Full-waveform laser radar target echo signal detection method and terminal

Abstract

The invention relates to a full-waveform laser radar target echo signal detection method and a terminal. The method comprises the steps of obtaining an echo signal, decomposing the echo signal to obtain a plurality of waveform components, performing feature extraction on each waveform component, and obtaining a waveform component feature corresponding to each waveform component; screening the waveform components according to the waveform component characteristics of the waveform components to obtain a target waveform component; and generating a point cloud according to the target waveform component. According to the method, the features of each waveform component of the echo signals are extracted, the echo signals are screened according to the features of each waveform component, and the screened echo signals are adopted to generate the point cloud, so that the extraction of the target point cloud is prevented from being interfered by non-target echo signals, and the existence of interference point clouds is reduced.

Description

Technical Field

[0001] The present invention relates to the field of laser signal processing technology, and in particular to a full-waveform laser radar target echo signal detection method, terminal and readable storage medium. Background Art

[0002] Please refer to Figure 1 As shown in the figure, airborne laser bathymetry (ALB) technology utilizes the transmission properties of the blue-green light band (450nm~580nm) and avoids the adverse effects of the operating environment on the regional environmental monitoring effect through active non-contact measurement. It has obvious advantages in describing the underwater topography and geomorphology characteristics of shallow sea areas, constructing three-dimensional underwater terrain models, and extracting the optical characteristics of water bodies.

[0003] Airborne laser bathymetry technology uses a laser radar system (hereinafter referred to as full-waveform laser radar) with digital full-waveform detection technology. By collecting and analyzing the continuous change characteristics of the laser echo waveform signal intensity during the transmission process, it can obtain the propagation time of multiple backscattered cross-section echo signals in the detection target area, and then obtain the target spatial position and the laser signal intensity change process through system calculation.

[0004] The full-waveform bathymetric laser echo signal completely records the spatial position of the target object's reflection cross section and water quality information during the laser propagation process, providing a reliable data basis for monitoring and analyzing regional environmental characteristics and changes. However, due to the complex spatial structure of the detection area, the non-uniformity of the propagation medium, and the error of the signal receiving system, the full-waveform echo signal usually contains different echo signal components such as target echoes of different intensities, water scattering, and noise interference. The laser signal intensity, waveform broadening, and time position of the components between different echo components are superimposed and coupled, making it difficult to distinguish the target reflection echo. This problem has a very adverse effect on the accurate calculation of the laser signal penetrating multiple surfaces, extracting the multi-target reflection components in the laser optical path, and solving the multi-target position in the optical path.

[0005] Please refer to Figure 2The usual processing method shown is to directly perform point cloud solution on the laser echo containing complex signal components, and transform the problem of screening the target echo components of the generated point cloud into a point cloud filtering or classification extraction problem. The spatial filtering algorithm is classified according to the different principles it is based on, such as slope-based, morphology-based, surface fitting-based, irregular triangulated network-based, etc. These methods are based on the relatively clear relative position of the point cloud and its spatial structural characteristics, and have good application effects in land target detection based on infrared laser radar. However, for the depth-sounding laser radar using the visible light band, the noise intensity or noise frequency in the signal is high, and when the spatial structural characteristics of the detection target are complex or distributed in multiple layers along the laser propagation direction, the extraction of the target point cloud is interfered by the non-target echo signal, resulting in the inability to effectively screen the target echo component. It relies on a large amount of manual interactive deletion, and it is difficult to accurately determine the spatial position of the target echo surface. Summary of the invention

[0006] The purpose of the present invention is to provide a full-waveform laser radar target echo signal detection method and terminal, aiming to solve the problem that the extraction of target point cloud in the prior art is interfered by non-target echo signals, resulting in the inability to effectively screen target echo components.

[0007] The technical solution adopted by the present invention to solve the technical problem is as follows: The present invention provides a full-waveform laser radar target echo signal detection method, the full-waveform laser radar target echo signal detection method comprising: Acquire an echo signal, decompose the echo signal to obtain a plurality of waveform components, extract features of each of the waveform components, and acquire waveform component features corresponding to each of the waveform components; Screening each of the waveform components according to the waveform component characteristics of each of the waveform components to obtain a target waveform component; A point cloud is generated based on the target waveform components.

[0008] Furthermore, the extracting features of each of the waveform components to obtain waveform component features corresponding to each of the waveform components specifically includes: Fitting each of the waveform components to obtain Gaussian parameters of each of the waveform components; Obtaining the peak intensity, wave width, peak time position and correlation coefficient of each of the waveform components according to each of the Gaussian parameters; The peak intensity, the wave width, the peak time position and the correlation coefficient are used as the waveform component features.

[0009] Further, the filtering of each of the waveform components according to the waveform component characteristics of each of the waveform components to obtain the target waveform component specifically includes: Obtaining characteristic function results of each of the waveform components according to the characteristics of each of the waveform components; Calculating the probability that each of the waveform components is a target waveform component and the probability that each of the waveform components is not a target waveform component according to the characteristic function results of each of the waveform components; Whether each waveform component is a target waveform component is determined based on the probability that each waveform component is a target waveform component and the probability that each waveform component is not a target waveform component.

