CN114563771A - Double-threshold laser radar cloud layer detection algorithm based on cluster analysis - Google Patents

Double-threshold laser radar cloud layer detection algorithm based on cluster analysis Download PDF

Info

Publication number
CN114563771A
CN114563771A CN202111618360.4A CN202111618360A CN114563771A CN 114563771 A CN114563771 A CN 114563771A CN 202111618360 A CN202111618360 A CN 202111618360A CN 114563771 A CN114563771 A CN 114563771A
Authority
CN
China
Prior art keywords
cloud
layer
cloud layer
laser radar
signals
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111618360.4A
Other languages
Chinese (zh)
Inventor
常建华
陈思成
王博业
周妹
孟园园
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing University of Information Science and Technology
Original Assignee
Nanjing University of Information Science and Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Nanjing University of Information Science and Technology filed Critical Nanjing University of Information Science and Technology
Priority to CN202111618360.4A priority Critical patent/CN114563771A/en
Publication of CN114563771A publication Critical patent/CN114563771A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/48—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00
    • G01S7/4802—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S17/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
    • G01S17/88—Lidar systems specially adapted for specific applications
    • G01S17/95—Lidar systems specially adapted for specific applications for meteorological use
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The invention discloses a cluster analysis-based dual-threshold laser radar cloud layer detection algorithm, which comprises the following steps: preprocessing a laser radar echo signal; screening the preprocessed laser radar echo signals by using a layer peak-to-layer bottom ratio threshold value, and extracting cloud layer signals; extracting the missed thin cloud layer signals from the preprocessed laser radar echo signals by using a background noise threshold value; acquiring differential zero points of each cloud layer by using a differential zero crossing method and calculating the heights of the cloud top and the cloud bottom in an inversion manner; and performing cluster analysis on the relative humidity and the vertical height of each cloud layer differential zero point by using an ISODATA clustering algorithm, clustering samples with similar relative humidity characteristics in the same height range into one class, obtaining a plurality of classified differential zero point sets, and judging the cloud layer structure to obtain cloud layer information. The method effectively eliminates the interference of aerosol signals, can accurately screen out cloud layer signals, and realizes the accurate processing of cloud layering.

