Method for extracting short time-span growth volume of eucalypt plantations based on UAV data

Abstract

The present application provides a method for extracting short-time-span growth of eucalyptus plantations based on UAV data, and relates to the field of forest resource investigation and forestry quantitative remote sensing research. The method comprises the following steps: field investigation and UAV data collection are carried out on selected eucalyptus plantation sample plots; DEM and monthly DOM and DSM of the sample plot are obtained, and monthly CHM is obtained by subtracting monthly DSM from DEM; a training set is divided from a data set composed of 12-month CHM, each tree crown on the CHM is labeled to obtain a labeled training set, and single tree segmentation is carried out by using the training set data and deep learning algorithm to obtain a monthly single tree segmentation vector diagram; single tree height, single tree crown and DOM spectral reflectance index of the sample plot are extracted; a single tree volume inversion model of the sample plot is constructed; single tree volume of eucalyptus plantations of the sample plot is extracted every month, and the sum is obtained to obtain stand volume, so as to realize short-time-span growth extraction of single tree and stand of eucalyptus plantations.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of forest resource investigation and forestry quantitative remote sensing research, and particularly relates to a method for extracting short-time-span growth of eucalyptus plantation based on UAV data. BACKGROUND

[0002] Eucalyptus not only grows rapidly and has good wood properties, but also plays an important role in ecological safety, global climate change regulation and national timber security, and is the most important source of short fibers in pulp and paper production.

[0003] The growth of single tree is the growth of single tree volume, and the single tree volume refers to the volume of the trunk above the root neck. The growth of stand volume is the sum of the growth of single tree volume in the region. The growth of traditional eucalyptus plantation is usually determined by field measurement to determine the diameter at breast height, tree height and crown width of the tree, and then the growth of single tree volume is calculated to obtain the growth of stand volume. The results are accurate, but time-consuming, costly and inefficient. The method of obtaining single tree and stand growth data through traditional field measurement represented by national forest resource inventory has a long cycle and slow data update frequency, while eucalyptus plantations are mostly short-rotation management, and are harvested and utilized in 3-5 years. Therefore, the existing forest resource inventory data cannot provide timely data support for eucalyptus plantation planning management and policy making, and therefore it is urgent to use advanced technical means to quickly and accurately obtain eucalyptus plantation resources and their dynamic change information.

[0004] The successful application of remote sensing technology to some extent makes up for the shortcomings of traditional field measurement. In order to realize the inversion of eucalyptus plantation stand volume, researchers have studied different types of remote sensing data. Although airborne laser radar data has better inversion results, the high cost of data acquisition and the difficulty of processing make optical satellite images still the most commonly used remote sensing data source for regional eucalyptus plantation volume inversion. Berra et al. extracted band reflectance and vegetation index from medium resolution multispectral Landsat TM image, and estimated the volume of Brazilian eucalyptus plantation based on multiple linear regression. Gebreslasie et al. extracted texture information from high spatial resolution IKONOS image, and estimated the volume of eucalyptus plantation based on multiple linear regression. These existing studies all use satellite remote sensing technology to invert the stand volume of large-area eucalyptus plantations. Since the spatial resolution of free satellite images available on the Internet is generally 10m, 30m or even lower, it is difficult to accurately invert the single tree volume of eucalyptus plantations based on satellite remote sensing technology.

[0005] With the development of motion recovery structure and multi-view stereo vision technology, the UAV (Unmanned Aerial Vehicle) image can quickly extract high-density point cloud data, which greatly increases the forestry application of UAV image digital photogrammetry. In recent years, due to the characteristics of high spatial resolution, low flight height, flexible operation and low cost of UAV, the method based on UAV data has gradually become an effective choice for the research of single tree structure parameters of artificial forest. The collection of high time resolution UAV data provides an economical and feasible alternative solution for the research of temporal growth change of eucalyptus plantation. The time span of the existing research on the extraction of forest single tree and plot scale stand volume growth based on UAV data is mostly years or longer, and the research on the extraction of single tree and plot scale stand growth with a monthly time span is less.