[0010] Further, obtaining the characteristic function results of each waveform component according to the characteristics of each waveform component specifically includes: Acquire the wave width characteristic function result of each of the waveform components according to the wave width of each of the waveform components; Obtaining intensity characteristic function results of each of the waveform components according to the peak intensity of each of the waveform components; Obtaining similarity characteristic function results of each of the waveform components according to the correlation coefficient and peak time position of each of the waveform components; Acquire the transfer characteristic function result of each of the waveform components according to the peak intensity, wave width and peak time position of each of the waveform components; The wavelength characteristic function result, the intensity characteristic function result, the similarity characteristic function result and the transfer characteristic function result are taken as the characteristic function results.

[0011] Furthermore, the wave width characteristic function result of each waveform component is obtained according to the wave width of each waveform component, specifically, the wave width characteristic function result is obtained according to the following formula: ; in, is the characteristic function of the wave width, For the Waveform components The wave width, is the transmission waveform width, Indicates Waveform components The wave width characteristic function results.

[0012] Further, the obtaining of the intensity characteristic function result of each of the waveform components according to the peak intensity of each of the waveform components specifically includes: De-attenuating the peak intensity of each of the waveform components to obtain the de-attenuated peak intensity of each of the waveform components; The intensity characteristic function result of each of the waveform components is obtained according to the de-attenuated peak intensity of each of the waveform components:

[0013] in, Indicates waveform components The de-attenuated peak intensity, is the intensity characteristic function, represents the noise standard deviation, Indicates Waveform components The intensity characteristic function results of .

[0014] Further, the obtaining of similarity characteristic function results of each of the waveform components according to the correlation coefficient and peak time position of each of the waveform components specifically includes: Obtain the time position of the maximum value of the correlation coefficient in the correlation coefficient of each of the waveform components ; The similarity characteristic function results of each of the waveform components are obtained according to the time position of the maximum value of the correlation coefficient and the peak time position of each of the waveform components: ; in, is the similarity feature function, is the time position of the maximum value of the correlation coefficient, is the transmission waveform width, For the Waveform components The peak time position of is the transmission waveform width, Indicates Waveform components The similarity feature function result of .

[0015] Further, the step of obtaining the transfer characteristic function result of each of the waveform components according to the peak intensity, wave width and peak time position of each of the waveform components specifically includes: According to the peak intensity of each waveform component, the peak intensity ratio of adjacent echo components of each waveform component is obtained: ; in, For the Waveform components The peak intensity ratio of adjacent echo components, For the Waveform components The peak intensity, For the Waveform components The peak intensity of Obtaining the first multiple of each of the waveform components according to the wave width of each of the waveform components: ; in, For the Waveform components The first multiple of For the Waveform components The wave width, For the Waveform components The wave width, is the transmitting waveform width; Obtain the second multiple of each of the waveform components according to the peak time position of each of the waveform components: ; Among them, the For the Waveform components The second multiple of For the Waveform components The peak time position of For the Waveform components The peak time position of Acquire a first critical point for resolution and a second critical point for resolution of each of the waveform components according to the first multiple and the second multiple of each of the waveform components; Obtain the resolution parameters of each of the waveform components according to the peak intensity ratio of adjacent echo components, the first multiple, the second multiple, the first resolution critical point and the second resolution critical point of each of the waveform components: ; in, Indicates Waveform components The resolution parameters, is the resolution model coefficient, represents the first critical resolution point, represents the second critical resolution point; According to the resolution parameter, the second multiple, the first resolution critical point and the second resolution critical point of each of the waveform components, the transfer characteristic function result of each of the waveform components is obtained: ; in, represents the transfer characteristic function, Indicates Waveform components The transfer characteristic function result is Indicates Waveform components The transfer characteristic function results.

[0016] In addition, to achieve the above-mentioned purpose, the present invention also provides a terminal, which includes: a memory, a processor, and a full-waveform lidar target echo signal detection program stored in the memory and run on the processor. When the full-waveform lidar target echo signal detection program is executed by the processor, it controls the terminal to implement the steps of the full-waveform lidar target echo signal detection method as described above.

[0017] In addition, to achieve the above-mentioned purpose, the present invention also provides a readable storage medium, which stores a full-waveform lidar target echo signal detection program. When the full-waveform lidar target echo signal detection program is executed by a processor, the steps of the full-waveform lidar target echo signal detection method as described above are implemented.

[0018] The present invention adopts the above technical solution to achieve the following effects: The present invention extracts the characteristics of each waveform component of the echo signal, then screens the echo signal according to the characteristics of each waveform component, and uses the screened echo signal to generate a point cloud, thereby preventing the extraction of the target point cloud from being interfered by non-target echo signals and reducing the existence of interference point clouds. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a schematic diagram of the integrated land and water detection of laser radar in the prior art; Figure 2 This is a schematic diagram of the water body scattering interference point cloud; Figure 3 It is a flowchart of the steps of a full-waveform laser radar target echo signal detection method in a preferred embodiment of the present invention; Figure 4 It is a data processing flow chart of a full-waveform laser radar target echo signal detection method in a preferred embodiment of the present invention; Figure 5 It is a schematic diagram of fitting of the transmitting waveform and the receiving waveform in a preferred embodiment of the present invention; Figure 6 It is a schematic diagram of target echo signal marking and identification in a preferred embodiment of the present invention; Figure 7 It is a schematic diagram of training data processing and training of the conditional random field model in a preferred embodiment of the present invention; Figure 8 Schematic diagram of a first cross section and a second cross section in a preferred embodiment of the present invention; Fig. 9 This is a schematic diagram of the comparison results of the land-water interface area in a preferred embodiment of the present invention; Fig.10 is a schematic diagram of the comparison results of the water area in a preferred embodiment of the present invention; Fig.11 A schematic structural diagram of a preferred embodiment of the terminal of the present invention. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0021] Embodiment 1 See also Figure 3 and Figure 4 Embodiment 1 of the present application is a full-waveform laser radar target echo signal detection method, comprising the steps of: S1. Acquire an echo signal, decompose the echo signal to obtain multiple waveform components, extract features from each of the waveform components, and acquire waveform component features corresponding to each of the waveform components.

