A laser radar waveform identification method and device

Abstract

The application discloses a laser radar waveform identification method and device. The method comprises the following steps: obtaining a plurality of waveform data to be identified; calculating a plurality of waveform characteristic values corresponding to each waveform data, so as to obtain a waveform characteristic set corresponding to each waveform data; clustering each waveform data based on the waveform characteristic set by using a Gaussian mixture model, so as to obtain a first waveform category containing a plurality of waveform data and a second waveform category containing a plurality of waveform data; and identifying each waveform data in the first waveform category as a zero-return type saturated waveform, each waveform data in the second waveform category as a non-saturated waveform, or each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return type saturated waveform based on the waveform characteristic values corresponding to each waveform data in the first waveform category and the second waveform category. The application can quickly and accurately identify saturated waveforms.

Description

Technical Field

[0001] This invention relates to the field of lidar technology, and in particular to a lidar waveform recognition method and apparatus. Background Technology

[0002] Laser radar waveform data processing has been extensively studied by scholars both domestically and internationally. However, most research focuses on waveform data decomposition, waveform noise filtering, and applications of waveform data in polar and forest environments. Comparatively, research on waveform saturation phenomena in waveform data is relatively limited. However, saturation in echo waveforms is unavoidable, making waveform identification crucial.

[0003] Therefore, there is an urgent need for a laser radar waveform recognition method to solve the problem that existing technologies cannot quickly and accurately identify saturated waveforms. Summary of the Invention

[0004] In view of this, the present invention provides a laser radar waveform recognition method and device, the main purpose of which is to solve the current problem of being unable to quickly and accurately identify saturated waveforms.

[0005] To address the above problems, this application provides a laser radar waveform recognition method, comprising:

[0006] Acquire several waveform data to be identified;

[0007] For each waveform data, calculate one or more waveform feature values ​​from the following: rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor, to obtain the waveform feature set corresponding to each waveform data.

[0008] Using a Gaussian mixture model based on each waveform feature set, cluster each waveform data to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data;

[0009] Based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, it is identified that each waveform data in the first waveform category is a zero-return saturated waveform and each waveform data in the second waveform category is a non-saturated waveform, or it is identified that each waveform data in the first waveform category is a non-saturated waveform and each waveform data in the second waveform category is a zero-return saturated waveform.

[0010] Optionally, acquiring the waveform data to be identified specifically includes:

[0011] The full-waveform lidar data LAS file is parsed to obtain the aforementioned waveform data.

[0012] Optionally, the step of calculating the rising edge-falling edge slope factor corresponding to each waveform data specifically includes:

[0013] For the same waveform data, obtain the number of lines connecting the adjacent start and end points of the rising and falling edges corresponding to the waveform data, the slope of the lines connecting the adjacent start and end points of the rising and falling edges, and the minimum half-width.

[0014] Based on the number of lines connecting the adjacent start and end points of the rising and falling edges corresponding to each waveform data, the slope of the lines connecting the adjacent start and end points of the rising and falling edges, and the minimum half-wave width, the rising edge-falling edge slope factor corresponding to each waveform data is calculated using the first calculation formula.

[0015] Optionally, the step of calculating the jitter-intensity factor corresponding to each waveform data specifically includes:

[0016] For the same waveform data, the first intersection point and the maximum waveform intensity are determined by using the first amplitude threshold line corresponding to the waveform data.

[0017] Based on the number of intersections and the maximum waveform intensity corresponding to each waveform data, the jitter-intensity factor corresponding to each waveform data is calculated using the second calculation formula.

[0018] Optionally, the step of calculating the oscillation factor corresponding to each waveform data for each waveform data specifically includes:

[0019] For the same waveform data, the peaks and troughs in the waveform data are used to determine the corresponding maximum and minimum points, so as to determine the number of maximum and minimum points.

[0020] Based on the number of maximum and minimum points corresponding to each waveform data, the oscillation factor corresponding to each waveform data is calculated using the third calculation formula.

[0021] Optionally, the step of calculating the saturation jitter factor corresponding to each waveform data specifically includes:

[0022] For the same waveform data, using the second amplitude threshold line and the third threshold line corresponding to the waveform data, determine the first waveform intensity corresponding to the intersection of the waveform data and the second amplitude threshold line, the second waveform intensity corresponding to the intersection of the waveform data and the third amplitude threshold line, the first time corresponding to the first waveform intensity, the second time corresponding to the second waveform intensity, the number of second intersection points of the waveform data and the second amplitude threshold line, and the number of third intersection points of the waveform data and the third amplitude threshold line; the amplitude of the second amplitude threshold line is less than the amplitude of the third amplitude threshold line.

[0023] Based on the first waveform intensity, second waveform intensity, first time, second time, second number of intersections, and third number of intersections corresponding to each waveform data, the saturation jitter factor corresponding to each waveform data is calculated using the fourth calculation formula.

