Linear array laser radar-based forest stand parameter inversion method

Abstract

The invention relates to the technical field of parameter inversion, and discloses a forest stand parameter inversion method based on a linear array laser radar, and the method is used for the inversion of felling or interfering forest stand vegetation parameters, and the inversion of the felling interfering forest stand vegetation parameters comprises the following steps: obtaining the point cloud data of forest stand vegetation through the vertical rotation scanning of a scanner; establishing a relation curve of the number of the point clouds, the intensity of the point clouds and the height, and extracting forest stand vertical structure parameters by using peak values in the relation curve; and calculating the forest stand canopy density by using the quantitative relationship between the point cloud echoes and the emission. According to the method, the accuracy of inversion of related parameters is high, vegetation growth and forest stand structure conditions of the interfered forest stand can be reflected, a scientific basis is provided for vegetation recovery, point cloud data does not need to be spliced, time cost and workload are reduced, a new method is provided for inversion of related parameters of the interfered forest stand, and the method is suitable for popularization and application. And a new direction is provided for the application of the linear array laser radar in forestry.

Description

Technical Field

[0001] The present invention relates to the technical field of parameter inversion, and more specifically, to a method for inverting stand parameters based on a linear array lidar. Background Art

[0002] As a huge natural resource of mankind, forests are the most widely distributed, most complex in composition and structure, and richest in species resources ecosystem on land, playing an important role in alleviating global warming, protecting biodiversity, preventing soil erosion and so on. If forest resources are damaged, it will seriously affect the exertion of its ecological effects. Therefore, it is crucial to maintain the health of forest ecosystems, which is related to whether forests can give full play to their multi-functional benefits and provide long-term services for mankind. In forest ecosystems, the destruction of forest resources caused by human and natural disturbances occurs from time to time. The growth of forest trees in the damaged forests will be restricted, and the stand structure will also change to varying degrees. Therefore, the research on parameter inversion of disturbed stands is particularly urgent. The traditional acquisition of vegetation parameters mainly relies on ground surveys. With the rapid development of remote sensing science and the continuous deepening of its application, remote sensing technology has played an excellent role in the inversion of stand vegetation parameters.

[0003] Lidar (Light Detection And Ranging, LiDAR), as an active remote sensing technology, can directly measure the distance between the laser scanner and the reflecting target, obtain high-precision three-dimensional point clouds, and quantitatively estimate forest parameters, playing an important role in forestry surveys. The inversion method of forest parameters is related to the type of LiDAR sensor. According to the platform height, the sensor can be divided into three types: spaceborne, airborne, and ground-based. Among them, terrestrial laserscanning (TLS) is a non-destructive high-resolution three-dimensional measurement method, which has the potential for automated data processing and can realize the automated or semi-automated acquisition of single-tree geometric structure parameters, providing the possibility for reconstructing the real three-dimensional scene of the forest. Compared with airborne lidar data, the accuracy of inverting stand internal factors and stand structure from ground-based lidar data is higher. If ground-based lidar is applied to the research on parameter inversion of disturbed stands, it can more accurately reflect the changes in forest structure during the recovery process and put forward more accurate measures and technical methods for the recovery of disturbed stands. Currently, commonly used laser scanners can be roughly divided into two categories: omnidirectional scanning and linear array (single-line or multi-line) scanning. Linear array laser scanners are small in size, light in weight, require a short scanning time, and are relatively low in price compared to other scanners, having more advantages in collecting data under the forest. If the required parameters can be accurately extracted based on linear array point cloud data, it is of great significance for the application of linear array lidar in forestry. Therefore, it is valuable to carry out research on parameter inversion of disturbed stands based on linear array lidar data. Summary of the Invention

[0004] In view of the problems in the related art, the present invention proposes a method for inverting stand parameters based on a linear array lidar to overcome the above-mentioned technical problems existing in the existing related art.

[0005] To this end, the specific technical solution adopted by the present invention is as follows:

[0006] According to one aspect of the present invention, there is provided a method for inverting stand parameters based on a linear array lidar for inverting the vegetation parameters of a harvested disturbed stand, including the following steps:

[0007] Obtain the point cloud data of the stand vegetation by vertical rotation scanning of the scanner, and preprocess the obtained point cloud data;

[0008] Based on the preprocessed point cloud data, establish a relationship curve between the point cloud quantity, point cloud intensity and height, and extract the stand vertical structure parameters by using the peak value in the relationship curve;

[0009] Calculate the stand canopy density by using the quantity relationship between the point cloud echo and the emission, and analyze the stand growth under different harvesting intensities according to the stand canopy density and the stand vertical structure parameters.

[0010] Further, the step of establishing a relationship curve between the point cloud quantity, point cloud intensity and height based on the preprocessed point cloud data, and extracting the stand vertical structure parameters by using the peak value in the relationship curve includes the following steps:

[0011] Based on the preprocessed point cloud data, establish a relationship curve between the point cloud quantity and height, and determine the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud quantity and height;

[0012] Based on the preprocessed point cloud data, establish a relationship curve between the point cloud intensity and height, and determine the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud intensity and height.

[0013] Further, the step of establishing a relationship curve between the point cloud quantity and height based on the preprocessed point cloud data, and determining the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height includes the following steps:

[0014] Use a correction coefficient to correct the assignment of the point cloud quantity in the preprocessed point cloud data, where the correction coefficient is the square of the distance value from the point cloud to the scanner;

[0015] Divide the vertical height of the stand into intervals, assign values to the number of point clouds after correction, count the sum of the number of point clouds at different vertical heights, establish a relationship curve between the number of point clouds and height, and perform smoothing processing;

[0016] Use the curve peak finding method to calculate the height values corresponding to the peaks and valleys in the relationship curve between the number of point clouds and height, and calculate the average height value of the stand. Among them, the average height value of the stand is the height value corresponding to when the sum of the number of point clouds from the lowest to the highest is 95% of the total number of point clouds;

[0017] Determine the number of vertical layers of the stand, the specific height within each layer, the distribution probability within each layer, and the average height of the stand according to the relationship curve between the number of point clouds and height and the calculation results. Among them, the height of each vertical layer of the stand is calculated from the mean value of the peak values within the same layer.

[0018] Furthermore, establishing a relationship curve between the point cloud intensity and height based on the preprocessed point cloud data, and determining the number of vertical layers of the stand, the specific height values within each layer, the distribution probability within each layer, and the average height of the stand according to the relationship curve between the point cloud intensity and height includes the following steps:

[0019] Use the correction coefficient to correct the assignment of the number of point clouds in the preprocessed point cloud data, and divide the vertical height of the stand into intervals;

[0020] According to the corrected assignment of the number of point clouds, count the intensity of the point clouds at different vertical heights, establish a relationship curve between the point cloud intensity and height, and perform smoothing processing;

[0021] Use the curve peak finding method to calculate the height values corresponding to the peaks and valleys in the relationship curve between the point cloud intensity and height, and calculate the average height value of the stand. Among them, the average height value of the stand is the height value corresponding to when the sum of the point cloud intensities from the lowest to the highest is 95% of the total point cloud intensity;

[0022] Determine the number of vertical layers of the stand, the specific height values within each layer, the distribution probability within each layer, and the average height of the stand according to the relationship curve between the point cloud intensity and height and the calculation results.