[0006] In summary, the existing research on the extraction of eucalyptus plantation growth still has the following problems:

[0007] 1. From the perspective of extracting eucalyptus plantation growth at single tree scale, the cost of using airborne laser radar technology to extract eucalyptus plantation single tree growth is too high; and when using optical satellites carrying optical sensors to extract eucalyptus plantation single tree growth, the images obtained are often blocked by clouds and rain, and the revisit period is long, so that the satellite images available for the same area may differ by several months, making it difficult to shorten the dynamic change monitoring of eucalyptus plantation single tree growth to a monthly scale.

[0008] 2. From the perspective of extracting eucalyptus plantation growth at stand scale, when using satellite remote sensing technology to invert eucalyptus plantation stand volume, the coverage area of a single satellite image is limited, and when the study area is large, multiple images are usually needed for stitching. The stitched images are usually from different months, and the time difference between the stitched images leads to a certain time difference in the eucalyptus plantation volume inverted based on the images, which can be up to several months, thereby leading to inaccurate annual dynamic change monitoring results in some areas, and the annual change may only be the change in a few months rather than the complete annual change. At the same time, the time difference between the stitched images also limits the time frequency of the dynamic change monitoring results to the annual scale and makes it difficult to achieve the monthly scale. SUMMARY

[0009] In view of the problems existing in the prior art, the present application provides a method for extracting short-time-span growth of eucalyptus plantation based on UAV data, which aims to shorten the extraction of eucalyptus plantation single tree and plot scale stand growth to a monthly scale, so as to realize the dynamic change monitoring of short-time-span single tree and plot scale stand growth of eucalyptus plantation, and provide data support for the short-rotation harvesting management of eucalyptus plantation.

[0010] The technical scheme of the present application is as follows:

[0011] The method for extracting short-time-span growth of eucalyptus plantations based on UAV data, characterized in that the method comprises the following steps:

[0012] Step 1: Select an eucalyptus plantation sample plot, and conduct field investigation and UAV data collection on the sample plot; the UAV data collection includes 1) collecting data on the selected sample plot every month for one year using a consumer-level camera-equipped UAV, and the obtained 12-period UAV data is referred to as Camera UAV data; 2) collecting data on the selected sample plot using a LiDAR sensor-equipped UAV, and the obtained UAV data is referred to as LiDAR UAV data;

[0013] Step 2: input the 12-period Camera UAV data of the sample plot into an aerial photogrammetry software for processing to obtain the digital orthophoto map (DOM) and digital surface model (DSM) of the sample plot in each period; process the LiDAR UAV data using LiDAR point cloud data processing software to generate the digital elevation model (DEM) of the sample plot;

[0014] Step 3: subtract the digital surface model (DSM) of each period of the sample plot from the digital elevation model (DEM) of the sample plot to obtain the canopy height model (CHM) of each period of the sample plot;

[0015] Step 4: divide a training set from the data set composed of the canopy height models (CHM) of all 12 periods of the sample plot, and label each tree crown in the training set of canopy height models (CHM) to obtain labeled training set data; use the labeled training set data and deep learning algorithm to perform single tree segmentation to obtain the single tree segmentation vector map of each period of the sample plot;

[0016] Step 5: according to the single tree segmentation vector map, digital orthophoto map (DOM) and canopy height model (CHM) of each period of the sample plot, use the local maximum value algorithm to extract the single tree height and single tree crown width of each period of the sample plot;

[0017] Step 6: according to the digital orthophoto map (DOM) of each period of the sample plot, extract the spectral reflectance index of the digital orthophoto map (DOM) of each period of the sample plot;

[0018] Step 7: Calculate the actual volume of each eucalyptus tree in the sample plot according to the corresponding single tree height and single tree diameter at breast height collected by field investigation, and form a data set consisting of the actual volume of each eucalyptus tree in the sample plot, the corresponding single tree height, the single tree crown width and the actual volume of each eucalyptus tree in the sample plot collected by field investigation, and divide the training set from the data set; take the single tree volume in the training set as the dependent variable, and take the single tree height, the single tree crown width and the spectral reflectance index of the sample plot in the same month as the independent variables extracted in steps 5 and 6, and use the training set obtained in this step and the random forest algorithm to construct the single tree volume inversion model of the sample plot;

[0019] Step 8: Based on the UAV data of 12 periods, the single tree volume of the eucalyptus plantation in the sample plot is extracted every month by using the extracted single tree height, single tree crown width and spectral reflectance of each period of the sample plot and the single tree volume inversion model of the sample plot, and then the sum of all extracted single tree volumes is obtained to obtain the stand volume of the sample plot every month, and then the single tree volume change and the stand volume change of the sample plot every month are obtained, that is, the single tree growth and the stand growth of the eucalyptus plantation in the sample plot every month, so as to realize the extraction of the short time span growth of the single tree and the sample plot of the eucalyptus plantation.