[0022] In this embodiment, in order to ensure that the spatial detection range is expanded as much as possible under the condition of ensuring the detection capability of the system, the airborne depth-sounding laser radar system records signal strength data of a longer time series to ensure the reception of target echo signals within a larger distance range during the laser propagation process, thereby improving the detection range of the system in the ranging direction. The signal intensity time series recorded at a specific frequency is the full-waveform laser echo waveform data. After the full-waveform laser echo waveform data is obtained, the digital signal is preprocessed to obtain the echo signal.

[0023] The preprocessing includes two steps: extracting the effective echo signal segment and filtering out the random noise of the echo signal.

[0024] Among them, intercepting the signal segment reflecting the target echo according to the signal change characteristics helps to reduce the data redundancy in the original signal collected by the receiving system, and ensures that the relevant processing methods focus on the signal segment containing the target echo. In this embodiment, a signal interception method based on signal statistical distribution is specifically adopted, which can improve the accurate extraction of echo signals and the integrity of signal segments.

[0025] After completing the effective echo signal segment interception and obtaining the effective echo signal segment, the signal random noise is filtered out of the effective echo signal segment. Specifically, the echo waveform can be considered as the superposition of the echo signal obtained from different reflection cross sections in the same propagation path and Gaussian white noise. If the echo pulse energy of the returned effective echo signal segment is assumed to obey the Gaussian distribution, the full waveform data can be expressed as follows: ; in, express Full waveform data at the moment, express The moment Waveform components, Indicates Waveform components The parameters of the Gaussian function , and A collection of express Obey the set , express Gaussian white noise at time, express Obey is a Gaussian distribution with parameters, Indicates the total number of waveform components.

[0026] According to the laser radar echo signal model, the echo signals recorded by different receiving systems usually have different signal contents, and are affected by environmental scattering or the uncertainty generated by the system's own circuits, resulting in the echo waveform usually containing a large number of "pseudo-signals". In order to ensure the subsequent laser echo signal processing results, the high-frequency noise in the signal can first be filtered as a whole by low-pass filtering to reduce the high-frequency noise interference in the signal and obtain the echo signal.

[0027] After that, feature extraction is performed on the echo signal to obtain echo signal features. First, the echo signal is decomposed. According to the representation of the full waveform data, the laser echo waveform can be decomposed into a combination of multiple functions by using a waveform decomposition method after preprocessing. In this embodiment, a mixed increasing Gaussian function is specifically used to decompose the waveform. Different waveform components are separated in sequence according to the order in which the echo signals are received, thereby reducing the peak offset problem caused by waveform superposition.

[0028] The advantage of waveform decomposition is that the characteristics of each component can be decomposed into quantitative parameters, which can be used to identify the target echo component in the full waveform. The specific value of the parameter reflects the special properties of the target reflected echo in terms of optical response to a certain extent. Referring to the formula of the effective echo signal segment above, there are three related parameters in the echo component: , and , the present invention uses Represents a set of three parameters, so obtaining more accurate waveform components means obtaining more accurate In order to improve the fitting accuracy of the waveform components, the present invention uses the Levenberg-Marquardt (LM) method to optimize the parameters of the waveform components. , to improve the decomposition accuracy of each waveform component. The LM method is based on the shape of the full waveform signal characteristics, so it can achieve better fitting accuracy. The LM algorithm obtains more accurate parameters through multiple iterations, and the formula for each iteration is as follows: ; in, Indicates the first The parameters of the waveform components, Indicates the first The parameters of the waveform components, represents the Jacobian matrix of the Gaussian mixture model, represents the damping coefficient, represents the diagonal identity matrix, Indicates the received The true value of the waveform components, Indicates transpose.

[0029] The termination condition of the iteration is , is to set the optimization termination threshold. Finally, the optimized Three parameters of the waveform components , and .

[0030] Considering that the received signal is essentially the convolution of the backscattering cross-section action functions of the transmitted signal at different propagation stages, the echo component after the decomposition of the full waveform echo signal should have a high correlation with the transmitted waveform. Therefore, the correlation characteristics between the target echo component and the transmitted waveform can be applied to the judgment of the target echo component, that is, the ASDF method is used for calculation, and the similarity coefficient between the echo waveform and the transmitted waveform is used as the response characteristic of the target echo component.

[0031] by Represents the emission waveform, and the correlation coefficient between the blue-green lidar emission signal and the echo signal is calculated using the mean square difference discriminant method (ASDF): ; in, For the Waveform components The correlation coefficient of For the The transmitted waveform at the time of sampling, represents the sampling interval, Indicates the number of sampling times, express The waveform component , Indicates the first Subsampling, .

