Self-adaptive filtering method based on signal-to-noise ratio

Abstract

The invention provides an adaptive filtering method based on a signal-to-noise ratio and a mean filter. According to the method, a smooth window is dynamically adjusted through an SNR (Signal to Noise Ratio) quantitative index. And in a near-field region with strong signals, the SNR is high, a small window is automatically adopted for filtering, and signal details are reserved preferentially, so that fine structures such as a cloud layer and a boundary layer top are reserved to the greatest extent. And in a far-field area with weak signals, the SNR is low, a large window is automatically adopted, the filtering task is preferential, noise is strongly suppressed, and the availability of far-field data is improved. According to the method, the size of the filter window can be adaptively adjusted through the signal-to-noise ratio of the laser radar echo signal, the high operation speed is kept, few operation resources are occupied, high costs such as manpower and equipment do not need to be additionally paid, and the method is very suitable for real-time online inversion of the atmospheric laser radar.

Description

Technical Field

[0001] This invention relates to the field of atmospheric sounding lasers, and more specifically to an adaptive filtering method based on signal-to-noise ratio. Background Technology

[0002] Atmospheric lidar emits laser beams of specific wavelengths into the atmosphere and receives the backscattered signals, thereby retrieving the distribution and characteristics of atmospheric components such as aerosols and clouds. The 1064nm band, due to its strong scattering response to atmospheric particles, is widely used for detecting the optical properties of aerosols and clouds. However, in actual detection processes, the echo signals received by the radar not only contain effective atmospheric scattering information but also inevitably contain various types of noise, such as background radiation noise, detector thermal noise, and signal quantization noise. These noises severely degrade signal quality, affecting the accuracy and reliability of subsequent optical parameter inversion.

[0003] Traditional fixed-window filtering methods such as mean filtering and median filtering have inherent limitations. Once the filter window size is set, it cannot be changed throughout the filtering process. These methods are only suitable for filtering stationary time series and cannot handle non-stationary sequences like atmospheric lidar echo profiles, lacking adaptability to local signal characteristics. While using larger filters can effectively suppress noise in the far field, it can lead to over-smoothing of high signal-to-noise ratio signals in the near field, resulting in the loss of useful details such as fine cloud base structures and aerosol layer gradients. Conversely, while smaller filters can preserve near-field details, noise in weak signals in the low signal-to-noise ratio far field is not effectively suppressed, increasing the uncertainty of inversion parameters (such as optical parameters and depolarization ratio) and potentially leading to erroneous results.

[0004] Besides fixed-window filtering methods, other filtering methods used in the industry include frequency decomposition and reconstruction (FDRC) and deep learning methods. FDRC suffers from a large number of hyperparameters; the same set of parameters cannot be applied to LiDAR echo signals under different environments. Parameter values ​​need to be adjusted empirically based on different echo signal states to achieve a good filtering effect, making it unsuitable for real-time online inversion equipment. Deep learning methods, on the other hand, have limited applicability due to the difficulty in creating datasets and the high cost of training (requiring graphics cards and other high-cost components). Summary of the Invention

[0005] One or more embodiments of this application describe an adaptive filtering method based on signal-to-noise ratio, which can at least partially overcome the above-mentioned technical problems.

[0006] Firstly, an adaptive filtering method based on signal-to-noise ratio is provided, which includes: Acquire the echo signal from the atmospheric sounding lidar; For each data point of the echo signal, the signal-to-noise ratio (SNR) of that data point is calculated to obtain an SNR curve with the same length as the echo signal, and the maximum SNR value on the SNR curve is determined. and minimum signal-to-noise ratio ; Set the minimum window size for adaptive mean filtering. and maximum window size ; Establish a linear mapping relationship between signal-to-noise ratio and window size: ; in, Indicates the first The window size corresponding to each data point Indicates the first The signal-to-noise ratio corresponding to each data point; right Perform a rounding operation to obtain ; For each data point of the echo signal, according to the corresponding window size Create a mean filter to perform mean filtering on the data points.

[0007] In one optional implementation, the method further includes: Extract the tail data segment of the echo signal, calculate the standard deviation of the tail data segment, and use the standard deviation as a noise benchmark to calculate the signal-to-noise ratio of the data point.

[0008] In one alternative implementation, for Perform a rounding operation to obtain Specifically, it includes: right Perform a rounding operation. If the rounded value is even, add 1 to the rounded value to make it odd.

[0009] In one alternative implementation, wherein, .

[0010] In an optional implementation, the method further includes: The echo signal is mirror-symmetrically extended at both ends.

[0011] Specifically, the echo signal is mirror-symmetrically extended at both ends, including: Select the head of the echo signal For each data point, the selected data points are mirrored and inverted to obtain header extension data, which is then added to the header of the echo signal. Select the tail of the echo signal Each data point is mirrored to obtain tail extension data, which is then added to the tail of the echo signal. This indicates the number of the last data point of the echo signal.

