A multi-method fusion laser radar atmospheric boundary layer height identification method

Abstract

The application discloses a multi-method fusion laser radar atmospheric boundary layer height identification method, belongs to the field of laser radar atmospheric detection, and is used for solving the problem that residual layers, cloud layers and atmospheric boundary layer structures are difficult to distinguish. The method processes atmospheric echo signals into distance square correction echo signal data sets, converts into time-high two-dimensional matrix images, and processes to obtain morphological edge detection original gray scale images. After threshold segmentation and morphological operation, a preliminary boundary layer height is obtained by using a region growing method, and edge line types are differentiated in time periods. Gradient method and wavelet covariance method are fused to obtain final atmospheric boundary layer height estimation, final residual layer height estimation and final cloud layer height estimation. All-day data are processed to generate time-high profile graphs containing height data and layer attributes. The application can accurately distinguish atmospheric boundary layers, cloud layers and residual layers, dynamically select wavelet scales, and fuse gradient method, so that the classification precision and identification stability of the atmospheric boundary layer, the cloud layer and the residual layer are improved.

Description

Technical Field

[0001] This invention belongs to the field of atmospheric detection technology, specifically relating to a multi-method fusion lidar method for atmospheric boundary layer height identification. Background Technology

[0002] Boundary layer height is a crucial parameter of the boundary layer, typically ranging from hundreds to thousands of meters and dynamically evolving with time and space. Studying boundary layer height is essential for optimizing regional climate models, improving weather forecast accuracy, and understanding pollutant dispersion mechanisms. Lidar, using aerosols as tracers, can indirectly reflect the dynamic changes of the atmospheric boundary layer and accurately reflect the vertical distribution and temporal and spatial evolution of pollutants, making it an ideal tool for atmospheric boundary layer detection. Common methods for retrieving boundary layer height based on lidar include the gradient method, the variance method, and the wavelet covariance transform method. The gradient method determines the boundary layer height by analyzing the vertical gradient of the echo signal, suitable for scenarios with significant signal changes, but its accuracy may be limited when the signal gradient is relatively flat. The wavelet covariance transform method shows significant potential in boundary layer height retrieval, but a unified standard for selecting its key parameters has not yet been established. The existing technologies mainly suffer from the following problems: First, single algorithms are easily affected by noise, clouds, or residual layers, resulting in unstable recognition results; second, there is a lack of a systematic approach that combines image structural features with traditional algorithms; and third, insufficient consideration is given to the differences in boundary layer types caused by day and night, making it impossible to automatically distinguish between convective boundary layers, stable boundary layers, and residual layers. This leads to significant limitations in boundary layer recognition using lidar in actual monitoring, necessitating a comprehensive atmospheric boundary layer height recognition method that can combine multi-source features and possesses high robustness and high accuracy. Summary of the Invention

[0003] This invention provides a multi-method fusion lidar atmospheric boundary layer height identification method to solve the problems of large noise interference and difficulty in distinguishing residual layers from clouds and atmospheric boundary layer structures in existing traditional lidar atmospheric boundary layer height identification methods.

[0004] A multi-method fusion lidar method for atmospheric boundary layer height identification includes the following steps:

[0005] S1. The atmospheric boundary region to be measured is vertically and continuously probed using lidar to obtain backscattered echo signals. After integrity checks and normalization preprocessing, these signals are then converted into lidar range-squared corrected echo signals. This forms a standardized dataset;

[0006] S2. Construct a time-high two-dimensional matrix image from the standardized dataset through color mapping, and then perform grayscale image conversion, filtering and noise reduction, and contrast enhancement to obtain the original grayscale image for morphological edge detection;

[0007] S3. Threshold segmentation is performed on the original grayscale image to obtain two types of grayscale regions: target region and background region. The edge lines are extracted by performing shrinkage-extension morphology operation on the target region. Then, the edge lines of the target region at each time time are extracted by combining the region growing method to obtain the preliminary atmospheric boundary layer height position. The entire day time period is divided, the target regions for which the preliminary atmospheric boundary layer height has been obtained are excluded, and the edge lines of the remaining target regions are differentially judged.

