A method for extracting canopy height by adaptive waveform decomposition of spaceborne lidar
By using an adaptive waveform decomposition method based on spaceborne lidar to dynamically remove noise, smooth signals, and build a model, the problem of extracting forest canopy height under complex terrain was solved, achieving higher accuracy and efficiency in obtaining forest canopy height.
Patent Information
- Application Number
- CN202311521535.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-15
- Publication Date
- 2026-08-25
- Estimated Expiration
- 2043-11-15
AI Technical Summary
In areas with complex forest stands and rugged terrain, traditional methods struggle to accurately determine forest canopy height, especially since the last echo is a mixture of ground, shrubs, and low-lying trees, making it difficult to obtain ground location information.
An adaptive waveform decomposition method using spaceborne lidar is employed to automatically extract forest canopy height through dynamic noise removal, signal smoothing, establishing a forest echo transmission process model, calculating total echo energy and decomposition parameters.
It improves the accuracy and efficiency of forest canopy height extraction, reduces the waveform broadening effect caused by slope issues, and can accurately obtain forest canopy height in complex terrain areas.
Smart Images
Figure CN117538847B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of waveform decomposition technology, and particularly relates to a method for extracting canopy height through adaptive waveform decomposition of spaceborne lidar. Background Technology
[0002] Forests are a vital component of terrestrial ecosystems and the largest carbon sink on land. Forest canopy height reflects forest growth and is a crucial indicator for assessing forest health and ecosystem services. Rapid and accurate acquisition of forest canopy height data can better assess forest carbon balance and net carbon emissions, providing a scientific basis for developing climate change adaptation and mitigation measures.
[0003] Compared with traditional ground surveys of forest canopy height, using full-waveform spaceborne lidar to acquire forest canopy height is faster, and the waveform data within its light spot can extract continuously arranged information on the vertical structure of the forest, which provides important data support for the study of forest canopy height and other forest vertical structure parameters.
[0004] In forest echo information, the crest of the first echo is generally considered the starting point of the forest echo, and the crest of the last echo is considered the ground. The forest canopy height is obtained by calculating the relative distance between the two crests. However, in areas with complex forest stands and rugged terrain, the last echo is a mixture of ground, shrubs, and low-lying trees, making it difficult to obtain accurate ground location information. Waveform decomposition can break down continuous waveform signals into individual waveform information and is often used for forest echo decomposition and forest canopy height extraction. In the field of signal decomposition, Gaussian decomposition is the most widely used method. Compared with other decomposition methods, Gaussian decomposition has the highest decomposition accuracy given the number of crests. However, when applied to forestry, due to the complexity of the forest canopy, it is difficult to obtain the vertical distribution of the forest in advance to know the number of forest echo crests, and it is also unsuitable for large-scale forest light spot processing. Summary of the Invention
[0005] To address the aforementioned technical problems, this invention proposes a method for extracting canopy height through adaptive waveform decomposition using a spaceborne lidar.
[0006] To achieve the above objectives, the present invention provides a method for extracting canopy height through adaptive waveform decomposition using a spaceborne lidar, comprising:
[0007] Acquire the echo signal, and perform dynamic noise removal on the echo signal to obtain the effective signal;
[0008] The effective signal is smoothed, and the smoothed effective signal is analyzed to obtain several wavelet signals.
[0009] The total echo energy is calculated based on several of the wavelet signals, and the decomposition parameters are determined based on the total echo energy.
[0010] The smoothed echo signal is decomposed based on the decomposition parameters to obtain the canopy height.
[0011] Optionally, during the process of acquiring the echo signal and performing dynamic noise removal on the echo signal to obtain the effective signal, dynamic noise removal is performed based on the following formula:
[0012] rxwaveform final =rxwaveform original -(noise mean +n*noise θ )
[0013] In the formula, rxwaveform final This represents the processed echo signal, rxwaveform original Indicates the initial echo signal, noise. mean The noise level represents the mean noise, and n represents the multiple of the noise standard deviation, where n = 1, 2, 3, ..., noise. θ This represents the standard deviation of noise.