[0023] Furthermore, calculating the canopy density of the stand using the quantitative relationship between the point cloud echo and the emission, and analyzing the growth situation of the stand under different harvesting intensities according to the canopy density and the vertical structure parameters of the stand includes the following steps:

[0024] Place the scanner in a preset environment for scanning, count the number of echoes, and determine the total number of laser pulses emitted by the scanner each time. Among them, the preset environment is an environment that can receive all echoes;

[0025] Calculate the porosity based on the point clouds within different zenith angle ranges to obtain the porosity within a preset horizontal range centered on the site. Here, the zenith angle is the azimuth angle of the collected point clouds, and the porosity calculation formula is:

[0026] P(θ) = 1 - N t / N s

[0027] In the formula, θ is the zenith angle, N t is the number of pulses reflected by the vegetation, and N s is the total number of emitted pulses;

[0028] Calculate the porosity values and the single - station average within different zenith angle ranges for each single station in the sample plot, and statistically calculate the average value of all stations within the same sample plot to obtain the canopy porosity of the sample plot;

[0029] According to the relationship between the canopy density and the porosity in the vertical direction, calculate the stand canopy density based on the point clouds at different emission angles, and analyze the stand growth conditions under different harvesting intensities according to the stand canopy density and the stand vertical structure parameters.

[0030] According to another aspect of the present invention, there is provided a method for inverting stand parameters based on a linear array lidar for inverting the vegetation parameters of a fire - disturbed stand, including the following steps:

[0031] Based on the point cloud data collected by single - station scanning, use a single - tree recognition algorithm for linear array lidar data to extract the average breast diameter of the stands in the burned and control plots of different fire years;

[0032] Extract the porosity based on the point cloud data collected by mobile scanning, and calculate the leaf area index according to the Lambert - Beer law;

[0033] Analyze the vegetation growth conditions and stand structure conditions of the post - fire stand according to the average breast diameter, porosity, and leaf area index of the stand.

[0034] Furthermore, the step of using a single - tree recognition algorithm for linear array lidar data to extract the average breast diameter of the stands in the burned and control plots of different fire years based on the point cloud data collected by single - station scanning includes the following steps:

[0035] Obtain the point cloud data collected by single - station scanning, and perform filtering and noise reduction processing on the collected point cloud data;

[0036] Use a single - tree recognition algorithm based on distance and trunk characteristics to identify the single trees in the stands of the burned and control plots of different fire years, obtain the single - tree trunk skeleton points and trunk point clouds, and fit a circle by combining the random sample consensus algorithm to obtain the position and breast diameter of each tree, and calculate the average breast diameter of the stands in the burned and control plots of different fire years based on the single - tree breast diameters.

[0037] Furthermore, the single-tree recognition algorithm based on distance and trunk characteristics is used to identify single trees in the stands of burned and control plots in different burned years, obtaining the single-tree trunk skeleton points and trunk point clouds. Combining with the random sample consensus algorithm to fit a circle to obtain the position and diameter at breast height (DBH) of each tree, and calculating the average DBH of the stands in the burned and control plots in different burned years based on the single-tree DBH includes the following steps:

[0038] Extract the trunk point cloud based on the distance relationship between the single-tree trunk and other surrounding objects to the scanning station, and draw a two-dimensional graph with the negative value of the distance from all point clouds to the scanning station as the vertical axis; compare the two-dimensional graphs before and after filtering. When the distance between adjacent points is greater than the spacing threshold, this point is the boundary between the trunk and the surrounding objects, and determine the starting and ending positions of each trunk curve according to the characteristics of the points on the trunk after filtering;

[0039] Extract the relative coordinates of the midpoints of the trunk curves according to the starting and ending positions of each trunk curve, and set the points selected by the relative coordinates of the curve midpoints as the trunk skeleton points; remove the discrete points and the points close to the crown or trunk branches based on the set horizontal distance and height thresholds, and search for the skeleton points within the preset threshold range from any trunk skeleton point, and combine the qualified skeleton points to obtain the single-tree trunk skeleton;

[0040] Calculate the zenith angle and azimuth angle of the trunk tilt direction based on the trunk skeleton points, and determine the rotation matrix through the zenith angle and azimuth angle, and rotate all the trunk point clouds to the vertical direction according to the rotation matrix; calculate the standard deviation of the relative position coordinates of all the trunk point clouds after rotation, remove the point clouds of irregular trunks, and perform circle fitting on the trunk point clouds of different scan line arrays based on the trunk skeleton points and trunk point clouds through the trunk fitting algorithm;

[0041] Classify the trunk point clouds according to the scan line array number, and use the random sample consensus algorithm to fit a circle for the trunk point clouds in the same line array, count the number of fitted circles for each trunk, and generate the trunk through the continuously fitted circles; rotate and translate the trunk after circle fitting to the original azimuth according to the corresponding rotation matrix and translation formula;

[0042] Calculate the mean value of the position coordinates of the fitted circles at the preset height to obtain the single-tree position coordinates. For the trunks with more than two fitted circles, take the diameter at the preset height as the DBH of the tree. For the trunks with less than three fitted circles, take the average value of the diameters of the fitted circles based on the single-tree position as the DBH of the tree;

[0043] Calculate the average DBH of the stand according to the total number of trees in the stand, single-tree positions and single-tree DBHs. The calculation formula is:

[0044]

[0045] Wherein, D g is the average breast height diameter of the stand, N is the total number of trees in the stand, and d i is the breast height diameter of the i-th tree.

[0046] Furthermore, extracting the porosity based on the point cloud data collected by mobile scanning and calculating the leaf area index according to the Lambert-Beer law includes the following steps:

[0047] Convert the recorded longitude and latitude values into geographic coordinates, draw the walking route of the scanner in the sample plot, and perform format conversion and outlier processing on the point cloud data collected by mobile scanning;

[0048] Obtain point cloud data in the stand by a backpack mobile scanner, divide the zenith angle using a preset interval, calculate the porosity values within different zenith angle ranges for each frame of point cloud in the sample plot, and statistically calculate the mean of the porosity values within different zenith angle ranges in the same sample plot to obtain the canopy porosity of the sample plot;

[0049] Calculate the leaf area index according to the Lambert-Beer law in combination with the porosity value and the clumping index.

[0050] Furthermore, the calculation formula for the leaf area index is:

[0051]

[0052]

[0053] Wherein, LAI is the leaf area index, θ is the zenith angle, α is the leaf inclination angle, P(θ) represents the porosity at the zenith angle of θ, G(θ,α) is the projection function, Ω is the clumping index, is the mean of the porosities of each ring, is the average value of the natural logarithms of the porosities of each ring.

[0054] The beneficial effects of the present invention are:

[0055] 1) The present invention can not only extract the parameters related to the vertical structure of the stand and the stand canopy density based on the point cloud data obtained by the scanner rotating 360° vertically, so as to analyze the growth situation of the stand under different cutting intensities, but also extract the average breast height diameter of the stands in the burned and control sample plots of different burned years based on the point cloud data collected by single-station scanning, and extract the porosity based on the point cloud data collected by mobile scanning, and calculate the leaf area index according to the Lambert-Beer law, so as to analyze the vegetation growth situation and stand structure status of the stand after fire.