[0024] Optionally, the step of calculating the peak-area factor corresponding to each waveform data specifically includes:

[0025] For the same waveform data, determine the peak value of each peak, and use the third amplitude threshold line corresponding to the waveform data to determine the minimum area among the peak areas formed by the third amplitude threshold line and each peak.

[0026] Based on the peak value and minimum area of ​​each peak corresponding to each waveform data, the peak-area factor corresponding to each waveform data is calculated using the fifth calculation formula.

[0027] Optionally, the step of using a Gaussian mixture model to cluster the waveform data based on each waveform feature set to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data specifically includes:

[0028] Based on each of the waveform feature sets, using an initial Gaussian mixture model containing two Gaussian distributions, the initial probability of each waveform data belonging to each Gaussian distribution is calculated respectively, and the initial Gaussian mixture model is updated based on each of the initial probabilities to train and obtain the current Gaussian mixture model.

[0029] Based on the current Gaussian mixture model, the current probability of each waveform data belonging to each Gaussian distribution is reconstructed and calculated using each waveform feature set. The current Gaussian mixture model is then updated based on each current probability until the updated current Gaussian mixture model meets the predetermined training stopping condition, thereby obtaining the target Gaussian mixture model and the target probability of each waveform data belonging to each Gaussian distribution.

[0030] Based on the target probability of each waveform data belonging to each Gaussian distribution, each waveform data is classified to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

[0031] Optionally, the step of identifying each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identifying each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, specifically includes:

[0032] Based on the same waveform feature value corresponding to each waveform data in the first waveform category and the predetermined waveform feature threshold, the waveform data in the first waveform category are identified as zero-return saturated waveforms or non-saturated waveforms, and the waveform data in the second waveform category are identified as non-saturated waveforms or zero-return saturated waveforms.

[0033] Alternatively, based on the same waveform feature value corresponding to each waveform data in the second waveform category and a predetermined waveform feature threshold, each waveform data in the second waveform category can be identified as a zero-return saturated waveform or a non-saturated waveform, and each waveform data in the first waveform category can be identified as a non-saturated waveform or a zero-return saturated waveform.

[0034] To address the above problems, this application provides a laser radar waveform recognition device, comprising:

[0035] The acquisition module is used to acquire several waveform data to be identified;

[0036] The calculation module is used to calculate one or more waveform feature values, such as rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor, for each waveform data, so as to obtain the waveform feature set corresponding to each waveform data.

[0037] The clustering module is used to cluster the waveform data based on the waveform feature sets using a Gaussian mixture model to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

[0038] The identification module is used to identify, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or to identify each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform.

[0039] The lidar waveform recognition method in this application extracts waveform feature values ​​related to each waveform data, and then uses Gaussian mixture model clustering method to cluster these waveform feature values. This can quickly and accurately classify similar waveform data into the same category. Subsequently, based on the waveform feature values ​​corresponding to the waveform data in each category, it can be determined whether the waveform data in that category is a zero-return saturated waveform or a non-saturated waveform, thus solving the problem in the prior art that it is impossible to quickly and accurately identify saturated waveforms.

[0040] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0041] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0042] Figure 1 This is a flowchart of a laser radar waveform recognition method according to an embodiment of this application;

[0043] Figure 2 This is a structural block diagram of a laser radar waveform recognition device according to another embodiment of this application. Detailed Implementation

[0044] Various embodiments and features of this application are described herein with reference to the accompanying drawings.

[0045] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this application will be apparent to those skilled in the art.

[0046] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.

[0047] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.

[0048] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application.

[0049] The above and other aspects, features and advantages of this application will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.

[0050] Specific embodiments of this application are described thereafter with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of this application, which can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely serve as the basis and representative basis for the claims to teach those skilled in the art to use this application in a variety of substantially any suitable detailed structures.

[0051] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in other embodiments,” all of which may refer to one or more of the same or different embodiments according to this application.

[0052] This application provides a method for laser radar waveform recognition, such as... Figure 1 As shown, it includes the following steps:

[0053] Step S101: Obtain several waveform data to be identified;

[0054] In practice, this step involves parsing the full-waveform lidar data Las file to obtain several waveform data points. Specifically, based on the Las file format, code is used to acquire information such as coordinates and intensity, thereby obtaining several waveform data points, i.e., several lidar waveforms.

[0055] Step S102: For each waveform data, calculate any one or more waveform feature values ​​among the rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor corresponding to each waveform data, so as to obtain the waveform feature set corresponding to each waveform data.

[0056] In this step, to make the recognition more accurate and reliable, five waveform feature values ​​can be calculated simultaneously, which will facilitate subsequent waveform recognition based on the five waveform feature values.

[0057] Step S103: Using a Gaussian mixture model based on each waveform feature set, cluster each waveform data to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

[0058] Step S104: Based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, identify each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identify each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform.