[0012] In one optional implementation, the mean filter corresponding to each data point is represented as a filter of length [length missing]. A vector, wherein each element of the vector is .

[0013] Specifically, for each data point of the echo signal, according to the corresponding window size Creating a mean filter to perform mean filtering on the data points specifically includes: For each data point of the echo signal, find the data point centered at that data point. 1 data point, as the data segment to be smoothed; The data segment to be smoothed is convolved with the vector of the mean filter to obtain a scalar value, which is the filtering result of the corresponding data point of the echo signal.

[0014] Furthermore, the method also includes: Create an all-zero matrix with the same length as the echo signal; The scalar value corresponding to each data point of the echo signal is used to replace the element value at the corresponding position in the all-zero matrix to obtain the updated result matrix; After all data points of the echo signal are subjected to mean filtering, the data stored in the result matrix is ​​used as the mean filtering result.

[0015] Secondly, a mean filter is provided, which uses the aforementioned adaptive filtering method based on signal-to-noise ratio to perform mean filtering on the acquired target echo signal.

[0016] Beneficial Effects: One or more embodiments of this application provide an adaptive filtering method based on signal-to-noise ratio (SNR). This method dynamically adjusts the smoothing window using SNR as a quantization metric. In the near-field region with strong signals and high SNR, a small window is automatically used for filtering, prioritizing the preservation of signal details and maximizing the retention of fine structures such as clouds and boundary layer tops. In the far-field region with weak signals and low SNR, a large window is automatically used, prioritizing filtering, strongly suppressing noise, and improving the usability of far-field data.

[0017] This method provides a real-time noise estimate that matches the current data acquisition conditions. It exhibits greater adaptability and robustness to changing observation conditions.

[0018] This method can adaptively adjust the filter window size based on the signal-to-noise ratio of the lidar echo signal, while maintaining high computing speed and consuming less computing resources. It does not require additional high costs such as manpower and equipment, making it very suitable for real-time online inversion of atmospheric lidar.

[0019] It should be noted that one or more embodiments of this application provide a mean filter that also has the above-mentioned beneficial effects. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an adaptive filtering method based on signal-to-noise ratio involved in one or more embodiments of this application.

[0022] Figure 2 This is a schematic diagram of the structure of a mean filter involved in one or more embodiments of this application. Detailed Implementation

[0023] First, it should be noted that the terminology used in the embodiments of this invention is for the purpose of describing specific embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” used in the embodiments of this invention and the appended claims are also intended to include the plural forms, unless the context clearly indicates otherwise.

[0024] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0025] It should be noted that the steps of the corresponding methods in other embodiments are not necessarily performed in the order shown and described in this application. In some other embodiments, the methods may include more or fewer steps than those described in this application. Furthermore, a single step described in this application may be broken down into multiple steps in other embodiments; and multiple steps described in this application may be combined into a single step in other embodiments.

[0026] The adaptive filtering method based on signal-to-noise ratio described in this application will be further described in detail below with reference to the accompanying drawings and specific embodiments. However, this detailed description does not constitute a limitation on the embodiments of this application.

[0027] Please refer to Figure 1 , Figure 1 A schematic diagram illustrating the process of an adaptive filtering method based on signal-to-noise ratio is shown. Figure 1 As shown, the method includes steps S100 to S110.

[0028] S100: Acquire the echo signal from the atmospheric sounding lidar.

[0029] The raw echo signal from the atmospheric sounding lidar is sampled to obtain discrete echo signals.

[0030] S102: For each data point of the echo signal, calculate the signal-to-noise ratio (SNR) of the data point to obtain an SNR curve with the same length as the echo signal, and determine the maximum and minimum SNR values ​​on the SNR curve.

[0031] For each data point of the echo signal The signal-to-noise ratio (SNR) is calculated using the SNR formula: ; This yields a signal-to-noise ratio curve with the same length as the echo signal, where... Representing data points The signal-to-noise ratio.

[0032] In the above formula for calculating the signal-to-noise ratio, This is the noise standard deviation. In this step, to match this noise standard deviation with the current real-time data acquisition conditions, the tail segment of the echo signal is selected (this is because the signal has attenuated to near noise levels at the tail). The standard deviation of the tail segment is calculated and used as the noise benchmark to calculate the signal-to-noise ratio of the data points.

[0033] After obtaining the signal-to-noise ratio (SNR) curve, determine the maximum SNR value on the curve. And determine the maximum signal-to-noise ratio value on the signal-to-noise ratio curve. and minimum signal-to-noise ratio .

[0034] S104: Set the minimum and maximum window sizes for adaptive mean filtering.

