A tree root sampling device

Abstract

The present invention discloses a tree root sampling device, which relates to the technical field of tree root sampling. The device comprises: a distribution identification module for identifying the distribution of trees in a target area and generating a tree point distribution map; a grid division module for dividing the sampling grids according to the coordinates of the tree point distribution map and determining Q sampling grids; a sampling quantity distribution module for allocating Q sampling quantities based on the tree distribution density of each sampling grid; a ground scanning module for scanning the Q grids using a portable GPR device, inverting the root distribution, and determining dense areas, sparse areas, and interlaced areas; and a sampling point identification module for identifying a set of sampling points based on the sampling quantity and root distribution, and extracting root samples at the sampling points. This achieves the technical effect of improving the scientific nature of sampling point distribution and sampling quantity allocation, and enhancing sampling accuracy and efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of tree root sampling, in particular to a tree root sampling device. Background Art

[0002] In tree root system research, traditional sampling methods usually rely on manual experience or randomly selected sampling points, lack a systematic scientific basis, and often cannot fully cover the root distribution characteristics in the target area. They are also easily affected by human factors, resulting in uneven distribution of sampling points and unreasonable distribution of sampling volume, which affects the representativeness and scientificity of the sampling results. Summary of the Invention

[0003] The present invention provides a tree root sampling device to solve the technical problems in the prior art such as uneven distribution of sampling points, unreasonable distribution of sampling volume, and impact on the rationality and efficiency of sampling, thereby achieving the technical effect of improving the scientific nature of sampling point distribution and sampling volume distribution, and enhancing sampling accuracy and efficiency.

[0004] The present invention provides a tree root sampling device, comprising:

[0005] The distribution recognition module is used to identify the point distribution of trees in the target area and determine the tree point distribution map.

[0006] The grid division module is used to divide the sampling grid according to the tree point coordinates in the tree point distribution map, and determine Q sampling grids, where Q is a positive integer.

[0007] The sampling amount distribution module is used to distribute the sampling amount according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts.

[0008] The ground scanning module is used to use a portable GPR device to perform ground scanning on Q sampling grids respectively, and perform root distribution inversion based on the scanning data to determine Q root-dense areas, Q root-sparse areas and Q staggered distribution areas.

[0009] The sampling point identification module is used to identify the sampling points of the Q sampling grids based on the Q sampling amounts, the Q root-dense areas, the Q root-sparse areas, and the Q staggered distribution areas, determine Q sampling point sets, and extract root samples at the Q sampling point sets.

[0010] In a feasible implementation, sampling grids are divided according to the tree point coordinates in the tree point distribution map to determine Q sampling grids. The steps of executing the grid division module include:

[0011] According to the first sampling scale and the tree point distribution map, random sampling tree points of a preset identification sampling amount are screened to determine M first random sampling tree points and M second random sampling tree points, where M is a positive integer and M is less than or equal to Q.

[0012] Perform pairwise distance interval authentication on the M first randomly sampled tree locations and the M second randomly sampled tree locations, respectively. Based on the authentication results, determine M target randomly sampled tree locations. Divide the tree location distribution map into sampling grids based on the M target randomly sampled tree locations to obtain the Q sampling grids.

[0013] In a feasible implementation, M target randomly sampled tree locations are determined, and the steps of executing the grid division module further include:

[0014] Determine whether the distance intervals of the pairwise random sampling tree point combinations of the M first randomly sampled tree points and the M second randomly sampled tree points are all greater than or equal to a preset distance interval threshold for the pairwise random sampling tree point combinations. If so, calculate the sampling dispersion coefficients of the M first randomly sampled tree points and the M second randomly sampled tree points to obtain a first sampling dispersion coefficient and a second sampling dispersion coefficient.

[0015] When the first sampling dispersion coefficient is greater than the second sampling dispersion coefficient, the M first randomly sampled tree locations are used as the M target randomly sampled tree locations.

[0016] In a feasible implementation, the Q sampling grids are obtained, and the execution step of the grid division module further includes:

[0017] The M target randomly sampled tree points are respectively used as sampling grid centers, and M initial division grids are constructed according to a preset grid construction scale.

[0018] The M initial divided grids are diffused once according to a preset diffusion scale to obtain M primary diffused divided grids.

[0019] When the grid density of the M initial division grids is less than or equal to the grid density of the M first-diffusion division grids, the M first-diffusion division grids are used as M stage diffusion division grids, and secondary diffusion is continued according to the preset diffusion scale until the diffusion stop constraint is satisfied, thereby obtaining M target diffusion division grids.

[0020] The M target diffusion division grids are interactively overlapped and analyzed in combination with the tree point distribution map, and the missing tree points are supplemented with grids to obtain the Q sampling grids.

[0021] In a feasible implementation, the diffusion stop constraint is that the number of diffusions meets a preset maximum number of diffusions and / or the difference in grid density between two adjacent diffusions is less than or equal to a preset difference threshold.

[0022] In a feasible implementation, the M target diffusion grids are interactively overlapped and analyzed in combination with the tree point distribution map, and missing tree points are supplemented to obtain the Q sampling grids. The execution steps of the grid division module further include:

[0023] Based on the tree point distribution map and the regional positions of the M target diffusion division grids, an interactive overlapping area is determined to obtain a set of M grid interactive overlapping areas.