[0059] The lidar waveform recognition method in this application extracts waveform feature values ​​related to each waveform data, and then uses Gaussian mixture model clustering method to cluster these waveform feature values. This can quickly and accurately classify similar waveform data into the same category. Subsequently, based on the waveform feature values ​​corresponding to the waveform data in each category, it can be determined whether the waveform data in that category is a zero-return saturated waveform or a non-saturated waveform, thus solving the problem in the prior art that it is impossible to quickly and accurately identify saturated waveforms.

[0060] Based on the above embodiments, another embodiment of this application provides a laser radar waveform recognition method. In this embodiment, the calculation process of each waveform feature value during step S102 is as follows:

[0061] (I) When calculating the waveform characteristic value of the Rise-Fall Slope Factor (RFSF), for the same waveform data, the number of lines connecting adjacent start and end points of the rising and falling edges, the slope of these lines, and the minimum half-width can be obtained. Based on the number of lines connecting adjacent start and end points of the rising and falling edges, the slope of these lines, and the minimum half-width, the Rise-Fall Slope Factor (RFSF) corresponding to each waveform data is calculated using the first calculation formula. The first calculation formula is as follows:

[0062]

[0063] Where RFSF represents the rising-falling edge slope factor; n represents the number of adjacent start and end points of the waveform (the number of adjacent start and end points of the rising and falling edges); K r The slope of the line connecting the adjacent start and end points of the r-th (r = 1, 2, 3, ..., n) rising and falling edges is represented by W; W represents the half-width at half-maximum (WW), defined as the width of the wave corresponding to half the peak value, where each peak value and its adjacent rising and falling edges form an echo. min This represents the minimum half-width among the various half-widths.

[0064] In this implementation, during the target measurement process by the full-waveform lidar, the waveform amplitude will return to zero if it exceeds the maximum intensity threshold limited by the radar system. Therefore, the waveform amplitude will jump between the threshold and zero, exhibiting an oscillating phenomenon. Consequently, the waveform will have multiple rising and falling edges, i.e., multiple starting and ending points for rising and falling edges. In addition, the relatively dense peak points of the waveform result in a small half-width. Therefore, the rising-falling edge slope factor, a waveform characteristic, can be calculated based on the number of lines connecting adjacent starting and ending points of rising and falling edges, the slope of the lines connecting adjacent starting and ending points of rising and falling edges, and the minimum half-width.

[0065] (II) When calculating the waveform characteristic value of Jitter-Amplitude Factor (JAF), for the same waveform data, the number of first intersection points and the maximum waveform intensity can be determined using the first amplitude threshold line corresponding to the waveform data. Based on the number of intersection points and the maximum waveform intensity corresponding to each waveform data, the JAF corresponding to each waveform data is calculated using the second calculation formula. The first amplitude threshold line is a threshold line parallel to the horizontal axis set at 20% of the maximum amplitude. In this embodiment, the second calculation formula is:

[0066] JAF=T*I max

[0067] Where T represents the number of first intersection points between the waveform data and the first amplitude threshold line (i.e., the number of times the waveform line crosses the first amplitude threshold line); I represents the intensity value of the waveform; I max This indicates the maximum waveform intensity.

[0068] In this embodiment, since the amplitude of the zero-return saturated waveform will return to zero when it exceeds the maximum intensity threshold limited by the radar system, the waveform amplitude will jump between the threshold and zero, exhibiting an oscillating phenomenon. A first amplitude threshold line parallel to the horizontal axis is set at 20% of the maximum amplitude. This line will cross the waveform multiple times, and the maximum intensity value of the zero-return saturated waveform will not be less than that of the unsaturated waveform. Based on these characteristics, the jitter-intensity factor waveform feature can be calculated using the number of first intersection points and the maximum waveform intensity.

[0069] (III) When calculating the oscillation intensity factor (OIF), a waveform characteristic value, for the same waveform data, the peaks and troughs in the waveform data can be used to determine the corresponding maxima and minima, thus determining the number of maxima and minima. Based on the number of maxima and minima corresponding to each waveform data, the oscillation intensity factor corresponding to each waveform data is calculated using the third calculation formula. The third calculation formula is as follows:

[0070]

[0071] Where a represents the number of maxima in the waveform; b represents the number of minima in the waveform.

[0072] In this embodiment, since the amplitude of the zero-return saturated waveform will return to zero when it exceeds the maximum intensity threshold limited by the radar system, the waveform amplitude will jump between the threshold and zero, exhibiting an oscillating phenomenon. Therefore, the waveform will have multiple peaks and troughs, i.e., multiple maxima and minima. Based on these characteristics, the oscillation factor, a waveform feature, can be calculated using the number of maxima and minima.