Description

Double-threshold laser radar cloud layer detection algorithm based on cluster analysis
Technical Field
The invention relates to a double-threshold laser radar cloud layer detection algorithm based on cluster analysis, and belongs to the technical field of laser radar detection.
Background
The cloud covers 60% of the global area, and participates in large-scale circulation and global water circulation by influencing solar radiation, latent heat release and the like, and has an important role in global climate change research. And the laser radar has high detection precision and strong sustainability, and is widely applied to cloud detection. The laser cloud measuring instrument adopts a laser radar as main equipment, emits laser pulses to the atmosphere when in work, and can invert cloud layer information such as cloud bottom height, cloud top height and the like by analyzing received laser echo signals, so that the accurate judgment of the cloud layer signals is very critical.
In view of the real-time, efficient and continuous characteristics of the laser radar in the aspect of high space-time resolution cloud structure detection, the laser radar becomes an effective means for detecting space-time distribution of atmospheric physical parameters, cloud, aerosol and the like gradually, and is widely applied to continuous monitoring of space distribution of medium and low-altitude atmospheric aerosol and cloud layers. The differential zero crossing method is a common inversion method for cloud parameter inversion based on laser radar echo signals, but the method is easy to cause misjudgment of cloud layer signals under the condition of low signal-to-noise ratio, and the detection precision is influenced. Typically, the cloud is much denser than the aerosol in the tropospheric atmosphere. However, when a thin cloud layer signal similar to the aerosol layer signal or a thick aerosol layer signal similar to the cloud layer signal interferes, it is difficult to extract cloud layer information directly from the lidar echo signal. Although the algorithm has been improved in recent years, methods such as fitting data, judging in combination with the time of the front and back neighborhoods of the echo signal, combining a threshold value and the like are adopted. However, the following disadvantages still exist: 1. interference of aerosol signals exists, and correct cloud layer information cannot be provided; 2. when background noise in a scene fluctuates greatly, the method can generate large errors; 3. the setting of the threshold is subjective and is very prone to deviation in practical application.
The existence of the problems seriously restricts the working efficiency of the laser radar for measuring the cloud height, and the realization of accurate measurement of cloud layer signals is the key point for solving the problems.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a double-threshold laser radar cloud layer detection algorithm based on cluster analysis, so that the interference of aerosol signals is effectively eliminated and the cloud layer signals are accurately screened out.
The invention specifically adopts the following technical scheme to solve the technical problems:
a double-threshold laser radar cloud layer detection algorithm based on cluster analysis comprises the following steps:
step 1, preprocessing a laser radar echo signal;
step 2, processing the preprocessed laser radar echo signals by adopting a double-threshold value: screening the preprocessed laser radar echo signals by using a layer peak-to-layer bottom ratio threshold value, and extracting cloud layer signals; extracting the missed thin cloud layer signals from the preprocessed laser radar echo signals by using a background noise threshold value;
step 3, based on the extracted cloud layer signals and thin cloud layer signals, carrying out differential processing by using a differential zero crossing method to obtain differential zero points of the extracted cloud layer signals, and carrying out inversion to calculate the heights of the cloud top and the cloud bottom;
step 4, carrying out cluster analysis on the differential zero points of the cloud layer signals by using an ISODATA (inverse discrete cosine transform) clustering algorithm according to the relative humidity and the vertical height, and clustering samples with similar relative humidity characteristics in the same height range into a class, namely each class of differential zero points can be regarded as the differential zero points of the same layer of cloud, so as to obtain a plurality of classified differential zero point sets; and judging the cloud layer structure according to the classified multiple differential zero point sets to obtain corresponding cloud layer information.
Further, as a preferred technical solution of the present invention, the preprocessing of the laser radar echo signal in step 1 includes: and denoising the initial laser radar echo signal by adopting an LMD-ITM algorithm, and performing 5-point moving average processing on the denoised signal.
Further, as a preferred technical solution of the present invention, in the step 2, the preprocessed laser radar echo signal is screened by using a layer-to-peak-to-layer bottom ratio threshold, and a formula is adopted:
Figure BDA0003434912880000021
wherein X is a layer peak-to-layer bottom ratio threshold; r is a radical of hydrogenm、rbThe positions of the cloud layer peak and the layer bottom, P (r)b)、P(rm) Respectively the laser radar echo signal intensity of the cloud layer peak and the layer bottom.
Further, as a preferred technical solution of the present invention, in the step 2, the thin cloud layer signal that is missed to be selected is extracted from the preprocessed laser radar echo signal by using a background noise threshold, specifically:
when the laser radar detects the distance r<At 5km, the background noise threshold Y is set to [ P (r) ]m)-P(rb)]>3I, wherein P (r)b)、P(rm) Respectively the intensity of the laser radar echo signals of the cloud layer peak and the layer bottom, and I is the intensity of background noise;