[0020] Further, according to the eucalyptus plantation short time span growth extraction method based on UAV data, the deep learning algorithm is a DeeplabV3+ model.

[0021] Further, according to the eucalyptus plantation short time span growth extraction method based on UAV data, the method for extracting the single tree height and the single tree crown width of each period of the sample plot in step 5 is: in the ArcGIS software, the single tree segmentation vector graph of each period of the sample plot, the canopy height model CHM of each period of the sample plot and the digital orthographic image DOM are opened through three layers, and the position of the crown of each eucalyptus tree in each period of the sample plot is determined by manual visual interpretation by combining the corresponding single tree segmentation vector graph and the digital orthographic image DOM. The local maximum value algorithm is used to extract the single tree height and the single tree crown width of each period of the sample plot.

[0022] Further, according to the eucalyptus plantation short time span growth extraction method based on UAV data, the spectral reflectance index includes normalized red-blue index NDRB, normalized red-green index NDRG and normalized green-blue index NDGB, wherein the normalized red-blue index NDRB is calculated according to formula (1); the normalized red-green index NDRG is calculated according to formula (2); and the normalized green-blue index NDGB is calculated according to formula (3):

[0023]

[0024]

[0025]

[0026] In the above formula, R is the corresponding pixel value of the normalized red wave band, G is the corresponding pixel value of the normalized green wave band, and B is the corresponding pixel value of the normalized blue wave band.

[0027] Compared with the prior art, the technical scheme provided by the present application has the following beneficial effects:

[0028] (1) The UAV data acquisition is carried out by using a relatively low-cost unmanned aerial vehicle, a simple operation, low-cost and high-precision eucalyptus plantation single tree and sample plot stand growth dynamic change monitoring method is provided for realizing the extraction of eucalyptus plantation short-time span growth, and the time frequency of eucalyptus plantation single tree and sample plot stand growth dynamic change monitoring can be shortened to the monthly scale.

[0029] (2) The extraction result of the monthly scale eucalyptus plantation single tree growth can correct the inaccurate result of the annual dynamic change of the large-area scale eucalyptus plantation stand volume caused by the time difference of the spliced image, provide data support for the time difference correction of the large-area scale eucalyptus plantation stand volume, and further realize the monthly scale dynamic change monitoring of the large-area eucalyptus plantation stand volume.

[0030] (3) The traditional field measurement method for obtaining eucalyptus plantation stand volume is time-consuming, laborious, high in cost, limited in range and long in data acquisition period, and the time required for eucalyptus plantation single tree and sample plot stand growth investigation is greatly reduced.

[0031] (4) The method realizes high-time-frequency growth extraction, improves the update frequency of eucalyptus plantation dynamic change data, and is beneficial to providing timely data support for single eucalyptus plantation forest management and policy making, and providing data support for short-rotation eucalyptus plantation management. BRIEF DESCRIPTION OF DRAWINGS

[0032] In order to more clearly illustrate the specific ways in the embodiments of the present application, the related drawings involved in the embodiments will be briefly described below, and the following drawings are only preferred embodiments of the present application, and other drawings can be obtained according to these drawings without creative changes for those skilled in the art.