[0035] Let the minimum window size be... The maximum window size is Considering the symmetry of the filtering window, these two values ​​are generally odd numbers, used to represent the length of the filtering window, and .

[0036] S106: Establish a linear mapping relationship between signal-to-noise ratio and window size.

[0037] To use a small window (to preserve details) at high signal-to-noise ratios and a large window (to smooth noise) at low signal-to-noise ratios, an inverse linear mapping relationship is established: ; in, Indicates the first The window size corresponding to each data point Indicates the first The signal-to-noise ratio corresponding to each data point.

[0038] S108: Round the calculated window size.

[0039] The calculation at this time It is in floating-point form, therefore it needs to be... Perform a rounding operation to obtain .

[0040] The specific rounding operation is as follows: right Perform a rounding operation. If the rounded value is even, add 1 to the rounded value to make it odd. Simultaneously, ensure... That is, after the above rounding and even-numbering processes, if The last value Greater than Then the last one Subtract 2, making fall into . This indicates the number of the last data point in the echo signal.

[0041] S110: For each data point of the echo signal, according to the corresponding window size Create a mean filter to perform mean filtering on the data points.

[0042] In some implementations, before performing mean filtering on the echo signal, the two ends of the echo signal can be mirrored symmetrically extended to define the boundaries of the echo signal.

[0043] Specifically, select the head of the echo signal. For each selected data point, a mirror image is created to obtain header extension data, which is then appended to the header of the echo signal.

[0044] Select the tail of the echo signal Selected data points are mirrored to obtain tail extension data, which is then appended to the tail of the echo signal.

[0045] Before performing mean filtering, the window size corresponding to each data point is determined. Create a mean filter, which is represented as a filter of length [length missing]. A vector, where each element is .

[0046] Next, for each data point of the echo signal Find the data point Centered 10 data points are used as the data segment to be smoothed.

[0047] Convolve the data segment to be smoothed with the vector of the mean filter to obtain a scalar value. This scalar value is... The filtering results.

[0048] In some implementations, a zero matrix (one-dimensional matrix) with the same length as the echo signal can be created to store the filtered result.

[0049] The algorithm iterates through each data point of the echo signal. After performing mean filtering on each data point, the corresponding element in the all-zero matrix is ​​replaced with the scalar value corresponding to each data point in the echo signal, resulting in an updated matrix. Once mean filtering has been performed on all data points of the echo signal, the data stored in the resulting matrix is ​​used as the mean filtering result.

[0050] The above describes an adaptive filtering method based on signal-to-noise ratio (SNR) provided in this embodiment. This method dynamically adjusts the smoothing window using the SNR quantization metric. In the near-field region with strong signals and high SNR, a small window is automatically used for filtering, prioritizing the preservation of signal details and maximizing the retention of fine structures such as clouds and boundary layer tops. In the far-field region with weak signals and low SNR, a large window is automatically used, prioritizing filtering and strongly suppressing noise to improve the usability of far-field data.

[0051] This method provides a real-time noise estimate that matches the current data acquisition conditions. It exhibits greater adaptability and robustness to changing observation conditions.

[0052] This method can adaptively adjust the filter window size based on the signal-to-noise ratio of the lidar echo signal, while maintaining high computing speed and consuming less computing resources. It does not require additional high costs such as manpower and equipment, making it very suitable for real-time online inversion of atmospheric lidar.

[0053] Corresponding to the above-mentioned adaptive filtering method based on signal-to-noise ratio, this embodiment also provides a mean filter, which is used to perform mean filtering processing on the acquired target echo signal using the above-mentioned adaptive filtering method based on signal-to-noise ratio.

[0054] Please refer to Figure 2 , Figure 2 A schematic diagram of the structure of a mean filter is shown. For example... Figure 2 As shown, the mean filter includes: The data acquisition module is used to acquire the echo signals from the atmospheric sounding lidar.

[0055] The signal-to-noise ratio (SNR) calculation module calculates the SNR for each data point of the echo signal, obtains an SNR curve with the same length as the echo signal, and determines the maximum SNR value on the SNR curve. and minimum signal-to-noise ratio .

[0056] The window size calculation module is used to set the minimum window size for adaptive mean filtering. and maximum window size And establish a linear mapping relationship between signal-to-noise ratio and window size: ; in, Indicates the first The window size corresponding to each data point Indicates the first The signal-to-noise ratio corresponding to each data point.

[0057] The data processing module is used for... Perform a rounding operation to obtain .

[0058] The filtering module is used to filter each data point of the echo signal according to the corresponding window size. Create a mean filter to perform mean filtering on the data points.

[0059] Optionally, the aforementioned signal-to-noise ratio calculation module is specifically used to extract the tail data segment of the echo signal, calculate the standard deviation of the tail data segment, and use the standard deviation as a noise benchmark to calculate the signal-to-noise ratio of the data points.