[0014] Optionally, after acquiring a valid signal, the process may also include:
[0015] A first threshold is preset. The minimum amplitude and maximum amplitude of the effective signal are extracted. The ratio of the minimum amplitude to the maximum amplitude is calculated. If the ratio is less than the first threshold, the denoising is completed. Otherwise, the denoising operation is repeated until the ratio is less than the first threshold.
[0016] Optionally, the process of smoothing the effective signal includes:
[0017] The signal is processed using a Gaussian filter to remove burst interference and high-frequency components from the effective signal.
[0018] Extract the changing trend and time-domain features of the processed effective signal, including the maximum amplitude, the number of smoothed peaks, and the elevation range.
[0019] Optionally, the process of analyzing the smoothed effective signal to obtain several wavelet signals includes:
[0020] A forest echo transmission process model is constructed, and the time-domain characteristics of the smoothed echo signal are analyzed based on the forest echo transmission process model to simplify the types of forest echoes.
[0021] The smoothed effective signal is divided into several wavelet signals, including: canopy echo, mid-level echo and ground echo;
[0022] The forest echo propagation process model is as follows:
[0023] Signal total =Signal canpy +∑Signal midst +Signal ground
[0024] In the formula, Signal total Indicates the total echo signal, Signal canpy Indicates canopy echo, Signal midst Indicates mid-level echo, Signal ground This indicates a ground echo.
[0025] Optionally, the process of calculating the total echo energy based on several wavelet signals is as follows: calculate the energy of each wavelet signal and sum them into the total energy, as shown in the following formula:
[0026] E total =∫Signal canpy +∑∫Signal midst +∫Signal ground +E θ
[0027] In the formula, E total E represents the total energy of the echo. θ This represents the energy loss term.
[0028] Optionally, the process of determining the decomposition parameters includes:
[0029] By summarizing the time-domain characteristics of the echo signal and determining the maximum number of decompositions and the energy loss term through a preset energy threshold, the system can effectively summarize the characteristics of the echo signal.
[0030] Optionally, the process of decomposing the smoothed echo signal based on the decomposition parameters includes:
[0031] The echo signal is decomposed based on the determined decomposition parameters. During the decomposition process, the solution with the minimum energy loss term is calculated based on the following formula.
[0032]
[0033] In the formula, t represents the elevation of the wave crest, and t0 and t t Let A represent the elevation of the first peak position and the elevation of the last peak position, respectively. Let A represent the wavelet constant term, σ represent the standard deviation of the wavelet amplitude, and H represent the canopy height.
[0034] Optionally, after waveform decomposition is completed, the decomposition results are also evaluated to ensure that the decomposition results meet the expected time-domain characteristics.
[0035] Technical effects of the invention:
[0036] This invention discloses a method for extracting canopy height using adaptive waveform decomposition from a spaceborne lidar system. This method has higher extraction efficiency and, compared to traditional waveform decomposition methods, can automatically and dynamically remove background noise and extract forest height. It can also reduce the waveform broadening effect caused by slope issues to a certain extent and improve the extraction accuracy of forest canopy height. Attached Figure Description
[0037] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:
[0038] Figure 1 This is a flowchart of the dynamic noise reduction process in an embodiment of the present invention;
[0039] Figure 2 This is a schematic diagram comparing the noise reduction effect in an embodiment of the present invention, where (a) represents before processing and (b) represents after processing;
[0040] Figure 3 This is a schematic diagram comparing the smoothing results in an embodiment of the present invention;
[0041] Figure 4 This is a comparative schematic diagram of the forest echo transmission process in an embodiment of the present invention, where (a) represents simple terrain and (b) represents complex terrain;
[0042] Figure 5 The following diagram illustrates the results of four decomposition methods in this embodiment of the invention, where (a) represents the adaptive Gaussian decomposition method, (b) represents the ordinary Gaussian decomposition method, (c) represents the wavelet decomposition method, and (d) represents the deconvolution integral solution method.
[0043] Figure 6 This is a schematic diagram comparing the tree height extraction accuracy in an embodiment of the present invention, where (a) represents simple terrain, (b) represents normal terrain, and (c) represents complex terrain.