[0073] (iv) When calculating the saturation jitter factor (SJF), a waveform characteristic value, for the same waveform data, the second amplitude threshold line and the third amplitude threshold line corresponding to the waveform data can be used to determine the first waveform intensity corresponding to the intersection of the waveform data and the second amplitude threshold line, the second waveform intensity corresponding to the intersection of the waveform data and the third amplitude threshold line, the first time corresponding to the first waveform intensity, the second time corresponding to the second waveform intensity, the number of second intersection points of the waveform data and the second amplitude threshold line, and the number of third intersection points of the waveform data and the third amplitude threshold line; the amplitude of the second amplitude threshold line is less than the amplitude of the third amplitude threshold line; based on the first waveform intensity, second waveform intensity, first time, second time, number of second intersection points, and number of third intersection points corresponding to each waveform data, the saturation jitter factor corresponding to each waveform data is calculated using the fourth calculation formula. The second amplitude threshold line is a threshold line parallel to the horizontal axis set at 10% of the maximum amplitude. The third amplitude threshold line is a threshold line parallel to the horizontal axis set at 90% of the maximum amplitude. The fourth calculation formula is:

[0074]

[0075] Where i represents the intensity value at the intersection of the waveform line and the second amplitude threshold line, i.e., the first waveform intensity. I represents the intensity value at the intersection of the waveform line and the third amplitude threshold line, i.e., the second waveform intensity. t represents the time value on the horizontal axis corresponding to the intersection of the waveform line and the second amplitude threshold line, i.e., the first time. T represents the time value on the horizontal axis corresponding to the intersection of the waveform line and the third amplitude threshold line, i.e., the second time. m represents the number of intersections between the waveform line and the third amplitude threshold line;

[0076] q represents the number of intersections between the waveform line and the second amplitude threshold line. I1 represents the intensity value at the l-th (l = 1, 2, 3, ..., m) intersection point between the waveform line and the third amplitude threshold line; i h This represents the intensity value at the h-th (h = 1, 2, 3, ..., q) intersection point of the line connecting the waveforms and the second amplitude threshold line. |Tt| min This represents the minimum difference between the second time and the first time.

[0077] In this embodiment, since the amplitude of a zero-return saturated waveform will return to zero when it exceeds the maximum intensity threshold limited by the radar system, the waveform amplitude will jump between the threshold and zero, exhibiting an oscillating phenomenon. A threshold line parallel to the horizontal axis is set at 10% and 90% of the maximum amplitude. These two threshold lines will repeatedly cross the line connecting the waveforms, resulting in multiple intersection points. The values ​​of the horizontal and vertical axes at these intersection points can then be extracted. Based on these characteristics, the waveform feature of saturation jitter factor can be calculated.

[0078] (V) When calculating the peak-area factor (PAF), a waveform characteristic value, the peak value of each peak can be determined for the same waveform data. The minimum area among the peak areas formed by the third amplitude threshold line corresponding to the waveform data and each peak can be determined. Based on the peak value and minimum area of ​​each peak corresponding to each waveform data, the peak-area factor corresponding to each waveform data is calculated using the fifth calculation formula. The third amplitude threshold line is a threshold line parallel to the horizontal axis set at 90% of the maximum amplitude. The fifth calculation formula is as follows:

[0079]

[0080] Where V represents the peak value of the wave. c This represents the peak value of the c-th peak in the waveform.

[0081] p represents the number of peaks generated due to the waveform returning to zero. ε represents the area formed by the line connecting the portion above the third amplitude threshold line and the waveform, that is, the area of ​​the peaks formed by the third amplitude threshold line and the peaks.

[0082] εmin It represents the smallest area among the areas formed by the line connecting the part above the third amplitude threshold line and the waveform, that is, the smallest area among the areas of each wave peak.

[0083] In this embodiment, since the amplitude of the zero-return saturated waveform will return to zero when it exceeds the maximum intensity threshold limited by the radar system, the waveform amplitude will jump between the threshold and zero, exhibiting an oscillating phenomenon. Therefore, multiple peaks will appear in the waveform, corresponding to multiple peak values. A threshold line parallel to the horizontal axis (i.e., the third amplitude threshold line) is set at 90% of the maximum amplitude, and this threshold line intersects with the waveform line. Based on the above characteristics, the peak-area factor waveform feature can be calculated.

[0084] In this embodiment, by using the above method, the waveform feature values ​​can be calculated quickly and accurately, laying the foundation for subsequent clustering of waveform data based on waveform feature values.

[0085] Based on the above embodiments, another embodiment of this application provides a laser radar waveform recognition method. In this embodiment, when performing step S103, a Gaussian mixture model is used to cluster each waveform data based on each waveform feature set to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data. The specific process is as follows:

[0086] Based on the waveform feature sets, an initial Gaussian mixture model (GMM) containing two Gaussian distributions is used to calculate the initial probability that each waveform data belongs to each Gaussian distribution. The initial GMM is then updated based on these initial probabilities to train and obtain the current GMM. Based on the current GMM, the current probability that each waveform data belongs to each Gaussian distribution is reconstructed using the waveform feature sets. The current GMM is then updated again based on these current probabilities until the updated GMM meets a predetermined training stopping condition, thus obtaining the target GMM and the target probability that each waveform data belongs to each Gaussian distribution. Based on the target probability that each waveform data belongs to each Gaussian distribution, the waveform data is categorized to obtain a first waveform category containing several waveform data points and a second waveform category containing several waveform data points.