[0024] The tree points in the M grid interactive overlapping area sets are respectively retained in the target diffusion division grids closest to them, to obtain M cleaning grid interactive overlapping area sets.

[0025] The M target diffusion partitioning grids are updated based on the M sets of interactive overlapping regions of the cleaning grids to obtain M updated diffusion partitioning grids.

[0026] The tree points in the tree point distribution map except the M updated diffusion divided grids are summarized to obtain a missing tree point set.

[0027] Perform nearest neighbor partitioning on the missing tree point set to obtain N missing partitioning grids, where N is a positive integer.

[0028] The M updated diffusion partitioning grids and the N missing partitioning grids are aggregated to obtain the Q sampling grids, where Q=M+N.

[0029] In a feasible implementation, the missing tree point set is divided according to a preset equal division strategy to obtain the N missing division grids.

[0030] In a feasible implementation, Q sampling point sets are determined, and the steps executed by the sampling point identification module include:

[0031] The proportions of the Q dense root areas, the Q sparse root areas, and the Q staggered distribution areas in the Q sampling grids are obtained to obtain Q dense proportions, Q sparse proportions, and Q staggered distribution proportions.

[0032] The Q dense ratios, Q sparse ratios, and Q interleaved distribution ratios are multiplied by the Q sampling amounts to determine Q dense sampling amounts, Q sparse sampling amounts, and Q interleaved distribution sampling amounts.

[0033] Based on the Q dense sampling amounts, the Q sparse sampling amounts and the Q interactively distributed sampling amounts, the Q root-dense areas, the Q root-sparse areas and the Q interlaced distribution areas are randomly sampled to determine Q sampling point sets.

[0034] In a feasible implementation, the method further includes:

[0035] The failure sample acquisition unit is used to obtain a sampling failure sample log set.

[0036] The failure point updating unit is used to traverse and obtain the sampling environment information of the set of Q sampling points, perform feature comparison on the sampling environment information with the set of sampling failure sample logs, and if the feature comparison result meets the preset comparison threshold, mark the corresponding sampling point as a failed sampling point and randomly select a new sampling point.

[0037] In a feasible implementation, the sampling amount is distributed according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts. The execution steps of the sampling amount distribution module include:

[0038] The total number of trees in the Q sampling grids is counted respectively, and the statistical results are respectively divided by the area of ​​the Q sampling grids to obtain the Q tree point distribution densities.

[0039] The Q tree point distribution densities are divided by the sum of the Q tree point distribution densities to obtain Q sampling amount distribution coefficients.

[0040] A preset total sampling amount is obtained, and the Q sampling amount distribution coefficients are traversed and multiplied by the preset total sampling amount to obtain the Q sampling amounts.

[0041] The present invention discloses a tree root sampling device, comprising: a distribution identification module that identifies the distribution of trees in a target area and generates a tree point distribution map; a grid division module that divides the sampling grid according to the tree point coordinates in the distribution map and determines Q sampling grids, where Q is a positive integer; a sampling quantity distribution module that distributes the sampling quantity according to the distribution density of the tree points in each sampling grid and obtains Q corresponding sampling quantities; a ground scanning module that uses a portable GPR device to perform ground scanning on each of the Q sampling grids and inverts the root system based on the scanned data. The invention discloses a tree root sampling device which solves the technical problems of uneven sampling point distribution, unreasonable sampling amount distribution and influence on the rationality and efficiency of sampling, and realizes the technical effect of improving the scientificity of sampling point distribution and sampling amount distribution, and improving sampling accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a structural schematic diagram of a tree root sampling device of the present invention.

[0043] Figure 2 The figure is a schematic diagram of the execution flow of a sampling point identification module in a tree root sampling device of the present invention.

[0044] Description of the accompanying drawings: distribution identification module 11, grid division module 12, sampling quantity distribution module 13, ground scanning module 14, sampling point identification module 15. DETAILED DESCRIPTION

[0045] The above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods of the specification to better understand the above technical solution. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments used only to explain the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. In addition, it should be noted that, for the convenience of description, only the parts related to the present invention, rather than all, are shown in the drawings.

[0046] Example, Figure 1 The present invention is a schematic structural diagram of a tree root sampling device, wherein the tree root sampling device comprises:

[0047] The distribution identification module 11 is used to identify the point distribution of trees in the target area and determine a tree point distribution map.

[0048] Specifically, first, the trees in the target area are identified through the distribution identification module 11 to determine the tree point distribution map, wherein the tree point distribution map reflects multiple groups of data including the coordinates, distribution density, etc. of the trees, and each point corresponds to an individual tree.

[0049] Optionally, the distribution identification module 11 identifies the locations of trees in the target area through methods such as remote sensing technology, drone mapping, and manual marking to generate a tree point distribution map: first, the location information of the trees is obtained using remote sensing equipment or manual methods; then, the obtained location information is input into a geographic information system (GIS) to generate a visual map, thereby providing basic data support for subsequent grid division and sampling point identification.

