Method and system for generating vertical profile layers of atmospheric lidar based on cloth simulation

Abstract

The present application relates to a method and system for generating a vertical profile layer of an atmospheric lidar based on cloth simulation, and belongs to the field of lidar technology. The method comprises: pre-processing the original echo signal to generate a target signal, and flipping the target signal in the vertical direction; constructing a cloth model, simulating the cloth falling to the flipped target signal to generate a fitting surface, and flipping the fitting surface in the vertical direction to generate a target surface; and generating a vertical profile layer of an atmospheric lidar based on the target signal and the target surface. By simulating the dynamic process of cloth falling and fitting under the action of gravity and elastic internal force, a target curve that can naturally respond to sudden changes in boundaries is constructed in the atmospheric vertical signal. By comparing the target curve and the echo signal, the jump area of ​​the echo signal is effectively captured, thereby accurately detecting the vertical profile layer of the atmospheric lidar, and solving the problems of poor layer detection accuracy and low robustness.

Description

Technical Field

[0001] The present application relates to the field of laser radar technology, and in particular to a method and system for generating a vertical profile layer of an atmospheric laser radar based on cloth simulation. Background Art

[0002] In fields such as climate change, environmental monitoring, and numerical weather forecasting, the vertical distribution information of clouds and aerosols is considered critical basic data. LiDAR, due to its high sensitivity, high resolution, and all-day detection capabilities, has become the primary remote sensing method for obtaining this type of vertical structure information. LiDAR echo signals typically exhibit obvious jump characteristics in the vertical direction, especially at the boundaries of the aerosol layer, cloud base, and cloud top. However, under actual observation conditions, due to multiple factors such as signal attenuation, background light interference, atmospheric scattering instability, and detector performance limitations, the vertical profile data obtained are often noisy and have a large dynamic range, making it difficult to stably and accurately extract the layer structure boundaries. This problem has become a key bottleneck restricting radar inversion algorithms, the quality of remote sensing products, and the accuracy of subsequent physical modeling.

[0003] Since the 1990s, atmospheric lidar layer detection technology has undergone multiple phases of development, with different methods demonstrating unique advantages and limitations in improving accuracy and adaptability. In early research, Pal et al. innovatively employed the first-order derivative method in 1992, capturing the zero point where the slope of the echo signal transitions from negative to positive to locate cloud base. However, this method is sensitive to noise interference and prone to misjudgment due to signal jitter. The threshold method, a classic detection scheme, provides stable detection by identifying whether the signal continuously exceeds a preset threshold sequence. It is widely used in cloud and aerosol products from ground-based and spaceborne lidar. Vaughan proposed the SIBYL algorithm, which combines the attenuated scattering ratio (ASR) with adaptive thresholding and multi-resolution spatial averaging, significantly improving the processing efficiency of CALIPSO data. However, subsequent research revealed that it can miss weak signal layers. In 2016, Lewis et al. achieved a breakthrough in high-level cloud detection by fusing the slope method with the threshold method in low signal-to-noise ratio environments. However, the identification of mid- and low-level clouds remains limited.

[0004] Recent methodological innovations have focused on breaking through traditional threshold dependence. Mao et al. proposed a multi-scale detection method. Furthermore, the multi-scale Bernoulli probability model, proposed in 2024, effectively circumvents the threshold selection challenge through a probabilistic framework, enhancing the ability to capture weak signals. However, new challenges remain in non-stationary signal processing, such as strong dependence on statistical assumptions and sensitivity to window parameters. Notably, while deep learning technology demonstrates the potential for autonomous feature learning, its development is constrained by challenges with reliable training data.

[0005] In summary, the existing atmospheric lidar layer detection technology has problems of poor detection accuracy and low robustness in layer detection. Summary of the Invention

[0006] Based on this, it is necessary to provide a method and system for generating the vertical profile layer of an atmospheric lidar based on cloth simulation to address the above technical problems. The idea of ​​cloth simulation modeling is introduced. By simulating the dynamic process of cloth falling and fitting under the action of gravity and elastic internal force, a target curve that can naturally respond to the sudden change boundary is constructed in the atmospheric vertical signal. By comparing the target curve and the echo signal, the jump area of ​​the echo signal can be effectively captured, thereby accurately detecting the vertical profile layer of the atmospheric lidar and solving the problems of poor layer detection accuracy and low robustness.