[0087] In other words, clustering specifically includes the following three steps:

[0088] Step 1: Preprocess the waveform data to better describe the data distribution.

[0089] Step 2: Initialize Gaussian Mixture Model Parameters: Select a cluster size of 2, i.e., two Gaussian distributions. Randomly initialize the parameters of the two Gaussian distributions, including the mean, covariance matrix, and mixing coefficients.

[0090] Step 3: Iterative optimization using the Expectation-Maximization (EM) algorithm: Estimate the parameters of the Gaussian mixture model iteratively.

[0091] Expectation step (E step): Calculate the posterior probability of each sample belonging to each Gaussian distribution. Based on the current Gaussian distribution parameters and Bayes' theorem, calculate the posterior probability of each sample corresponding to each Gaussian distribution.

[0092] Maximization step (M step): Update the parameters of the Gaussian distribution based on the posterior probabilities calculated in the E step. Specifically, update the mean, covariance matrix, and mixing coefficients of each Gaussian distribution using a weighted average of the posterior probabilities of all samples. Repeat the E and M steps until the model converges or the predetermined number of iterations is reached.

[0093] After the iterations are complete, each sample has a posterior probability corresponding to each Gaussian distribution. The cluster affiliation of each sample can be determined based on the posterior probability, i.e., clustering is complete.

[0094] In this embodiment, by using a Gaussian mixture model to cluster each waveform data using the waveform feature set corresponding to each waveform data, the clustering results can be made more accurate. This allows waveform data with similar characteristics to be grouped into one category, laying the foundation for subsequent determination of whether the waveform data in each category is a zero-return saturated waveform or a non-saturated waveform.

[0095] Based on the above embodiments, another embodiment of this application provides a laser radar waveform recognition method. In this embodiment, based on the waveform feature values ​​corresponding to each waveform data in a first waveform category and a second waveform category, the method identifies each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identifies each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform. The specific process is as follows: For the same waveform feature, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, a first mean value corresponding to the first waveform category and a second mean value corresponding to the second waveform category are determined; based on the first mean value corresponding to the first waveform category and the second mean value corresponding to the second waveform category, the method identifies each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identifies each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform.

[0096] In other words, after clustering to obtain the first and second categories, the first category is assigned the label "1" and the second category is assigned the label "2". Then, the mean of the rising-falling slope factor "RFSF" in labels "1" and "2" can be further calculated; the mean of the first feature corresponding to label "1" and the mean of the second feature corresponding to label "2" can be obtained.

[0097] Step 2: Determine the mean of the first feature corresponding to label "1" and the mean of the second feature corresponding to label "2". If the mean of "RFSF" (mean of the first feature) corresponding to "1" is larger, then "1" is a zero-return saturated waveform and "2" is a non-saturated waveform; if the mean of "RFSF" (mean of the first feature) corresponding to "1" is smaller, then "1" is a non-saturated waveform and "2" is a zero-return saturated waveform. Similarly, the mean value of the first feature corresponding to the first category and the mean value of the second feature corresponding to the second category can also be obtained by calculating the mean values ​​of other waveform feature values. For example, the mean values ​​of the jitter-intensity factor corresponding to the first category and the jitter-intensity factor corresponding to the second category can be calculated to obtain the mean values ​​of the first and second features; or the mean values ​​of the oscillation factor corresponding to the first category and the oscillation factor corresponding to the second category can be calculated to obtain the mean values ​​of the first and second features; or the mean values ​​of the saturation jitter factor corresponding to the first category and the saturation jitter factor corresponding to the second category can be calculated to obtain the mean values ​​of the first and second features; or the mean values ​​of the peak-area factor corresponding to the first category and the peak-area factor corresponding to the second category can be calculated to obtain the mean values ​​of the first and second features. Finally, the waveform data in the first and second categories can be determined as follows based on the mean values ​​of the first and second features corresponding to the first and second categories.

[0098] In this embodiment, classification results are obtained for zero-return saturated waveforms and unsaturated waveforms. The classification of unsaturated waveforms and zero-return saturated waveforms is based on features with smaller and larger eigenvalues, while the overall clustering result is based on cluster labels for five features. This classification method helps to better understand and analyze waveform data, providing classification results based on eigenvalues ​​and overall features.