[0032] In summary, we get the Waveform components Echo response characteristics , , and .

[0033] Waveform decomposition will generate many potential echoes along the laser propagation path, but only some of them are related to the target, and the others may be caused by other uncertain factors. Therefore, it is necessary to set a characteristic function according to the response characteristics of the target echo signal for judging the target echo. In this embodiment, the characteristic function is constructed from the characteristics of the echo component intensity, waveform width, the time position relationship of adjacent components, and the correlation of the transmitted signal to obtain the state label of the echo component. Please refer to Figure 5 , Figure 5 (a) is the result of Gaussian fitting of the emission waveform. represents the emission model signal, Indicates the peak intensity of the transmitted waveform, Indicates the peak time position of the emission waveform, are the parameters of the transmitting waveform, is the transmission waveform width. According to the result of Gaussian fitting of the transmission waveform, the peak intensity, width and time position of each waveform component are obtained. The width is expressed as 2 times the mean error. Figure 5 (b) is the result of waveform decomposition of the echo signal, and Gaussian fitting is also performed to obtain the peak intensity, wave width and time position.

[0034] The peak intensity, wave width, time position and correlation coefficient are used as waveform component features.

[0035] S2. Screening each of the waveform components according to the waveform component characteristics of each of the waveform components to obtain a target waveform component.

[0036] The waveform component characteristics are brought into the wave width characteristic function, the intensity characteristic function, the similarity characteristic function and the transfer characteristic function to obtain the output of each characteristic function.

[0037] Among them, regarding the wave width characteristic function, the ALB system usually adopts the method of recording the laser emission waveform in order to record the accurate emission state of the laser and the laser pulse emission time. By processing the original emission waveform, the accurate laser emission time can be obtained, and the target echo component can be screened by the waveform width of the emission waveform: ; in, is the characteristic function of the wave width, for The pulse width of the target response is due to pulse broadening caused by target scattering. Should be wider than the transmitted wave If the width of the echo component is less than or equal to the transmit waveform width, the echo component is considered not to be the target echo component, and the corresponding width label .

[0038] Regarding the intensity characteristic function, since the signal propagates in the propagation medium, its return energy often exhibits exponential decay. In order to use the signal intensity to screen the target echo component, it is necessary to first remove the influence of the exponential decay trend in the continuously recorded full waveform signal on the intensity change analysis, and then remove the first echo component. The signal segment between the first maximum value and the last maximum value of the waveform after the echo component intensity is greater than the threshold is fitted with an exponential function, and the fitting result is removed as the approximate trend of signal attenuation. The echo component with an intensity within the confidence interval is the time position where the target echo may exist. The effective target echo should be significantly greater than the background noise. The noise standard deviation is estimated using the data at the tail of the full waveform signal If the de-attenuated peak intensity of the waveform component is greater than If it is three times larger, it means that the corresponding echo signal has a 99.73% confidence probability of being a non-noise signal.

[0039] The intensity characteristic function of the waveform component is: ; in, Indicates waveform components The de-attenuated peak intensity, is the intensity characteristic function.

[0040] Regarding the similarity characteristic function, the time position corresponding to the extreme value in the correlation coefficient sequence is the possible position of the target echo. In order to avoid the influence of noise in the signal, improve the accuracy of target position judgment, and consider the actual data processing effect, this embodiment uses the positions corresponding to the first three extreme values ​​of the maximum correlation coefficient as the time position where the target echo may appear, and the similarity characteristic function is: ; in, is the similarity feature function, is the time position of the maximum value of the correlation coefficient, is the waveform component after decomposition Peak time position. Although the correlation coefficient sequence can eliminate the influence of noise, it also causes a certain delay in the corresponding position of the echo. Based on the above analysis, the transmission waveform width is used here. as the confidence radius.

[0041] Regarding the transfer characteristic function, since the interval between adjacent waveform components is relatively close, whether they can be distinguished by signal detection is usually significantly correlated with the echo characteristics between the adjacent waveform components. In the past, in full waveform signal processing, the intensity, wave width and relative time interval of the echo components were usually examined as separate characteristics. However, in fact, there is a significant coupling effect between adjacent echo components with different signal characteristics. In order to improve the accuracy of target echo signal screening under waveform superposition conditions, this embodiment studies the problem of decreased echo resolution caused by mutual superposition between adjacent echo components.

[0042] In order to conduct quantitative analysis, the relative relationship between the characteristics of adjacent echo components was constructed to measure whether the adjacent components are distinguishable. The three indicators are the ratio of the peak intensities of adjacent echo components, the ratio of the wave widths of adjacent echo components relative to the transmitted wave width, and the ratio of the peak intensities of adjacent echo components to the transmitted wave width. The multiple of the peak position time interval of adjacent echo components is relative to multiples of.

[0043] Among them, the waveform component The peak intensity ratio of adjacent echo components for: ; in, is the waveform component The peak intensity, is the waveform component The peak intensity.

[0044] Waveform Components The width ratio of adjacent echo components relative to the transmitted width The first multiple of for: ; in, is the waveform component The wave width, is the waveform component The wave width.

[0045] Waveform Components The time interval between the peak positions of adjacent echo components is relative to the transmission wave width. The second multiple of for: ; in, is the waveform component The peak time position of is the waveform component The peak time position.