[0056] 2) The accuracy of inverting relevant parameters based on the algorithm of the present invention is relatively high, which can reflect the vegetation growth and stand structure status of the disturbed forest stand, providing a scientific basis for vegetation restoration. At the same time, the algorithm of the present invention does not require stitching point cloud data, reducing the time cost and workload, providing a new method for inverting relevant parameters of the disturbed forest stand, and also providing a new direction for the application of linear array lidar in forestry.

[0057] 3) Based on the characteristics of single-frame linear array lidar data, the present invention extracts the average breast diameter of the forest stand in the burned area at different recovery periods, realizes the estimation of the growth status of the forest stand relying on sparse point clouds, and at the same time, based on the number of laser pulse emissions and echoes, extracts the canopy density of the forest stand after logging and the leaf area index of the vegetation after fire, providing a reference for the structural changes of the forest stand in the plots disturbed by logging and fire. Description of the Drawings

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0059] Figure 1 is a flowchart of a method for inverting forest stand parameters based on linear array lidar according to an embodiment of the present invention;

[0060] Figure 2 is a flowchart of a method for inverting forest stand parameters based on linear array lidar according to another embodiment of the present invention. Detailed Embodiments

[0061] To further illustrate the embodiments, the present invention provides drawings. These drawings are part of the disclosure of the present invention, mainly used to illustrate the embodiments, and can be combined with the relevant descriptions in the specification to explain the operation principle of the embodiments. With reference to these contents, those of ordinary skill in the art should be able to understand other possible implementation manners and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are usually used to represent similar components.

[0062] Embodiment 1

[0063] As Figure 1 shown, a method for inverting forest stand parameters based on linear array lidar, used for inverting the vegetation parameters of the forest stand disturbed by logging, the method includes the following steps:

[0064] S101. Obtain the point cloud data of the forest stand vegetation through vertical rotation scanning by the scanner, and preprocess the obtained point cloud data;

[0065] Data acquisition is as follows: In the control, light cutting, and moderate cutting plots, data was collected using a VLP-16 scanner in a single-station scanning mode. The scanning stations were set at the intersections of each small quadrat, with a total of 16 stations. Due to the complex understory vegetation in the cutting plots, it was relatively difficult for personnel and instruments to enter the plots. Therefore, a sampling method was used to select the intersections of small quadrats as the station locations. The number of stations in the light and moderate cutting plots was 11 and 8 respectively. During data acquisition, the scanner was placed vertically and rotated around the scanning station for scanning, basically covering a 360° range in the horizontal and vertical directions.

[0066] Data preprocessing is as follows: The collected raw point cloud data was in the.pcap format. To facilitate subsequent processing, the data was first converted into a.csv format file in the VeloView software. One.csv file represents a single-frame point cloud data. Secondly, since the scanner was placed vertically during data acquisition, the coordinate system inside the scanner does not represent the actual azimuth information. Therefore, the coordinate system needs to be transformed, and the z-axis is rotated to the vertical direction, which represents the tree height direction. Finally, to ensure the accuracy of the subsequent processing effect and eliminate the influence of occlusion between trees, the horizontal range of each single-frame point cloud data was limited within 20 m from the scanning station, and denoising processing was performed in the cloudcompare software.

[0067] S102. Establish the relationship curves of the point cloud quantity, point cloud intensity, and height based on the preprocessed point cloud data, and extract the vertical structure parameters of the forest stand using the peaks in the relationship curves;

[0068] Specifically, S102 includes two parts: parameter extraction based on point cloud quantity statistics and parameter extraction based on point cloud intensity statistics.

[0069] Parameter extraction based on point cloud quantity statistics is as follows:

[0070] The point cloud is the laser points reflected back after the emitted laser pulses irradiate the object surface. It is assumed that all the point clouds in the forest stand are reflected back from the vegetation surface. When the VLP-16 scanner is placed vertically, it will generate point cloud data with a vertical field of view of 360° and a horizontal field of view of 30° in the actual azimuth. During the rotation scanning process centered on the station, the horizontal field of view will gradually overlap, and after rotating half a circle or even a full circle, a 360° omnidirectional scan centered on the scanning station will be formed. In order to extract the vertical structure of the forest stand, in most current studies, the point cloud data will be stitched to form the three-dimensional structure of the entire forest stand, and based on this, the vertical stratification of the forest stand will be analyzed. On the one hand, this method will increase the work difficulty and workload, and the stitching accuracy also has a certain impact on the extraction results; on the other hand, the operation of a large amount of point cloud data requires a relatively high computer, and the running time is relatively long. In order to obtain the vertical structure of the forest stand more quickly and conveniently, based on the spatial distribution of the point cloud, a method for extracting the vertical structure of the forest stand based on the quantity statistics of the point cloud in the height direction is proposed. The specific steps are as follows:

[0071] (1) Establish a relationship curve between the point cloud quantity and height. The point clouds in the forest stand are all reflected back from the vegetation surface. The position where the point cloud is located can indicate the presence of vegetation at that place, and the number of point clouds can reflect the amount of vegetation. Therefore, counting the number of point clouds at different heights can show the distribution of vegetation in the vertical direction. The established relationship curve is similar in external form to the waveform data, but is completely different in essence, so it is called a "pseudo-waveform".

[0072] Since there are many reflecting objects in the forest stand, there is a problem of penetration rate during the transmission of laser pulses. For the convenience of data processing, in this embodiment, it is assumed that the laser only undergoes single reflection, and the influence of the laser penetration index is not considered. In addition, due to the existence of the divergence angle, as the distance increases, the point cloud density at the same distance from the scanning station will decrease. Under the same vegetation distribution, the farther away from the scanning station, the fewer the point clouds reflected by the vegetation, which cannot reflect the true situation of the vegetation in the vertical height. Each point cloud represents the point cloud quantity at that place as 1. To eliminate the influence of distance on the point cloud quantity statistics, a corresponding coefficient is multiplied to the value of the point cloud quantity. Since the change of the point cloud density is inversely proportional to the square of the distance, the correction coefficient for each point cloud is the square of the distance value from this point to the scanner.

[0073] After the quantity assignment of the point cloud is corrected, the number of point clouds at different vertical heights is counted. When counting the quantity, the height needs to be divided at a certain interval, and the sum of the number of point clouds within different height intervals is calculated, and then the relationship curve of the number of point clouds - height is established. When dividing the interval, if the interval is too small, there will be too many peaks in the curve, making the effective information mixed in the invalid waveforms and increasing unnecessary analysis; if the interval is too large, some effective waveforms will disappear, affecting the subsequent extraction results. Therefore, it is necessary to find the most suitable interval to divide the overall height. The point cloud data of some stations are selected in each plot, and the relationship curves are established at intervals of 0.1m, 0.2m... 0.9m, 1.0m respectively. By observing the waveforms and peak conditions of each curve, 0.6m is selected as the division interval to retain the effective waveforms and remove the invalid waveforms to the greatest extent.