[0099] The lidar waveform recognition method in this embodiment, by identifying saturated waveforms, can help promptly detect anomalies in signals or systems, such as hardware failures or signal distortion. Accurate identification of saturated waveforms improves the accuracy of fault detection and allows for appropriate maintenance measures to ensure normal system operation. Furthermore, the presence of saturated waveforms often leads to signal quality degradation, such as increased distortion and noise. Identifying saturated waveforms allows for signal quality assessment and optimization of signal processing algorithms based on the identification results, thereby improving signal quality and accuracy. Saturated waveform identification provides crucial information for adaptive signal processing algorithms. By identifying saturated waveforms, algorithm parameters can be dynamically adjusted or corresponding control strategies can be adopted to adapt to signal changes, improving the system's adaptability and robustness. In addition, saturated waveform identification can help detect anomalies in signals, such as illegal interference, attacks, or data tampering. Timely detection and identification of saturated waveforms enhances system security and protects the integrity of communication data and information. Therefore, research on saturated waveform identification has significant theoretical and practical implications and can provide strong support for applications in related fields. Saturated waveform identification is a broad research area with numerous techniques and tools available to support this task.

[0100] Based on the above embodiments, another embodiment of this application provides a laser radar waveform recognition device, such as... Figure 2 As shown, it includes:

[0101] Acquisition module 11 is used to acquire several waveform data to be identified;

[0102] Calculation module 12 is used to calculate one or more waveform feature values ​​among rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor for each waveform data, so as to obtain a waveform feature set corresponding to each waveform data.

[0103] Clustering module 13 is used to cluster each waveform data based on each waveform feature set using a Gaussian mixture model to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

[0104] The identification module 14 is used to identify, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or to identify each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform.

[0105] In this embodiment, the acquisition module is specifically used to: parse the full-waveform lidar data Las file to obtain the several waveform data.

[0106] In this embodiment, the calculation module is specifically used to: for the same waveform data, obtain the number of lines connecting adjacent start and end points of the rising and falling edges corresponding to the waveform data, the slope of the lines connecting adjacent start and end points of the rising and falling edges, and the minimum half-wave width; based on the number of lines connecting adjacent start and end points of the rising and falling edges corresponding to each waveform data, the slope of the lines connecting adjacent start and end points of the rising and falling edges, and the minimum half-wave width, calculate the rising edge-falling edge slope factor corresponding to each waveform data using the first calculation formula.

[0107] In this embodiment, the calculation module is specifically used to: for the same waveform data, using the first amplitude threshold line corresponding to the waveform data, determine the first intersection point and the maximum waveform intensity of the waveform data intersecting with the first amplitude threshold line; based on the number of intersection points and the maximum waveform intensity corresponding to each waveform data, calculate the jitter-intensity factor corresponding to each waveform data using the second calculation formula.

[0108] In this embodiment, the calculation module is specifically used to: for the same waveform data, use the peaks and troughs in the waveform data to determine the maximum and minimum points corresponding to the waveform data, so as to determine the number of maximum and minimum points; based on the number of maximum and minimum points corresponding to each waveform data, use the third calculation formula to calculate the oscillation factor corresponding to each waveform data.

[0109] In this embodiment, the calculation module is specifically used to: for the same waveform data, using the second amplitude threshold line and the third threshold line corresponding to the waveform data, determine the first waveform intensity corresponding to the intersection of the waveform data and the second amplitude threshold line, the second waveform intensity corresponding to the intersection of the waveform data and the third amplitude threshold line, the first time corresponding to the first waveform intensity, the second time corresponding to the second waveform intensity, the number of second intersection points of the waveform data and the second amplitude threshold line, and the number of third intersection points of the waveform data and the third amplitude threshold line; the amplitude of the second amplitude threshold line is less than the amplitude of the third amplitude threshold line; based on the first waveform intensity, second waveform intensity, first time, second time, number of second intersection points, and number of third intersection points corresponding to each waveform data, calculate the saturation jitter factor corresponding to each waveform data using the fourth calculation formula.

[0110] In this embodiment, the calculation module is specifically used to: determine the peak value of each peak for the same waveform data, and determine the minimum area among the peak areas formed by the third amplitude threshold line corresponding to the waveform data and each peak using the third amplitude threshold line and each peak; based on the peak value and minimum area of ​​each peak corresponding to each waveform data, calculate the peak-area factor corresponding to each waveform data using the fifth calculation formula.

[0111] In this embodiment, the clustering module is specifically used for: calculating the initial probability that each waveform data belongs to each Gaussian distribution based on each waveform feature set and using an initial Gaussian mixture model containing two Gaussian distributions; updating the initial Gaussian mixture model based on each initial probability to train and obtain the current Gaussian mixture model; reconstructing and calculating the current probability that each waveform data belongs to each Gaussian distribution based on the current Gaussian mixture model and each waveform feature set; updating the current Gaussian mixture model again based on each current probability until the updated current Gaussian mixture model meets a predetermined training stopping condition to obtain the target Gaussian mixture model and the target probability that each waveform data belongs to each Gaussian distribution; classifying each waveform data based on the target probability that each waveform data belongs to each Gaussian distribution to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

[0112] In this embodiment, the identification module is used to: for the same waveform feature, determine the first mean value corresponding to the first waveform category and the second mean value corresponding to the second waveform category based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category; based on the first mean value corresponding to the first waveform category and the second mean value corresponding to the second waveform category, identify each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identify each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform.