[0046] As the above three indicators change, the distinguishability of adjacent echo components also changes. This section constructs a transfer characteristic function to characterize the actual distinguishability of adjacent target echo components in a one-dimensional signal. In practice, the state characteristic and The two-dimensional space composed of different intensity ratio conditions After the adjacent echo components are superimposed, the waveform is first-order differentiated to extract the waveform peak value, and the two peaks are obviously and The value of is taken as the critical condition for the resolution of echo components under specific intensity ratio conditions. The piecewise equation is used to simulate the critical condition. Curve. and The two-dimensional space composed of the waveform components can be divided into three sections, namely , and .

[0047] , under certain intensity conditions, the current waveform component and the previous echo component merge with each other and become indistinguishable. is the first critical resolution point.

[0048] :The coupling relationship between echo component characteristics no longer has a significant impact on the resolution of echo components, and adjacent components can be clearly distinguished. is the second critical resolution point.

[0049] :The echo component recognition varies with the difference in the intensity and width of adjacent superimposed waveforms. The critical condition curve under the conditions of intensity and width variation is calculated using a cubic polynomial. Fitting is performed. The relationship between adjacent echo components that can be distinguished is obtained as follows: ; in, is the resolution model coefficient. For the Waveform components The resolution parameter is When the current waveform and the previous waveform can still be effectively distinguished, When adjacent Gaussian components are not distinguishable.

[0050] Please refer to Table 1. Different adjacent echo component peak intensity ratios will lead to different first resolution critical points, second resolution critical points and resolution model coefficients. This embodiment is based on the application of measured data and is set at equal intervals. The applicable scope of the model is determined, and the relevant resolution model coefficients are discretized for reference.

[0051] Table 1: Characteristic coupling of adjacent Gaussian components and their critical conditions for resolution

[0052] According to the above analysis, the present invention designs the following transfer characteristic function, and the transfer parameter reflects the relationship between adjacent echo components.

[0053] ; in, represents the transfer characteristic function, Indicates waveform components The transfer characteristic function results.

[0054] Please refer to Figure 6 , the determination of the target echo component is converted into a conditional probability problem of solving the marker sequence under the given observation sequence. By calculating the type of echo component when the probability is maximum, the determination and extraction of underwater effective echo components are realized. Corresponding to a Gaussian waveform component, according to the set The overall echo response feature conditional random field is used to identify the corresponding label , is the attribute label corresponding to each Gaussian component, which is used to mark the two types of target and non-target. In this embodiment, the object of the signal echo can be set to a digital value here ,when When , it is represented as a non-target component, When is represented as the target component, The number of elements of The same and one-to-one correspondence.

[0055] In this embodiment, whether a waveform component is a target waveform component is determined based on the width feature, intensity feature, similarity feature and transfer feature, and then the part that is not the target waveform component is filtered out, and the target waveform component is retained.

[0056] Please refer to Figure 7After the full waveform data is decomposed, the empirical conditional probability and weight coefficients of the conditional random field (CRF) model are obtained through training, and the full waveform data obtained by the ALB system in the area are predicted based on the CRF model to achieve target screening based on laser echo signals, thereby avoiding the interference caused by reflections at different heights and realizing the extraction of target reflection surfaces under the conditions of target echo surfaces with complex spatial structures.

[0057] Specifically, the set of echo components within a single waveform is , the target attributes corresponding to each waveform component are marked as According to the object of the signal echo, it can be set to digital value. , indicating that the corresponding waveform component is a non-target waveform component, Indicates that the corresponding waveform component is the target waveform component.

[0058] The wave width feature, intensity feature, similarity feature and transfer feature are substituted into the conditional random field model that can be used to distinguish the full waveform target echo signal. It can be expressed as: ; in, The conditional random field model parameters are In the case of Tagged with The probability of is the conditional random field model parameter The four sub-parameters are:

[0059] For the waveform component , No. Characteristic function Value and If the consistency is If not, it is 1; if not, it is 0: ; Once the CRF conditional random field model of the target echo signal in the detection area is established, the problem of identifying the target echo signal can be transformed into the problem of labeling the implicit attributes of different waveform components in each waveform signal. The target components in each echo signal are labeled using the determined CRF model.

[0060] Among them, the parameters of the CRF model The CRF model is learned and trained through estimation and optimization methods.

[0061] Therefore, in actual processing, a certain amount of observation data is first extracted for model training. The observation data should include different types of targets within the water and land space as much as possible and facilitate the identification of related targets. In this embodiment, the learning loss function is: ; in, represents the loss value, is the joint empirical probability distribution of samples, Indicates the sequence of components in the input waveform In the case of probability.

[0062] The loss function can be expanded as: ; in, Indicates waveform components The real label, For the waveform component , No. Characteristic function Value and consistency, Indicates waveform components The label is the true label The empirical probability distribution of Indicates waveform components Tagged with The empirical probability distribution of .

[0063] The Taylor series is used to solve the conditional random field in Expand, set Represents the number of iterations, and the following iteration relationship can be obtained: ; in, For the The conditional random field model parameters of the iteration , For the Iterated Conditional Random Field Model Parameters , is the inverse of the Hessian matrix, for The approximate substitution matrix of for: ; in, To update the function: ; in, is the marginal empirical probability, the approximate substitution matrix It can be calculated by the following iterative method: ; in, For the Iterative approximate substitution matrix, , , in the above iterative process, when The process terminates when , and the parameters obtained at this time are the parameters of the conditional random field screening model of the target echo signal in the area, where is the iteration stopping threshold.