[0074] (2) Numerical extraction based on the relationship curve. After the curves are established at an interval of 0.6m for all single stations in each plot, the curves need to be smoothed. Common smoothing filtering methods include median filtering, Gaussian filtering, moving average method, etc. After comparing the use of various filtering methods, one-dimensional median filtering is selected to smooth the curves. The waveform of the filtered curve is relatively clear, and the peaks and valleys are relatively obvious. The peak represents a large number of point clouds at that height and relatively dense vegetation, which may be the position of a forest layer in the vertical structure of the forest stand. In order to determine the height and range of each forest layer, it is necessary to automatically find the values of the peaks and valleys through corresponding algorithms. Using the method of finding the curve peaks in MATLAB, the height values corresponding to the peaks and valleys in each curve are calculated. At the same time, the height value corresponding to when the sum of the number of point clouds from the lowest to the highest is 95% of the total number of point clouds is calculated, and this represents the position of the average height of the forest stand.

[0075] (3) Extraction of stand vertical structure parameters. Based on the curves and values established from the data of each single station in the sample plot, determine the number of vertical layers of the stand in each sample plot, the specific heights within each layer, the distribution probabilities within each layer, and the average height of the stand, etc. To facilitate the subsequent comparison of the stand vertical structure, according to the actual situation of the stand, the stand vertical height is artificially stratified to statistically analyze the vertical structure parameters extracted from each single station. During the on-site investigation of the sample plot, it is found that the vertical layers of the stand in the sample plot are relatively rich. Combining the characteristics of the coniferous and broad-leaved mixed forest, a total of five layers are divided, with height ranges of 0 - 5m, 5 - 10m, 10 - 15m, 15 - 20m, and 20 - 25m respectively. Usually, the vegetation below 5m in height is defined as shrubs. In addition, the stand also includes saplings, sub-dominant trees, dominant trees, etc. Therefore, the ranges of 0 - 25m correspond to the heights of the shrub layer regeneration layer, low-layer forest, middle-layer forest, and high-layer forest in sequence. The number of vertical layers of the stand is determined by whether the peak values in all single stations of each sample plot meet the height ranges corresponding to each layer. Then, calculate the average value of the peak values within the same layer range to obtain the specific heights in each vertical layer. The distribution probability of the vegetation within each layer is the ratio of the number of stations within the same layer to the total number of stations. The height corresponding to the first peak value in the curve graph is the ground position. Calculate the difference between the height value corresponding to 95% of the point cloud quantity and the ground height to obtain the average height of the stand. After extracting the relevant parameters in the three sample plots, explore the differences in the stand vertical structure under different harvesting intensities.

[0076] Parameter extraction based on point cloud intensity statistics is as follows:

[0077] When the surface materials of the reflecting objects are the same, the magnitude of the point cloud intensity is related to the number of reflecting objects. In the stand, the greater the point cloud intensity at a certain position, the more vegetation there is at that place. Therefore, statistically analyzing the magnitude of the point cloud intensity at different heights can also show the vertical structure distribution of the stand. The method for extracting parameters based on point cloud intensity statistics is similar to the above, and the specific steps are as follows:

[0078] First, establish the relationship curve between the point cloud intensity and the height. The point cloud intensity is the sum of all point cloud intensities at the same height. Therefore, it is also necessary to assign a value to the point cloud quantity and multiply it by a correction coefficient, which is also the distance from the point cloud to the scanning station, and then statistically analyze the magnitude of the point cloud intensity. Next, it is necessary to determine an appropriate interval height to facilitate the statistical analysis of the point cloud intensity at different heights. Establish relationship curves at intervals of 0.1m, 0.2m... 0.9m, and 1.0m respectively.

[0079] Secondly, the point cloud intensity and height curves established by all single-station data are smoothed and peaks are extracted. After comparing several common filtering methods, one-dimensional median filtering is selected to smooth the curve. The height values ​​corresponding to the peaks and troughs are determined using the MATLAB algorithm, and the height value corresponding to the point cloud intensity from low to high is calculated when the sum of the point cloud intensity from low to high is 95% of the total point cloud intensity.

[0080] Finally, the above method was used to extract the vertical structural parameters of the forest stand. The forest stand was divided into five layers in the vertical direction, with an interval of 5m, and 0-25m was divided into 5 height ranges. Each height range corresponds to the shrub layer, low-level forest, middle-level forest, high-level forest, etc. The layer ranges that the peak values ​​of all single stations in each sample plot meet were counted, the number of vertical layers was determined, and the mean value of the peak values ​​in the same layer was calculated to obtain the height of each vertical layer of the forest stand. Based on the extracted parameters, the vertical structure of the forest stand in each felled sample plot was compared. The average height of the forest stand is calculated by the difference between the height value corresponding to 95% of the point cloud number and the ground height.

[0081] S103. Calculate the stand canopy density by using the quantitative relationship between the point cloud echo and the emission, and analyze the stand growth under different felling intensities based on the stand canopy density and stand vertical structure parameters.

[0082] Specifically, the stand density is extracted as follows:

[0083] In traditional plot surveys, the forest stand canopy density is generally obtained by analyzing and processing hemispherical photos taken of the canopy by the head-up method or a fisheye lens. When using ground-based lidar data to extract canopy density, the canopy porosity is also extracted first, and then the canopy density is calculated based on the relationship between canopy density and porosity. When TLS data is used to extract porosity, most of them are based on the principle of hemispherical photos. The spliced ​​point cloud data is voxelized, and a hemispherical image is generated through coordinate transformation, which is then divided into several concentric rings to calculate the porosity. This method is widely used, but the workflow is cumbersome, and the voxel size also has a certain influence on the porosity inversion. Therefore, this embodiment proposes a new method for porosity extraction. There is no need to voxelize the point cloud data and convert the coordinate system. The zenith angle information of the three-dimensional point cloud is used to count the ratio of the number of "sky" point clouds within different zenith angle ranges to the total number of emitted point clouds.

[0084] Place the scanner in an environment where all echoes can be received, count the number of echoes, and determine the total number of 16-line array laser pulses emitted by the VLP-16 scanner each time. When scanning in a forest stand, assume that the laser pulses that do not receive echoes are emitted into the sky, and the received echoes are all from vegetation reflections. Assign a value of 1 to the pulses that detect vegetation, and count the ratio of the number of laser pulses that detect vegetation to the total number of pulses emitted. 1 minus this ratio is the canopy porosity. The porosity calculation formula is:

[0085] P(θ)=1-Nt / N s

[0086] where θ is the zenith angle, N t is the number of pulses reflected by the vegetation, and N s is the total number of pulses emitted;