[0044] Figure 7 This is a schematic diagram of the method for extracting canopy height by adaptive waveform decomposition of spaceborne lidar in an embodiment of the present invention. Detailed Implementation
[0045] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.
[0046] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.
[0047] like Figure 7 As shown, this embodiment provides a method for extracting canopy height using adaptive waveform decomposition of a spaceborne lidar, used to extract forest canopy height. The principle is to re-divide the forest echo signal composition to obtain the location of the average ground elevation, thereby extracting a forest canopy height close to the actual situation. Its advantages include dynamic noise removal, no need to preset the number of peaks for Gaussian waveform decomposition, and the ability to decompose ground echo signals in areas with steep slopes, thus quickly and accurately extracting the forest canopy height within the light spot. The main steps include:
[0048] 1) Dynamic noise estimation and effective signal extraction: Determine that noise will be generated during the propagation of the pulse, and extract the signal of the main vibration part;
[0049] 2) Signal smoothing: Remove burst interference and high-frequency components from the signal, and extract the signal's variation trend and time-domain characteristics;
[0050] 3) Establish a forest echo transmission process model: analyze the time-domain characteristics of the echo signal and simplify the types of forest echoes;
[0051] 4) Determine the total echo energy: Calculate the energy of each wavelet signal and sum them up to the total energy;
[0052] 5) Determine the decomposition parameters: Determine the maximum number of decompositions and the energy threshold to be set during decomposition;
[0053] 6) Waveform Gaussian Decomposition: Decompose the echo signal, calculate the number of peaks that meet the energy threshold, and guide the decomposition.
[0054] 7) Evaluation of decomposition results: After the waveform decomposition is completed, it is usually necessary to evaluate it to ensure that the decomposition results meet the expected time domain characteristics.
[0055] The process of dynamic estimation of background noise is as follows:
[0056] Due to factors such as the atmosphere, noise will be generated during the propagation of the pulse. The high-frequency part of the signal is extracted from the echo signal and the noise mean and noise standard deviation are calculated. According to Equation (1), the noise in the waveform signal is removed and the signal of the main vibration part is extracted.
[0057] rxwaveform final =rxwaveform original -(noise mean +n*noise θ (1)
[0058] In the formula, rxwaveform final This represents the processed echo signal, rxwaveform original Indicates the initial echo signal, noise. mean The noise level represents the mean, and n represents the coefficient of the noise standard deviation, where n = 1, 2, 3, ..., noise. θ This represents the standard deviation of noise.
[0059] However, in actual denoising processes, the standard deviation coefficient n often needs to be determined based on the specific circumstances and cannot be fixed, because a dynamic denoising method is employed. For example... Figure 1 As shown, the minimum amplitude Amp is extracted from the effective signal extracted each time. min With maximum amplitude Amp max The degree of denoising is judged by calculating their ratio k. When k is less than a preset threshold (0.05), the noise removal effect is considered good, and the effective signal is successfully extracted. An example of denoising effect is shown below. Figure 2 As shown.
[0060] The process of signal smoothing is as follows:
[0061] The waveform signal exhibits an approximately normal distribution during propagation, but due to factors such as multi-blade interference and water vapor, "spiking" occurs. To remove sudden interference and high-frequency components from the signal, a Gaussian filter is used to smooth the signal. The kernel function is shown in equation (2), thereby extracting the signal's trend and time-domain characteristics, facilitating subsequent waveform decomposition operations. Figure 3 As shown.
[0062]
[0063] In the formula, t represents the elevation of the waveform position, and σ represents the standard deviation of the waveform amplitude.
[0064] Forest echo propagation process model:
[0065] During the transmission of forest echo pulses, the pulse signal first contacts the top of the forest canopy, and then the ground, such as... Figure 5 In (a), the signal composition of the echo pulse during transmission conforms to:
[0066] Signaltotal =Signal canpy +Signal ground (3)
[0067] In the formula, Signal total Indicates total echo, Signal canpy Indicates canopy echo, Signal ground This indicates a ground echo.