when the detection distance r of the laser radar is more than or equal to 5km, setting the background noise threshold Y in the daytime as [ P (r)m)-P(rb)]>1.5I, set late background noise threshold Y ═ P (r)m)-P(rb)]>25I。
Further, as a preferred technical solution of the present invention, in the step 4, performing cluster analysis on the differential zero point of each cloud layer signal according to relative humidity and vertical height by using an ISODATA clustering algorithm, specifically including:
constructing a cloud layer differential zero point set;
carrying out judgment classification on the constructed cloud layer differential zero point set, including splitting and merging, and in the iterative judgment process, if the number Nc of the zero point initial clustering centers is less than or equal to K/2 and K is the expected number of the clustering centers, namely the number of the zero point initial clustering centers is less than or equal to half of the expected number of the clustering centers, splitting the existing clusters; if the iterative operation times are even times or Nc is more than or equal to 2K, merging processing is carried out until the maximum iterative times are reached;
each differential zero set is a differential zero of the same cloud layer, and the cloud bottom and the cloud top height of each cloud layer are the minimum value and the maximum value of the cloud bottom height calculated by a differential zero crossing method in the differential zero set; in the operation, the iteration times are added by 1, so that the cloud layer differential zero point clustering is completed, and a plurality of classified differential zero point sets are obtained.
By adopting the technical scheme, the invention can produce the following technical effects:
according to the method, according to the characteristics of the cloud layer signal and the aerosol signal, the laser radar echo signal is preprocessed to enable the zero crossing point to be reduced remarkably, and the quality of the laser radar echo signal is effectively improved; in addition, the improved layer peak-to-bottom ratio threshold value and the background noise threshold value are combined with a differential zero crossing method, so that cloud layer signal misjudgment caused by noise signals and aerosol signals is effectively avoided, and the screening precision of the laser radar cloud layer signals is improved; and combining a clustering algorithm and a differential zero crossing method for the first time, performing clustering analysis on the differential zero points of the cloud layer signals by adopting an ISODATA algorithm according to the vertical distribution characteristics of the cloud layer and the vertical height and the relative humidity of the signal differential zero points of the cloud layer signals to obtain a plurality of classified differential zero point sets, adaptively integrating and dividing the detected multiple sections of cloud layer signals, getting rid of the subjective limitation of layering processing of the traditional threshold method cloud, and particularly accurately detecting the thin cloud layer signals and the high-level cloud signals.
Therefore, the method effectively eliminates the interference of the aerosol signal and can accurately screen out the cloud layer signal. Meanwhile, the clustering algorithm is applied to cloud layer structure analysis for the first time, and accurate cloud layering processing is achieved. Experiments are carried out by adopting ARM data, and experimental results show that the algorithm achieves the detection accuracy of 93.62% of low-level clouds, 92.78% of middle-level clouds and 93.03% of high-level clouds, and is greatly improved compared with the traditional algorithm.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a schematic diagram showing the comparison between the original laser radar echo signal and the laser radar echo signal after the pre-processing.
Fig. 3 is a schematic diagram of the basic principle of the differential zero crossing method of the present invention.
FIG. 4 is a schematic diagram of simulated laser radar echo signals and corresponding algorithm processing results in the present invention.
Fig. 5 is a schematic diagram of a laser radar echo signal and a cloud layer part amplification signal in the invention.
FIG. 6 is a schematic diagram of an improved algorithm inversion result and a cloud layer signal amplification.
FIG. 7 is a schematic diagram of the detection results of the differential zero crossing method and the improved algorithm of the present invention.
Detailed Description
The present invention is further illustrated by the following examples, which are intended to be purely exemplary and are not intended to limit the scope of the invention, as various equivalent modifications of the invention will occur to those skilled in the art upon reading the present disclosure and fall within the scope of the appended claims.
As shown in fig. 1, the invention relates to a cluster analysis-based dual-threshold lidar cloud layer detection algorithm, which comprises the following steps:
step 1, preprocessing a laser radar echo signal.
The invention adopts LMD-ITM algorithm to the laser radar echo signal pretreatment. Denoising the initial laser radar echo signal by an LMD-ITM algorithm, and performing 5-point moving average processing on the denoised signal, wherein the laser radar echo signal before and after preprocessing is shown in FIG. 2. As can be seen from FIG. 2, the zero crossing point of the preprocessed laser radar echo signal is significantly reduced, and the quality of the laser radar echo signal is effectively improved.
Step 2, in order to effectively select cloud layer signals from interference signals, the invention adopts double thresholds to process preprocessed laser radar echo signals, and the method specifically comprises the following steps:
1) layer peak layer bottom ratio threshold X: screening the preprocessed laser radar echo signals by using a layer peak-to-layer bottom ratio threshold value X, and extracting cloud layer signals; repeated tests prove that the cloud layer signal and the thick aerosol signal are distinguished by selecting the ratio of the layer peak layer bottom correction signal as an initial threshold value X by Morille et al, and the expression is as follows:
Figure BDA0003434912880000041
wherein r ism、rbThe positions of the cloud layer peak and the layer bottom, P (r)b)、P(rm) Respectively cloud layer peak rmLayer bottom rbThe laser radar echo signal strength of (1). When the ratio threshold value X of the middle-layer peak layer bottom is more than or equal to 4, the cloud layer information can be judged, the value of 4 is the optimal value selected after multiple tests, otherwise, X is the optimal value<4, it can be judged as thick aerosol signal. Therefore, an obvious cloud layer signal is extracted through screening of a layer peak layer bottom ratio threshold value X.
2) Background noise threshold Y: and extracting the missed thin cloud layer signals from the preprocessed laser radar echo signals by using a background noise threshold value. Effective cloud layer signals which are missed to be judged may exist in the signals after the layer peak bottom ratio threshold value X is screened, and the thin cloud layer signals need to be distinguished from the aerosol signals by adopting a more accurate threshold value. In order to make the judgment of cloud base and cloud peak more accurate, a background noise threshold value proposed by Pal and the like is introduced for judgment. Pal considers that the cloud peak in the laser radar echo signal is regarded as an effective cloud base and a cloud peak only when the difference value between the cloud peak and the corresponding cloud base intensity is more than 2 times of the background noise intensity, and the formula is as follows:
Y＝[P(rm)-P(rb)]>2I (2)
where I is the background noise intensity, and considering that there is a certain difference between the background noise intensity existing in the daytime and in the evening, an improvement is made to the threshold 2I proposed by Pal et al.
Through multiple experiments, the method determines the detection distance r of the laser radar<At 5km, the background noise threshold Y is set to [ P (r) ]m)-P(rb)]>3I; laserWhen the detection distance r is more than or equal to 5km, setting the background noise threshold Y in the daytime as [ P (r)m)-P(rb)]>1.5I, set background noise threshold Y at night ═ P (r)m)-P(rb)]>25I。
Therefore, the processed signal is subjected to improved background noise threshold value Y processing, the thin cloud layer signal which is missed is selected, and the missed judgment and the misjudgment are effectively prevented.
And 3, based on the extracted cloud layer signals and the extracted thin cloud layer signals, carrying out differential processing by using a differential zero crossing method to obtain differential zero points of the extracted cloud layer signals, and carrying out inversion to calculate the heights of the cloud top and the cloud bottom.
The basic principle of the differential zero crossing method is shown in fig. 3, which specifically includes the following steps:
when laser light irradiates the cloud layer boundary, because the density of the cloud is far higher than that of the aerosol, the strength P (r) of the atmospheric back scattering echo signal is rapidly increased, and when the laser light passes through the cloud body, P (r) is rapidly reduced, and a cloud layer echo signal is generated. As shown in fig. 3, the curve (a) is the strength p (r) of the atmospheric backscattered echo signal, and the curve (b) is the simulated lidar echo signal, where r isb、rtRespectively represents the positions of the cloud bottom and the cloud top of the cloud layer, rmRepresenting the cloud peak position of the cloud layer. As can be seen from fig. 3, a cloud lidar echo signal may be represented as a set of adjacent peaks and troughs on its corresponding first-order differential signal, and the cloud base height rbHeight of cloud rmCloud top height rtI.e. the zero point where its differential signal intersects the height axis. The cloud layer information is judged through the zero point on the differential signal, the inversion process has high requirements on the quality of the laser radar echo signal, and the laser radar is inevitably polluted by various noises such as sunlight in the actual use process. In addition, the aerosol layer and the cloud layer can cause the change of laser radar echo signals, and the thick aerosol layer or the thin cloud layer is easily judged as the aerosol layer by mistake by using a differential zero crossing method. In order to reduce noise signals in laser radar echo signalsThe interference caused by the interference needs to effectively preprocess the original echo signal, and particularly, the laser radar echo signal is weak at a far field, so that the dP/dr is easy to fluctuate around zero. Therefore, the invention adopts the dual-threshold technology to process the echo signal on the basis of the differential zero crossing method.
Step 4, carrying out cluster analysis on the differential zero points of the cloud layer signals by using an ISODATA (inverse discrete cosine transform) clustering algorithm according to the relative humidity and the vertical height, and clustering samples with similar relative humidity characteristics in the same height range into a class, namely each class of differential zero points can be regarded as the differential zero points of the same layer of cloud, so as to obtain a plurality of classified differential zero point sets; and (3) judging the cloud layer structure according to the classified multiple differential zero point sets to obtain corresponding cloud layer information, which is as follows:
the adjacent cloud layers can be classified into the same cloud layer from the physical characteristics of relative humidity and the like, so that the cloud layers with approximately the same relative humidity in the same height range can be effectively classified together by clustering analysis of the relative humidity and the height of the differential zero point of each cloud layer signal obtained in the step 3 through the ISODATA clustering algorithm. The ISODATA clustering algorithm can cluster the constructed differential zero point set, and samples with similar relative humidity characteristics are clustered into one class, namely, the samples can be regarded as the same layer of cloud. The distance threshold value between every two layers of clouds is set to be 0.05km, and if the distance threshold value does not meet the condition, the cloud is regarded as one layer of clouds. The cloud layering processing process based on the ISODATA clustering algorithm is as follows:
(1) and constructing a cloud layer differential zero point set. Constructing a two-dimensional data point set x according to the relative humidity and the vertical height of the laser radar detected at each zero pointi＝{(ri，bi) 1, 2, N, where r isiIs a vertical height, biAnd the relative humidity is the constructed cloud layer differential zero set. Preselection of NcInitial clustering center of zero point z1，z2，...zNcAnd fifthly, randomly selecting an initial position. Preselecting K to 3 as the expected number of cluster centers, and setting the minimum number of zero points theta in each cluster domain zero point set N2. Minimum distance theta between two cluster center zeroscWhen less than 50, thetacThe two clusters need to be combined, and the maximum combinable cluster center in one iteration operation is 2; the iteration number M is set to 30. Assigning N pattern samples to the nearest cluster SjIf:
Dj＝min{||x-zi||，i＝1，2，...，Nc} (3)
i.e. | | x-zjThe distance of | is minimal, where j is denoted as the jth cluster and x denotes a single sample in a certain cluster, where z isjDenoted as jth cluster SjIs then x ∈ SjIf clustering SjNumber of samples S inj＜θNThen the sample subset is cancelled, at which point the zero initial cluster center NcMinus 1.
Correcting the zero point of each cluster center, wherein NjRepresents the jth cluster SjThe number of samples of (a):
Figure BDA0003434912880000061
calculating each clustering zero point domain BjAverage distance between the medium pattern sample and each cluster center:
Figure BDA0003434912880000062
calculating the total average distance between all sample zero points and the corresponding clustering center zero points:
Figure BDA0003434912880000063
(2) and carrying out discrimination classification on the constructed multiple cloud layer differential zero point sets. The part mainly comprises splitting and merging, and in the process of iterative discrimination, if the number Nc of zero initial clustering centers is less than or equal to K/2, and K is the expected number of the clustering centers, namely the number of the clustering centers is less than or equal to half of the expected number of the clustering centers, splitting the existing clusters; if the iterative operation times are even times or Nc is more than or equal to 2K, merging processing is carried out until the maximum iterative times are reached.
Splitting treatment (i): calculating the standard deviation of the sample distance in each cloud layer differential zero set:
σj＝(σ1j，σ2j，...，σnj)T (7)
in the formula, σjThe matrix is a distance standard deviation matrix of cloud layer differential zero concentration samples; sigmanjThe standard deviation of the distance of each specific sample in the cloud layer differential zero set; where the components of the vector are:
Figure BDA0003434912880000071
in the formula, i is 1, 2, N is the dimension of the sample feature vector, and j is 1, 2cIs the number of clusters, NjFor the jth clustering zero set SjNumber of samples in (1). x is a radical of a fluorine atomikIs the ith zero, z, in the kth one of the prospective clustersijIs the cluster center of the jth cluster of the preselected cluster centers.
Splitting treatment (ii): calculate the distance standard deviation vector { sigma over standard deviation of each standard samplej，j＝1，2，...，NcMaximum component of { sigma }, by { sigma }jmax，j＝1，2，...，NcTake the example, at any maximum component set { σ }jmax，j＝1，2，...，NcIn the specification, if there is ajmax＞θSAnd satisfy
Figure BDA0003434912880000072
Nj＞(θN+1) or
Figure BDA0003434912880000073
Then zero is clustered to the center zjSplitting into two new zero point clustering centers, the number of the zero point initial clustering centers NcAnd adding 1. If the judgment condition is not met, the result is that the center z of the cluster is in the center zjPreselecting two zeros on both sidesIs a new clustering center and removes the original zero clustering center zj。
Merging treatment (i): calculating the distance of all cluster centers:
Dij＝||zi-zj||，i＝1，2，…，Nc-1，j＝i+1，…，Nc (9)
wherein z isiIs the zero cluster center of the ith cluster, zjIs the zero cluster center of the jth cluster;
merging treatment (ii): comparison DijAnd thetacA value of Dij＜θcAre arranged in ascending order of minimum distance, i.e. Di1j1，Di2j2，...，DjLjLIn which D isi1j1＜Di2j2＜...＜DiLjL. If the minimum distance theta between the zero points of two cluster centerscWhen less than 50, thetacThe two clusters need to be merged; if D isijGreater than thetacThe case (2) indicates that the two clusters are independent from each other and do not need to be merged.
Merging treatment (iii): will be at a distance DikjkTwo zero cluster centers Z in the kth cluster of (2)i k And Zj k And (3) merging to obtain a new zero clustering center as follows:
Figure BDA0003434912880000081
in the formula, L is the maximum logarithm of the cluster centers that can be merged in one iteration. Wherein, the two merged zero clustering center vectors are weighted by the number of samples in the clustering domain, so that a new clustering center zk *Is a true average vector.
(3) And judging the cloud layer structure according to the classified differential zero point sets of the plurality of cloud layer signals to obtain corresponding cloud layer information. Judging the cloud layer structure of each classified differential zero point set, wherein each differential zero point set is the differential zero point of the same cloud layer, and the cloud bottom height and the cloud top height of each cloud layer are the minimum value and the maximum value of the cloud top height calculated by a differential zero crossing method in the differential zero point set; in the operation, the iteration number M is added by 1, so that all cloud layer differential zero point clustering is completed, and finally a plurality of classified differential zero point sets are obtained.