[0007] In a first aspect, the present application provides a method for generating a vertical profile layer of an atmospheric lidar based on cloth simulation, comprising:

[0008] Preprocess the original echo signal to generate the target signal, and flip the target signal in the vertical direction;

[0009] Construct a cloth model, simulate the cloth falling to the target signal after flipping to generate a fitting surface, and flip the fitting surface in the vertical direction to generate the target surface;

[0010] Generate atmospheric lidar vertical profile layer based on target signal and target surface.

[0011] In one embodiment, the original echo signal is pre-processed, including: performing distance square correction processing on the original echo signal to generate PR 2 Signal; PR 2 The signal is normalized; the normalized PR 2 The signal is smoothed and denoised.

[0012] In one embodiment, the process of building a cloth model includes:

[0013] Construct a one-dimensional horizontal cloth model. The cloth model consists of equally spaced particles, each with mass, displacement, and velocity properties. The initial velocity of each particle is set to 0. Adjacent particles are connected by springs that follow Hooke's law and have an elastic coefficient of k. All particles are driven by the downward force of gravity g.

[0014] The cloth is initially suspended above the target signal after flipping. The initial height of the cloth is greater than the maximum value of the target signal after flipping. All particles fall downward at the same time in the simulation.

[0015] In one embodiment, generating a fitting surface from a target signal after the cloth is simulated to fall and flip, includes:

[0016] All the mass points of the cloth are allowed to fall freely downward simultaneously. During the falling process, the displacement of two adjacent mass points is iteratively adjusted using the preset spring internal force constraint mechanism until the spring force and external force acting on the cloth are balanced.

[0017] Generate a fitting surface based on the shape of the cloth after force balance.

[0018] In one embodiment, a preset spring internal force constraint mechanism is used to continuously iteratively adjust the displacement of two adjacent mass points, including: if two adjacent mass points connected by a spring are both movable, the two mass points are moved in opposite directions by the same amount; if one mass point is immovable and the other mass point is movable, the other mass point is moved; wherein, when a mass point is immovable, the mass point falls to the target curve data position.

[0019] In one embodiment, the displacement calculation formula of the particle is as follows:

[0020]

[0021]

[0022] in, m is the state of the particle, a value of 1 indicates that the particle can be moved, and a value of 0 indicates that the particle cannot be moved; is the current state of the particle; is the state of the adjacent particles; is the elastic constant of the spring; is the displacement of the particle; is the current position of the particle; is the position of the adjacent particles.

[0023] In one embodiment, generating an atmospheric lidar vertical profile layer based on a target signal and a target surface includes:

[0024] Calculate the overall offset of the target signal relative to the target curve and the local offset of the target signal's local transition region relative to the corresponding region of the target curve. When the local offset is greater than a set multiple of the overall offset, determine that the current local transition region is a candidate hierarchical structure.

[0025] Generate atmospheric lidar vertical profile layers based on the candidate hierarchy.

[0026] In one embodiment, generating an atmospheric lidar vertical profile layer based on a candidate hierarchy includes:

[0027] Calculate the difference between the upper and lower boundaries of the candidate hierarchy ,when When , the candidate hierarchy is retained as a valid layer; when When , the candidate hierarchy is an invalid layer and is eliminated; is the minimum upper and lower boundary difference threshold set; is the weight coefficient based on the current signal-to-noise ratio change.

[0028] In one embodiment, is defined as follows:

[0029]

[0030] Where SNR is the current signal-to-noise ratio.