[0008] S4. The gradient is calculated along the height direction, and the gradient method is used to obtain the altitude corresponding to the extreme value near the initial atmospheric boundary layer height. and Corresponding gradient magnitude, according to The corresponding gradient magnitude completes the differential determination of the edge lines of the remaining target areas in S3; the expansion factor is dynamically adjusted according to the initial atmospheric boundary layer height position. The height was obtained using the wavelet covariance method. fusion and The final atmospheric boundary layer height estimate is obtained; the residual layer height and cloud layer height are both fused using the gradient method and wavelet covariance method to obtain the corresponding final height estimates. The constraints of the fusion weights are consistent with the constraints of the fusion weights of the atmospheric boundary layer height estimate. Data for all times of the day are processed according to S1-S3, and the gradient calculation, height determination and fusion estimation operations are performed in S4 to generate a time-height profile containing height data and layer attributes.

[0009] Furthermore, the backscattered echo signal is valid data within 5km of the ground; the integrity check includes identifying missing data frames, outliers, and backscattered echo signal data with insufficient signal-to-noise ratio, and removing or correcting defective backscattered echo signal data.

[0010] Furthermore, the lidar range-squared corrected echo signal It is calculated using the following formula:

[0011] ;

[0012] In the formula, For lidar system factors; To match the detection altitude Relevant geometric factors; The wavelength of the laser; To detect altitude Atmospheric backscattering coefficient at that location; Atmospheric transmittance; To detect altitude.

[0013] Furthermore, the color mapping is a jet(64) color mapping, including the first row as dark blue, the ninth row as pure dark blue, the seventh row as bright blue, the twentieth row as cyan, the thirtieth row as yellow-green, the 41st row as yellow, the 49th row as orange, the 57th row as pure red, and the 64th row as dark red. Rows not explicitly set will be interpolated based on the colors of the rows above and below to make the color transition smooth.

[0014] Furthermore, the grayscale image is a 256-level grayscale image, including row 1 being pure black, row 64 being dark gray, row 128 being medium gray, row 192 being light gray, and row 256 being pure white. Rows not explicitly set will be interpolated based on the colors of the rows above and below to ensure smooth color transitions. Filtering and noise reduction employs Gaussian weighted filtering to reduce random noise interference, including: weighted averaging based on the grayscale values ​​of the pixel to be filtered and its neighboring pixels, effectively filtering out superimposed noise in the image while preserving image contour information. Contrast enhancement processing includes: using a linear stretching method to make the brightness changes of the atmospheric boundary layer and clouds in the image clearer, improving visual separability and layer discrimination ability.

[0015] Furthermore, threshold segmentation is performed on the original grayscale image, with pixels above the threshold designated as target regions and pixels below the threshold designated as background regions. A shrink-extend morphological operation is then performed on the target regions to extract edge lines. This involves deleting some pixels from the target region to shrink the corresponding grayscale image in the direction of decreasing grayscale value, then adding some pixels from the target region boundary to extend the corresponding grayscale image in the direction of increasing grayscale value. The grayscale image after shrink-extend processing is then compared pixel-by-pixel with the original grayscale image from the morphological edge detection. The pixel with the largest difference in grayscale value is extracted as the edge point, and the edge points are connected to obtain the edge line of the target region.

[0016] Furthermore, the entire day is divided into periods of active convection and other periods. During this division, target areas for which preliminary atmospheric boundary layer heights have been obtained are excluded. Differential determinations are applied to the edge lines of the remaining target areas. During periods of active convection, the edge line with the smallest height value among all target area edge lines extracted using morphological operations combined with region growing is designated as the low-altitude target area edge line, and its height is determined to be the atmospheric boundary layer height. During periods of active convection, all other edge lines higher than the low-altitude target area edge line are determined to be cloud heights. During other periods, the edge line with the smallest height value among all target area edge lines is designated as the low-altitude target area edge line, and its height is determined to be the atmospheric boundary layer height. For target area edge lines in other periods with heights higher than the low-altitude target area edge line, the type is not directly determined. The edge lines for other periods are reserved for distinguishing between residual layers and clouds.

[0017] Furthermore, the distinction between the residual layer and the cloud layer at the edge line in other time periods includes the existence of multiple edge lines in the target area at the corresponding time in other time periods, combined with... The gradient magnitude is differentiated and determined accordingly; the gradient magnitude at the height of the target region's edge is less than... The height of the residual layer is determined by the gradient magnitude of the target area and the height of its edge line below 2km above the ground. The gradient magnitude at the height of the target area's edge line is not less than... The height of the cloud layer is determined by the gradient magnitude or the height of the edge of the target area, which is at least 2km above the ground.

