stages, and select the component combination with the greatest similarity as the optimal component combination; Reconstruct low-precision lidar point cloud data using optimal component combination.

7. The method for improving the accuracy of laser radar point cloud data according to claim 6, characterized in that: The low-precision lidar point cloud data after interpolation is converted into a linear curve and decomposed through adaptive Fourier decomposition to extract M components, including: Use Adaptive Basis Each basis function B n is called the modified Blaschke product and is defined as: where a n ∈D={z∈C:|z|<1}, C is the complex plane; Initialization: Set the initial signal to f0(t); Iterative decomposition: In each iteration, an optimal parameter a is selected n Make the energy converge as fast as possible, and calculate the next signal f through the recursive formula n+1 (t): f n+1 (t)=f n (t)-<f n ,B n >B n in, <f n ,B n > represents the signal f n (t) and basis function B n (t) Inner product.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the method for improving the accuracy of lidar point cloud data as described in any one of claims 1 to 7 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to execute the method for improving the accuracy of lidar point cloud data as described in any one of claims 1 to 7.

Images