[0033] Figure 1 The flowchart of the eucalyptus plantation short-time span growth extraction method based on UAV data of the present embodiment. DETAILED DESCRIPTION

[0034] For the purposes of this application, a more complete understanding of the application can be obtained by reference to the following description taken in connection with the accompanying drawings. The drawings are described below. The preferred embodiments of the application are illustrated in the drawings. However, the application can be implemented in many different forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0035] Figure 1 is a flowchart of the short-time-span growth extraction method of eucalyptus plantations based on UAV data according to the embodiment, and the short-time-span growth extraction method of eucalyptus plantations based on UAV data according to the embodiment includes the following steps:

[0036] Step 1: Select a eucalyptus plantation plot, and conduct field plot investigation and UAV data collection on the plot; the UAV data collection includes 1) collecting data on the selected plot every month for one year using a UAV equipped with a consumer camera, and the obtained 12-period UAV data is referred to as Camera UAV data;

[0037] 2) collecting data on the selected plot using a UAV equipped with a LiDAR sensor, and the obtained UAV data is referred to as LiDAR UAV data;

[0038] In the implementation process of this step, a piece of eucalyptus plantation is arbitrarily selected as the research object. Because the subsequent steps need to use the real single-tree volume data to establish a single-tree volume inversion model, field plot investigation needs to be conducted on the plot. Since the real single-tree volume must be used together with the parameters (single-tree height, single-tree crown width, and spectral reflectance) extracted based on UAV data in the same time period to establish a single-tree volume inversion model, field plot investigation is conducted once when collecting UAV data in any period. The plot investigation mainly includes the coordinates of all single trees in the plot, the height of all single trees, the diameter at breast height of all single trees, the crown width of all single trees, and the number of single trees. When conducting field plot investigation, tools such as height meters, diameter tapes, and tapes are used to measure the height of all eucalyptus trees in the plot, the diameter at breast height of all eucalyptus trees in the plot, and the crown width of all eucalyptus trees in the plot. Because the signal is disturbed under the forest, positioning equipment (such as RTK (Real-time kinematic, real-time dynamic carrier phase difference technology)) cannot work normally, so the coordinates of single trees are obtained by manually processing the data obtained by a light UAV equipped with RTK indoors, and the number of single trees can also be obtained.

[0039] In the implementation of this step, the method for collecting UAV data in the selected sample plot is as follows: a clear, windless and cloudless noon period is selected. On the one hand, the RTK light unmanned aerial vehicle carrying a cheap consumer camera is used to collect data in the sample plot continuously for one year based on one month, and 12 periods of high-resolution Camera UAV data are obtained. The operation mode of the UAV is five-direction flight, and one orthographic direction route and four oblique direction routes are planned, a total of five routes are operated in each period. Among them, the gimbal angle of the UAV in the orthographic direction route is 90°, and the gimbal angle of the UAV in the oblique direction route is 45°. The collected Camera UAV data is used to generate digital orthophoto map (DOM) and digital surface model (DSM) in the subsequent step. On the other hand, in order to obtain the terrain information of the sample plot, the RTK light unmanned aerial vehicle carrying the LiDAR (Light Laser Detection and Ranging) sensor is used to collect data in the selected sample plot. Since the terrain usually does not change, it is only necessary to collect LiDAR UAV data in the selected sample plot once. In this embodiment, a clear, windless and cloudless noon period is selected, and the light unmanned aerial vehicle carrying the LiDAR sensor is used to operate under the condition that the flight height is set to 60 m, the flight speed is set to 6 m / s, the echo mode is 3 echoes, the sampling frequency is 240 KHZ, the scanning mode is set to repeated scanning, and the lateral overlap is 80%. The collected LiDAR UAV data is used to generate the digital elevation model (DEM) of the selected sample plot in the subsequent step.

[0040] Step 2: The 12 periods of Camera UAV data of the sample plot are input into the aerial photogrammetry software for processing to obtain the digital orthophoto map (DOM) and digital surface model (DSM) of the sample plot in each period. The LiDAR UAV data is processed by using the LiDAR point cloud data processing software to generate the digital elevation model (DEM) of the sample plot.