[0031] In a second aspect, the present application provides a system for generating vertical profile layers of an atmospheric lidar based on cloth simulation, comprising:

[0032] The signal preprocessing module is used to preprocess the original echo signal to generate a target signal and flip the target signal in the vertical direction;

[0033] The cloth simulation module is used to build a cloth model, simulate the cloth falling to the flipped target signal to generate a fitting surface, and flip the fitting surface in the vertical direction to generate the target surface;

[0034] The vertical profile layer generation module is used to generate the atmospheric lidar vertical profile layer based on the target signal and target surface.

[0035] This application adopts the above-mentioned method and system for generating vertical profile layers of atmospheric lidar based on cloth simulation, which has the following beneficial effects:

[0036] 1. By introducing the concept of cloth simulation modeling, the dynamic process of cloth falling and fitting under the action of gravity and elastic internal forces is simulated. A target curve that can naturally respond to the jump boundary is constructed in the atmospheric vertical signal. By comparing the target curve with the echo signal, the jump area of ​​the echo signal is effectively captured, thereby accurately detecting the vertical profile layer of the atmospheric lidar and improving detection accuracy and robustness.

[0037] 2. This application applies the particle-spring cloth model in computer graphics to the layer identification task of the vertical profile of the lidar, simulating the natural conformation of the cloth to the signal jump boundary under the action of gravity and elastic internal forces. From the perspective of physical modeling, it enhances the response capability to the atmospheric boundary jump. Different from traditional empirical or statistical algorithms, it has a more intuitive and reasonable explanation, and significantly improves the detection capability of weak signal layers and multi-layer structures.

[0038] 3. This application proposes a detection mechanism that is highly sensitive to local disturbances in the signal. By fitting the dynamic process of the cloth model and introducing the local deviation energy index (LDEI), it can effectively identify weak scattering layers such as thin clouds and sparse aerosols, significantly reducing the missed detection rate. It is particularly suitable for boundary extraction tasks in complex multi-layer structure environments, and enhances the robustness of the algorithm to environments with different signal-to-noise ratios.

[0039] 4. This application designs a difference judgment mechanism based on the minimum upper and lower boundaries of the signal-to-noise ratio adaptation, which can achieve sensitive detection and noise suppression in high SNR and low SNR scenarios respectively, thereby improving the versatility and stability under various observation conditions and reducing parameter dependence and the degree of manual intervention.

[0040] 5. The core algorithm of this application relies on physical model drive, has few parameter settings, strong versatility, and strong adaptability, which is conducive to promotion to lidar systems of different types and configurations. The algorithm is simple to implement, has high computational efficiency, and has good engineering practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 The distance correction signal PR provided by one embodiment 2 Schematic diagram of the structure;

[0042] Figure 2 A schematic structural diagram of a mass-spring cloth model provided in one embodiment;

[0043] Figure 3 A schematic diagram of a method for generating a vertical profile layer of an atmospheric lidar based on cloth simulation provided by one embodiment;

[0044] Figure 4 A schematic diagram of a signal flip structure provided by an embodiment;

[0045] Figure 5 A schematic diagram of a spring internal force constraint mechanism in cloth simulation provided by one embodiment;

[0046] Figure 6 PR provided for an embodiment 2 Schematic diagram of the signal and the corresponding target curve;

[0047] Figure 7 A time-height profile of the 532 nm lidar received signal intensity (1440 profiles stitched together) provided for one embodiment.

[0048] Figure 8 A time-height profile of the 532 nm lidar received signal intensity processed by the fabric algorithm provided in one embodiment (1440 profiles stitched together);

[0049] Figure 9 A schematic diagram of test results of different SNRs provided by an embodiment. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0051] In general, the elastic scattering lidar equation is as follows:

[0052]

[0053] Where R is the detection distance; is the echo power received by the detector; C is the radar system constant; is the optical overlap factor; A is the effective receiving area of ​​the telescope mirror; is the atmospheric volume backscattering coefficient; is the atmospheric extinction coefficient.