Then, cloud layer structures are judged according to the classified differential zero point sets, and cloud layer information such as cloud bottoms, cloud tops, cloud peak position heights and the like of the cloud layers is obtained.
Therefore, the method can effectively eliminate the interference of the aerosol signal, can accurately screen out the cloud layer signal and realizes the accurate processing of cloud layering.
In order to verify the applicability of the method, a simulation experiment and a real data test are respectively carried out.
First verification example,
The simulation experiment is shown in fig. 4, wherein (a) is a simulated low cloud lidar echo signal; (b) processing results for simulating a low cloud signal algorithm; (c) simulating a high cloud signal; (d) the result is processed by an algorithm for simulating a high cloud signal. The heights of the simulated low cloud bottom and the simulated high cloud top are respectively 120m and 644m, and the heights of the simulated high cloud bottom and the simulated high cloud top are respectively 7654m and 8823 m. According to experimental results, the improved differential zero crossing method can effectively extract cloud layer signals aiming at cloud layers with different heights, and the heights of the cloud bottom and the cloud top can be well determined.
The second verification example,
The method provided by the invention is adopted to carry out inversion on the actually measured laser radar echo data profile. Taking the actual measurement data of a laser radar from an ARM website Southern Great Plains (SGP), Lamont, Oklahoma (Southern Plains, Lamont, usa) as an example, the monitoring time is selected to be 2021 year, 4 month, 19 day 11: 30, laser radar return signal profile. It can be seen from the laser radar echo signal diagram that within the detectable range of the laser radar, within the range of 2.5-4.5km and within the range of 4-6.5km, obvious cloud layer signals are provided, after the signals at the height of 7-10km are amplified, weak cloud layer signals mixed in noise signals can be found within the range of 8-10km, the laser radar echo signal diagram and local amplification signals are shown in fig. 5, the signals are processed by using the improved algorithm provided by the invention, and the processed results and local amplification are shown in fig. 6.
In fig. 5, the heights of the bottom and the top of the first layer of cloud are 3.12km and 3.94km, the heights of the bottom and the top of the second layer of cloud are 4.47km and 6.13km, and the heights of the bottom and the top of the third layer of cloud are 8.48km and 9.50 km. And (3) processing echo signals by using a traditional differential zero crossing method for echo data of the same site at the same moment to obtain cloud bottom height data and cloud top height data.
According to cloud layer height data obtained by experimental inversion, the improved differential zero crossing method has more excellent performance in measuring the cloud base and cloud top height of the cloud layer than the traditional differential zero crossing method.
In order to further verify the superiority of the algorithm, the actual measurement data of the laser radar from Southern Great Plains (SGP), Lamont, Oklahoma (Southern Great Plains, Lamont, Rusk Holland) 2021, 4, month, 19 and day 4, month and 20 are selected for analysis, the echo signal of the laser radar comprises a time span of 20 hours from 12:00 at 19 days 4, month and 19 at 2021 to 08:00 at 20 days 4, month and 20 at 2021, and 667 groups of signal profiles are measured in the period. The differential zero crossing method and the detection result of the present algorithm are shown in fig. 7. From the block a area in fig. 7 (a), it can be found that there is an intermittent weak cloud layer signal, and the breakpoint occurring in the block a area is caused by misjudgment. Comparing the area of block B in fig. 7 (B) with the corresponding area in fig. 7 (a), it can be seen that the missing selection phenomenon is caused in the original algorithm in fig. 7 (a) because the number of points of the cloud layer signal dP/dr greater than zero does not meet the requirement of the threshold, and the improved algorithm correctly detects the cloud layer information existing in the area by combining the layer peak-to-layer bottom ratio threshold and the background noise threshold. Comparing the area of the block C in fig. 7 (b) with the corresponding area in fig. 7 (a), it can be found that, due to poor accuracy of the original algorithm, the continuously existing cloud layer information is difficult to accurately identify, so that the detection rate of the cloud layer is low and the continuity is poor, which causes selection omission and misjudgment. As can be seen from the figure, the cloud layer detection results of the two algorithms keep better consistency in the period of 02:00-06: 00. Experiments show that compared with ARM cloud layer data, the detection accuracy of the cloud layer by the improved algorithm provided by the invention achieves 93.62% of detection accuracy of the low-layer cloud, 92.78% of detection accuracy of the medium-layer cloud and 93.03% of detection accuracy of the high-layer cloud.
In conclusion, the method of the invention effectively eliminates the interference of aerosol signals and can accurately screen out cloud layer signals. Meanwhile, the clustering algorithm is applied to cloud layer structure analysis for the first time, and accurate cloud layering processing is achieved.
The embodiments of the present invention have been described in detail with reference to the drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of those skilled in the art without departing from the gist of the present invention.