[0041] In the implementation process of this step, the 12-phase Camera UAV data of the sample plot is input into the aerial photogrammetry software (such as Pix4Dmapper, ContextCapture, PhotoScan, Inpho, Tiangong Godwork) for processing to obtain the digital orthographic image DOM and digital surface model DSM of each phase of the sample plot. In this embodiment, the 12-phase Camera UAV data collected by the RTK light unmanned aerial vehicle carrying a cheap consumer camera is put into the Pix4Dmapper software, and the initialization processing, generation of three-dimensional dense point cloud and generation of DOM operation are sequentially performed in the Pix4Dmapper software to obtain the digital orthographic image DOM of each phase of the sample plot; the 12-phase Camera UAV data collected by the RTK light unmanned aerial vehicle carrying a cheap consumer camera is put into the ContextCapture software, and the aerial triangulation settlement, three-dimensional grid generation and DSM generation operation are sequentially performed in the ContextCapture software, so that the digital surface model DSM of each phase of the sample plot can be obtained; the LiDAR UAV data collected by the RTK light unmanned aerial vehicle carrying a LiDAR sensor is put into the LiDAR point cloud data processing software (such as DJI ZhiTu software, LiDAR360, LASTools, TerraSolid series software), and in this embodiment, the LiDAR UAV data is put into the DJI ZhiTu software, and the digital elevation model DEM of the sample plot can be obtained by using the DEM generation function.

[0042] Step 3: Subtracting the digital surface model DSM of each phase of the sample plot obtained in step 2 from the digital elevation model DEM of the sample plot to obtain the canopy height model CHM of each phase of the sample plot.

[0043] In the implementation process of this step, the digital surface model DSM and the digital elevation model DEM obtained in step 2 are opened in the ArcGIS software, and the raster calculator in the map algebra function is used to subtract the raster values of the digital surface model DSM of each phase of the sample plot from the raster values of the digital elevation model DEM of the sample plot, so that the canopy height model CHM (Canopy Height Model) of each phase of the sample plot can be obtained.

[0044] Step 4: Dividing the training set from the data set composed of the canopy height models CHM of all 12 phases of the sample plot, and labeling each tree crown on the canopy height model CHM in the training set to obtain the training set data with labels; using the training set data with labels and the deep learning algorithm to perform single tree segmentation to obtain the single tree segmentation vector diagram of each phase of the sample plot;

[0045] In the implementation process of this step, first, the canopy height model CHM of the entire 12 periods of the sample plot is subjected to clipping segmentation processing, and then the data set composed of the CHM after the clipping segmentation processing is divided into a training set, a test set and a validation set according to a ratio of 6:2:2; then the canopy height model CHM in the training set is labeled using the labelme software, each tree crown in the canopy height model CHM in the training set is drawn and labeled as a tree crown, and is put into a deep learning algorithm for training of a single tree segmentation model. The deep learning algorithm of this embodiment adopts a DeeplabV3+ model, the test set data is used to test the single tree segmentation model after the preliminary training, so as to adjust the hyperparameters in the model and further improve the single tree segmentation model to obtain a trained single tree segmentation model. Finally, the validation set data is put into the trained single tree segmentation model for prediction, so that the single tree segmentation result of each period of the sample plot can be obtained, and the single tree segmentation vector diagram of each period of the sample plot is obtained.

[0046] Step 5: According to the single tree segmentation vector diagram of each period of the sample plot, the digital orthophoto map DOM and the canopy height model CHM, the single tree height and the single tree crown width of each period of the sample plot are extracted by using the local maximum value algorithm.

[0047] In the ArcGIS software, the single tree segmentation vector diagram of each period of the sample plot, the canopy height model CHM of each period of the sample plot and the digital orthophoto map DOM are opened in three layers respectively, and the position of each eucalyptus tree crown in each period of the sample plot is determined by combining the corresponding single tree segmentation vector diagram and the digital orthophoto map DOM through artificial visual interpretation. The single tree height and the single tree crown width of each period of the sample plot are extracted by using the local maximum value algorithm on the canopy height model CHM of each period of the sample plot.

[0048] Step 6: According to the digital orthophoto map DOM of each period of the sample plot, the spectral reflectance index of the digital orthophoto map DOM of each period of the sample plot is extracted.

[0049] In ArcGIS, open the digital orthophoto map DOM of each period of the sample plot, first normalize each pixel value in the digital orthophoto map DOM: Since the DOM has only three bands of red, green and blue, the digital orthophoto map DOM is an RGB image. Each pixel value in the RGB image is between 0 and 255, therefore, divide all pixel values in the digital orthophoto map DOM of each period of the sample plot by 255 to complete the normalization of the digital orthophoto map DOM; then extract the spectral reflectance index of the DOM of each period of the sample plot by calculation. The spectral reflectance index includes NDRB normalized red-blue index, NDRG normalized red-green index and NDGB normalized green-blue index, wherein the NDRB normalized red-blue index is calculated according to formula (1); the NDRG normalized red-green index is calculated according to formula (2); and the NDGB normalized green-blue index is calculated according to formula (3).