[0068] However, in areas with steep terrain slopes and complex forest stands, the mixed waveforms at the same horizontal elevation, with the terminal peak being the lowest point on the ground within the light spot, can lead to an overestimation of the forest canopy height. Figure 4 (b) In this case, the signal composition of the echo pulse during transmission is as follows:
[0069] Signal total =Signal canpy +ΣSignal c&t +ΣSugnal trunk +ΣSIgnal t&g +Signal ground (4)
[0070] In the formula, Signal total Indicates the total echo signal, Signal canpy Indicates canopy echo, Signal trunk Indicates tree trunk echo, Signal c&t Indicates coronal-interstitial mixed echo, Signal t&g Indicates a mixed echo between the tree trunk and the ground, Signal ground This represents ground echoes. When calculating forest canopy height using waveform data, only the initial and final peak positions of the echo signal are needed. Therefore, a simplified model of waveform signal transmission in the forest is constructed:
[0071] Signal total =Signal canpy +∑Signal midst +Signal ground (5)
[0072] In the formula, Signal total Indicates total echo, Signal canpy Indicates canopy echo, Signal midst Indicates mid-level echo, Signal ground This indicates a ground echo.
[0073] In the simplified transmission process model of forest waveform signals, the echo signal is divided into three parts: canopy echo, mid-level echo, and ground echo. The most idealized case is when the ground slope is 0. The mid-level echo (Signal)... midst When the ground slope increases or the complexity of the forest stand increases, Signal becomes 0, and equation (5) is consistent with equation (3). canpy With Signal ground Distribution decreases, Signal midst As the distribution increases, the composition of forest echo signals gradually approaches equation (4), that is, several mixed echoes of the canopy and trunk layers, and echoes of the trunk and ground appear.
[0074] The process of determining the echo energy is as follows:
[0075] The integral function of the echo signal is used instead of energy calculation because the greater the signal energy, the larger its amplitude, i.e., the larger the peak, which is undoubtedly closely related to the integral of the echo function. This also introduces an energy loss term, E. θ This facilitates the explanation of energy transfer losses in practical situations, as shown in equation (6):
[0076] E total =∫Signal canpy +∑∫Signal midst +∫Signal ground +E θ (6)
[0077] In the formula, E total Signal represents the total energy of the echo. canpy Indicates canopy echo, Signal midst Indicates mid-level echo, Signal ground E represents ground echo. θ This represents the energy loss term.
[0078] The process of determining the decomposition parameters is as follows:
[0079] Before starting the decomposition, the maximum number of decompositions and the energy loss term E are determined based on the statistical time-domain characteristics of the echo signals (maximum amplitude of the echo signals, number of smoothed peaks, elevation range, etc.). θ Threshold a. Typically, even in complex forests, the vertical structure rarely exhibits more than six layers. Excessive mid-level echoes are attributed to subtle vibrations caused by the canopies of trees growing at different rates within the forest. Therefore, mid-level echoes (Signal) midst It should not exceed 5 layers, and merging is allowed.
[0080] Calculate the energy of each wavelet, sum them into the total energy, and then calculate the wavelet energy and error loss term after each decomposition to determine the number of decompositions.
[0081] In determining the maximum number of decompositions, each decomposition is accompanied by energy loss. An energy threshold is set beforehand, and by iteratively decomposing the number of decompositions, the maximum energy loss is found. That is, when the energy loss is less than the given threshold, the iteration is completed, thus determining the maximum number of decompositions.
[0082] Waveform decomposition to extract canopy average height
[0083] The forest echo is decomposed using the set parameters, and E is solved according to equation (7). θ Find the solution with the smallest value, and after decomposition, check whether the forest canopy height and altitude distribution meet the actual situation.
[0084]
[0085] In the formula, t represents the elevation of the wave crest, and t0 and t t Let A represent the elevation of the first peak position and the elevation of the last peak position, respectively. Let A represent the wavelet constant term, σ represent the standard deviation of the wavelet amplitude, and H represent the canopy height.
[0086] The evaluation process includes: analyzing the smoothness of the decomposed echo, the changes in the echo, the processing of edge signals, and the fluctuations in the peaks. This step is generally a qualitative analysis.
[0087] This invention provides an adaptive Gaussian decomposition method for spaceborne lidar to extract forest canopy height. It offers higher extraction efficiency compared to traditional waveform decomposition methods, automatically removing background noise and extracting forest height. It can also reduce waveform broadening caused by slope issues to some extent, improving the accuracy of forest canopy height extraction. The Adaptive Gaussian Decomposition (AGD) algorithm is used to decompose the waveform signals within forest light spots under simple (slope less than 10°), normal (slope 10°-20°), and complex (slope greater than 20°) conditions to extract forest canopy height. The results are compared with those from Normal Gaussian Decomposition (NGD), Wavelet Decomposition (WD), and Deconvolution Decomposition (DD), and validated using airborne laser scanning data. Figure 5The results of the four decomposition methods show that the AGD method is more refined in handling the beginning and end positions of the waveform signal. For example... Figure 6 As shown, the proposed method achieved the best results in extracting forest canopy height under three different conditions. This invention can improve the efficiency of waveform data utilization and is expected to promote the application of waveform lidar data in forestry, ecology and other fields.
[0088] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for extracting canopy height through adaptive waveform decomposition using a spaceborne lidar, characterized in that, include: Acquire the echo signal, and perform dynamic noise removal on the echo signal to obtain the effective signal; The effective signal is smoothed, and the smoothed effective signal is analyzed to obtain several wavelet signals. The total echo energy is calculated based on several of the wavelet signals, and the decomposition parameters are determined based on the total echo energy. The smoothed echo signal is decomposed based on the decomposition parameters to obtain the canopy height. During the process of acquiring the echo signal and performing dynamic noise removal on the echo signal to obtain a valid signal, dynamic noise removal is performed based on the following formula: In the formula, This indicates the processed echo signal. Indicates the initial echo signal. Indicates the noise mean. Indicates a multiple of the noise standard deviation. , Indicates the standard deviation of noise; The process of analyzing a smoothed effective signal to obtain several wavelet signals includes: A forest echo transmission process model is constructed, and the time-domain characteristics of the smoothed echo signal are analyzed based on the forest echo transmission process model to simplify the types of forest echoes. The smoothed effective signal is divided into several wavelet signals, including: canopy echo, mid-level echo and ground echo; The forest echo propagation process model is as follows: In the formula, Indicates the total echo signal. Indicates canopy echo, Indicates mid-level echo, Indicates ground echo; The process of calculating the total echo energy based on several wavelet signals is as follows: calculate the energy of each wavelet signal and sum them into the total energy. The calculation formula is shown in the following formula: In the formula, This represents the total energy of the echo. Represents the energy loss term; The process of determining the decomposition parameters includes: The time-domain characteristics of the echo signal are summarized, and the maximum number of decompositions and energy loss terms are determined by a preset energy threshold. The process of decomposing the smoothed echo signal based on the decomposition parameters includes: The echo signal is decomposed based on the determined decomposition parameters. During the decomposition process, the solution with the minimum energy loss term is calculated based on the following formula: 。 2. The method for extracting canopy height using adaptive waveform decomposition from a spaceborne lidar as described in claim 1, characterized in that, After obtaining a valid signal, the following steps are also included: A first threshold is preset. The minimum amplitude and maximum amplitude of the effective signal are extracted. The ratio of the minimum amplitude to the maximum amplitude is calculated. If the ratio is less than the first threshold, the denoising is completed. Otherwise, the denoising operation is repeated until the ratio is less than the first threshold.
3. The method for extracting canopy height using adaptive waveform decomposition of spaceborne lidar as described in claim 1, characterized in that, The process of smoothing the effective signal includes: The signal is processed using a Gaussian filter to remove burst interference and high-frequency components from the effective signal. Extract the changing trend and time-domain features of the processed effective signal, including the maximum amplitude, the number of smoothed peaks, and the elevation range.
4. The method for extracting canopy height using adaptive waveform decomposition from a spaceborne lidar as described in claim 1, characterized in that, After waveform decomposition is completed, the decomposition results are evaluated to ensure that the decomposition results meet the expected time-domain characteristics.