Claims (5)

1. The double-threshold laser radar cloud layer detection algorithm based on cluster analysis is characterized by comprising the following steps of:
step 1, preprocessing a laser radar echo signal;
step 2, processing the preprocessed laser radar echo signals by adopting double thresholds: screening the preprocessed laser radar echo signals by using a layer peak-to-layer bottom ratio threshold value, and extracting cloud layer signals; extracting a thin cloud layer signal which is missed to be selected by using a background noise threshold value to the preprocessed laser radar echo signal;
step 3, based on the extracted cloud layer signals and thin cloud layer signals, carrying out differential processing by using a differential zero crossing method to obtain differential zero points of the extracted cloud layer signals, and carrying out inversion to calculate the heights of the cloud top and the cloud bottom;
step 4, carrying out cluster analysis on the differential zero points of the cloud layer signals by using an ISODATA (inverse discrete cosine transform) clustering algorithm according to the relative humidity and the vertical height, and clustering samples with similar relative humidity characteristics in the same height range into a class, namely each class of differential zero points can be regarded as the differential zero points of the same layer of cloud, so as to obtain a plurality of classified differential zero point sets; and judging the cloud layer structure according to the classified multiple differential zero point sets to obtain corresponding cloud layer information.
2. The cluster analysis based dual-threshold lidar cloud detection algorithm of claim 1, wherein the preprocessing of the lidar return signals in step 1 comprises: and denoising the initial laser radar echo signal by adopting an LMD-ITM algorithm, and performing 5-point moving average processing on the denoised signal.
3. The cluster analysis-based dual-threshold lidar cloud layer detection algorithm of claim 1, wherein in step 2, the preprocessed lidar echo signals are filtered by using a layer-to-peak layer-to-bottom ratio threshold, and a formula is adopted:
Figure FDA0003434912870000011
wherein X is a layer peak-to-layer bottom ratio threshold; r ism、rbThe positions of the cloud layer peak and the layer bottom, P (r)b)、P(rm) Respectively the laser radar echo signal intensity of the cloud layer peak and the layer bottom.
4. The cluster analysis-based dual-threshold lidar cloud detection algorithm of claim 1, wherein the step 2 extracts the missing thin cloud layer signals from the preprocessed lidar echo signals by using a background noise threshold, specifically:
when the laser radar detects the distance r<At 5km, the background noise threshold Y is set to [ P (r) ]m)-P(rb)]>3I, wherein P (r)b)、P(rm) Respectively obtaining the intensity of laser radar echo signals of the cloud layer peak and the layer bottom; i is background noise intensity;
when the detection distance r of the laser radar is more than or equal to 5km, setting the background noise threshold Y in the daytime as [ P (r)m)-P(rb)]>1.5I, set background noise threshold Y at night ═ P (r)m)-P(rb)]>25I。
5. The cluster analysis-based dual-threshold lidar cloud layer detection algorithm according to claim 1, wherein the cluster analysis of the differential zero point of each cloud layer signal by using the ISODATA clustering algorithm in step 4 according to the relative humidity and the vertical height specifically comprises:
constructing a cloud layer differential zero point set;
carrying out judgment classification on the constructed cloud layer differential zero point set, including splitting and merging, and in the iterative judgment process, if the number Nc of the zero point initial clustering centers is less than or equal to K/2 and K is the expected number of the clustering centers, namely the number of the zero point initial clustering centers is less than or equal to half of the expected number of the clustering centers, splitting the existing clusters; if the iterative operation times are even times or Nc is more than or equal to 2K, merging processing is carried out until the maximum iterative times are reached;
each classified differential zero point set is a differential zero point of the same cloud layer, and the cloud bottom height and the cloud top height of each cloud layer are the minimum value and the maximum value of the cloud bottom height calculated by a differential zero crossing method in the differential zero point set; in the operation, the iteration times are added by 1, so that the cloud layer differential zero point clustering is completed, and a plurality of classified differential zero point sets are obtained.
CN202111618360.4A 2021-12-27 2021-12-27 Double-threshold laser radar cloud layer detection algorithm based on cluster analysis Pending CN114563771A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111618360.4A CN114563771A (en) 2021-12-27 2021-12-27 Double-threshold laser radar cloud layer detection algorithm based on cluster analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111618360.4A CN114563771A (en) 2021-12-27 2021-12-27 Double-threshold laser radar cloud layer detection algorithm based on cluster analysis