[0050]

[0051]

[0052]

[0053] In the above formula, R is the corresponding pixel value of the normalized red band, G is the corresponding pixel value of the normalized green band, and B is the corresponding pixel value of the normalized blue band.

[0054] Step 7: Calculate the actual value of the single tree volume according to the corresponding single tree height and single tree diameter at breast height of each eucalyptus tree in the sample plot collected by the field sample plot investigation, and divide the data set composed of the corresponding single tree height, single tree crown width and actual value of the single tree volume of each eucalyptus tree in the sample plot collected by the field sample plot investigation into a training set; take the single tree volume in the training set as the dependent variable, and take the single tree height, single tree crown width and spectral reflectance index of the sample plot in the same month extracted in step 5 and step 6 as the independent variables, and use the training set obtained in this step and the random forest algorithm to construct the single tree volume inversion model of the sample plot.

[0055] In the implementation process of this step, the corresponding single tree height and single tree diameter at breast height of each eucalyptus tree in the sample plot collected by the field sample plot investigation are used, and the actual value of the single tree volume is calculated by using the Guangxi Eucalyptus Binary Volume Table compiled by Huang Dao Nian and Liao Zezhao (1986):

[0056] V = 0.0434785 - 6.75245 x 10 -3 D 2 + 5.02044 x 10 -4 DH + 1.54609 x 10 -5 D 2 H - 3.35291 x 10-3 H (4) In the above formula, V represents the single tree volume of the eucalyptus plantation; D represents the diameter at breast height; and H represents the tree height.

[0057] The data set is composed of the actual values of the single tree height, the single tree crown width and the single tree volume of each eucalyptus tree in the sample plot obtained from the field investigation. 75% of the data set is selected as the training set, and the remaining 25% is selected as the test set. When the random forest algorithm is used, the actual value of the single tree volume of the eucalyptus plantation is taken as the dependent variable, the single tree height, the single tree crown width and the spectral reflectance index of the sample plot in the same month as the field investigation are extracted based on steps 5 and 6, which are called characteristic factors, and the characteristic factors are taken as the independent variables. The training set data obtained in this step is put into the random forest algorithm for regression training to obtain a regression model, i.e., the single tree volume inversion model of the sample plot. The test set data is used to evaluate the prediction accuracy of the single tree volume inversion model of the eucalyptus plantation.

[0058] Step 8: Based on the UAV data of 12 periods, the single tree volume of the eucalyptus plantation in each month of the sample plot is extracted by using the single tree height, the single tree crown width and the spectral reflectance of each period of the sample plot and the single tree volume inversion model of the sample plot. Then, the sum of all the extracted single tree volumes in the sample plot is calculated to obtain the stand volume of the sample plot. Then, the change amount of the single tree volume and the change amount of the stand volume of the sample plot in each month are obtained, i.e., the single tree growth and the stand growth of the eucalyptus plantation in each month of the sample plot, so as to realize the extraction of the short-time-span growth of the single tree and the sample plot of the eucalyptus plantation.

[0059] In the implementation process of this step, based on the UAV data of 12 periods, the single tree height, the single tree crown width and the spectral reflectance index of the sample plot are extracted, and the single tree volume inversion model of the eucalyptus plantation is established. Then, the single tree volume of the eucalyptus plantation in each month of the sample plot is extracted. Then, the sum of all the extracted single tree volumes in the sample plot is calculated to obtain the stand volume of the sample plot. Then, the change amount of the single tree volume and the change amount of the stand volume of the sample plot in each month are obtained, i.e., the single tree growth and the stand growth of the eucalyptus plantation in each month of the sample plot, so as to realize the extraction of the short-time-span growth of the single tree and the sample plot of the eucalyptus plantation.

[0060] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not limited thereto. Although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that the technical solutions described in the foregoing examples can be modified, or some or all of the technical features can be replaced by equivalents. Therefore, these modifications or replacements do not change the essence of the corresponding technical solutions beyond the scope defined by the claims of the present application.