[0064] In terms of target echo signal prediction, the target attribute corresponding to each waveform component can be recursively calculated based on the idea of ​​the Viterbi algorithm, that is, the unnormalized probability of whether it is a target echo component: ; ; in, The first waveform component is The unnormalized probability of For the The waveform components are The unnormalized probability of For the The waveform components are The unnormalized probability of for A collection of: ; Calculate the non-standardized maximum value of each echo component under different annotation conditions, and record the attribute annotation sequence of the non-standardized probability maximum value in turn: ; in, Indicates that the The largest .

[0065] until , the recursion terminates, and the maximum value of the non-normalized probability can be obtained at this time And the optimal sequence of attribute annotations ,in: ; According to the above random field model establishment and signal labeling process, the attribute labeling of each echo component in the full waveform data can be realized.

[0066] S3. Generate a point cloud according to the target waveform component.

[0067] In determining the target waveform component The target waveform components extracted from The peak time position , and the propagation distance of the blue-green laser is calculated accordingly. The preliminary calculation of the point cloud position is similar to the solution method of the scanning detection position of the existing system, that is, based on the signal processing results, the position and direction of the system to the target echo surface are obtained according to the system's optical-mechanical structure and its scanning method, and then merged with the position and attitude data obtained by the system to obtain the point cloud space position in the spatial geodetic coordinate system framework. The specific coordinate conversion method should be calculated according to the specific settings of each sensor: ; in, is the coordinate of the laser foot point in the instantaneous laser beam coordinate system, is the laser distance, is the azimuth of the laser projection on the plane, and the point cloud coordinate solution result can be expressed as: ; in, is the point cloud coordinates obtained by settlement, is the horizontal coordinate of the point cloud, is the vertical coordinate of the point cloud, is the point cloud height coordinate, and the placement deviation of the laser scanning system in the inertial navigation system coordinate system is , whose exact value can be obtained through measurement and system calibration. is the rotation matrix from the laser scanning coordinate system to the inertial platform reference coordinate system, The rotation matrix constructed by the three real-time observations of IMU: Heading, Pitch, and Roll. is the position deviation from the inertial platform reference center to the GNSS antenna phase center. In addition, is the rotation matrix for transforming the system to the geodetic coordinate system. The rotation angle needs to refer to the longitude and latitude of the local horizontal coordinate system in the geodetic coordinate system. is the spatial position of the GNSS antenna phase center in the geodetic coordinate system. , and The specific angles and composition forms of the three rotation matrices should be determined according to the placement direction of each system.

[0068] After the above calculations, the real-time data collected by the system is converted into laser point cloud data in the geodetic coordinate system, the three-dimensional position of the laser foot point space is solved, and a preliminary point cloud is generated. After the preliminary point cloud is generated, the preliminary point cloud can be further filtered according to the existing filtering algorithm to obtain a more accurate point cloud.

[0069] In this solution, a certain area is used as the actual measurement area and the method of the present invention is applied to perform measurement. Specifically, in this embodiment, the measurement is performed in an area near a certain island.

[0070] First, we construct a dataset by extracting multiple profile point clouds containing complex targets from the scan data and their corresponding waveform data. The training samples are generally required to cover all terrain features for model parameter estimation. The selected two profile spatial positions are as follows: Figure 8 By establishing a data index, the interactive annotation of the attributes of the specific waveform decomposition results is realized, and then the training data representing the echo characteristics of the main targets in the scanning detection area is obtained.

[0071] Figure 8 The figure shows two cross-section images selected for conditional random field (CRF) model training in the experiment, including the original point cloud and the manually selected target point cloud.

[0072] There are obvious differences between the results of conditional processing using different screening conditions. In order to understand the effect of conditional random field processing in this paper and accurately evaluate the accuracy of the trained conditional random field model, two accuracy indicators are constructed here: the error rate of the waveform echo component (CER) and the error rate of the laser beam (Laser Beam Error Rate, LBER). CER can show the accuracy of each waveform component: ; LBER can be used to measure the change in overall data within the detection area after waveform processing based on target echo identification.

[0073] ; in, Indicates the total number of echo components that need to be labeled in the training data, is the number of echo components whose annotation results are inconsistent with the training data. is the total number of beams or waveforms in the training data, Indicates waveform data whose annotation results are different from the echo positions and quantities of the corresponding training data.

[0074] Please refer to Table 2 and Table 3. Table 2 gives Figure 8 Table 3 shows the processed data of the first section in the middle and upper part. Figure 8 Processed data for the second section in the lower center.

[0075] Table 2: Target echo recognition error rate for different feature combinations of the first profile point cloud

[0076] Table 3: Target echo recognition error rate of the second profile point cloud for different feature combinations

[0077] Based on the point cloud profile data extracted in the target area for model training, the changes in indicators in the table show the differences in point cloud generation results between conditional screening of single echo signal features and constructing conditional random fields through echo signal features in echo signal processing.