[0087] When the VLP-16 scanner is placed vertically, the azimuth angle of the collected point cloud is the zenith angle. By selecting the point clouds within different zenith angle ranges to calculate the porosity, the porosity conditions within a certain horizontal range centered on this site can be obtained. To improve the calculation accuracy of the subsequent canopy density, the point clouds within the 70° zenith angle range are selected for porosity calculation, and the point cloud data is divided at intervals of 10°. The 0-70° zenith angle is divided into 7 intervals. At the same time, the horizontal range of the point cloud data is controlled within 10 m from the scanning site. Calculate the porosity values and the single-site average values within different zenith angle ranges for each single site in the sample plot, and statistically analyze the average value of all sites within the same sample plot to obtain the canopy porosity of this sample plot. Generally, it is considered that the sum of the porosity and the canopy density is 1 in the vertical direction, and based on this, the stand canopy density value is obtained. However, the emission angles of the 16 linear arrays in the VLP-16 scanner are not exactly the same in the vertical direction, and the emitted point clouds are inclined to the ground. To explore whether the inclined point cloud data has an impact on the extraction of the canopy density, the canopy density values are extracted based on the point clouds with different emission angles. In addition, the point clouds on the same linear array are also inclined to the ground. Considering that the number of sites set in this embodiment is small, if only the point clouds in the vertical direction are selected for parameter extraction, the data volume will be greatly reduced, so the inclined point clouds on the same linear array are not removed. In addition, this embodiment can also consider continuously obtaining lidar data at the diagonal position in the sample plot and selecting the point clouds in the vertical direction to extract the canopy density.

[0088] In addition, in this embodiment, the research objects are the logging plots with different logging intensities in Jiaohe Forestry Experimental Area Administration Bureau of Jilin Province. The VLP-16 lidar scanner is placed vertically, and the point cloud data of different stations in each plot is obtained by rotating scanning. In order to invert parameters more quickly and conveniently, a new extraction method is proposed to obtain the relevant parameters of the vertical structure of the forest stand in the logging plot and the value of the forest stand canopy density, and then analyze the vegetation growth situation of the forest stand after logging. The results show that in terms of the vertical structure of the forest stand, the difference between the logging plot and the control plot lies in whether there is vegetation distribution in the higher stratification range and the quantity of vegetation distribution in each stratification. Generally speaking, the vegetation in the logging plot is mainly distributed in the height range of 0-15m, while the vegetation in the control plot is mainly distributed in the height range of 5-20m. The vegetation height in the logging plot has not reached the level of the control plot yet. Especially in the moderately logged plot, the vegetation distribution in the height range of 10-20m is less than that in the lightly logged plot. However, the overall growth situation of the forest stand in the logging plot is good, but it needs more time to make the vertical structure more complete. In terms of the forest stand canopy density, after six years of growth and recovery, the canopy density of each logging plot has increased compared with that in the year of logging. The increase value of the canopy density in the moderately logged plot is the largest, followed by the lightly logged plot. The difference in canopy density between the logging plot and the control plot has also decreased a lot compared with that in the year of logging, but the gap still exists, and the difference in the moderately logged plot is larger. It shows that the canopy degree of the logged forest stand is continuously increasing and it still needs some time to reach the level of the unlogged plot. The vertical structure of the forest stand and the forest stand canopy density respectively reflect the growth situation of the forest stand in the vertical and horizontal directions, and reflect the overall growth situation and spatial structure of the forest stand.

[0089] Embodiment 2

[0090] As Figure 2 shown, a method for inverting forest stand parameters based on a linear array lidar is used for inverting the vegetation parameters of the forest stand affected by fire disturbance. The method includes the following steps:

[0091] S201. Based on the point cloud data collected by single-station scanning, use the single-tree recognition algorithm for linear array lidar data to extract the average breast diameter of the forest stand in the burned and control plots of different fire years;

[0092] The acquisition of single-station lidar data is as follows: The data acquisition of the lidar is synchronized with the field investigation. The instrument for collecting lidar data is Velodyne VLP-16, which is a lidar scanner with a relatively light weight (830g). This scanner is based on the principle of time of flight by pulse method and has a total of 16 scanning linear arrays. When collecting single-station point cloud data, the centers of 9 small sample plots are used as scanning stations, and the station numbers are sequentially 1-9 from left to right and from bottom to top. There are no living trees in some small sample plots of the burned sample plots, so the number of scanning stations in some plots is less than 9.

[0093] The data preprocessing is as follows: The Cloth Simulation Filter (CSF) algorithm is used to filter the point cloud data of each single station. The ground points separated by filtering are used to generate the Digital Elevation Model (DEM). By subtracting the DEM value of the corresponding coordinates from the height values of all point clouds, the influence of the ground slope can be eliminated, and the height normalization of the point cloud can be achieved. Since there are many shrubs blocking in natural forests, in order to facilitate the extraction of subsequent relevant parameters, most of the noise points are removed using the denoising function in CloudCompare software. Then, according to the point cloud intensity information and height values, the point clouds with weak intensity, tree trunks, branches, and their upper parts are removed respectively to complete the data preprocessing.

[0094] The extraction of the single-tree diameter at breast height is as follows: The diameter at breast height is a basic parameter reflecting the growth of trees, and the single-tree diameter at breast height has a certain impact on the single-tree crown width and the stand basal area. According to the characteristics of the line-scanning lidar data, in this embodiment, an algorithm based on distance and tree trunk characteristics is used to identify single trees in the stand, extract the single-tree trunk skeleton points and trunk point clouds, and combine the Random Sample Consensus (RANSAC) algorithm to fit a circle to finally obtain the position and diameter at breast height of each tree. The specific algorithm is as follows:

[0095] (1) Trunk skeleton extraction. The prerequisite for extracting the trunk skeleton is to extract the point cloud representing the trunk from each scan line array. In this embodiment, the trunk point cloud is extracted based on the distance relationship between the single-tree trunk and other surrounding objects to the scan station. A two-dimensional graph is drawn with the negative value (negative distance) of the distance from all point clouds to the scan station as the vertical axis. Since the distances from the point clouds on the trunk of the same tree to the scan station are approximately the same, while the distances from other objects scanned through the gaps between different trunks to the scan station vary greatly, and the point clouds on the trunk are closer to the scan station than other surrounding objects, the point clouds on the trunk appear as several peaks of approximately arc-shaped curves in the two-dimensional graph, and the whole graph is similar to the profile contour line of the Digital Surface Model (DSM).

[0096] Select a window size of 1×3 and perform filtering processing using the template [1, -2, 1]. After that, the arc-shaped curve representing the trunk and the irregular broken line representing other objects can be distinguished by certain rules. If the distance difference between two adjacent points is greater than 2, it is considered that this point may be the boundary between the trunk and the surrounding objects. Then, according to the characteristics of the points on the trunk after filtering, the starting and ending positions of each trunk curve are determined, and then the relative coordinates of the midpoints of the trunk curve are extracted. The midpoint of the trunk curve is the point on each line array of the trunk that is closest to the scan station, and the point cloud selected according to the coordinates of this point is set as the trunk skeleton point. Set the horizontal distance and height thresholds to remove the discrete points far from the trunk and the points close to the crown or trunk branches. The filtered skeleton points are used for the extraction of the trunk skeleton. Taking any trunk skeleton point as the center, search for the skeleton points within a certain threshold range from this point, and the points that meet the conditions are combined to form the trunk skeleton of a tree.