Claims

1. A method for extracting short-span growth of eucalyptus plantations based on UAV data, characterized in that, The method includes the following steps: Step 1: Select a eucalyptus plantation plot and conduct field surveys and UAV data collection on the plot. The UAV data collection includes: 1) collecting data on the selected plot monthly for one year using a drone equipped with a consumer-grade camera. The obtained 12 periods of UAV data are called Camera UAV data; 2) collecting data on the selected plot using a drone equipped with a LiDAR sensor. The obtained UAV data are called LiDAR UAV data. Step 2: Input the 12 periods of Camera UAV data of the quadrat into aerial photogrammetry software for processing to obtain the digital orthophoto (DOM) and digital surface model (DSM) of the quadrat for each period; use LiDAR point cloud data processing software to process the LiDARUAV data to generate the digital elevation model (DEM) of the quadrat. Step 3: Subtract the Digital Surface Model (DSM) and Digital Elevation Model (DEM) of each quadrat to obtain the Canopy Height Model (CHM) of each quadrat. Step 4: Divide the training set from the dataset consisting of the canopy height model CHM of all 12 phases of the quadrat, and label each canopy on the canopy height model CHM in the training set to obtain labeled training set data; use the labeled training set data and deep learning algorithm to perform single tree segmentation to obtain the single tree segmentation vector map of each phase of the quadrat. Step 5: Based on the single-tree segmentation vector map, digital orthophoto DOM, and canopy height model CHM of each sample plot, extract the single-tree height and single-tree canopy width of each sample plot using the local maximum algorithm; Step 6: Extract the spectral reflectance index of the digital orthophoto DOM of each sample plot based on the DOM of each sample plot. Step 7: Calculate the actual value of the timber volume per eucalyptus tree based on the individual tree height and diameter at breast height (DBH) of each eucalyptus tree in the quadrat collected from the field survey. The dataset is composed of the individual tree height, crown width, and timber volume of each eucalyptus tree in the quadrat collected from the field survey, and a training set is extracted from this dataset. Using the timber volume per eucalyptus tree in the training set as the dependent variable and the individual tree height, crown width, and spectral reflectance index of the quadrat in the same month as the field survey extracted in Steps 5 and 6 as independent variables, construct the timber volume inversion model of the quadrat using the training set obtained in this step and the random forest algorithm. Step 8: Based on the 12 periods of UAV data, using the extracted tree height, crown width, and spectral reflectance of each plot, and the plot's individual timber volume inversion model, extract the monthly eucalyptus plantation individual timber volume for each plot. Then, sum all the extracted individual timber volumes to obtain the monthly stand volume of the plot, thereby obtaining the monthly change in individual timber volume and stand volume, i.e., the monthly growth of individual trees and the monthly growth of the eucalyptus plantation in the plot. This achieves the extraction of short-term growth of individual trees and plot stands in eucalyptus plantations.

2. The method for extracting short-span growth of eucalyptus plantations based on UAV data according to claim 1, characterized in that, The deep learning algorithm is the DeeplabV3+ model.

3. The method for extracting short-span growth of eucalyptus plantations based on UAV data according to claim 1, characterized in that, The method for extracting the individual tree height and individual tree crown width of each sample plot in step 5 is as follows: In ArcGIS software, the individual tree segmentation vector map, the crown height model (CHM), and the digital orthophoto DOM of each sample plot are opened as three separate layers. Combining the corresponding individual tree segmentation vector map and digital orthophoto DOM, the position of each eucalyptus tree crown in each sample plot is determined by manual visual interpretation. The individual tree height and individual tree crown width of each sample plot are extracted by using the local maximum algorithm on the crown height model (CHM).

4. The method for extracting short-span growth of eucalyptus plantations based on UAV data according to claim 1, characterized in that, The spectral reflectance index includes the normalized red-blue index NDRB, the normalized red-green index NDRG, and the normalized green-blue index NDGB, wherein the normalized red-blue index NDRB is calculated according to equation (1); the normalized red-green index NDRG is calculated according to equation (2); and the normalized green-blue index NDGB is calculated according to equation (3). In the above formula, R is the normalized pixel value of the red band, G is the normalized pixel value of the green band, and B is the normalized pixel value of the blue band.

Images