[0113] The lidar waveform recognition device in this embodiment extracts waveform feature values ​​related to each waveform based on each waveform data, and then uses the Gaussian mixture model clustering method to cluster these waveform feature values. This can quickly and accurately classify similar waveform data into the same category. Subsequently, based on the waveform feature values ​​corresponding to the waveform data in each category, it can be determined whether the waveform data in that category is a zero-return saturated waveform or a non-saturated waveform, thus solving the problem in the prior art that it is impossible to quickly and accurately identify saturated waveforms.

[0114] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A laser radar waveform recognition method, characterized in that, include: Acquire several waveform data to be identified; For each waveform data, calculate one or more waveform feature values ​​from the following: rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor, to obtain the waveform feature set corresponding to each waveform data. Using a Gaussian mixture model based on each waveform feature set, cluster each waveform data to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data; Based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, it is identified that each waveform data in the first waveform category is a zero-return saturated waveform and each waveform data in the second waveform category is a non-saturated waveform, or it is identified that each waveform data in the first waveform category is a non-saturated waveform and each waveform data in the second waveform category is a zero-return saturated waveform. The step of using a Gaussian mixture model to cluster the waveform data based on each waveform feature set to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data specifically includes: Based on each of the waveform feature sets, using an initial Gaussian mixture model containing two Gaussian distributions, the initial probability of each waveform data belonging to each Gaussian distribution is calculated respectively, and the initial Gaussian mixture model is updated based on each of the initial probabilities to train and obtain the current Gaussian mixture model. Based on the current Gaussian mixture model, the current probability of each waveform data belonging to each Gaussian distribution is reconstructed and calculated using each waveform feature set. The current Gaussian mixture model is then updated based on each current probability until the updated current Gaussian mixture model meets the predetermined training stopping condition, thereby obtaining the target Gaussian mixture model and the target probability of each waveform data belonging to each Gaussian distribution. Based on the target probability of each waveform data belonging to each Gaussian distribution, each waveform data is classified to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

2. The method as described in claim 1, characterized in that, The acquisition of several waveform data to be identified specifically includes: The full-waveform lidar data LAS file is parsed to obtain the aforementioned waveform data.

3. The method as described in claim 1, characterized in that, The step of calculating the rising edge-falling edge slope factor for each waveform data specifically includes: For the same waveform data, obtain the number of lines connecting the adjacent start and end points of the rising and falling edges corresponding to the waveform data, the slope of the lines connecting the adjacent start and end points of the rising and falling edges, and the minimum half-width. Based on the number of lines connecting adjacent start and end points of rising and falling edges corresponding to each waveform data, the slope of the lines connecting adjacent start and end points of rising and falling edges, and the minimum half-wave width, the rising edge-falling edge slope factor corresponding to each waveform data is calculated using the first calculation formula. The first calculation formula is: ； in, Indicates the rising-falling slope factor; This indicates the number of lines connecting adjacent start and end points of the waveform, i.e., the number of lines connecting adjacent start and end points of the rising and falling edges; Indicates the first The slope of the line connecting the adjacent starting and ending points of the rising and falling edges, that is, the slope of the line connecting the adjacent starting and ending points of the rising and falling edges. The half-width is defined as the width of a waveform that is half the width of a peak along with its adjacent rising and falling edges. This represents the minimum half-width among the various half-widths.

4. The method as described in claim 1, characterized in that, The step of calculating the jitter-intensity factor corresponding to each waveform data specifically includes: For the same waveform data, the first intersection point and the maximum waveform intensity are determined by using the first amplitude threshold line corresponding to the waveform data. Based on the number of intersections and the maximum waveform intensity corresponding to each waveform data, the jitter-intensity factor corresponding to each waveform data is calculated using the second calculation formula. The second calculation formula is: ； in, This represents the number of first intersection points where the waveform data intersects with the first amplitude threshold line, i.e., the number of times the waveform connection line crosses the first amplitude threshold line; Indicates the intensity value of the waveform; This indicates the maximum waveform intensity.

5. The method as described in claim 1, characterized in that, The calculation of the oscillation factor corresponding to each waveform data specifically includes: For the same waveform data, the peaks and troughs in the waveform data are used to determine the corresponding maximum and minimum points, so as to determine the number of maximum and minimum points. Based on the number of maximum and minimum points corresponding to each waveform data, the oscillation factor corresponding to each waveform data is calculated using the third calculation formula. The third calculation formula is: ； in, Indicates the number of maxima in the waveform; This indicates the number of local minima in the waveform.