[0054] In practical applications, when laser pulses propagate through the atmosphere, they are affected by aerosol and molecular scattering and absorption effects, causing the echo signal to significantly attenuate with distance. In order to highlight the signal characteristics at different heights, the echo power is often multiplied by the square of the detection distance R. 2 Distance correction is performed to form a PR² profile, whose value is approximately proportional to the atmospheric aerosol concentration at that altitude. It represents the intensity of the lidar atmospheric echo signal at each altitude. The height-corrected signal is proportional to the atmospheric aerosol concentration at that altitude.

[0055] Reference Figure 1 When the laser pulse encounters a cloud or aerosol layer, due to its backscattering coefficient The larger the PR² value, the more pronounced it is. Based on this physical property, the core principle of hierarchical structure detection lies in monitoring the locations where echo signal strength changes. To achieve this, a detection method suitable for identifying signal transitions is required.

[0056] Cloth Simulation Filtering (CSF) is a point cloud data filtering method based on physical modeling. It was originally used to automatically separate ground and non-ground points from 3D LiDAR point clouds. Its core lies in the introduction of a mass-spring system. Figure 2As shown, cloth is modeled as a two-dimensional grid structure composed of a large number of mass points, interconnected by virtual springs and elastically constrained according to Hooke's law. Each mass point has physical properties such as mass and position, and the entire system simulates the deformation of real cloth under the influence of gravity. Another key concept of the algorithm lies in the vertical inversion of point cloud data. The algorithm flips the original 3D point cloud vertically, shifting high points to low points, thereby simulating the process of cloth "covering" the terrain surface from above. As the cloth continues to fall in the gravitational field, the mass points are driven by gravity while being constrained by spring forces (internal forces) to prevent excessive deformation. At the same time, they gradually come into contact and interact with the inverted point cloud. Ultimately, the cloth stabilizes at the "high points" of the point cloud surface, forming a fitted surface that approximates the ground. By detecting the contact boundary between the cloth and the original point cloud, CSF can automatically distinguish between ground points and non-ground points. Compared to traditional filtering algorithms, CSF offers advantages such as fewer parameters, clear physical meaning, and strong adaptability. It has been widely used for ground extraction tasks in various complex terrain environments.

[0057] The present invention extends the physical simulation mechanism of CSF to the identification of vertical profile layer structure of atmospheric lidar, and utilizes its high sensitivity to "mutation boundaries" to realize the automatic detection of atmospheric hierarchical structures such as cloud layers and aerosol layers.

[0058] First, based on the above principles, refer to Figure 3 , this application provides a method for generating vertical profile layers of atmospheric lidar based on cloth simulation, comprising:

[0059] S100 , pre-processing the original echo signal to generate a target signal, and flipping the target signal in a vertical direction.

[0060] In one embodiment, the original echo signal is pre-processed, including: performing distance square correction processing on the original echo signal to generate PR 2 Signal; PR 2 The signal is normalized; the normalized PR 2 The signal is smoothed and denoised.

[0061] In order to eliminate the natural attenuation of echo power with distance, the original echo signal is first corrected by the square of the distance, that is, the signal is multiplied by the square of the detection distance R 2 , generating the PR² signal, i.e., the PR² profile data. Then, to improve the algorithm's versatility and numerical stability, the PR² signal is normalized to the [0, 1] range, and smoothing and denoising preprocessing is performed to suppress some high-frequency noise, generating the target signal.

[0062] In one embodiment, the smoothing and denoising preprocessing may use a three-point sliding average.

[0063] Reference Figure 4 To simulate the process of fabric fitting downward from above under the influence of gravity, the target signal is vertically flipped. Specifically, the higher-intensity portions of the original signal (corresponding to clouds or aerosol layers) are projected downward, forming low-lying areas. This creates a "topographic depression" for the fabric's natural descent, facilitating precise fitting at sudden changes in layer boundaries.

[0064] S200, constructing a cloth model, simulating the cloth falling to a target signal after flipping to generate a fitting surface, and flipping the fitting surface in a vertical direction to generate a target surface.