[0050] The grid division module 12 is used to divide the sampling grids according to the tree point coordinates in the tree point distribution map, and determine Q sampling grids, where Q is a positive integer.

[0051] Specifically, the target area is divided into multiple sub-areas based on the coordinates of the tree points, where each sampling grid is a sub-area. During the division process, it is necessary to ensure that the tree distribution characteristics within each grid are relatively uniform to avoid the trees in the grid being too dense or sparse. In other words, the tree distribution characteristics obtained in each sub-area can be considered relatively uniform.

[0052] Through scientific grid division, the target area can be divided into multiple representative sub-areas, avoiding the problem of uneven distribution of sampling points in traditional methods. At the same time, the tree distribution characteristics within each grid are relatively uniform, ensuring the rationality of subsequent sampling volume allocation.

[0053] In some embodiments, sampling grids are divided according to the tree point coordinates in the tree point distribution map to determine Q sampling grids. The steps executed by the grid division module 12 include:

[0054] According to the first sampling scale and in combination with the tree point distribution map, a preset identified sampling amount of random sampling tree points is screened to determine M first random sampling tree points and M second random sampling tree points, where M is a positive integer and M is less than or equal to Q; pairwise distance interval authentication is performed on the M first random sampling tree points and the M second random sampling tree points, and based on the authentication results, M target random sampling tree points are determined; based on the M target random sampling tree points, the tree point distribution map is divided into sampling grids to obtain the Q sampling grids.

[0055] Specifically, the first sampling scale is the preset sampling density or sampling interval, which is used to control the scope and number of random sampling. The preset identification sampling volume refers to the number of sampling points pre-set according to research needs. The target random sampling tree sites are random sampling sites that meet the preset conditions after distance interval certification.

[0056] Specifically, according to the first sampling scale and the preset identification sampling quantity, M first randomly sampled tree points and M second randomly sampled tree points are randomly selected from the tree point distribution map, where M is the preset identification sampling quantity; then, the selected M first randomly sampled tree points and M second randomly sampled tree points are respectively subjected to pairwise distance interval authentication to ensure that the distance between each pair of points is greater than or equal to the preset distance interval threshold, thereby avoiding the selection of points that are too close and do not meet the expectations; then, based on the authentication results, M target randomly sampled tree points that meet the distance requirements are selected.

[0057] Furthermore, the tree point distribution map is divided according to preset rules with the M target randomly sampled tree points as the center, thereby obtaining Q sampling grids. Through the above steps, the scientific and uniform division of the sampling grids is ensured, thereby avoiding the unreasonable grid division problem caused by human selection in traditional methods.

[0058] In some implementations, after determining M target randomly sampled tree locations, the gridding module 12 may further execute the following steps:

[0059] Determine whether the distance intervals of any two-way random sampling tree point combinations of the M first randomly sampled tree points and the M second randomly sampled tree points are all greater than or equal to a preset distance interval threshold for any two-way random sampling tree point combinations. If so, calculate the sampling dispersion coefficients of the M first randomly sampled tree points and the M second randomly sampled tree points to obtain a first sampling dispersion coefficient and a second sampling dispersion coefficient. When the first sampling dispersion coefficient is greater than the second sampling dispersion coefficient, use the M first randomly sampled tree points as the M target random sampling tree points.

[0060] Specifically, the distance interval between any two randomly sampled tree locations is the straight-line distance between any two sampling points. This value is used to determine whether the samples are clustered. The preset distance interval threshold is a minimum distance control parameter used to constrain the distribution density between sample points and prevent excessive sample concentration. The sampling dispersion coefficient is an indicator used to quantify the spatial dispersion of the sampling points; a larger value indicates a more evenly distributed sample.

[0061] Specifically, first, any two points from the M first randomly sampled tree locations (and the M second randomly sampled tree locations) are combined to form M(M-1) / 2 point pairs, and the spatial distances (such as GPS distances) between these point pairs are calculated one by one. If the obtained M(M-1) / 2 distance values ​​are all greater than or equal to the preset distance interval threshold, the sampling dispersion coefficients of the M first randomly sampled tree locations and the M second randomly sampled tree locations are further calculated. Exemplarily, the sampling dispersion coefficients are calculated in the following ways: the standard deviation of the distances between points, which reflects the concentration of the point distribution; the average distance of each point to several neighboring points, and then taking the overall average as a uniformity measure; constructing a Voronoi diagram, and quantifying the discreteness according to the standard deviation of the polygon areas generated by the distribution of each point.

[0062] Specifically, if the points are more evenly distributed across the entire park boundary and central area, the sampling dispersion coefficient will be larger. Conversely, if they are concentrated in a corner, the sampling dispersion coefficient will be smaller. Based on the first sampling dispersion coefficient and the second sampling dispersion coefficient, a quantitative comparison of the spatial coverage capabilities of the two schemes can be made.

[0063] Furthermore, if the first sampling dispersion coefficient is greater than the second sampling dispersion coefficient, it can be considered that the M first randomly sampled tree locations have stronger spatial coverage and better uniformity, and the M first randomly sampled tree locations are selected as the final M target randomly sampled tree locations.