[0018] Furthermore, the wavelet covariance method yields high... This includes using the Haar wavefunction formula and the wavelet covariance formula. The Haar wavefunction formula is as follows:

[0019] ;

[0020] The wavelet covariance formula is:

[0021] ;

[0022] In the formula, It is the Haar wave function; Expansion factor; As a variable; The vertical height detected by the lidar; The numerical range corresponds to the effective detection altitude range of the lidar from 0km to 5km. and These are the upper and lower limits of integration, corresponding to the upper and lower limits of the effective detection height of the lidar, respectively; For high altitude The square-corrected lidar backscattered echo signal is equivalent to ;

[0023] Search near the initial atmospheric boundary layer height The maximum point of the result will be When the maximum value is reached, the corresponding Value as height .

[0024] Furthermore, expansion factor This includes: based on the preliminary atmospheric boundary layer height obtained from S3, when the preliminary atmospheric boundary layer height is less than 500m, selecting 2-3 times the height resolution of the lidar as the expansion factor. When the initial atmospheric boundary layer height is 500–800m, an expansion factor of 5–6 times the height resolution of the lidar is selected. When the initial atmospheric boundary layer height is higher than 800m, an expansion factor of 8-9 times the height resolution of the lidar is selected. ;

[0025] Final Atmospheric Boundary Layer Height Estimation It is obtained from the following formula:

[0026] ;

[0027] In the formula, and These are the weighting coefficients, and both satisfy the weighting constraint conditions. .

[0028] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0029] This invention enhances the visibility of layered structures through image processing, making the changes in the atmospheric boundary layer and clouds clearer. Combining image recognition technology with observation time characteristics, it distinguishes between the atmospheric boundary layer, clouds, and residual layer, effectively improving classification accuracy. In the wavelet covariance method, the wavelet scale parameter is dynamically selected based on the image recognition results, successfully avoiding the difficulty of wavelet scale selection in traditional methods. Furthermore, by integrating this method with the gradient method, it can quickly and accurately locate the boundary layer height while improving the stability and physical consistency of the recognition results. In addition, this invention can effectively identify the boundary layer, residual layer, and clouds, adapting to complex multi-layered atmospheric structures, and can generate time-altitude profile maps, providing a complete and intuitive display of various altitude information, thus facilitating practical applications. Attached Figure Description

[0030] Figure 1 This is a time-height two-dimensional color mapping diagram of the lidar range-squared corrected echo signal in an embodiment of the present invention;

[0031] Figure 2 This is a time-high two-dimensional original grayscale image of the lidar echo signal after noise suppression and contrast enhancement in this embodiment of the invention;

[0032] Figure 3 This is a time-altitude profile of the atmospheric boundary layer-residual layer-cloud layer generated in the embodiment of the present invention;

[0033] Figure 4 This is a flowchart of the present invention. Detailed Implementation

[0034] The present invention will be further illustrated below with reference to embodiments. These embodiments are for illustrative purposes only and are not intended to limit the invention in any way. It should be understood that the described embodiments are merely some, not all, of the embodiments described in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0035] Example

[0036] like Figure 4 As shown, a multi-method fusion lidar method for atmospheric boundary layer height identification includes the following steps:

[0037] S1. A lidar device is used to continuously probe the atmospheric area under test in the vertical direction, obtaining the lidar backscattered echo signal. The backscattered echo signal is valid data within 5 km of the ground. The obtained backscattered echo signal data is then subjected to an integrity check, which includes identifying missing data frames, outliers, and backscattered echo signal data with different signal-to-noise ratios. Defective backscattered echo signal data is discarded or corrected. The echo signals after integrity checks are normalized to obtain a preprocessed signal set. The preprocessed dataset is then processed to obtain the lidar range-squared corrected echo signal. , It is obtained from the following formula:

[0038] ;

[0039] In the formula, For lidar system factors; To match the detection altitude Relevant geometric factors; The wavelength of the laser; To detect altitude Atmospheric backscattering coefficient at that location; Atmospheric transmittance; To detect altitude;

[0040] obtained through This allows us to obtain a standardized dataset of lidar range-squared corrected echo signals.

[0041] S2. The standardized dataset of the lidar range-squared corrected echo signal obtained in S1 is constructed into a time-height two-dimensional matrix image using jet(64) color mapping. Therefore, for a complete day (24h), the data should contain 1440 sets, corresponding to the time range of 1 to 1440 on the horizontal axis; the vertical axis represents the altitude data range of 0 to 5km. The intensity of the echo signal is used as the pixel color value, and jet(64) color mapping is adopted. The jet(64) color mapping includes the first row as dark blue and the ninth row as pure dark blue. The colors are: blue (row 17 is bright blue), cyan (row 25 is cyan), yellowish-green (row 33 is yellowish-green), yellow (row 41 is yellow), orange (row 49 is orange), pure red (row 57 is pure red), and dark red (row 64). Rows not explicitly set will be interpolated based on the colors of the rows above and below to ensure smooth color transitions. During image generation, the continuity of the temporal dimension and the consistency of the height dimension resolution are ensured, and all images undergo uniform standardization processing to ensure consistent brightness ranges across different time periods, guaranteeing the recognition of subsequent boundary layers. Figure 1 As shown, Figure 1 This is a time-height two-dimensional color mapping of the range-squared corrected echo signal of the lidar in the embodiment; Figure 1 The image is mainly suitable for human observation. Through jet(64) color mapping, the observer can quickly distinguish the strong signal area near the ground and intuitively perceive the time-altitude distribution pattern of the atmospheric structure. The gradual transition of colors can also help identify the continuous change of signal intensity, which is convenient for preliminary judgment of atmospheric layers. However, color mapping will reduce the sensitivity to edge direction. Therefore, the jet(64) color mapping is modified to a 256-level grayscale image. The 256-level grayscale image includes the first row as pure black, the 64th row as dark gray, the 128th row as medium gray, the 192nd row as light gray, and the 256th row as pure white. The rows that are not explicitly set will be interpolated according to the colors of the two rows above and below to make the color transition smooth. The grayscale image of the echo signal is plotted, and then Gaussian weighted filtering is applied to reduce random noise interference. The core principle of Gaussian weighted filtering to reduce random noise interference is to perform a weighted average of the grayscale values ​​of the pixel to be filtered and its neighboring points according to certain parameter rules, effectively filtering out superimposed noise in the image while maintaining the image contour information. After noise suppression, the image is then subjected to contrast enhancement processing. Contrast enhancement processing includes using a linear stretching method to make the brightness changes of the atmospheric boundary layer and clouds in the image clearer, improving visual separability and layer discrimination ability. The image obtained after Gaussian weighted filtering for noise reduction and linear stretching for contrast enhancement is used as the original grayscale image for morphological edge detection. Figure 2 As shown, Figure 2 This is the time-high two-dimensional original grayscale image of the lidar echo signal after noise suppression and contrast enhancement in the embodiment; from Figure 2It can be seen that the atmospheric boundary layer and cloud outlines are relatively clear from 10 to 24 hours, and the brightness levels at different altitudes are highly distinguishable.

[0042] S3. Based on S2, the color gradient changes of pixels in the original grayscale image are obtained and divided into two grayscale regions (target region and background region). A threshold is selected to distinguish the two regions by matching the grayscale distribution characteristics of the current image. The selected threshold is 0.55, which corresponds to the dark gray position in the 256-level grayscale color bar. Pixels with a threshold greater than 0.55 are classified as target regions, and pixels with a threshold less than 0.55 are classified as background regions. Then, a shrink-stretch morphology operation is performed on the target region to extract the edge lines. The shrink-stretch morphology operation to extract the edge lines includes deleting... In addition to a subset of pixels within the target region (selected based on grayscale values ​​greater than a preset threshold, causing the corresponding grayscale image to shrink towards lower grayscale values), pixels at the target region boundaries are added. These pixels are selected based on grayscale values ​​less than a preset threshold, causing the corresponding grayscale image to extend towards higher grayscale values. The grayscale image after shrinkage-extension processing is then compared pixel-by-pixel with the original grayscale image from morphological edge detection. The pixel with the largest difference in grayscale value is extracted as the edge point. Connecting these edge points yields the edge lines of each target region. Based on this, to facilitate subsequent observation and accurate recognition, the image is re-processed using... Figure 1 The background is mapped with color, and the extracted edge lines are marked on the image. Then, the region growing analysis method is used to identify targets in the image regions of each time slice, extract the edge lines of the target regions at each time point, and obtain the preliminary atmospheric boundary layer height. Then, the edge line type is determined by combining the position of the edge line in the image with the observation time information. Finally, the edge lines are adjusted according to the subsequent processing results to determine the accurate atmospheric boundary layer height. Based on the correspondence between the height distribution characteristics of the edge lines and the observation time, the edge line types of the target regions are divided into the active convection period and the rest of the time period. The target regions for which the preliminary atmospheric boundary layer height has been obtained are excluded, and the edge lines of the remaining target regions are differentiated.

[0043] The methods for determining each time period are as follows:

[0044] Active convection periods typically correspond to the convective boundary layer (CBL). During this period, the edge line with the smallest height value among all the target region edge lines extracted by morphological operations combined with the region growing method is taken as the low-altitude target region edge line. The height of the edge line of the low-altitude target region is determined as the atmospheric boundary layer height, and the heights of the other edge lines above the target region are determined as cloud heights.

[0045] During the remaining time periods, the edge line with the smallest height value among all target area edge lines extracted by morphological operations combined with the region growing method is taken as the edge line of the low-altitude target area. The height of the edge line of the low-altitude target area is determined as the atmospheric boundary layer height. Since there may be a residual layer (RL), the heights of other edge lines above the edge line of the low-altitude target area cannot be determined as cloud heights. Edge lines are reserved for determination at other times.

[0046] S4. For the corresponding time... The gradient is calculated along the height direction. Using the gradient method, the abrupt change points and maxima of the gradient are searched near the initial atmospheric boundary layer height obtained in S3 to obtain the height corresponding to the extreme value at the current moment. and obtain The corresponding gradient magnitude; based on The determination of the attributes of multiple edge lines in the target area based on the gradient magnitude and time period division rules includes: in other time periods, there are multiple edge lines in the target area at the corresponding time, combined with... The gradient magnitude is differentiated and determined accordingly; the gradient magnitude at the height of the target region's edge is less than... The height of the residual layer is determined by the gradient magnitude of the target area and the height of its edge line below 2km above the ground. The gradient magnitude at the height of the target area's edge line is not less than... The height of the cloud layer is determined by the gradient magnitude or the height of the edge of the target area, which is at least 2km above the ground.

[0047] The expansion factor is dynamically determined based on the preliminary atmospheric boundary layer height obtained from S3. expansion factor The rules for determining the value include: if the preliminary atmospheric boundary layer height obtained from S3 is less than 500m, then 2-3 times the height resolution of the lidar should be selected as the expansion factor. If the initial atmospheric boundary layer height is 500–800m, then an expansion factor of 5–6 times the height resolution of the lidar should be selected. If the initial atmospheric boundary layer height is higher than 800m, then an expansion factor of 8-9 times the height resolution of the lidar should be selected. Height obtained by wavelet covariance method The height was obtained using the wavelet covariance method. This includes the Haar wavefunction formula and wavelet covariance formula used. The Haar wavefunction formula is as follows:

[0048] ;

[0049] The wavelet covariance formula is as follows:

[0050] ;

[0051] In the formula, It is the Haar wave function; Expansion factor; As a variable; The vertical height detected by the lidar; The numerical range corresponds to the effective detection altitude range of the lidar from 0km to 5km; and These are the upper and lower limits of integration, corresponding to the upper and lower limits of the effective detection height of the lidar, respectively; For high altitude The square-corrected lidar backscattered echo signal is equivalent to ;

[0052] Searching near the initial atmospheric boundary layer height obtained in S3 The maximum point of the result will be When the maximum value is reached, the corresponding The value is determined as height. ;

[0053] Fusion and The final atmospheric boundary layer height estimate is obtained. It is obtained from the following formula:

[0054] ;

[0055] In the formula, and These are the weighting coefficients, and both satisfy the weighting constraint conditions. This embodiment preferably adopts... By using the gradient method and wavelet covariance method with equal weighted fusion, a stable and physically consistent final atmospheric boundary layer height estimate for the current moment is obtained.