[0097] (2) Hierarchical fitting of circles. Calculate the standard deviation of the relative position coordinates (x, y) of all tree trunk point clouds, and remove the point clouds of extremely irregular tree trunks; based on the extracted skeleton points and tree trunk point clouds, perform circle fitting on the tree trunk point clouds of different scan line arrays through the tree trunk fitting algorithm. Considering the inclined growth of the tree trunks, calculate the inclination direction of the tree trunks, that is, the zenith angle and azimuth angle, based on the skeleton points, determine the rotation matrix through the calculated zenith angle and azimuth angle, and rotate all the tree trunk point clouds to the vertical direction according to the rotation matrix for convenient circle fitting. The rotation matrix is as follows:

[0098]

[0099]

[0100] In the formula, Ty is the rotation matrix for rotation around the y-axis, Tz is the rotation matrix for rotation around the z-axis, α is the zenith angle, and β is the azimuth angle.

[0101] Classify the tree trunk point clouds according to the scan line array number. The tree trunk point clouds of the same line array are fitted with circles using the Random Sample Consensus (RANSAC) algorithm, count the number of circles fitted for each tree trunk, and generate the tree trunk through the continuously fitted circles. The tree trunk after circle fitting is rotated and translated to the original azimuth according to the corresponding rotation matrix and translation formula, and then the position and diameter at breast height of the tree trunk are extracted.

[0102] (3) Extraction of the position and diameter at breast height of individual trees. Extract the position coordinates of the fitted circles at a height of 1.3 m of the same tree, and calculate their mean value as the position coordinates of the individual tree. For tree trunks with more than two fitted circles, perform robust regression on the diameter of the fitted circles and the height of the tree trunk, and extract the diameter at a height of 1.3 m as the diameter at breast height of the tree; for tree trunks with less than three fitted circles, take the diameter of the fitted circles based on the position of the individual tree and calculate the average value as the diameter at breast height of the tree.

[0103] The calculation of the average diameter at breast height of the stand is as follows: The average diameter at breast height of the stand is a basic indicator reflecting the thickness of forest trees, defined as the diameter at breast height corresponding to the average basal area of the stand. The calculation formula is as follows:

[0104]

[0105] In the formula, is the average basal area of the stand, D g is the average diameter at breast height of the stand, N is the total number of forest trees in the stand, G is the total basal area of the stand, g i is the basal area of the i-th forest tree, d i is the diameter at breast height of the i-th forest tree.

[0106] The location coordinates of the scanning sites are (0, 0). Among the forest trees extracted at each scanning site, individual trees with the absolute value of the location coordinates within 5 are selected, and the total number of forest trees is obtained by counting the sum of the number of individual trees that meet the requirements at all sites. Then, the average breast diameter of the stand is calculated.

[0107] S202. Extract the porosity based on the point cloud data collected by mobile scanning, and calculate the leaf area index according to the Lambert-Beer law;

[0108] Specifically, the mobile lidar data is obtained as follows: The mobile acquisition device is a backpack-type lidar scanning system, which consists of a VLP-16 and a mobile phone holder. The power supply, interface box, etc. for the laser scanner to work in this system are integrated in a hard box, and there is a properly sized notch on the box; one end of a telescopic straight rod is fixed in the box, and the other end is connected to the laser scanner outside the box through the notch using a screw connection, and the height of the scanner can be adjusted according to the length of the telescopic straight rod; the mobile phone equipped with the AndroSensor application software is fixed on the mobile phone holder directly below the laser scanner, and data such as longitude, latitude, acceleration, gyroscope, and orientation angle can be recorded in this software. When collecting data in the sample plot, the scanner is placed almost vertically, and the three-dimensional information of the surrounding environment is recorded and stored in real time. When the operator walks in the sample plot, the body movement of the operator himself and the looseness of the system on the operator's back will cause a certain degree of shaking of the scanner. Since the shaking of the mobile phone and the scanner is consistent, the offset and rotation values recorded by the AndroSensor software can be used to adjust the laser scanner to the vertical placement state and reproduce the data collection route.

[0109] The data collection route is as follows: According to the longitude and latitude values recorded by the AndroSensor software, the longitude and latitude values are converted into geographic coordinates using the deg2utm algorithm in MATLAB, and the walking route in the sample plot is drawn. However, in a relatively closed forest stand, the GPS signal will be interfered, and there are offsets at some positions in the drawn walking route. To avoid signal interference, an attempt is made to reproduce the route based on the IMU principle using known data. According to the recorded acceleration values in different time directions, the speed and the distance traveled within adjacent times are calculated, and then the relative coordinates between adjacent points are calculated using the azimuth angle to obtain the route of the entire traveling process. The shape of this route is good, but the system drift is large, and it is calibrated with the GPS coordinates with the best signal, and the final route map is obtained after rotation and translation.

[0110] The point cloud preprocessing is as follows: The original point cloud data collected is in the.pcap format. For the convenience of subsequent processing, the data is converted into a.csv file in the VeloView software. One.csv file represents one frame of point cloud data. During the walking process, the AndroSensor software records the position information every 0.05 seconds. According to the time, the information recorded by the mobile phone is corresponding to the point cloud data, and the tilt angles in the X-axis direction (left and right) and Y-axis direction (front and back) of each frame of point cloud are found. According to the X-axis tilt angle, the zenith angle of the point cloud is added with an offset, and the corrected zenith angle is used for the next parameter extraction. The point cloud with a tilt angle greater than 5° in the Y-axis direction is directly removed, and the point cloud with a tilt angle less than 5° is retained without processing. To avoid the offset deviation caused by the recording information error, finally, the unleveled point cloud is manually adjusted by visual interpretation.

[0111] The porosity extraction is as follows: The method of extracting porosity is the same as the method proposed above, and it is calculated based on the relationship between the number of laser pulses emitted and the number of point cloud echoes. When the backpack mobile scanner obtains data in the forest stand, the azimuth angle of the collected point cloud is the zenith angle, and the actual azimuth angle is related to the forward direction. Using the zenith angle value of the point cloud, the porosity is calculated by selecting the point cloud within different ranges, and this value can reflect the porosity situation in the horizontal azimuth during the traveling process. To obtain higher-precision results, the zenith angle of 0 - 90° is divided into 18 intervals, that is, the point cloud data is divided at intervals of 5°. Calculate the porosity values of each frame of point cloud within different zenith angle ranges in the sample plot, and the mean value is statistically obtained as the canopy porosity of the sample plot. In addition, the porosity values are also extracted based on the above method for the point cloud data with different numbers. Different numbers represent different vertical angles when the linear array laser pulses are emitted, and then explore whether the linear array emission angle has an impact on the parameter inversion.

[0112] The leaf area index calculation is as follows: The leaf area index (LAI) is an important indicator for estimating the ability of the vegetation canopy to interact with the external environment. The LAI can be calculated from the porosity value, and the calculation formula is:

[0113]

[0114]

[0115] In the formula, LAI is the leaf area index, θ is the zenith angle, α is the leaf inclination angle, P(θ) represents the porosity when the zenith angle is θ, G(θ,α) is the projection function, Ω is the clumping index, is the mean value of the porosity of each ring, is the average value of the natural logarithm of the porosity of each ring.