[0064] By introducing two quality indicators, point distance threshold and sampling dispersion, on the basis of initial randomness, the problems of point clustering and coverage blind spots in traditional random sampling are effectively solved, the spatial representativeness and stability of the sampling data are improved, and repeated sampling of the same area can be avoided, the monitoring coverage can be expanded, and more extensive and scientific basic data can be provided for subsequent vegetation growth analysis, carbon flux assessment and environmental anomaly detection.

[0065] In some implementations, to obtain the Q sampling grids, the grid division module 12 may further execute the following steps:

[0066] Taking the M target randomly sampled tree points as sampling grid centers, M initial division grids are constructed according to a preset grid construction scale; the M initial division grids are diffused once according to a preset diffusion scale to obtain M primary diffusion division grids; when the grid density of the M initial division grids is less than or equal to the grid density of the M primary diffusion division grids, the M primary diffusion division grids are used as M stage diffusion division grids, and secondary diffusion is continued according to the preset diffusion scale until the diffusion stop constraint is satisfied, thereby obtaining M target diffusion division grids; in combination with the tree point distribution map, interactive overlapping analysis is performed on the M target diffusion division grids, and grids are supplemented for missing tree points to obtain the Q sampling grids.

[0067] Specifically, the initial gridding refers to a basic sampling area established with the target randomly sampled tree site as the center, based on a standard size (preset grid construction scale) such as 5m×5m or 10m×10m. The number of such areas is M, which corresponds to the number of target randomly sampled tree sites. The diffusion scale refers to the range or step size of the outward expansion of the grid, which is used to adjust the coverage of the grid and control the growth rate of the grid. The grid density refers to the distribution density of tree sites within the grid, usually defined as the number of valid sample points contained per unit area or the proportion of sampling points covered by the grid.

[0068] Specifically, the stage diffusion mesh refers to the intermediate mesh formed after a single diffusion process. The target diffusion mesh is the final mesh after multiple diffusion processes and satisfying the diffusion stop constraint. The diffusion stop constraint is a condition that limits the termination of the diffusion process, such as the maximum number of diffusions, the maximum coverage area, or the density saturation value.

[0069] Specifically, in this implementation, M target randomly sampled tree points are first obtained, and M initial division grids are generated according to the preset grid construction scale with each point as the center to ensure that the area around each representative point is initially sampled and covered; then, a diffusion operation is performed on each initial grid, that is, based on the current grid boundary, one layer of grid is expanded outward to generate M first-time diffusion division grids, and the grid density of the initial grid and the diffusion grid is compared. If the density after diffusion is not lower than the original density, it means that the diffusion operation has improved the coverage efficiency or maintained the information density. At this time, the first-time diffusion grid is used as the interim result.

[0070] Furthermore, iterative diffusion is performed with the same diffusion scale to form secondary and tertiary diffusion grids until the diffusion stop constraint is met (such as each grid covers at least 3 trees), and finally a stable and well-covered diffusion partition grid of M targets is formed.

[0071] Furthermore, combined with the existing tree point distribution map, an interactive overlap analysis is performed on the M target diffusion grids obtained to identify redundant areas and missing areas (areas not yet covered by any grid but still containing sample trees). Grids are then added to the missing points to ensure complete and balanced spatial coverage. Ultimately, Q sampling grids are obtained, which serve as the spatial basic units for subsequent data sampling, modeling, or monitoring.

[0072] For example, in a forest resource survey project, M=50 representative tree points were obtained as the initial sampling core. With each point as the center, an initial grid was generated according to the construction scale of 20 meters × 20 meters. Subsequently, a diffusion operation was performed, with each diffusion scale being 10 meters. After one diffusion, it was found that the grid density increased by an average of 12%, and the diffusion was continued until the density increase was less than 1% after the fourth diffusion, triggering the diffusion stop condition. 50 target diffusion grids were obtained. Furthermore, combined with the existing remote sensing tree distribution map, it was found that there were vacancies in some areas, and then 12 grids were automatically supplemented in these areas, and finally Q=62 sampling grids were generated, with a coverage rate of 98.5%.

[0073] Through the grid construction process based on "target point as the center - diffusion - overlap verification - filling in the gaps", samples in edge or sparse areas are automatically completed, thereby improving the representativeness of the sampling; the diffusion process enables each grid to cover more available samples, avoiding resource waste caused by low grid density and improving sampling efficiency; the diffusion and supplementation mechanism ensures automatic adaptation to diverse terrains or urban patterns; the final Q sampling grids can be seamlessly superimposed on the GIS map system, facilitating subsequent data collection and route planning.

[0074] In some implementations, the diffusion stop constraint is that the number of diffusions satisfies a preset maximum number of diffusions and / or the difference in mesh density between two adjacent diffusions is less than or equal to a preset difference threshold.