[0056] The residual layer height and cloud height are fused using the gradient method and wavelet covariance method with equal weights (the weight constraints are still satisfied). Preferred This process obtains a stable and physically consistent final residual layer height estimate and final cloud layer height estimate for the current moment. Data from all times of the day is processed according to steps S1-S3, followed by gradient calculation, height determination, and fusion estimation operations in step S4. Finally, a time-height profile containing height data and layer attributes is generated. Figure 3 As shown, Figure 3 This is a time-altitude profile of the atmospheric boundary layer-residual layer-cloud layer generated in the embodiment of the present invention. Figure 3The pink curve represents the final atmospheric boundary layer height measured on that day, the white curve represents the final cloud height measured on that day, and the orange-yellow curve represents the final residual top height measured on that day. This image visually presents the diurnal variation structure of the atmospheric boundary layer.

[0057] This invention will The signal is converted into a time-height two-dimensional matrix image. After filtering enhancement, threshold segmentation, and morphological operations to extract edge lines, the preliminary atmospheric boundary layer height is obtained by combining the region growing method. On the other hand, the whole day is divided into convection-active periods and other periods. Differentiated edge line judgment rules are formulated to reduce the risk of misjudgment from the source. The gradient method and adaptive wavelet covariance method are fused to accurately distinguish the atmospheric boundary layer, cloud layer and residual layer, and generate a time-height profile map containing height data and layer attributes. The application value and robustness are significant.

[0058] This invention achieves accurate identification of multi-layered structures in complex atmospheric scenes through image screening, time segmentation, and multi-algorithm fusion, effectively improving identification accuracy and robustness, and providing reliable data support for atmospheric environment assessment and air quality forecasting.

[0059] Of course, the above description is not intended to limit the present invention, and the present invention is not limited to the examples given above. Any changes, modifications, additions or substitutions made by those skilled in the art within the scope of the present invention should also fall within the protection scope of the present invention.

Claims

1. A multi-method fusion lidar method for atmospheric boundary layer height identification, characterized in that, Includes the following steps: S1. The atmospheric boundary region to be measured is vertically and continuously probed using lidar to obtain backscattered echo signals. After integrity checks and normalization preprocessing, these signals are then converted into lidar range-squared corrected echo signals. This forms a standardized dataset; S2. Construct a time-high two-dimensional matrix image from the standardized dataset through color mapping, and then perform grayscale image conversion, filtering and noise reduction, and contrast enhancement to obtain the original grayscale image for morphological edge detection; S3. Threshold segmentation is performed on the original grayscale image to obtain two types of grayscale regions: target region and background region. The edge lines are extracted by performing shrinkage-extension morphology operation on the target region. Then, the edge lines of the target region at each time time are extracted by combining the region growing method to obtain the preliminary atmospheric boundary layer height position. The day is divided into periods of active convection and other periods. Excluding target areas where the preliminary atmospheric boundary layer height has been obtained, differential determination is performed on the edge lines of the remaining target areas. During periods of active convection, the edge line with the smallest height value among all target area edge lines extracted by morphological operations combined with region growing method is taken as the edge line of the low-altitude target area, and the height of the edge line of the low-altitude target area is determined as the atmospheric boundary layer height. During periods of active convection, the edge lines of other target areas with heights higher than the edge line of the low-altitude target area are determined as cloud heights. During other periods, the edge line with the smallest height value among all target area edge lines is taken as the edge line of the low-altitude target area, and the height of the edge line of the low-altitude target area is determined as the atmospheric boundary layer height. For target area edge lines with heights higher than the edge line of the low-altitude target area during other periods, the type is not directly determined, and the edge lines of other periods reserve the distinction between residual layer and cloud layer. S4. The gradient is calculated along the height direction, and the gradient method is used to obtain the altitude corresponding to the extreme value near the initial atmospheric boundary layer height. and Corresponding gradient magnitude, according to The corresponding gradient magnitude completes the distinction between the reserved residual layer and the cloud layer at the edge line in the other time periods of S3, including the existence of multiple edge lines in the target area at the corresponding time in the other time periods, combined with The gradient magnitude is differentiated and determined accordingly; the gradient magnitude at the height of the target region's edge is less than... The height of the residual layer is determined by the gradient magnitude of the target area and the height of its edge line below 2km above the ground. The gradient magnitude at the height of the target area's edge line is not less than... The cloud height is determined by the gradient amplitude or the height of the edge of the target area at a height of not less than 2 km above the ground; the expansion factor is dynamically adjusted according to the initial atmospheric boundary layer height. The height was obtained using the wavelet covariance method. fusion and The final atmospheric boundary layer height estimate is obtained; the residual layer height and cloud layer height are both fused using the gradient method and wavelet covariance method to obtain the corresponding final height estimates. The constraints of the fusion weights are consistent with the constraints of the fusion weights of the atmospheric boundary layer height estimate. Data for all times of the day are processed according to S1-S3, and the gradient calculation, height determination and fusion estimation operations are performed in S4 to generate a time-height profile containing height data and layer attributes.

2. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, Backscattered echo signals are valid data within 5km of the ground; integrity checks include identifying missing data frames, outliers, and backscattered echo signal data with insufficient signal-to-noise ratio, and removing or correcting defective backscattered echo signal data.

3. The method for identifying atmospheric boundary layer height using lidar based on multi-method fusion according to claim 1, characterized in that, LiDAR range-squared corrected echo signal It is calculated using the following formula: ； In the formula, For lidar system factors; To match the detection altitude Relevant geometric factors; The wavelength of the laser; To detect altitude Atmospheric backscattering coefficient at that location; Atmospheric transmittance; To detect altitude.

4. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, The color mapping is a jet(64) color mapping, including the first row as dark blue, the ninth row as pure dark blue, the seventh row as bright blue, the twentieth row as cyan, the thirtieth row as yellow-green, the 41st row as yellow, the 49th row as orange, the 57th row as pure red, and the 64th row as dark red. Rows not explicitly set will be interpolated based on the colors of the rows above and below to make the color transition smooth.

5. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, The grayscale image is a 256-level grayscale image, including row 1 as pure black, row 64 as dark gray, row 128 as medium gray, row 192 as light gray, and row 256 as pure white. Rows not explicitly set will be interpolated based on the colors of the rows above and below to make the color transition smooth. The filtering and noise reduction uses Gaussian weighted filtering to reduce random noise interference, including weighted averaging based on the grayscale values ​​of the pixel to be filtered and its neighboring pixels, effectively filtering out superimposed noise in the image while maintaining the image contour information. Contrast enhancement processing includes: using a linear stretching method to make the brightness changes of the atmospheric boundary layer and clouds in the image clearer, thereby improving visual separability and the ability to distinguish layers.

6. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, The original grayscale image is segmented by thresholding, with pixels larger than the threshold defined as the target region and pixels smaller than the threshold defined as the background region. The target region is subjected to shrink-extend morphological operations to extract edge lines. This includes deleting some pixels in the target region to shrink the corresponding grayscale image in the direction of decreasing grayscale value, and then adding some pixels at the boundary of the target region to extend the corresponding grayscale image in the direction of increasing grayscale value. The grayscale image after shrink-extend processing is compared pixel by pixel with the original grayscale image of morphological edge detection. The pixel point with the largest difference in grayscale value is extracted as the edge point. The edge points are connected to obtain the edge line of the target region.

7. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, Wavelet covariance method yields high This includes using the Haar wavefunction formula and the wavelet covariance formula. The Haar wavefunction formula is as follows: ； The wavelet covariance formula is: ； In the formula, It is the Haar wave function; Expansion factor; As a variable; The vertical height detected by the lidar; The numerical range corresponds to the effective detection altitude range of the lidar from 0km to 5km. and These are the upper and lower limits of integration, corresponding to the upper and lower limits of the effective detection height of the lidar, respectively; For high altitude The square-corrected lidar backscattered echo signal is equivalent to ; Search near the initial atmospheric boundary layer height The maximum point of the result will be When the maximum value is reached, the corresponding Value as height .

8. The multi-method fusion lidar atmospheric boundary layer height identification method according to claim 1, characterized in that, expansion factor This includes: based on the preliminary atmospheric boundary layer height obtained from S3, when the preliminary atmospheric boundary layer height is less than 500m, selecting 2-3 times the height resolution of the lidar as the expansion factor. When the initial atmospheric boundary layer height is 500–800m, an expansion factor of 5–6 times the height resolution of the lidar is selected. When the initial atmospheric boundary layer height is higher than 800m, an expansion factor of 8-9 times the height resolution of the lidar is selected. ; Final Atmospheric Boundary Layer Height Estimation It is obtained from the following formula: ； In the formula, and These are the weighting coefficients, and both satisfy the weighting constraint conditions. .

Images