[0116] S203. Analyze the vegetation growth situation and stand structure status of the post-fire forest stand according to the average breast diameter, porosity, and leaf area index of the forest stand.

[0117] Specifically, in this embodiment, the burned areas in the Genhe Forest Region of Inner Mongolia with different fire years were used as the research objects. Based on the VLP-16 linear array laser scanner, two data acquisition methods, namely single-station scanning and mobile scanning, were adopted to obtain the point cloud data of the burned areas with different fire years. A parameter extraction algorithm was proposed for the characteristics of the linear array lidar point cloud to obtain the relevant parameters of the burned forest stands, and the inversion research of the post-fire forest stand vegetation parameters was realized. The results showed that the average breast diameters of the forest stands in the control plot and the burned plot in 2003 were 21.05 cm and 17.02 cm respectively, with differences of 4.94% and 7.79% from the measured values. The estimation errors of the average breast diameters of the forest stands in the burned plots in 2010 and 2015 were 5.75% and 9.29% respectively, and the relative errors of the control plots were relatively large, being 39.27% and 41.58%. Compared with the measured values measured by the fisheye photos, the porosity values extracted from the linear array point cloud had a very high linear regression fitting degree, with R 2 as high as 0.9827. When calculating the leaf area index, the absolute error between the leaf area index calculated based on the linear array point cloud data and the measured results of the fisheye photos was between -0.18 and +1.39. A linear regression relationship between the two was established, and the determination coefficient R 2 of the fitted straight line was 0.7368. The estimated values of the average breast diameters of the forest stands in the control plots in 2010 and 2015 deviated greatly from the measured values, mainly because the understory shrubs were severely blocked, resulting in noise points being misidentified as trunk point clouds during trunk extraction, causing the estimated values to be much larger. The estimated value of the leaf area index of the control plot in 2003 differed greatly from the measured value, mainly because the leaf area index of the plot itself was relatively large, and it would seem much larger when expressed by the absolute error. Overall, the accuracy of inverting the relevant parameters based on the algorithm of this embodiment was relatively high, which reflected the vegetation growth and forest stand structure conditions in different post-fire recovery periods to a certain extent. Through the comparative analysis of the control plots and the burned plots with different fire years, it was found that with the increase of the fire recovery time, both the growth situation and the forest stand structure of the forest stands were gradually recovering to the level of the unburned forest stands.

[0118] To sum up, by means of the above technical solutions of the present invention, the present invention can not only extract the relevant parameters of the forest stand vertical structure and the forest stand canopy density based on the point cloud data obtained by the 360° vertical rotation scanning of the scanner, so as to analyze the growth situation of the forest stands under different logging intensities, but also extract the average breast diameters of the forest stands in the burned and control plots with different fire years based on the point cloud data collected by single-station scanning, and extract the porosity based on the point cloud data collected by mobile scanning, and calculate the leaf area index according to the Lambert-Beer law, so as to analyze the vegetation growth situation and the forest stand structure conditions of the post-fire forest stands.

[0119] In addition, the accuracy of inverting relevant parameters based on the algorithm of the present invention is relatively high, which can reflect the vegetation growth and stand structure conditions of the disturbed forest stand, providing a scientific basis for vegetation restoration. At the same time, the algorithm of the present invention does not require stitching point cloud data, reducing the time cost and workload, providing a new method for inverting relevant parameters of the disturbed forest stand, and also providing a new direction for the application of linear array lidar in forestry.

[0120] In addition, based on the characteristics of single-frame linear array lidar data, the present invention extracts the average breast diameter of the forest stand in the burned area at different restoration periods to realize the estimation of the growth status of the forest stand relying on sparse point clouds. At the same time, based on the number of laser pulses emitted and echoes, the canopy density of the forest stand after logging and the leaf area index of the vegetation after fire are extracted, providing a reference for the structural changes of the forest stand in the plots disturbed by logging and fire.

[0121] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for inverting stand parameters based on a linear array lidar, which is used for inverting the vegetation parameters of a logged disturbed stand, Characterized in that, The method for inverting stand parameters based on a linear array lidar includes the following steps: Obtain the point cloud data of the stand vegetation by vertical rotation scanning with a scanner, and preprocess the obtained point cloud data; Based on the preprocessed point cloud data, establish a relationship curve between the point cloud quantity, point cloud intensity and height, and extract the stand vertical structure parameters by using the peak value in the relationship curve; Calculate the stand canopy density by using the quantitative relationship between the point cloud echo and the emission, and analyze the growth situation of the stand under different logging intensities according to the stand canopy density and the stand vertical structure parameters.

2. The method for inverting stand parameters based on a linear array lidar according to claim 1, Characterized in that, The step of establishing a relationship curve between the point cloud quantity, point cloud intensity and height based on the preprocessed point cloud data, and extracting the stand vertical structure parameters by using the peak value in the relationship curve includes the following steps: Based on the preprocessed point cloud data, establish a relationship curve between the point cloud quantity and height, and determine the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud quantity and height; Based on the preprocessed point cloud data, establish a relationship curve between the point cloud intensity and height, and determine the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud intensity and height.

3. The method for inverting stand parameters based on a linear array lidar according to claim 2, Characterized in that, The step of establishing a relationship curve between the point cloud quantity and height based on the preprocessed point cloud data, and determining the number of stand vertical layers, the specific height values within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud quantity and height includes the following steps: Use a correction coefficient to correct the assignment of the point cloud quantity in the preprocessed point cloud data, where the correction coefficient is the square of the distance value from the point cloud to the scanner; Divide the stand vertical height at intervals, and count the sum of the point cloud quantities at different vertical heights according to the corrected point cloud quantity assignment, establish a relationship curve between the point cloud quantity and height and perform smoothing processing; Use the curve peak finding method to calculate the height values corresponding to the peaks and valleys in the relationship curve between the point cloud quantity and height, and calculate the stand average height value, where the stand average height value is the height value corresponding to when the sum of the point cloud quantities from the lowest to the highest is 95% of the total point cloud quantity; Determine the number of stand vertical layers, the specific height within each layer, the distribution probability within each layer and the stand average height according to the relationship curve between the point cloud quantity and height and the calculation results, where the height of each vertical layer of the stand is calculated by the mean value of the peak values within the same layer.

4. The method for inverting stand parameters based on a linear array lidar according to claim 3, Characterized in that, The method for establishing the relationship curve between point cloud intensity and height based on the preprocessed point cloud data, and determining the number of vertical layers of the forest stand, the specific height values within each layer, the distribution probability within each layer, and the average height of the forest stand includes the following steps: Use the correction coefficient to correct the assignment of the number of points in the preprocessed point cloud data, and divide the vertical height of the forest stand at intervals; According to the corrected assignment of the number of points, count the intensity of the points at different vertical heights, establish the relationship curve between point cloud intensity and height, and perform smoothing processing; Use the curve peak finding method to calculate the height values corresponding to the peaks and valleys in the relationship curve between point cloud intensity and height, and calculate the average height value of the forest stand. Among them, the average height value of the forest stand is the height value corresponding to when the sum of the point cloud intensities from the lowest to the highest is 95% of the total point cloud intensity; Determine the number of vertical layers of the forest stand, the specific height values within each layer, the distribution probability within each layer, and the average height of the forest stand according to the relationship curve between point cloud intensity and height and the calculation results.