[0065] In one embodiment, the process of constructing a cloth model includes: constructing a one-dimensional horizontal cloth model; wherein the cloth model is composed of equally spaced mass points, each with mass, displacement, and velocity properties; the initial velocity of each mass point is set to 0, and adjacent mass points are connected by springs that follow Hooke's law and have an elastic coefficient k; all mass points are driven by a downward gravity g; the cloth is initially suspended above a target signal after flipping, with the initial height of the cloth greater than the maximum value of the target signal after flipping, and all mass points are simultaneously subjected to downward free fall during the simulation.

[0066] In one embodiment, the elastic coefficient k is 0.5; the gravity g is 9.8 m / s 2 .

[0067] In one embodiment, a fitting surface is generated based on a target signal after the cloth is simulated to fall and flip, including: simultaneously causing all particles of the cloth to freely fall downward; during the falling process, using a preset spring internal force constraint mechanism to continuously and iteratively adjust the displacement of two adjacent particles until the spring force and external force acting on the cloth are balanced; and generating a fitting surface based on the shape of the cloth after the force balance.

[0068] Reference Figure 5 To ensure that the cloth model doesn't deform excessively or move unrestricted while conforming to the target signal data surface, we introduced a spring internal force constraint mechanism into the cloth model. The core idea behind this mechanism is to use springs connecting adjacent mass points to simulate the elastic deformation properties of cloth and constrain mass point movement. Therefore, two mass points with different heights will attempt to move to the same horizontal plane.

[0069] Specifically, if two adjacent mass points connected by a spring are both movable, the two mass points are moved in opposite directions by the same amount; if one mass point is immovable and the other is movable, the other mass point is moved; when the mass point is immovable, the mass point falls to the target curve data position. The moving direction is as follows: Figure 5 shown.

[0070] As the cloth falls, the position of each mass point is continuously updated based on the forces acting on it. This internal spring force constraint mechanism iteratively adjusts the displacements of two adjacent mass points, ultimately stabilizing the cloth's shape. This balances the cloth's spring force and external forces.

[0071] It should be noted that when a particle falls to the target signal data location, it is set to an immovable state. Therefore, during the falling process of the cloth, there are two situations: one is that two adjacent particles are still falling and both are in a movable state, or one particle is falling and in a movable state, while the other particle has already fallen to the target signal data location and is in an immovable state.

[0072] Reference Figure 6 After the fabric's shape stabilizes, the surface fitting curve can smoothly conform to the clean air region of the PR2 profile, while effectively identifying local areas of significant transitions. After the fabric system reaches stable convergence, a vertical flip operation is performed to generate the target curve, ensuring spatial consistency with the PR2 profile.

[0073] In one embodiment, the displacement calculation formula of the particle is as follows:

[0074]

[0075]

[0076] in, m is the state of the particle, a value of 1 indicates that the particle can be moved, and a value of 0 indicates that the particle cannot be moved; is the current state of the particle; is the state of the adjacent particles; is the elastic constant of the spring; is the displacement of the particle; is the current position of the particle; is the position of the adjacent particles.

[0077] S300: Generate an atmospheric lidar vertical profile layer based on the target signal and the target surface.

[0078] The characteristics of the target signal are further quantitatively analyzed based on the target surface. In order to evaluate the degree of deviation between the target signal and the target surface, the global deviation energy E is introduced, and the calculation formula is as follows:

[0079]

[0080] in, is the value of a data point of the target signal in the selected analysis area, is the fitting value of the target surface at the corresponding position,N is the number of data points within the selected analysis area.

[0081] In one embodiment, based on the above-mentioned global deviation energy E, step S300 includes: calculating the overall offset degree of the target signal relative to the target curve and the local offset degree of the local jump area of ​​the target signal relative to the corresponding area of ​​the target curve; when the local offset degree is greater than a set multiple of the overall offset degree, the current local jump area is determined to be a candidate hierarchical structure; and generating an atmospheric lidar vertical profile layer based on the candidate hierarchical structure.

[0082] Specifically, when the calculation range covers the entire target signal curve, E represents the overall deviation of the target signal, denoted as MSE, and is used to evaluate the global fitting effect; when the calculation area is limited to the local jump area, E is used as a local volatility measurement indicator, defined as the Local Deviation Energy Index (LDEI), which is used to characterize the instantaneous fluctuation energy of the target signal relative to the target curve.