6. The method as described in claim 1, characterized in that, The calculation of the saturation jitter factor corresponding to each waveform data specifically includes: For the same waveform data, using the second amplitude threshold line and the third threshold line corresponding to the waveform data, determine the first waveform intensity corresponding to the intersection of the waveform data and the second amplitude threshold line, the second waveform intensity corresponding to the intersection of the waveform data and the third amplitude threshold line, the first time corresponding to the first waveform intensity, the second time corresponding to the second waveform intensity, the number of second intersection points of the waveform data and the second amplitude threshold line, and the number of third intersection points of the waveform data and the third amplitude threshold line; the amplitude of the second amplitude threshold line is less than the amplitude of the third amplitude threshold line. Based on the first waveform intensity, second waveform intensity, first time, second time, second number of intersections, and third number of intersections corresponding to each waveform data, the saturation jitter factor corresponding to each waveform data is calculated using the fourth calculation formula. The fourth calculation formula is: ； in, The intensity value at the intersection of the line connecting the waveforms and the second amplitude threshold line is also known as the intensity of the first waveform. The intensity value at the intersection of the line connecting the waveforms and the third amplitude threshold line is also known as the intensity of the second waveform. The time value on the horizontal axis corresponding to the intersection of the line connecting the waveforms and the second amplitude threshold line, which is also the first time; The second time is the time value on the horizontal axis corresponding to the intersection of the line connecting the waveforms and the third amplitude threshold line. This indicates the number of intersections between the waveform line and the third amplitude threshold line; This indicates the number of intersections between the waveform line and the second amplitude threshold line; The line connecting the waveform and the third amplitude threshold line represents the first... Intensity values ​​at each intersection point; The line connecting the waveform and the second amplitude threshold line represents the first... Intensity values ​​at each intersection point; This represents the minimum difference between the second time and the first time.

7. The method as described in claim 1, characterized in that, The step of calculating the peak-area factor corresponding to each waveform data for each waveform data specifically includes: For the same waveform data, determine the peak value of each peak, and use the third amplitude threshold line corresponding to the waveform data to determine the minimum area among the peak areas formed by the third amplitude threshold line and each peak. Based on the peak value and minimum area of ​​each peak corresponding to each waveform data, the peak-area factor corresponding to each waveform data is calculated using the fifth calculation formula. The fifth calculation formula is: ； in, The peak value represents the peak of a wave. Indicates the first wave in the waveform The peak value of each wave; This indicates the number of peaks generated due to the waveform returning to zero. It represents the area formed by the line connecting the part above the third amplitude threshold line and the waveform, that is, the area of ​​the wave crest formed by the third amplitude threshold line and the wave crest. It represents the smallest area among the areas formed by the line connecting the part above the third amplitude threshold line and the waveform, that is, the smallest area among the areas of each wave peak.

8. The method as described in claim 1, characterized in that, The step of identifying each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or identifying each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, specifically includes: For the same waveform feature, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, determine the first mean value corresponding to the first waveform category and the second mean value corresponding to the second waveform category; Based on the first mean value corresponding to the first waveform category and the second mean value corresponding to the second waveform category, the waveform data in the first waveform category is identified as a zero-return saturated waveform and the waveform data in the second waveform category is identified as a non-saturated waveform, or the waveform data in the first waveform category is identified as a non-saturated waveform and the waveform data in the second waveform category is identified as a zero-return saturated waveform.

9. A laser radar waveform recognition device, characterized in that, include: The acquisition module is used to acquire several waveform data to be identified; The calculation module is used to calculate one or more waveform feature values, such as rising edge-falling edge slope factor, jitter-intensity factor, oscillation factor, saturation jitter factor, and peak-area factor, for each waveform data, so as to obtain the waveform feature set corresponding to each waveform data. The clustering module is used to cluster the waveform data based on the waveform feature sets using a Gaussian mixture model to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data. The identification module is used to identify, based on the waveform feature values ​​corresponding to each waveform data in the first waveform category and the second waveform category, each waveform data in the first waveform category as a zero-return saturated waveform and each waveform data in the second waveform category as a non-saturated waveform, or to identify each waveform data in the first waveform category as a non-saturated waveform and each waveform data in the second waveform category as a zero-return saturated waveform. The clustering module is specifically used for: Based on the waveform feature sets, the waveform data are clustered to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data, specifically including: Based on each of the waveform feature sets, using an initial Gaussian mixture model containing two Gaussian distributions, the initial probability of each waveform data belonging to each Gaussian distribution is calculated respectively, and the initial Gaussian mixture model is updated based on each of the initial probabilities to train and obtain the current Gaussian mixture model. Based on the current Gaussian mixture model, the current probability of each waveform data belonging to each Gaussian distribution is reconstructed and calculated using each waveform feature set. The current Gaussian mixture model is then updated based on each current probability until the updated current Gaussian mixture model meets the predetermined training stopping condition, thereby obtaining the target Gaussian mixture model and the target probability of each waveform data belonging to each Gaussian distribution. Based on the target probability of each waveform data belonging to each Gaussian distribution, each waveform data is classified to obtain a first waveform category containing several waveform data and a second waveform category containing several waveform data.

Images