[0060] Optionally, the data processing module is specifically used for: right Perform a rounding operation. If the rounded value is even, add 1 to the rounded value to make it odd.

[0061] In addition, the data processing module also needs to ensure .

[0062] Optionally, the data processing module is also specifically used for: The echo signal is mirror-symmetrically extended at both ends. Specifically, the data processing module can use the following scheme to perform mirror-symmetric extension on both ends of the echo signal: Select the head of the echo signal For each data point, the selected data points are mirrored and inverted to obtain header extension data, which is then added to the header of the echo signal. Select the tail of the echo signal Each data point is mirrored to obtain tail extension data, which is then appended to the tail of the echo signal. This indicates the number of the last data point in the echo signal.

[0063] Optionally, the mean filter created by the filtering module is represented as a filter of length [length missing]. A vector, where each element is .

[0064] Specifically, for each data point in the echo signal, the filtering module finds the data point centered on the... 1 data point, as the data segment to be smoothed; Convolve the data segment to be smoothed with the vector of the mean filter to obtain a scalar value, which is the filtering result of the corresponding data point of the echo signal.

[0065] Optionally, the filtering module can also create an all-zero matrix with the same length as the echo signal; When performing mean filtering on the echo signal, the element value at the corresponding position in the all-zero matrix is ​​replaced with the scalar value corresponding to each data point of the echo signal to obtain the updated result matrix; After all data points of the echo signal are subjected to mean filtering, the data stored in the result matrix is ​​used as the mean filtering result.

[0066] It is understood that the structures illustrated in the embodiments of this application do not constitute a specific limitation on the system of the embodiments of this application. In other embodiments of the specification, the above system may include more or fewer components than illustrated, or combine some components, or split some components, or have different component arrangements. The illustrated components may be implemented in hardware, software, or a combination of software and hardware.

[0067] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the device embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0068] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0069] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

Claims

1. A method of adaptive filtering based on signal-to-noise ratio, characterized by, The method comprises: obtaining an echo signal of an atmospheric sounding laser radar; For each data point of the echo signal, a signal-to-noise ratio of the data point is calculated, a signal-to-noise ratio curve with the same length as the echo signal is obtained, and a maximum signal-to-noise ratio value on the signal-to-noise ratio curve is determined and a minimum signal-to-noise ratio value ; Setting minimum window size for adaptive mean filtering and maximum window size ; establishing a linear mapping relationship between the signal-to-noise ratio and the window size; ； wherein, represents the window size corresponding to the th data point, represents the signal-to-noise ratio corresponding to the th data point; ;​​ for each data point of the echo signal, according to a corresponding window size creating a mean filter, and performing mean filtering on the data points.

2. The method of claim 1, wherein, The method further comprises: truncating a tail data segment of the echo signal, calculating the standard deviation of the tail data segment, and taking the standard deviation as a noise reference to calculate the signal-to-noise ratio of the data points.

3. The method of claim 1, wherein, right Perform a rounding operation to obtain Specifically, it includes: to performing a rounding operation, and if the rounded value is even, then adding 1 to the rounded value to make the rounded value odd.

4. The method of claim 1, wherein, Wherein, 。 5. The method of claim 1, wherein, The method further comprises: mirror-symmetrically extending both ends of the echo signal.

6. The method of claim 5, wherein, Mirror-symmetrically extending both ends of the echo signal specifically comprises: selecting a data point at the head of the echo signal mirror-inverting the selected data point to obtain head extension data, and supplementing the head extension data to the head of the echo signal selecting a data point at the tail of the echo signal mirroring and inverting the selected data point to obtain tail extension data, and appending the tail extension data to the tail of the echo signal; indicates the number of the last data point of the echo signal.

7. The method of claim 1, wherein, Each of the data points corresponds to a mean filter represented as a vector of length , where each element of the vector is .

8. The method of claim 7, wherein, For each data point of the echo signal, according to the corresponding window size A mean filter is created, and the data points are subjected to mean filtering, specifically including: For each data point of the echo signal, find a data point centered around the data point as a data segment to be smoothed ; convolving the to-be-smoothed data segment with the vector of the mean filter to obtain a scalar value, the scalar value being the filtering result of the corresponding data point of the echo signal.

9. The method of claim 8, wherein, The method further comprises: creating a full-zero matrix with the same length as the echo signal; replacing the element values at the corresponding positions in the full-zero matrix with the scalar values corresponding to each data point of the echo signal to obtain an updated result matrix; after mean filtering processing is performed on all data points of the echo signal, taking the data stored in the result matrix as the mean filtering result.

10. A mean filter characterized by The mean filter adopts the method of any one of claims 1 to 9 to perform mean filtering processing on the obtained target echo signal.

Images