[0075] Specifically, during the grid diffusion process, to prevent the diffusion operation from expanding indefinitely, two types of stopping conditions are introduced as diffusion stopping constraints: Frequency control method (hard constraint): For example, if the maximum number of diffusions is set to three, if three diffusions have already been performed (initial → first → second → third), the diffusion is forced to terminate and no further diffusion is performed. Effect control method (soft constraint): In real time, the grid density after the previous and current diffusions is compared. If the density increase is very limited, below a set threshold (e.g., 0.05 trees / square meter), it indicates that the current grid has basically covered all key points and no further diffusion is required.

[0076] The two can be used independently or jointly, including "termination when any one of them is met" or "termination when both are met".

[0077] In some implementations, the M target diffusion grids are interactively overlapped and analyzed in combination with the tree point distribution map, and missing tree points are supplemented to obtain the Q sampling grids. The grid division module 12 further includes the following steps:

[0078] Based on the tree point distribution map and the regional positions of the M target diffusion partitioning grids, interactive overlapping areas are determined to obtain M grid interactive overlapping area sets; the tree points in the M grid interactive overlapping area sets are respectively retained in the target diffusion partitioning grids closest to them to obtain M cleaning grid interactive overlapping area sets; based on the M cleaning grid interactive overlapping area sets, the M target diffusion partitioning grids are updated to obtain M updated diffusion partitioning grids; the tree points in the tree point distribution map except the M updated diffusion partitioning grids are summarized to obtain a missing tree point set; the missing tree point set is subjected to nearest neighbor partitioning to obtain N missing partitioning grids, where N is a positive integer; the M updated diffusion partitioning grids and the N missing partitioning grids are summarized to obtain the Q sampling grids, where Q=M+N.

[0079] Specifically, the interactive overlapping area refers to the area where multiple target diffusion division grids overlap in space, which may cause tree points to be repeatedly covered by multiple grids; the grid interactive overlapping area set is a set composed of multiple grid overlapping parts; the updated diffusion division grid refers to the adjusted grid set after cleaning the interactive overlapping area.

[0080] Specifically, the missing tree point set refers to the set of tree points that are not covered in the diffusion partitioning grid. According to the location information of the missing tree points, they and their neighboring missing tree points are divided into a new grid, that is, neighbor partitioning, to ensure that all tree points are covered.

[0081] Specifically, in this implementation, after obtaining M target diffusion partitioning grids, combined with the tree point distribution map, we first determine whether there are interactive overlapping areas between these grids, that is, based on the spatial range of each grid, its intersection with other grids is superimposed and analyzed to obtain a set of M grid interactive overlapping areas; then, in order to avoid repeated sampling of tree points by multiple grids, the nearest distance retention strategy is implemented for the tree points in the interactive overlapping area, that is, each point is only retained in the target diffusion partitioning grid with the closest Euclidean distance to it, thereby forming a set of M cleaned interactive overlapping areas.

[0082] Specifically, based on the above cleaning results, the original M target diffusion partitioning grids are updated, that is, the redundant points superimposed on multiple grids are removed to obtain M updated diffusion partitioning grids, thereby ensuring that there is no overlapping redundancy between grids and improving data consistency and sampling efficiency.

[0083] Furthermore, the M updated diffusion grids are compared to the tree point distribution map, and differential tree points (i.e., points not covered by the updated grids) are screened to form a missing tree point set. A nearest neighbor partitioning algorithm (such as one based on K-means clustering, DBSCAN, or Voronoi partitioning) is then applied to the missing points to partition them into N spatial clusters. From these, N missing grids are constructed to complete the missing areas. Once the M updated diffusion grids and N missing grids are obtained, a summary operation is performed to merge these two types of grids to generate a complete set of sampling grids for subsequent data sampling, monitoring deployment, or model training.

[0084] Optionally, the aggregation operation includes not only the merging of spatial boundaries, but also the integration of grid attributes, such as the number of tree points in each grid, center coordinates, grid number, source type (diffusion grid or supplementary grid), etc., to ensure unified scheduling and analysis, and avoid data silos or duplicate processing.

[0085] This process ensures that all valid tree locations are covered by at least one sampling grid, achieving high-integrity sampling. Interactive overlap analysis avoids overlapping grids, improving sampling efficiency. Cleaning overlapping areas ensures unique attribution of tree locations, enhancing data cleaning quality. Nearest neighbor partitioning automatically identifies and completes missing areas, improving grid coverage integrity. Overall, this method based on interactive analysis and supplementary partitioning not only improves sampling comprehensiveness but also helps increase sampling efficiency, ensuring the accuracy and scientific nature of the sampling results.

[0086] In some implementations, the missing tree point set is divided according to a preset equal division strategy to obtain the N missing division grids.

[0087] Specifically, the equal division strategy is a method of regularizing the division of defective locations based on preset spatial scale parameters (such as distance thresholds), ensuring that the spatial range covered by each grid is roughly equal. It has the advantages of simple implementation, high computational efficiency, and uniform spatial distribution.

[0088] Specifically, after obtaining the set of missing tree points, clustering algorithms or complex spatial modeling methods are not used. Instead, based on a preset equal division strategy and a spatial distance threshold, the missing points are directly spatially gridded. For example, a distance threshold D (such as 50 meters) is set, and within the boundary of the missing point set, a regular grid structure (equilateral square grid or hexagonal grid, with the side length or diagonal length of each grid not exceeding D) is constructed according to this distance.