5. A method for inverting forest stand parameters based on a linear array lidar according to claim 1, characterized in that, The method for calculating the canopy density of the forest stand by using the quantitative relationship between the point cloud echo and the emission, and analyzing the growth situation of the forest stand under different cutting intensities according to the canopy density and the vertical structure parameters of the forest stand includes the following steps: Place the scanner in a preset environment for scanning, and count the number of echoes to determine the total number of laser pulses emitted by the scanner each time. Among them, the preset environment is an environment that can receive all echoes; Calculate the porosity according to the points within different zenith angle ranges, and obtain the porosity situation within a preset horizontal range centered on this site. Among them, the zenith angle is the azimuth angle of the collected point cloud, and the porosity calculation formula is: P(θ) = 1 - N t / N s where θ is the zenith angle, and N t is the number of pulses reflected by the vegetation, and N s is the total number of pulses emitted; Calculate the porosity values and the single-station average values within different zenith angle ranges of each single station in the sample plot, and count the average value of all stations in the same sample plot to obtain the canopy porosity of the sample plot; According to the relationship between the canopy density and the porosity in the vertical direction, calculate the canopy density of the forest stand based on the points with different emission angles, and analyze the growth situation of the forest stand under different cutting intensities according to the canopy density and the vertical structure parameters of the forest stand.

6. A method for inverting forest stand parameters based on a linear array lidar, used for inverting the vegetation parameters of a fire-disturbed forest stand, characterized in that, This method for inverting forest stand parameters based on a linear array lidar includes the following steps: Based on the point cloud data collected by single-station scanning, use the single-tree recognition algorithm for linear array lidar data to extract the average breast diameters of the forest stands in the burned and control sample plots of different fire years; Extract the porosity based on the point cloud data collected by mobile scanning, and calculate the leaf area index according to the Lambert-Beer law; Analyze the vegetation growth situation and the forest stand structure status of the post-fire forest stand according to the average breast diameter of the forest stand, the porosity, and the leaf area index.

7. A method for inverting forest stand parameters based on a linear array lidar according to claim 6, characterized in that, The method for extracting the average breast diameters of the forest stands in the burned and control sample plots of different fire years by using the single-tree recognition algorithm for linear array lidar data based on the point cloud data collected by single-station scanning includes the following steps: Obtain the point cloud data collected by single-station scanning, and perform filtering and noise reduction processing on the collected point cloud data; Use the single-tree recognition algorithm based on distance and trunk characteristics to identify single trees in the forest stands of burned and control plots in different burned years, obtain the single-tree trunk skeleton points and trunk point clouds, and combine the Random Sample Consensus (RANSAC) algorithm to fit a circle to obtain the position and diameter at breast height (DBH) of each tree. Calculate the average DBH of the forest stands in the burned and control plots in different burned years based on the single-tree DBH.

8. A method for inverting forest stand parameters based on a linear array lidar according to claim 7, characterized in that, the step of using the single-tree recognition algorithm based on distance and trunk characteristics to identify single trees in the forest stands of burned and control plots in different burned years, obtain the single-tree trunk skeleton points and trunk point clouds, and combine the Random Sample Consensus (RANSAC) algorithm to fit a circle to obtain the position and diameter at breast height (DBH) of each tree, and calculate the average DBH of the forest stands in the burned and control plots in different burned years based on the single-tree DBH includes the following steps: Extract the trunk point cloud based on the distance relationship between the single-tree trunk and other surrounding objects to the scanning station, and draw a two-dimensional graph with the negative value of the distance from all point clouds to the scanning station as the vertical axis; Compare the two-dimensional graphs before and after filtering. When the distance between adjacent two points is greater than the spacing threshold, this point is the boundary between the trunk and the surrounding objects, and determine the starting point and ending point positions of each trunk curve according to the characteristics of the points on the trunk after filtering; Extract the relative coordinates of the midpoints of the trunk curves according to the starting point and ending point positions of each trunk curve, and set the points selected by the relative coordinates of the midpoints of the curves as the trunk skeleton points; Remove the discrete points and the points close to the crown or trunk branches based on the set horizontal distance and height thresholds, and search for the skeleton points within the preset threshold range from any trunk skeleton point, and combine the qualified skeleton points to obtain the single-tree trunk skeleton; Calculate the zenith angle and azimuth angle of the trunk tilt direction based on the trunk skeleton points, and determine the rotation matrix through the zenith angle and azimuth angle. Rotate all the trunk point clouds to the vertical direction according to the rotation matrix; Calculate the standard deviation of the relative position coordinates of all the trunk point clouds after rotation, remove the point clouds of irregular trunks, and perform circle fitting on the trunk point clouds of different scanning linear arrays based on the trunk skeleton points and trunk point clouds through the trunk fitting algorithm; Classify the trunk point clouds according to the scanning linear array number, and use the Random Sample Consensus (RANSAC) algorithm to fit a circle for the trunk point clouds in the same linear array. Count the number of circles fitted for each trunk, and generate the trunk through the continuously fitted circles; Rotate and translate the trunk after circle fitting to the original azimuth according to the corresponding rotation matrix and translation formula; Calculate the mean value of the position coordinates of the fitted circles at the preset height to obtain the single-tree position coordinates. For the trunks with more than two fitted circles, take the diameter at the preset height as the DBH of the tree. For the trunks with less than three fitted circles, take the average value of the diameters of the fitted circles based on the single-tree position as the DBH of the tree; Calculate the average DBH of the forest stand according to the total number of trees in the forest stand, the single-tree positions and the single-tree DBHs. The calculation formula is: Where D g is the average diameter at breast height of the stand, N is the total number of trees in the stand, and d i is the diameter at breast height of the i-th tree.

9. A method for inverting forest stand parameters based on a linear array lidar according to claim 6, It is characterized in that extracting the porosity rate from the point cloud data collected by mobile scanning, and calculating the leaf area index according to the Lambert-Beer law includes the following steps: Converting the recorded longitude and latitude values into geographic coordinates, plotting the walking route of the scanner in the plot, and performing format conversion and outlier processing on the point cloud data collected by mobile scanning; Obtaining point cloud data in the forest by a backpack mobile scanner, dividing the zenith angle at a preset interval, calculating the porosity rate values within different zenith angle ranges of each frame of point cloud in the plot, and statistically averaging the porosity rate values within different zenith angle ranges in the same plot to obtain the canopy porosity rate of the plot; Calculating the leaf area index according to the Lambert-Beer law in combination with the porosity rate value and the clumping index.

10. The method for inverting stand parameters based on a linear array lidar according to claim 9, It is characterized in that the calculation formula of the leaf area index is: where LAI is the leaf area index, θ is the zenith angle, α is the leaf inclination angle, P(θ) represents the porosity when the zenith angle is θ, G(θ,α) is the projection function, Ω is the clumping index, is the mean value of the porosities of each ring, is the average value of the natural logarithms of the porosities of each ring.

Images