[0083] In this embodiment, when the LDEI value of the transition region is greater than twice the overall MSE, it indicates that the region has significant structural fluctuation characteristics and can be determined as a candidate hierarchical structure, while the fluctuation region below the threshold is regarded as noise and filtered out.

[0084] In one embodiment, generating an atmospheric lidar vertical profile layer based on a candidate hierarchy includes: calculating the difference between the upper and lower boundaries of the candidate hierarchy ,when When , the candidate hierarchy is retained as a valid layer; when When , the candidate hierarchy is an invalid layer and is eliminated; The difference threshold between the minimum upper and lower boundaries is set; is the weight coefficient based on the current signal-to-noise ratio change.

[0085] Under different signal-to-noise ratio conditions, layer structure recognition faces different challenges: when the signal-to-noise ratio is high, the noise impact is relatively small, and local fine structures (such as thin clouds) should be accurately identified; when the signal-to-noise ratio is low, local fluctuations may come from noise, and relying solely on LDEI may lead to misjudgment. Therefore, in this embodiment, the difference between the upper and lower boundaries of the candidate hierarchical structure is introduced. W , which is the difference between the upper and lower boundaries of the candidate hierarchy detected by CSF. When , the candidate hierarchy is retained as a valid layer. Thus, the difference between the upper and lower boundaries of the candidate hierarchy is W To further improve the accuracy of layer structure recognition and reduce the possibility of misjudgment.

[0086] In addition, in order to avoid misidentifying high-frequency noise as layer structure, this application introduces a minimum layer thickness judgment criterion. That is, when the height difference between the upper and lower boundaries of a candidate layer structure detected is less than the set threshold (e.g. 150m, corresponding to 20 sampling points), the candidate hierarchical structure will be considered an invalid structure and eliminated. This threshold can be automatically adjusted according to the signal-to-noise ratio of the data to adapt to different observation conditions.

[0087] In one embodiment, the difference threshold between the minimum upper and lower boundaries is set to 150 m, taking into account the vertical resolution of the laser radar of 7.5 m and the thickness characteristics of the typical atmospheric structure.

[0088] In one embodiment, is defined as follows:

[0089]

[0090] Where SNR is the current signal-to-noise ratio.

[0091] In practical applications, refer to Figure 7 and Figure 8 , using 1440 profiles from measured 532nm lidar data. Based on the aforementioned method for generating vertical profile layers from atmospheric lidar using cloth simulation, the measured data was processed and the layer detection accuracy for the 1440 profiles measured by the 532nm lidar was found to reach 95.7%. This excellent performance was also maintained in simulated data with low signal-to-noise ratios, verifying the accuracy and stability of the method and its strong engineering applicability.

[0092] Reference Figure 9 Using established physical modeling methods, they simulated and generated nine sets of radar signals under different SNR (signal-to-noise ratio) conditions. They set the signal layer structure (for example, from the 367th to the 453rd sampling points), added background light, dark counts, and quantum noise, and generated realistic signals. These signals were processed using a fabric-simulated atmospheric lidar vertical profile layer generation method. The algorithm was found to accurately identify layer boundaries under different SNR conditions, including SNRs of 10, 5, 2, and 1, with an error of less than three sampling points. The simulations were controllable and repeatable, validating the robustness and performance limits of the method.

[0093] In a second aspect, the present application provides a system for generating vertical profile layers of an atmospheric lidar based on cloth simulation, comprising:

[0094] The signal preprocessing module is used to preprocess the original echo signal to generate a target signal and flip the target signal in the vertical direction;

[0095] The cloth simulation module is used to build a cloth model, simulate the cloth falling to the flipped target signal to generate a fitting surface, and flip the fitting surface in the vertical direction to generate the target surface;

[0096] The vertical profile layer generation module is used to generate the atmospheric lidar vertical profile layer based on the target signal and target surface.