[0089] The sampling amount distribution module 13 is used to distribute the sampling amount according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts.

[0090] Specifically, tree point density refers to the number of trees per unit area and can be calculated by calculating the ratio of the number of trees in a grid to the grid area. Based on this density, the number of samples required for each sampling grid can be determined, ensuring that the sampling results accurately reflect the root distribution characteristics of the trees within the target area.

[0091] In some embodiments, the sampling amount is distributed according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts. The steps executed by the sampling amount distribution module 13 include:

[0092] The total number of trees in the Q sampling grids is counted respectively, and the statistical results are respectively divided by the area of ​​the Q sampling grids to obtain Q tree point distribution densities; the Q tree point distribution densities are respectively divided by the sum of the Q tree point distribution densities to obtain Q sampling volume allocation coefficients; a preset total sampling volume is obtained, and the Q sampling volume allocation coefficients are traversed and multiplied by the preset total sampling volume to obtain the Q sampling volumes.

[0093] Specifically, first, traverse the Q sampling grids and count the total number of tree points T in each grid. i ; Then, the total number of tree points in each grid T i Divide by the corresponding grid area A i , and obtain the distribution density D i =T i / A i ; Then, each distribution density D i Divide by the sum of all distribution densities to obtain the sampling distribution coefficient of the grid:

[0094] ;

[0095] Furthermore, each distribution coefficient α i Multiply it by the preset total sampling amount S to obtain the sampling amount S of each grid i Optional, according to actual needs, S i Round to an integer value and make a slight adjustment for rounding errors to ensure that the sum is still S.

[0096] For example, the following Table 1 shows the sampling amount distribution results after allocating the sampling amounts to 10 grids of a certain scene according to the above method steps.

[0097] Table 1 Example of sampling amount distribution table

[0098] Grid Number <![CDATA[Total amount of trees T i > <![CDATA[Area A i (m²)]]> <![CDATA[Density D i > <![CDATA[Partition coefficient α i > <![CDATA[Sampling amount S i > G1 120 600 0.2 0.08 80 G2 300 1000 0.3 0.12 120 ... ... ... ... ... ... G10 500 500 1 0.4 400

[0099] Through the above steps, dynamic sampling allocation based on actual point density is achieved, sampling representativeness is improved, oversampling in low-density areas and undersampling in high-density areas are avoided, and differentiated sampling needs of different types of areas (such as dense forests and sparse areas) are adapted.

[0100] The ground scanning module 14 is used to use a portable GPR device to perform ground scanning on Q sampling grids respectively, and perform root distribution inversion based on the scanning data to determine Q root dense areas, Q root sparse areas and Q staggered distribution areas.

[0101] Specifically, portable GPR (Ground Penetrating Radar) equipment can detect the distribution of underground objects by emitting and receiving electromagnetic waves. By analyzing the scanning data obtained by the GPR equipment, including radar wave reflection intensity, texture characteristics and depth distribution, combined with training models or expert rule bases, the distribution characteristics of underground roots (i.e., root distribution inversion) can be reconstructed, including the depth, density and distribution range of the roots.

[0102] Among them, the root-dense area refers to the area where the roots are densely distributed (the radar reflection signal is continuous and strong, and the texture is dense), the root-sparse area refers to the area where the roots are sparsely distributed (the reflection signal is weak or intermittent), and the staggered distribution area refers to the area where the roots are staggered with soil or other substances (the signal strength alternates and the texture changes are complex).

[0103] Using portable GPR equipment for ground scanning and root distribution inversion enables precise detection of root distribution, improving the scientific nature and efficiency of sampling. GPR equipment can non-destructively obtain underground root distribution information, eliminating the tedious excavation or drilling required by traditional methods. Furthermore, root distribution inversion can provide detailed insights into root distribution characteristics, providing precise guidance for subsequent sampling point identification and ensuring that sampling points cover different root distribution areas.

[0104] The sampling point identification module 15 is used to identify sampling points of the Q sampling grids based on the Q sampling amounts, the Q root-dense areas, the Q root-sparse areas, and the Q staggered distribution areas, determine Q sampling point sets, and extract root samples at the Q sampling point sets.

[0105] Specifically, based on the obtained Q root-dense areas, Q root-sparse areas, and Q staggered distribution areas, the sampling points can be further refined and identified and located by integrating the sampling volume and distribution area type, thereby realizing the on-demand distributed sampling point layout.

[0106] In some embodiments, as Figure 2As shown, to determine Q sampling point sets, the sampling point identification module 15 performs the following steps:

[0107] Obtain the proportions of the Q dense root areas, the Q sparse root areas, and the Q staggered distribution areas in the Q sampling grids, respectively, to obtain Q dense proportions, Q sparse proportions, and Q staggered distribution proportions; multiply the Q dense proportions, Q sparse proportions, and Q staggered distribution proportions by the Q sampling amounts to determine Q dense sampling amounts, Q sparse sampling amounts, and Q interactive distribution sampling amounts; randomly sample the Q dense root areas, Q sparse root areas, and Q staggered distribution areas based on the Q dense sampling amounts, Q sparse sampling amounts, and Q interactive distribution sampling amounts to determine Q sampling point sets.

[0108] Specifically, for the preset Q sampling grids, the sampling point identification module 15 first obtains the area ratios of the root-dense area, the root-sparse area, and the staggered distribution area in each grid, which are recorded as the dense ratio, the sparse ratio, and the staggered distribution ratio, respectively, thereby obtaining Q dense ratios, Q sparse ratios, and Q staggered distribution ratios. Then, the ratio of each area is added to the total sampling amount S of the grid. i Multiply them together to get the sampling amount of the three types of areas.

[0109] Optionally, the last sample is calculated by difference, ensuring that the sum is S i , to avoid quantitative deviations due to rounding errors.

[0110] Furthermore, random points are selected within each region according to the corresponding sampling quantity. Exemplary point selection strategies can include spatially uniform randomization, stratified randomization, or other representative sampling methods. The selected sampling points in the three regions are then merged to form a set of sampling points corresponding to each sampling grid. Ultimately, Q sets of sampling points are output, each corresponding to a sampling grid, for subsequent root sample extraction.

[0111] Through the above method, the reasonable distribution of sampling points in different root distribution areas is ensured, avoiding excessive concentration in a certain type of area; and the representativeness and diversity of the sampling points are improved.

[0112] In some embodiments, the tree root sampling device further comprises:

[0113] The failure sample acquisition unit is used to obtain a set of sampling failure sample logs; the failure point update unit is used to traverse and obtain the sampling environment information of the Q sampling point sets, and perform feature comparison with the sampling failure sample log set. If the feature comparison result meets the preset comparison threshold, the corresponding sampling point is marked as a failure sampling point, and the sampling point is randomly selected again.

[0114] Specifically, a failed sample log collection refers to failed sample logs recorded during past or current sampling processes, including the reasons for failure (e.g., soil and rock obstructions, shallow roots, sampling instrument interference) and corresponding environmental characteristics (e.g., soil conductivity, pH, GPS location, hardness level, etc.). Sampling environment information refers to the local geographic, soil, or vegetation environment of the current sampling point and is a key reference parameter affecting the feasibility of root sampling.

[0115] Specifically, first, the failure sample acquisition unit is called to retrieve the environmental feature logs of historical failure samples; then, the failure point update phase is entered, and environmental information is extracted from the Q selected sampling points one by one, and the current environmental information is matched with the failure sample log set. The similarity between the features (i.e., the matching degree) is calculated using methods such as Euclidean distance, cosine similarity, and KL divergence.

[0116] Specifically, if the similarity is higher than the set threshold, the point can be considered a "potential failure point" and marked; further, for the points marked as failed, the sampling points are replaced within the remaining optional range through a random reselection mechanism to ensure that the Q sampling points are as effective as possible.

[0117] In summary, the tree root sampling device provided by the present invention has the following technical effects:

[0118] The distribution identification module identifies the point distribution of trees in the target area and generates a tree point distribution map. Subsequently, the grid division module divides the sampling grid according to the tree point coordinates in the distribution map and determines Q sampling grids, where Q is a positive integer. The sampling quantity distribution module allocates the sampling quantity according to the distribution density of the tree points in each sampling grid and obtains Q corresponding sampling quantities. The ground scanning module uses a portable GPR device to perform ground scanning on these Q sampling grids respectively, and inverts the root distribution based on the scanning data to identify Q root-dense areas, Q root-sparse areas and Q staggered distribution areas. The sampling point identification module identifies the sampling points of the Q sampling grids respectively based on each sampling quantity and different root distribution types, determines a set of Q sampling points, and extracts root samples at these locations, thereby achieving the technical effect of improving the scientific nature of the sampling point distribution and sampling quantity allocation, and improving the sampling accuracy and efficiency.

[0119] It should be understood that the embodiments disclosed in the present invention and the above description can enable those skilled in the art to use the present invention to implement the present invention. At the same time, the present invention is not limited to the embodiments mentioned above. It should be understood that those skilled in the art can still modify the technical solutions described in the above embodiments or replace some of the technical features therein with equivalents; and such modifications or replacements do not deviate from the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention and are all included in the scope of protection of the present invention.

Claims

1. A tree root sampling device, characterized in that: The device includes: The distribution recognition module is used to identify the point distribution of trees in the target area and determine the tree point distribution map; A grid division module is used to divide the sampling grids according to the tree point coordinates in the tree point distribution map, and determine Q sampling grids, where Q is a positive integer; A sampling amount distribution module is used to distribute the sampling amount according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts; A ground scanning module is used to perform ground scanning on each of the Q sampling grids using a portable GPR device, and perform root distribution inversion based on the scanned data to determine Q root-dense areas, Q root-sparse areas, and Q staggered distribution areas; a sampling point identification module for identifying sampling points of the Q sampling grids based on the Q sampling amounts, the Q root-dense areas, the Q root-sparse areas, and the Q staggered distribution areas, determining Q sampling point sets, and extracting root samples at the Q sampling point sets; The sampling grid is divided according to the tree point coordinates in the tree point distribution map to determine Q sampling grids. The execution steps of the grid division module include: Screening a preset number of randomly sampled tree locations according to the first sampling scale and the tree location distribution map to determine M first randomly sampled tree locations and M second randomly sampled tree locations, where M is a positive integer and is less than or equal to Q; Based on the M first randomly sampled tree locations and the M second randomly sampled tree locations, pairwise random sampling tree location combinations are formed, and it is determined whether the distance interval between the two locations in each combination is greater than or equal to a preset distance interval threshold. If so, the sampling dispersion coefficients of the M first randomly sampled tree locations and the M second randomly sampled tree locations are respectively calculated to obtain a first sampling dispersion coefficient and a second sampling dispersion coefficient, wherein the sampling dispersion coefficient is calculated by constructing a Voronoi diagram and quantifying the dispersion according to the standard deviation of the area of ​​the polygons generated by the distribution of each point; When the first sampling dispersion coefficient is greater than the second sampling dispersion coefficient, the M first randomly sampled tree locations are used as the M target randomly sampled tree locations; To obtain the Q sampling grids, the grid division module further comprises the following steps: Taking the M target randomly sampled tree points as sampling grid centers, construct M initial division grids according to a preset grid construction scale; Performing a diffusion operation on the M initial divided grids according to a preset diffusion scale to obtain M primary diffused divided grids; When the grid density of the M initial division grids is less than or equal to the grid density of the M first-diffusion division grids, the M first-diffusion division grids are used as M stage diffusion division grids, and secondary diffusion is continued according to the preset diffusion scale until the diffusion stop constraint is satisfied, thereby obtaining M target diffusion division grids; The M target diffusion division grids are interactively overlapped and analyzed in combination with the tree point distribution map, and the missing tree points are supplemented with grids to obtain the Q sampling grids.

2. A tree root sampling device according to claim 1, characterized in that: The diffusion stop constraint is that the number of diffusions meets a preset maximum number of diffusions and / or the difference in grid density between two adjacent diffusions is less than or equal to a preset difference threshold.

3. A tree root sampling device according to claim 1, characterized in that: Interactive overlapping analysis is performed on the M target diffusion grids in combination with the tree point distribution map, and grids are supplemented for missing tree points to obtain the Q sampling grids. The execution steps of the grid division module further include: Determine the interactive overlapping area based on the tree point distribution map and the regional positions of the M target diffusion division grids, and obtain a set of M grid interactive overlapping areas; Retaining the tree points in the M grid interactive overlapping area sets to the target diffusion partitioning grid closest to them, and obtaining M cleaning grid interactive overlapping area sets; updating the M target diffusion partitioning grids based on the interactive overlapping area sets of the M cleaning grids to obtain M updated diffusion partitioning grids; Summarize the tree points in the tree point distribution map except for the M updated diffusion division grids to obtain a missing tree point set; Performing nearest neighbor partitioning on the missing tree point set to obtain N missing partitioning grids, where N is a positive integer; The M updated diffusion partitioning grids and the N missing partitioning grids are aggregated to obtain the Q sampling grids, where Q=M+N.

4. A tree root sampling device according to claim 3, characterized in that: include: The missing tree point set is divided according to a preset equal division strategy to obtain the N missing division grids.

5. A tree root sampling device according to claim 1, characterized in that: Determine a set of Q sampling points. The steps of executing the sampling point identification module include: Obtaining the proportions of the Q dense root areas, the Q sparse root areas, and the Q staggered distribution areas in the Q sampling grids, respectively, to obtain Q dense ratios, Q sparse ratios, and Q staggered distribution ratios; Multiplying the Q dense ratios, Q sparse ratios, and Q staggered distribution ratios by the Q sampling amounts to determine Q dense sampling amounts, Q sparse sampling amounts, and Q staggered distribution sampling amounts; Based on the Q dense sampling amounts, the Q sparse sampling amounts and the Q staggered distribution sampling amounts, the Q root dense areas, the Q root sparse areas and the Q staggered distribution areas are randomly sampled to determine Q sampling point sets.

6. A tree root sampling device according to claim 1, characterized in that: Also includes: A failure sample acquisition unit, used to acquire a sampling failure sample log set; The failure point updating unit is used to traverse and obtain the sampling environment information of the set of Q sampling points, perform feature comparison on the sampling environment information with the set of sampling failure sample logs, and if the feature comparison result meets the preset comparison threshold, mark the corresponding sampling point as a failed sampling point and randomly select a new sampling point.

7. A tree root sampling device according to claim 1, characterized in that: The sampling amount is distributed according to the tree point distribution density in the Q sampling grids to obtain Q sampling amounts. The execution steps of the sampling amount distribution module include: Counting the total number of trees in the Q sampling grids respectively, and dividing the statistical results by the area of ​​the Q sampling grids respectively to obtain the Q tree point distribution densities; Dividing the Q tree point distribution densities respectively by the sum of the Q tree point distribution densities to obtain Q sampling amount distribution coefficients; A preset total sampling amount is obtained, and the Q sampling amount distribution coefficients are traversed and multiplied by the preset total sampling amount to obtain the Q sampling amounts.

Images