[0078] Among the discrimination methods based on a single feature, the point cloud screening methods based on different echo component features can improve the accuracy of target echo signal extraction to a certain extent. These verify the feasibility of target detection and point cloud noise suppression in the signal processing stage. Among them, the similarity feature associates the echo signal feature with the system transmission pulse outside itself, and its screening error rate is the lowest. The effectiveness of the signal similarity feature is also reflected in the case of a feature conditional random field model. In Tables 2 and 3, the target echo component screening method based on conditional random fields fully considers the coupling relationship between different echo features with the help of machine learning methods, and the resulting CER and LBER are generally low. After adding the similarity feature, the improvement in processing effect is also very obvious.

[0079] The above experiments not only reflect the waveform processing effect, but also verify the accuracy of conditional random field model screening, providing an effective reference for data processing in subsequent regional detection.

[0080] The processing effects at different locations in the experimental area are compared and analyzed to verify the effectiveness of the method in practical applications. The results of the conventional echo signal strength detection in online processing and the conditional random field target echo screening based on echo signal characteristics proposed in this paper are shown in the figure. Fig. 9 As shown: from Fig. 9 It can be clearly seen from the water surface terrain detection results that the point cloud results obtained by directly using the signal detection method based on the intensity threshold contain a large number of non-target noise points. These noise points are mainly randomly distributed along the direction of laser propagation, and the main reason for their generation is the scattering interference in the laser optical path. Due to the spatial continuity of the topographic distribution, after the use of statistical random filtering, the filtering effect is obvious for the point cloud of landforms with obvious spatial structural characteristics, such as buildings, sea surface, beaches, etc. However, the filtering ability for non-target noise in areas such as trees and shrubs that are distributed in layers in the vertical direction is insufficient. In the results obtained by the echo signal processing method based on conditional random fields, the interference of non-target noise is significantly lower than that of the conventional signal strength detection method. In addition, for the multi-reflection echo layer in the elevation direction, such as the canopy of trees and shrubs, the resolution effect between the ground is also more obvious, and the effect after statistical random filtering is also better than the statistical filtering method.

[0081] Please refer to Fig.10 , Fig.10 The figure below shows the processing result of underwater terrain in the scanning flight strip. The strong scattering interference in the water body has a serious impact on the conventional peak detection results. Although the underwater terrain after SOR processing can initially show the terrain changes, the overall area still contains high-density random noise. After waveform processing using the conditional random field model, the extraction effect of underwater terrain is more obvious. Since the overall underwater terrain is relatively continuous, the non-target interference in the water body is greatly reduced after SOR processing. In addition, the above phenomenon can also be verified by the change in the number of laser foot points in the corresponding point cloud. Embodiment 2 See also Fig.11 Based on the above method, the present invention further provides a terminal, which includes a processor 10, a memory 20 and a display 30. However, it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0082] In some embodiments, the memory 20 may be an internal storage unit of the terminal, such as a hard disk or memory of the terminal. In other embodiments, the memory 20 may also be an external storage device of the terminal, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the terminal. Further, the memory 20 may also include both an internal storage unit of the terminal and an external storage device. The memory 20 is used to store application software and various types of data installed on the terminal, such as the program code of the installation terminal. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a full-waveform laser radar target echo signal detection program 40 is stored on the memory 20, and the full-waveform laser radar target echo signal detection program 40 can be executed by the processor 10, thereby realizing the terminal in the present application.

[0083] In some embodiments, the processor 10 may be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code or process data stored in the memory 20, such as executing relevant programs of the full-waveform lidar target echo signal detection method.

[0084] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light Emitting Diode) touch device, etc. The display 30 is used to display information on the terminal and to display a visual user interface.

[0085] In one embodiment, when the processor 10 executes the full-waveform lidar target echo signal detection program 40 in the memory 20, the steps of the full-waveform lidar target echo signal detection method as described above are implemented.

[0086] Embodiment 3 This embodiment provides a storage medium, which stores a full-waveform lidar target echo signal detection program. When the full-waveform lidar target echo signal detection program is executed by a processor, the steps of the full-waveform lidar target echo signal detection method as described above are implemented.

[0087] To summarize, the present invention extracts the features of each waveform component of the echo signal, then filters the echo signal according to the features of each waveform component, and uses the filtered echo signal to generate a point cloud, thereby preventing the extraction of the target point cloud from being interfered by non-target echo signals and reducing the existence of interference point clouds.

[0088] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or terminal including the element.

[0089] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium, and the program can include the processes of the above-mentioned method embodiments when executed. The storage medium can be a memory, a disk, an optical disk, etc.

[0090] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A full-waveform laser radar target echo signal detection method, characterized in that: The full-waveform laser radar target echo signal detection method comprises: Acquire an echo signal, decompose the echo signal to obtain a plurality of waveform components, extract features of each of the waveform components, and acquire waveform component features corresponding to each of the waveform components; Screening each of the waveform components according to the waveform component characteristics of each of the waveform components to obtain a target waveform component; A point cloud is generated based on the target waveform components.

2. A full-waveform laser radar target echo signal detection method according to claim 1, characterized in that: The extracting features of each of the waveform components to obtain waveform component features corresponding to each of the waveform components specifically includes: Fitting each of the waveform components to obtain Gaussian parameters of each of the waveform components; Obtaining the peak intensity, wave width, peak time position and correlation coefficient of each of the waveform components according to each of the Gaussian parameters; The peak intensity, the wave width, the peak time position and the correlation coefficient are used as the waveform component features.

3. A full-waveform laser radar target echo signal detection method according to claim 2, characterized in that: The screening of each of the waveform components according to the waveform component characteristics of each of the waveform components to obtain the target waveform component specifically includes: Obtaining characteristic function results of each of the waveform components according to the characteristics of each of the waveform components; Calculating the probability that each of the waveform components is a target waveform component and the probability that each of the waveform components is not a target waveform component according to the characteristic function results of each of the waveform components; Whether each waveform component is a target waveform component is determined based on the probability that each waveform component is a target waveform component and the probability that each waveform component is not a target waveform component.

4. A full-waveform laser radar target echo signal detection method according to claim 3, characterized in that: The step of obtaining the characteristic function results of each waveform component according to the characteristics of each waveform component specifically includes: Acquire the wave width characteristic function result of each of the waveform components according to the wave width of each of the waveform components; Obtaining intensity characteristic function results of each of the waveform components according to the peak intensity of each of the waveform components; Obtaining similarity characteristic function results of each of the waveform components according to the correlation coefficient and peak time position of each of the waveform components; Acquire the transfer characteristic function result of each of the waveform components according to the peak intensity, wave width and peak time position of each of the waveform components; The wavelength characteristic function result, the intensity characteristic function result, the similarity characteristic function result and the transfer characteristic function result are taken as the characteristic function results.

5. A full-waveform laser radar target echo signal detection method according to claim 4, characterized in that: The wave width characteristic function result of each waveform component is obtained according to the wave width of each waveform component, specifically, the wave width characteristic function result is obtained according to the following formula: ； in, is the characteristic function of the wave width, For the Waveform components The wave width, is the transmit waveform width, Indicates Waveform components The wave width characteristic function results.

6. A full-waveform laser radar target echo signal detection method according to claim 4, characterized in that: The step of obtaining the intensity characteristic function result of each of the waveform components according to the peak intensity of each of the waveform components specifically includes: De-attenuating the peak intensity of each of the waveform components to obtain the de-attenuated peak intensity of each of the waveform components; The intensity characteristic function result of each of the waveform components is obtained according to the de-attenuated peak intensity of each of the waveform components: in, Indicates waveform components The de-attenuated peak intensity, is the intensity characteristic function, represents the noise standard deviation, Indicates Waveform components The intensity characteristic function results of .

7. A full-waveform laser radar target echo signal detection method according to claim 4, characterized in that: The obtaining of similarity characteristic function results of each of the waveform components according to the correlation coefficient and peak time position of each of the waveform components specifically includes: Obtain the time position of the maximum value of the correlation coefficient in the correlation coefficient of each of the waveform components ; The similarity characteristic function results of each of the waveform components are obtained according to the time position of the maximum value of the correlation coefficient and the peak time position of each of the waveform components: ； in, is the similarity feature function, is the time position of the maximum value of the correlation coefficient, is the transmission waveform width, For the Waveform components The peak time position of is the transmission waveform width, Indicates Waveform components The similarity feature function result of .

8. The full-waveform laser radar target echo signal detection method according to claim 4, characterized in that: The step of obtaining the transfer characteristic function result of each waveform component according to the peak intensity, wave width and peak time position of each waveform component specifically includes: According to the peak intensity of each waveform component, the peak intensity ratio of adjacent echo components of each waveform component is obtained: ； in, For the Waveform components The peak intensity ratio of adjacent echo components, For the Waveform components The peak intensity, For the Waveform components The peak intensity of Obtaining the first multiple of each of the waveform components according to the wave width of each of the waveform components: ； in, For the Waveform components The first multiple of For the Waveform components The wave width, For the Waveform components The wave width, is the transmitting waveform width; Obtain the second multiple of each of the waveform components according to the peak time position of each of the waveform components: ； Among them, the For the Waveform components The second multiple of For the Waveform components The peak time position of For the Waveform components The peak time position of Acquire a first critical point for distinguishing each of the waveform components and a second critical point for distinguishing each of the waveform components according to the first multiple and the second multiple of each of the waveform components; Obtain the resolution parameters of each of the waveform components according to the peak intensity ratio of adjacent echo components, the first multiple, the second multiple, the first resolution critical point and the second resolution critical point of each of the waveform components: ； in, Indicates Waveform components The resolution parameters, is the resolution model coefficient, represents the first critical resolution point, represents the second critical resolution point; According to the resolution parameter, the second multiple, the first resolution critical point and the second resolution critical point of each of the waveform components, the transfer characteristic function result of each of the waveform components is obtained: ； in, represents the transfer characteristic function, Indicates Waveform components The transfer characteristic function result is Indicates Waveform components The transfer characteristic function results.

9. A terminal, characterized in that: The terminal includes: a memory, a processor, and a full-waveform lidar target echo signal detection program stored in the memory and executable on the processor. When the full-waveform lidar target echo signal detection program is executed by the processor, the terminal is controlled to implement the steps of the full-waveform lidar target echo signal detection method as described in any one of claims 1 to 8.

10. A readable storage medium, characterized in that: The readable storage medium stores a full-waveform lidar target echo signal detection program, and when the full-waveform lidar target echo signal detection program is executed by the processor, the steps of the full-waveform lidar target echo signal detection method as described in any one of claims 1-8 are implemented.

Images