[0097] Each module in the aforementioned cloth simulation-based atmospheric lidar vertical profile layer generation system can be implemented in whole or in part through software, hardware, or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in hardware form, or stored in a computer device memory in software form, so that the processor can call and execute the corresponding operations of each module.

[0098] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0099] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A method for generating vertical profile layers of atmospheric lidar based on cloth simulation, characterized in that: include: Preprocess the original echo signal to generate the target signal, and flip the target signal in the vertical direction; Construct a cloth model, simulate the cloth falling to the target signal after flipping to generate a fitting surface, and flip the fitting surface in the vertical direction to generate the target surface; Generate atmospheric lidar vertical profile layer based on target signal and target surface.

2. The method according to claim 1, characterized in that Preprocess the original echo signal, including: performing distance square correction on the original echo signal to generate PR 2 Signal; PR 2 The signal is normalized; the normalized PR 2 The signal is smoothed and denoised.

3. The method according to claim 1, characterized in that The process of building a cloth model includes: Construct a one-dimensional horizontal cloth model. The cloth model consists of equally spaced particles, each with mass, displacement, and velocity properties. The initial velocity of each particle is set to 0. Adjacent particles are connected by springs that follow Hooke's law and have an elastic coefficient of k. All particles are driven by the downward force of gravity g. The cloth is initially suspended above the target signal after flipping. The initial height of the cloth is greater than the maximum value of the target signal after flipping. All particles fall downward at the same time in the simulation.

4. The method according to claim 3, characterized in that Generate a fitting surface by simulating the cloth falling down and flipping over the target signal, including: All the mass points of the cloth are allowed to fall freely downward simultaneously. During the falling process, the displacement of two adjacent mass points is iteratively adjusted using the preset spring internal force constraint mechanism until the spring force and external force acting on the cloth are balanced. Generate a fitting surface based on the shape of the cloth after force balance.

5. The method according to claim 4, characterized in that The displacement of two adjacent mass points is continuously and iteratively adjusted using a preset spring internal force constraint mechanism, including: if two adjacent mass points connected by a spring are both movable, the two mass points are moved in opposite directions by the same amount; if one mass point is immovable and the other is movable, the other mass point is moved; among them, when a mass point is immovable, the mass point falls to the target curve data position.

6. The method according to claim 5, characterized in that The displacement calculation formula of the particle is as follows: in, m is the state of the particle, a value of 1 indicates that the particle can be moved, and a value of 0 indicates that the particle cannot be moved; is the current state of the particle; is the state of the adjacent particles; is the elastic constant of the spring; is the displacement of the particle; is the current position of the particle; is the position of the adjacent particles.

7. The method according to claim 4, characterized in that Generate atmospheric lidar vertical profile layer based on target signal and target surface, including: Calculate the overall offset of the target signal relative to the target curve and the local offset of the target signal's local transition region relative to the corresponding region of the target curve. When the local offset is greater than a set multiple of the overall offset, determine that the current local transition region is a candidate hierarchical structure. Generate an atmospheric lidar vertical profile layer based on the candidate hierarchy.

8. The method according to claim 7, characterized in that Generate atmospheric lidar vertical profile layers based on candidate hierarchies, including: calculating the difference between the upper and lower boundaries of the candidate hierarchies ， when When , the candidate hierarchy is retained as a valid layer; when When , the candidate hierarchy is an invalid layer and is eliminated; is the minimum upper and lower boundary difference threshold set; is the weight coefficient based on the current signal-to-noise ratio change.

9. The method according to claim 8, characterized in that is defined as follows: Where SNR is the current signal-to-noise ratio.

10. A system for generating vertical profile layers of atmospheric lidar based on cloth simulation, characterized in that: include: The signal preprocessing module is used to preprocess the original echo signal to generate a target signal and flip the target signal in the vertical direction; The cloth simulation module is used to build a cloth model, simulate the cloth falling to the flipped target signal to generate a fitting surface, and flip the fitting surface in the vertical direction to generate the target surface; The vertical profile layer generation module is used to generate the atmospheric lidar vertical profile layer based on the target signal and target surface.

Images