Publications (1)

Publication Number Publication Date
CN114563771A true CN114563771A (en) 2022-05-31

Family

ID=81712499

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111618360.4A Pending CN114563771A (en) 2021-12-27 2021-12-27 Double-threshold laser radar cloud layer detection algorithm based on cluster analysis

Country Status (1)

Country Link
CN (1) CN114563771A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116755112A (en) * 2023-08-18 2023-09-15 武汉大学 Multi-wavelength Raman laser radar inversion method and system based on layering and iteration

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116755112A (en) * 2023-08-18 2023-09-15 武汉大学 Multi-wavelength Raman laser radar inversion method and system based on layering and iteration
CN116755112B (en) * 2023-08-18 2023-10-27 武汉大学 Multi-wavelength Raman laser radar inversion method and system based on layering and iteration

Similar Documents

Publication Publication Date Title
CN110246112B (en) Laser scanning SLAM indoor three-dimensional point cloud quality evaluation method based on deep learning
CN110287932B (en) Road blocking information extraction method based on deep learning image semantic segmentation
CN103020978B (en) SAR (synthetic aperture radar) image change detection method combining multi-threshold segmentation with fuzzy clustering
CN105717556B (en) It is a kind of based on big data from parting radar pinch-reflex ion diode method
CN105869146B (en) SAR image change detection based on conspicuousness fusion
CN113838054B (en) Mechanical part surface damage detection method based on artificial intelligence
CN110658508B (en) K distribution sea clutter parameter estimation method based on characteristic quantity
CN113534240B (en) Microseism event detection and positioning method and system
Chawan et al. Automatic detection of flood using remote sensing images
CN104751185A (en) SAR image change detection method based on mean shift genetic clustering
CN104198998B (en) Clustering treatment based CFAR (Constant False Alarm Rate) detection method under non-uniform background
CN113189014A (en) Ozone concentration estimation method fusing satellite remote sensing and ground monitoring data
CN114563771A (en) Double-threshold laser radar cloud layer detection algorithm based on cluster analysis
CN114936201A (en) Satellite precipitation data correction method based on adaptive block neural network model
CN114663393A (en) Convolutional neural network-based detection method for defect of poor spacing between photovoltaic modules
CN113836808A (en) PM2.5 deep learning prediction method based on heavy pollution feature constraint
CN117251798A (en) Meteorological equipment anomaly detection method based on two-layer progressive process
CN116563205A (en) Wheat spike counting detection method based on small target detection and improved YOLOv5
CN112907113B (en) Vegetation change cause identification method considering spatial correlation
Woolford et al. Convergent data sharpening for the identification and tracking of spatial temporal centers of lightning activity
CN111768101B (en) Remote sensing cultivated land change detection method and system taking account of physical characteristics
CN114998719A (en) Forest fire prediction method based on deep learning and multi-source remote sensing data
CN107728121A (en) A kind of Local Good-fit test method based on variable window
Gill et al. Autoregressive moving average models of conifer crown profiles
CN113496159A (en) Multi-scale convolution and dynamic weight cost function smoke target segmentation method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination