Urban forest community carbon reserve assessment method and system based on multi-source data

Abstract

The invention is suitable for the technical field of forest assessment methods, and provides an urban forest community carbon reserve assessment method and system based on multi-source data, and the method comprises the steps: obtaining research region information; the method comprises the following steps of: obtaining a geometrical relationship between an orientation in a panoramic image and a point cloud center direction after equipment installation, generating individual tree data units according to a relationship model, and binding the individual tree data units with unique numbers; extracting contour features of image areas corresponding to the individual tree data units, identifying whether the contour features have contradictions, performing secondary stripping on abnormal contours to obtain final individual tree data units, and performing species classification according to the final individual tree data units; a structure weighting factor is constructed by combining the area of the sample plot, a corresponding grade is judged according to the total carbon storage value of the whole sample plot, a theoretical carbon storage upper limit is calculated, the relative deviation rate of the current carbon storage is calculated according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot, and the method has the advantages that structure expression is refined, model matching is flexible, and the distinguishable grade of an evaluation result can be improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of forest assessment methods, and particularly relates to a method and system for assessing the carbon storage of urban forest communities based on multi-source data. Background Art

[0002] In the context of the increasing attention paid to urban ecosystems, urban forests, as important green infrastructures, play a core role in aspects such as carbon sequestration functions, microclimate regulation, and ecological buffering. In order to scientifically evaluate their carbon storage capacity, the precise calculation of the carbon storage of forest communities has become the technical core of urban forestry planning and ecological management.

[0003] In the prior art, forest carbon storage assessment mainly relies on lidar point clouds to obtain vegetation structure data, and combines a unified biomass estimation model to perform carbon storage conversion on the trees in the sample plots. However, such methods have multiple limitations: on the one hand, a single data source is prone to misjudgment or missed judgment in areas with dense forest cover and intertwined structures, resulting in incomplete extraction of single-wood structure parameters; on the other hand, fixed general models ignore the differential effects of species, forest types, development stages, etc. on carbon storage composition, and cannot achieve a fine expression of "selecting a model according to the tree". In addition, the current assessment methods generally lack the participatory analysis of spatial structures, cannot reflect the distribution contribution of different structural areas to the overall carbon storage, and are also difficult to provide effective support for the improvement strategies of carbon sequestration potential.

[0004] At the same time, most of the existing solutions only stay at the level of "quantity calculation", lacking the grading and feedback ability of the assessment results, and unable to form a "evaluation-diagnosis-intervention" closed-loop mechanism for actual governance and optimization. Therefore, there is an urgent need to construct a new method for assessing urban forest carbon storage that integrates multi-source data collection, image-assisted recognition, dynamic model adaptation, structural zone integration, and grading judgment feedback, so as to improve the accuracy, adaptability, and management value of the assessment, and provide a decision-making basis for enhancing the carbon sequestration capacity of urban ecosystems. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a method for assessing the carbon storage of urban forest communities based on multi-source data, aiming to solve the problems proposed in the third part of the background art.

[0006] The embodiments of the present invention are implemented as follows. A method for assessing the carbon storage of urban forest communities based on multi-source data, the method includes: Obtain research area information, where the research area information includes the research area, research time, and research requirements, and the research time includes the data collection time and the duration of sample plot setting; Install the equipment according to the installation requirements. After the equipment is installed, obtain the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center. Generate single-tree data units according to the relationship model, and bind a unique number to each single-tree data unit. Extract the contour features of the image area corresponding to the single-tree data unit, identify whether there are contradictions in the contour features, perform secondary peeling on abnormal contours, obtain the final single-tree data unit, and perform species classification according to the final single-tree data unit. Construct a structure weighting factor in combination with the area of the sample plot, determine the corresponding grade according to the total carbon storage value of the entire sample plot, calculate the theoretical upper limit of carbon storage, and calculate the relative deviation rate of the current carbon storage according to the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot.

[0007] Preferably, the steps of installing the equipment according to the installation requirements, obtaining the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the equipment is installed, generating single-tree data units according to the relationship model, and binding a unique number to each single-tree data unit specifically include: Obtain equipment information, obtain installation requirements, and install the equipment according to the installation requirements. The installation requirements include installation height and calibration board. After the equipment is installed, obtain the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center. The point cloud is obtained by lidar scanning. Obtain the image recognition confidence and the acquisition integrity threshold, and evaluate the image recognition confidence and acquisition integrity of the panoramic image according to the threshold. Construct a projection envelope relationship model between the image and the point cloud according to the qualified geometric relationship. Generate single-tree data units according to the relationship model, and bind a unique number to each single-tree data unit. The single-tree data unit includes diameter at breast height, tree height, and crown width.

[0008] Preferably, the steps of extracting the contour features of the image area corresponding to the single-tree data unit, identifying whether there are contradictions in the contour features, performing secondary peeling on abnormal contours, obtaining the final single-tree data unit, and performing species classification according to the final single-tree data unit specifically include: Extract the contour features of the image area corresponding to the single-tree data unit. The contour features include crown shape, texture symmetry, and edge continuity. Identify whether there are contradictions in the contour features. The contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal. Perform secondary peeling on the abnormal contour. The secondary peeling is to construct an auxiliary cutting plane in the point cloud guided by the main trunk direction or the crown edge in the image, and re-divide the clustering block. After the division is completed, recalculate the diameter at breast height, tree height, and crown width of each sub-block to obtain the final single-tree data unit, and perform species classification according to the final single-tree data unit. Identify specific tree species through species classification.

[0009] Preferably, the steps of constructing a structural weighting factor based on the area of the combined sample plot, determining the corresponding grade according to the total carbon storage value of the entire sample plot, calculating the theoretical upper limit of carbon storage, and calculating the relative deviation rate of the current carbon storage based on the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot specifically include: Obtain the sample plot area, divide the sample plot into different hierarchical structural areas according to the final individual tree data units, where the structural areas include structurally homogeneous areas, structurally transitional areas, and structurally heterogeneous areas, and calculate the average individual tree carbon storage and the total number of trees within the structural area; Construct a structural weighting factor in combination with the area of the sample plot, integrate the total carbon storage value of the entire sample plot based on the structural weighting factor, obtain the carbon storage grade judgment rules for the sample plot, and determine the corresponding grade according to the total carbon storage value of the entire sample plot; Calculate the theoretical upper limit of carbon storage, calculate the relative deviation rate of the current carbon storage based on the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot, obtain the deviation rate threshold, and if the relative deviation rate is greater than the deviation rate threshold, obtain a list of improvement suggestions and send the list to the terminal.

[0010] Preferably, the device information includes lidar and panoramic cameras.

[0011] Another object of the embodiments of the present invention is to provide a multi-source data-based urban forest community carbon storage assessment system, where the system includes: A research preparation module that obtains research area information, where the research area information includes the research area, research time, and research requirements, and the research time includes the data collection time and the duration of sample plot setting; An individual tree data module that installs devices according to the installation requirements, obtains the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after device installation, generates individual tree data units according to the relationship model, and binds unique numbers to the individual tree data units; A secondary stripping module that extracts the contour features of the image area corresponding to the individual tree data units, identifies whether there are contradictions in the contour features, performs secondary stripping on the abnormal contours to obtain the final individual tree data units, and classifies the species according to the final individual tree data units; A grade determination module that constructs a structural weighting factor in combination with the area of the sample plot, determines the corresponding grade according to the total carbon storage value of the entire sample plot, calculates the theoretical upper limit of carbon storage, and calculates the relative deviation rate of the current carbon storage based on the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot.

[0012] Preferably, the individual tree data module includes: A device installation unit that obtains device information, obtains installation requirements, and installs devices according to the installation requirements, where the installation requirements include the installation height and calibration board; An information collection unit obtains the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the device is installed. The point cloud is obtained by lidar scanning, acquires the image recognition confidence and the acquisition integrity threshold, and evaluates the image recognition confidence and acquisition integrity of the panoramic image according to the threshold. A single-tree binding unit constructs a projection envelope relationship model between the image and the point cloud according to the qualified geometric relationship, generates a single-tree data unit according to the relationship model, and binds a unique number to the single-tree data unit. The single-tree data unit includes diameter at breast height, tree height, and crown width.

[0013] Preferably, the secondary stripping module includes: A contour feature unit extracts the contour features of the image area corresponding to the single-tree data unit. The contour features include crown shape, texture symmetry, and edge continuity, and identifies whether there are contradictions in the contour features. The contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal. A secondary stripping unit performs secondary stripping on the abnormal contour. The secondary stripping is to construct an auxiliary cutting plane in the point cloud guided by the main trunk direction or the crown edge in the image, and re-divide the clustering block. A species classification unit recalculates the diameter at breast height, tree height, and crown width of each sub-block after the division is completed to obtain the final single-tree data unit, and performs species classification according to the final single-tree data unit, and identifies specific tree species through species classification.

[0014] Preferably, the grade determination module includes: A structural area division unit obtains the sample area, divides the sample plot into different grade structural areas according to the final single-tree data unit. The structural areas include a structurally homogeneous area, a structurally transitional area, and a structurally heterogeneous area, and calculates the average single-tree carbon storage and the total number of trees in the structural area. A grade determination unit constructs a structural weighting factor in combination with the area of the sample plot, integrates the total carbon storage value of the entire sample plot according to the structural weighting factor, obtains the sample plot carbon storage grade evaluation rule, and determines the corresponding grade according to the total carbon storage value of the entire sample plot. A deviation rate unit calculates the theoretical carbon storage upper limit, calculates the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot, obtains the deviation rate threshold. If the relative deviation rate is greater than the deviation rate threshold, a list of improvement suggestions is obtained and sent to the terminal.

[0015] Preferably, the device information includes a lidar and a panoramic camera.

[0016] A method for evaluating the carbon storage of urban forest communities based on multi-source data provided by an embodiment of the present invention. In the data acquisition stage, by using the coaxial installation structure of a rotary handheld lidar and a 360° panoramic camera, combined with a QR code calibration board and differential GPS, real-time synchronous registration of point cloud and image data in space is achieved, and the projection envelope relationship between the image and the point cloud is constructed, providing an accurate spatial reference for subsequent individual tree recognition and species classification. In the individual tree segmentation stage, an image-assisted peeling mechanism is integrated. Based on the geometric consistency judgment between the image contour features and the point cloud clustering results, secondary segmentation is performed on the abnormally clustered parts, especially suitable for structurally complex areas such as shadow misalignment and trunk intersection in dense forest environments, greatly improving the accuracy and integrity of individual tree recognition.

[0017] In the carbon storage calculation stage, based on the species label, diameter at breast height data, and forest type classification information of individual trees, the most suitable biomass equation is dynamically matched, and the species or genus-level model is preferentially called. In the plot integration stage, a plot zoning strategy dominated by structure is proposed. The plot is divided into homogeneous areas, transition areas, and heterogeneous areas according to diameter at breast height diversity, mingling degree, and angle scale, and a structure weighting factor is constructed. The carbon storage is weighted and integrated according to the area ratio of each region to ensure that the carbon storage assessment can reflect the actual contribution of structural heterogeneity to ecological functions. Calculate the relative deviation rate of the current carbon storage, and introduce a management threshold for grading. When the deviation is significant, a list of improvement suggestions is automatically generated and pushed to the terminal, realizing a closed-loop management path from recognition to intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of a method for evaluating the carbon storage of urban forest communities based on multi-source data provided by an embodiment of the present invention; Figure 2 It is a flowchart of the steps for generating individual tree data units according to a relational model and binding unique numbers to the individual tree data units provided by an embodiment of the present invention; Figure 3 It is a flowchart of the steps for extracting the contour features of the image area corresponding to the individual tree data unit, performing secondary peeling on abnormal contours, and classifying species according to the final individual tree data unit provided by an embodiment of the present invention; Figure 4 It is a flowchart of the steps for determining the corresponding grade according to the total carbon storage value of the whole plot and calculating the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the plot provided by an embodiment of the present invention; Figure 5 It is an architecture diagram of a system for evaluating the carbon storage of urban forest communities based on multi-source data provided by an embodiment of the present invention; Figure 6 It is an architecture diagram of the individual tree data module provided by an embodiment of the present invention; Figure 7An architectural diagram of a secondary stripping module provided in an embodiment of the present invention; Figure 8 This is an architecture diagram of a level determination module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] It is understood that the terms "first", "second", etc. used in this application may be used herein to describe various elements, but unless otherwise specified, these elements are not limited by these terms. These terms are only used to distinguish a first element from another element. For example, without departing from the scope of this application, a first xx script may be referred to as a second xx script, and similarly, a second xx script may be referred to as a first xx script.

[0021] like Figure 1 As shown, a method for evaluating carbon storage of urban forest communities based on multi-source data is provided in an embodiment of the present invention, and the method includes: S100, obtaining research area information, wherein the research area information includes the research area, research time and research requirements, wherein the research time includes data collection time and sample plot setting duration.

[0022] In this step, the study area information is obtained to ensure the temporal and spatial consistency and ecological adaptability of subsequent data collection, sample plot layout and model selection. The study area information includes not only the geographical scope and spatial dimension, but also the data collection time window and sample plot setting cycle, as well as the time dimension, research focus, and functional dimension, thus providing complete background parameters for carbon storage modeling.

[0023] First, the administrative boundaries and ecological zoning of the research area need to be clear. For example, it can be set as "an urban forest area dominated by park-type green spaces within a main urban area". Geographically, it is necessary to obtain basic GIS data of the region, including vegetation coverage maps, DEM elevation maps, road and building vector boundaries, etc., for subsequent sample plot planning and point cloud registration. At the same time, the research time needs to be refined into two categories: data collection time and sample plot setting duration. The former refers to the specific date of field sampling or sensor deployment, such as the completion of all sample point cloud and image data collection from July to September 2025; the latter refers to the time period during which the sample plot remains unchanged under the same structural standard, ensuring that time series comparison, retesting and updating or long-term carbon storage evolution modeling can be carried out in the subsequent stages. For example, the sample plot setting maintenance time is 3 years, and the sample plot number and location will not be changed during this period.

[0024] In addition, the research requirements need to clarify the data emphasis of the collected area. For example, whether to emphasize structurally heterogeneous forest land, whether to focus on artificially regenerated forests or marginal disturbance areas with greater carbon storage growth potential, etc. Take a practical example: If the research requirement is to "prioritize the identification of the carbon storage improvement potential of fragmented forest land at the urban edge", the research area selection will tend to ecological transition zones such as boundary parks and residential green corridors, and the coverage ratio of structurally heterogeneous areas will be increased in the plot design.

[0025] S200, Install the equipment according to the installation requirements. Obtain the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the equipment is installed. Generate single-tree data units according to the relationship model, and bind unique numbers to the single-tree data units.

[0026] In this step, install the equipment according to the installation requirements. Build an integrated support platform at the center of the plot, and install the handheld lidar and the 360° panoramic camera coaxially, so that the collected data has a unified direction reference in physical space. By deploying high-contrast QR code calibration plates at the four corners of the plot and cooperating with high-precision coordinate acquisition, a clear mapping relationship between image pixels and point cloud coordinates is established, and then an image-point cloud projection envelope model is constructed to provide stable geometric support for subsequent image-assisted analysis.

[0027] Based on this registration model, the system jointly constructs the point cloud structure parameters and the image recognition results into "single-tree data units", and embeds unique numbers to achieve data traceability across stages and modules. Each single-tree data unit contains spatial position, geometric structure, species characteristics and image contour information, realizing the pairing of structural data and semantic information in the acquisition link. This mechanism not only significantly improves the data integrity and consistency, but also lays a standardized and traceable data foundation for subsequent single-tree segmentation, biomass matching and spatial partition integration, with strong system integration and engineering feasibility.

[0028] S300, Extract the contour features of the image area corresponding to the single-tree data unit, identify whether there are contradictions in the contour features, perform secondary peeling on abnormal contours, obtain the final single-tree data unit, and classify species according to the final single-tree data unit.

[0029] In this step, extract the contour features of the image area corresponding to the single-tree data unit. For each clustering block in the point cloud data, back-project it to the corresponding 360° panoramic image, and extract the crown contour of the area it occupies in the image space. This process uses a geometric algorithm based on spatial projection transformation and combines the known reference coordinates provided by the calibration plate to ensure that the contour extraction results have a stable geometric correspondence.

[0030] Subsequently, the CNN contour recognition model is called to extract features from each image contour region, obtain its edge connectivity, shape regularity, and boundary sharpness, and perform a consistency comparison with the crown space morphology of the point cloud clustering block. If it is found that there are multiple opposite crowns, occlusion and overlap, or structural breaks in the image contour, while they are classified as the same single tree in the point cloud, it is determined as abnormal clustering. At this time, a secondary peeling operation is triggered to re-divide the point cloud block based on the main direction projection rule until the image and the point cloud crown feature are highly consistent.

[0031] After completing the correction of the final single-tree data unit structure, the species classification operation is performed. This process combines the image feature vector and the trained model, and is supplemented by label library matching.

[0032] S400, construct a structure weighting factor in combination with the area of the sample plot, determine the corresponding level according to the total carbon storage value of the entire sample plot, calculate the theoretical carbon storage upper limit, and calculate the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot.

[0033] In this step, a structure weighting factor is constructed in combination with the area of the sample plot. The sample plot is divided into several structural areas, such as a structurally homogeneous area, a structurally transitional area, and a structurally heterogeneous area, and the area proportion A of each area is calculated i Combined with the structural complexity index of this area, the structural weight w can be comprehensively weighted and set by the diameter at breast height diversity Hd, mingling degree M, and angle scale W i Calculate a weighting factor that can reflect the structural complexity, which is used to correct the total carbon storage value, so that the evaluation result has stronger structural adaptability and ecological responsiveness.

[0034] After obtaining the weighted total carbon storage value, perform level division according to the mean and standard deviation of the regional statistical samples to form a carbon storage level evaluation system that is horizontally comparable and vertically traceable. At the same time, construct a theoretical carbon storage upper limit model, and estimate the maximum carbon sink potential of the sample plot under the current structure based on the optimal carbon storage density per unit area and the structure weighting factor. By comparing the actual carbon storage with the theoretical upper limit, calculate the relative deviation rate, and then identify the adaptation gap between the sample plot structure and the carbon storage.

[0035] As Figure 2 shown, as a preferred embodiment of the present invention, the step of installing the device according to the installation requirements, obtaining the geometric relationship between the orientation in the panoramic image and the point cloud center direction after the device is installed, and generating a single-tree data unit according to the relationship model and binding a unique number to the single-tree data unit specifically includes: S201, obtain device information, obtain installation requirements, and install the device according to the installation requirements, where the installation requirements include the installation height and the calibration board.

[0036] In this step, device information is obtained, including the core parameter information of the selected acquisition devices, such as the lidar model (e.g., LiGrip V100), the 360° panoramic camera model (e.g., Insta360 Pro), and the functional specifications, interface standards, and fixing methods of the differential GPS device (e.g., SR1 Pro). The device information should include the scanning frequency, field of view angle, and point cloud density of the lidar, the image resolution, synchronization mechanism, etc. of the panoramic camera to ensure that subsequent installation operations meet the technical compatibility requirements. At the same time, it is also necessary to obtain the power supply method and data synchronization interface standard of the device to provide a hardware foundation for integrated installation.

[0037] Before installation, clear installation requirements also need to be obtained. The installation height of the device should be uniformly set to 1.3 meters. The bracket structure should have a height locking and anti-offset function and cooperate with a level to complete installation correction to ensure that the device scanning axis is perpendicular to the ground and the pitch angle is 0°. High-contrast QR code calibration plates should be arranged at the four corners of the sample plot. The calibration plates should be made of high-strength anti-reflective materials and arranged at a height of about 1.2 - 1.5 meters.

[0038] Taking a 0.1-hectare urban sample plot as an example, a platform bracket is set up at the center of the sample plot, and the coaxial installation of LiGrip V100 and Insta360 Pro is completed. Calibration plates numbered A1 to A4 are arranged at the four corners of the sample plot respectively, and the GPS acquisition error does not exceed ±5 cm. After installation, each device enters the test run. After confirming that the image and point cloud outputs are synchronized and cover completely, the formal acquisition process can be entered.

[0039] S202, through the geometric relationship between the azimuth in the panoramic image obtained after device installation and the direction of the point cloud center, where the point cloud is obtained by lidar scanning, obtain the image recognition confidence and acquisition integrity threshold, and evaluate the image recognition confidence and acquisition integrity of the panoramic image according to the threshold.

[0040] In this step, through the geometric relationship between the azimuth in the panoramic image obtained after device installation and the direction of the point cloud center, after the coaxial installation of the lidar and the 360° panoramic camera is completed, a geometric correspondence relationship between the direction of each pixel in the panoramic image and the direction of the lidar point cloud center is established through the spatial calibration process. This correspondence relationship is based on the spatial mapping of the calibration plate coordinates and the image pixel points, and a projection transformation matrix between the image coordinate system and the point cloud coordinate system is constructed.

[0041] Introduce a dual evaluation mechanism for image recognition confidence and image acquisition integrity. The image recognition confidence refers to the classification probability output given by the species recognition result of each single-tree image region based on the deep learning model, denoted as P class. The acquisition integrity refers to whether the pixel coverage of the crown structure in the image space is complete, and whether there are situations such as occlusion, overexposure, underexposure, etc. that affect the extraction of structural information. It is usually comprehensively evaluated by edge connectivity, image entropy, and brightness histogram equalization, and a threshold T is set. c and T i are used to judge whether the recognition is credible and whether the image is complete respectively.

[0042] For example, in the image area of a single tree, if the confidence level given by the species recognition model is 0.92, and the image edge of its contour area is clear and the pixel distribution is complete, it can be considered that both the recognition confidence level and integrity meet the standards. On the contrary, if the confidence level is lower than 0.7, or the image edge is broken due to light, and the number of edge pixels is less than the preset threshold, then the image is regarded as an unreliable input and can be marked as "needs to be recollected", and subsequent weight reduction or reconstruction processing will be given in carbon storage calculation and species matching.

[0043] S203. According to the qualified geometric relationship evaluated, construct a projection envelope relationship model between the image and the point cloud, generate a single-tree data unit according to the relationship model, and bind a unique number to the single-tree data unit. The single-tree data unit includes diameter at breast height, tree height, and crown width.

[0044] In this step, according to the qualified geometric relationship evaluated, construct a projection envelope relationship model between the image and the point cloud. This model is used to describe the corresponding envelope area between the crown contour of a single tree in the image space and its three-dimensional structure in the point cloud space. The construction process depends on the geometric calibration matrix collected coaxially, projects the three-dimensional space information in the point cloud into the image space, and at the same time projects the crown contour extracted in the image back to the point cloud space in the reverse direction to form a spatially consistent single-tree contour closed area.

[0045] On this basis, use the segmented point cloud clustering blocks to perform projection overlap analysis with the image crown contour. For the image-point cloud matching pairs with an overlap degree greater than the set threshold (such as 85%), confirm them as single-tree instances, and then generate single-tree data units. The content of this data unit includes three core structural parameters: the diameter at breast height (D) is extracted from the horizontal section of the point cloud, the tree height (H) is obtained from the vertical range of the point cloud, and the crown width (W) is calculated jointly by the image contour and the horizontal projection of the point cloud to ensure spatial integrity and recognition accuracy.

[0046] To achieve the traceability of the data unit and the consistency of subsequent calculations, each single-tree data unit is automatically assigned a unique number. The numbering rule adopts a three-level structure: area number + plot number + tree number. For example, "XZ001-03-T12" represents the 12th tree in plot No. 3 in Xicheng District. All numbers and corresponding structural parameters are automatically written into the structured database, supporting direct indexing and retrieval in subsequent steps such as biomass model calling, species classification, and spatial structure division.

[0047] As Figure 3 shown, as a preferred embodiment of the present invention, the steps of extracting the contour features of the image region corresponding to the single-tree data unit, identifying whether there are contradictions in the contour features, performing secondary peeling on abnormal contours to obtain the final single-tree data unit, and performing species classification according to the final single-tree data unit specifically include: S301, extract the contour features of the image region corresponding to the single-tree data unit. The contour features include crown shape, texture symmetry, and edge continuity. Identify whether there are contradictions in the contour features. The contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal.

[0048] In this step, extract the contour features of the image region corresponding to the single-tree data unit. The contour features mainly include three aspects: crown shape, texture symmetry, and edge continuity. These features are extracted by a deep convolutional neural network in cooperation with an edge detection operator and represented in vector form.

[0049] During the identification process, the contour features extracted from the image are compared with the point cloud clustering results for spatial consistency verification. If it is found that a certain clustering block has a complete shape in the point cloud, but its corresponding image region shows obvious fragmentation, occlusion, or boundary expansion, or there is an obvious single-tree contour in the image while the corresponding structure is not detected in the point cloud, it is determined as a contradictory state where the image display is inconsistent with the point cloud clustering. Such contradictions indicate that there may be tree body merging in the current point cloud clustering, such as misidentifying two trees as one, or tree body omission, such as failure to identify due to occlusion.

[0050] For example, in a dense forest belt, a certain point cloud clustering block is identified as a single large tree in the LiDAR data, but the image region shows two clear crown boundaries and low texture continuity in the middle. This contradiction indicates that two adjacent trees may have been mismerged into one. At this time, the system automatically marks this clustering as "abnormal" and enters the subsequent image-assisted peeling process for secondary segmentation to ensure the independence and accuracy of single-tree identification.

[0051] S302, perform secondary peeling on the abnormal contour. The secondary peeling is to construct an auxiliary cutting plane in the point cloud guided by the trunk direction or the crown edge in the image and re-divide the clustering block.

[0052] In this step, for the secondary peeling of the abnormal contour, first analyze the structural features of the abnormal clustering region in the image, and preferentially extract its trunk direction, that is, the longitudinal direction from the crown top to the base and the contour trend of the crown edge.

[0053] Next, the extracted main image direction information is back-projected into the point cloud space, and an auxiliary cutting plane perpendicular to the main trunk direction is constructed in the point cloud. This cutting plane is generally based on the dividing line between the two crown sub-contours in the image or the smallest closed area of the edge opening angle to determine its spatial position and the direction of the normal vector. The auxiliary cutting operation realizes the re-division of the original clustering by intercepting the point set in the point cloud clustering block that is farther than the threshold from the cutting plane, so as to separate the two mis-merged tree bodies into independent units.

[0054] For example, in the image, it is found that an abnormal area shows a symmetric arrangement of double crowns, and the included angle between the crown edge contours is about 60°. After back-projection, two local density centers are located in the point cloud, and the cutting plane is established with the perpendicular bisector of the line connecting these two centers as the normal vector. Intercepting the point set around this profile line in the point cloud can complete the image-driven re-segmentation of the clustering block, enabling the two actually independent tree bodies to form new single-tree data units respectively, and re-binding the numbers and structural parameters.

[0055] S303. After the division is completed, the diameter at breast height, tree height, and crown width of each sub-block are recalculated to obtain the final single-tree data units. According to the final single-tree data units, species classification is carried out, and specific tree species are identified through species classification.

[0056] In this step, after the division is completed, the structural parameters of each re-divided sub-block are independently calculated to generate the final effective single-tree data units. A horizontal section is extracted at a height of 1.3 meters above the ground, and the point cloud slice density contour is used to fit a circular or elliptical cross-section to calculate the diameter. The Z-axis range is calculated between the highest point on the ground and the highest point of the crown. The sub-block is projected onto the horizontal plane, and the maximum envelope boundary of the point cloud in the X-Y plane is extracted to calculate the average diameters of the major axis and the minor axis; After the parameter calculation is completed, combined with the contour, texture, and color characteristics of the corresponding area of the sub-block in the image, further species classification and recognition are carried out. This recognition is based on a pre-trained convolutional neural network model, which generates feature vectors using the texture directionality, color distribution, and leaf morphological characteristics of the image, and compares the similarity with the standard species image vector library in the database. The output category is the species name, such as camphor tree, ginkgo tree, hackberry tree, etc., and at the same time, a recognition confidence score is given to judge the reliability of the recognition result.

[0057] For example, for the two sub-blocks obtained by one peeling, the diameter at breast height is calculated to be 24.3 cm and 17.5 cm respectively, the tree height is 12.7 m and 9.6 m respectively, and the crown widths are 5.1 m and 3.8 m respectively. Through image recognition, the confidence of the first sub-block is 0.94, matching the camphor tree, and the confidence of the second sub-block is 0.89, matching the hackberry tree. Both exceed the threshold and are confirmed as two valid single trees, and finally two independent data units are formed, written into the database and available for subsequent carbon storage model calls.

[0058] As Figure 4 shown, as a preferred embodiment of the present invention, the steps of constructing a structural weighting factor based on the area of the combined sample plot, determining the corresponding grade according to the total carbon storage value of the entire sample plot, calculating the theoretical upper limit of carbon storage, and calculating the relative deviation rate of the current carbon storage based on the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot specifically include: S401, obtain the sample plot area, divide the sample plot into different grade structure areas according to the final single-tree data unit, the structure areas include a structurally homogeneous area, a structurally transitional area, and a structurally heterogeneous area, and calculate the average single-tree carbon storage and the total number of trees in the structure area.

[0059] In this step, obtain the sample plot area, and divide the sample plot into spatial structure areas based on the distribution of single-tree structure characteristics. First, obtain the spatial boundary range of the current sample plot, and determine its area, shape, and geographical coordinate distribution. Subsequently, based on the structure parameters (diameter at breast height D, tree height H, crown width W) and spatial distribution coordinates in the single-tree data unit, scan the structure characteristics of the entire sample plot in a local statistical manner (such as a 3×3m or 5×5m sliding window); Divide the sample plot into three types of structure areas: Structurally homogeneous area: characterized by low H0, low M, and low W, with the tree species and sizes being highly uniform, commonly found in plantations or neat green belts; Structurally transitional area: having medium H0 and M, commonly found in partially updated or small-scale mixed areas; Structurally heterogeneous area: characterized by high H0, high M, and high W, with complex structures and mixed tree species, which are typical characteristics of natural secondary forests or areas with many years of natural growth; For each type of structure area, respectively count the total carbon storage and the total number of trees in all single-tree data units within it, and calculate the average single-tree carbon storage. The carbon storage calculation is based on the individual carbon storage value generated by the dynamic biomass matching mechanism (such as B j =a(D 2 H) b ×CF, where CF is the carbon factor).

[0060] The above H0 is the degree of diversity; M is the mingling degree, which is the degree of interlaced distribution of different tree species in a certain area; B j is the carbon storage of the jth tree; a and b are the parameters of the dynamic biomass model, corresponding to the regression parameters of a certain species, genus-level, or general carbon storage model, and need to be called from the regional model library; For example, in a sample plot of an urban green space, the structurally heterogeneous area contains 35 trees, with a total carbon storage of 6.3t and an average single-tree carbon storage of 0.18t; there are 46 trees in the structurally homogeneous area, with a total carbon storage of 5.2t and an average of 0.11t; the structurally transitional area is between the two. The above data provides a basis for subsequent regional weight construction and overall verification and evaluation.

[0061] S402. Construct a structure weighting factor in combination with the area of the sample plot, integrate the total carbon storage value of the entire sample plot according to the structure weighting factor, obtain the carbon storage level evaluation rules for the sample plot, and determine the corresponding level based on the total carbon storage value of the entire sample plot.

[0062] In this step, construct a structure weighting factor in combination with the area of the sample plot, and calculate the actual occupied area A of each type of structural area (such as structurally homogeneous area, structurally transitional area, structurally heterogeneous area) in the sample plot i and the total carbon storage C within this area i . According to the proportion w i = A i / A total of the total area of the entire sample plot, construct the structure weighting factor w i , and use this to weight and integrate the carbon storage values of each area: ; where C total is the weighted total carbon storage value of the sample plot, reflecting the differential contributions of different structural areas to the carbon storage capacity. After integrating the carbon storage values, it is necessary to introduce the carbon storage level evaluation rules. These rules are based on the historical sample plot statistical data of the same type of areas (such as urban green spaces, ecological isolation belts, secondary forests, etc.), and set the theoretical carbon storage upper limit C max , median C mid and bottom line C min . Compare the C total of the current sample plot with these reference values to divide the carbon storage level. For example: Grade 1 (excellent): C total ≥ 0.8 × C max ; Grade 2 (good): 0.6 × C max ≤ C total < 0.8 × C max ; Grade 3 (medium): 0.4 × C max ≤ C total < 0.6 × C max ; Grade 4 (poor): C total < 0.4 × C max ; For example, the area of a certain urban sample plot is 1 hectare, among which the structurally heterogeneous area accounts for 40%, the structurally transitional area accounts for 35%, and the structurally homogeneous area accounts for 25%. The total carbon storage of the three is 8.2t, 5.4t, and 3.1t respectively. Through weighted calculation, the weighted total carbon storage of the sample plot is 6.18t. If the historical carbon storage upper limit of this area is 8.5t, then 6.18t is approximately 73% of it, and it is judged as "good" level.

[0063] S403. Calculate the theoretical carbon storage upper limit, calculate the relative deviation rate of the current carbon storage based on the theoretical carbon storage upper limit and the total carbon storage value of the sample plot, obtain the deviation rate threshold. If the relative deviation rate is greater than the deviation rate threshold, obtain the list of improvement suggestions and send the list to the terminal.

[0064] In this step, calculate the theoretical carbon storage upper limit. According to the structural area division result, introduce the theoretical optimal single-tree carbon storage amount for each type of area, and then multiply it by the corresponding total number of trees to obtain the theoretical carbon storage upper limit C. max For example, the theoretical single-tree carbon storage in the structurally heterogeneous area is 0.28t, and the actual number of trees is 40, then the upper limit of this area is 11.2t.

[0065] Conduct a deviation analysis on the carbon storage performance of the current sample plot and calculate its relative deviation rate: ; This deviation rate reflects the distance between the carbon storage of the sample plot and its potential. To distinguish whether intervention is needed, introduce a deviation rate threshold (such as 25%) as the management trigger criterion. If the deviation rate R 偏离 >R 阈值 , the system determines that there is room for optimization in this sample plot.

[0066] On this basis, retrieve reference sample plots from the database that are structurally similar to the current sample plot but have excellent carbon storage performance, analyze their characteristics such as species composition, density, mixing ratio, and age structure, and automatically generate a list of improvement suggestions, such as increasing the mixing ratio, introducing dominant species with high carbon storage, such as Celtis sinensis and Pistacia chinensis, and optimizing the plant spacing. After being formatted, this list is automatically pushed to the forestry work terminal for managers to conduct feasibility assessment and restoration decision-making.

[0067] For example, the theoretical upper limit of a certain sample plot is 9.4t, the actual is 6.3t, the deviation rate is 32.9%, which is higher than the set threshold of 25%. The system determines it as "medium-low adaptation" and pushes the suggestion: replant 30 Cinnamomum camphora trees in the structural transition area to improve the mixing index, with an expected carbon storage gain of 0.9t.

[0068] As Figure 5 shown, a carbon storage assessment system for urban forest communities based on multi-source data provided by an embodiment of the present invention, the system includes: A research preparation module 100, used to obtain research area information, the research area information includes the research area, research time, and research requirements, and the research time includes the data collection time and the duration of sample plot setting.

[0069] In this system, the research preparation module 100 obtains the research area information to ensure the temporal and spatial unity and ecological adaptability of subsequent data collection, sample plot layout and model selection. The research area information includes not only the geographical scope and spatial dimension, but also the data collection time window and sample plot setting cycle, as well as the time dimension, research focus, and functional dimension, thereby providing complete background parameters for carbon storage modeling.

[0070] First, the administrative boundaries and ecological zoning of the research area need to be clear. For example, it can be set as "an urban forest area dominated by park-type green spaces within a main urban area". Geographically, it is necessary to obtain basic GIS data of the region, including vegetation coverage maps, DEM elevation maps, road and building vector boundaries, etc., for subsequent sample plot planning and point cloud registration. At the same time, the research time needs to be refined into two categories: data collection time and sample plot setting duration. The former refers to the specific date of field sampling or sensor deployment, such as the completion of all sample point cloud and image data collection from July to September 2025; the latter refers to the time period during which the sample plot remains unchanged under the same structural standard, ensuring that time series comparison, retesting and updating or long-term carbon storage evolution modeling can be carried out in the subsequent stages. For example, the sample plot setting maintenance time is 3 years, and the sample plot number and location will not be changed during this period.

[0071] In addition, the research requirements need to clarify the focus of the data collected in the area, such as whether to emphasize structurally heterogeneous forests, whether to focus on artificially regenerated forests or marginal disturbance areas with greater carbon storage growth potential, etc. To illustrate with a practical example: if the research requirement is to "prioritize the carbon storage enhancement potential of fragmented forests on the edge of the city", the research area selection will be biased towards ecological transition zones such as border parks and residential green corridors, and the coverage ratio of structurally heterogeneous areas will be enhanced in the sample site design.

[0072] The single tree data module 200 is used to install the equipment according to the installation requirements, obtain the geometric relationship between the position in the panoramic image and the center direction of the point cloud after the equipment is installed, generate a single tree data unit according to the relationship model, and bind the single tree data unit to a unique number.

[0073] In this system, the single tree data module 200 installs the equipment according to the installation requirements, builds an integrated bracket platform in the center of the sample plot, and installs the handheld laser radar and the 360° panoramic camera in a coaxial manner, so that the collected data has a unified direction reference in the physical space. By arranging high-contrast QR code calibration plates at the four corners of the sample plot and coordinating high-precision coordinate collection, a clear mapping relationship between image pixels and point cloud coordinates is established, and then an image-point cloud projection envelope model is constructed to provide stable geometric support for subsequent image-assisted analysis.

[0074] Based on this registration model, the system constructs the point cloud structure parameters and the image recognition results together into a "single-tree data unit", and embeds a unique number to achieve cross-stage and cross-module data traceability. Each single-tree data unit contains spatial position, geometric structure, species characteristics and image contour information, realizing the pairing of structural data and semantic information at the acquisition stage. This mechanism not only significantly improves data integrity and consistency, but also lays a standardized and traceable data foundation for subsequent single-tree segmentation, biomass matching and spatial partition integration, with strong system integration and engineering feasibility.

[0075] The secondary stripping module 300 is used to extract the contour features of the image region corresponding to the single-tree data unit, identify whether there are contradictions in the contour features, perform secondary stripping on abnormal contours to obtain the final single-tree data unit, and classify species according to the final single-tree data unit.

[0076] In this system, the secondary stripping module 300 extracts the contour features of the image region corresponding to the single-tree data unit. According to each clustering block in the point cloud data, it back-projects to the corresponding 360° panoramic image and extracts the crown contour of the area it occupies in the image space. This process uses a geometric algorithm based on spatial projection transformation and combines the known reference coordinates provided by the calibration board to ensure that the contour extraction result has a stable geometric correspondence.

[0077] Subsequently, the CNN contour recognition model is called to extract features from each image contour region, obtain its edge connectivity, shape regularity, and boundary sharpness, and compare them with the crown spatial morphology of the point cloud clustering block for consistency. If it is found that there are multiple opposite crowns, occlusion and overlap, or structural fractures in the image contour, while they are classified as the same single tree in the point cloud, it is determined as abnormal clustering. At this time, the secondary stripping operation is triggered, and the point cloud block is re-divided based on the main direction projection rule until the crown features of the image and the point cloud are highly consistent.

[0078] After completing the structure correction of the final single-tree data unit, the species classification operation is performed. This process combines the image feature vector and the trained model, and is supplemented by label library matching.

[0079] The grade determination module 400 is used to construct a structure weighting factor in combination with the area of the sample plot, determine the corresponding grade according to the total carbon storage value of the entire sample plot, calculate the theoretical carbon storage upper limit, and calculate the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot.

[0080] In this system, the grade determination module 400 constructs a structure weighting factor in combination with the area of the sample plot, divides the sample plot into several structural areas, such as structurally homogeneous areas, structurally transitional areas, and structurally heterogeneous areas, and calculates the area proportion A of each area respectively. i, combined with the structural complexity index of this area, the structural weight w can be comprehensively and weightedly set by the diameter at breast height diversity Hd, mingling degree M, and angle index W. i , a weighted factor that can reflect the structural complexity is calculated and used to correct the total carbon storage value, so that the evaluation result has stronger structural adaptability and ecological responsiveness.

[0081] After obtaining the weighted total carbon storage value, it is classified according to the mean and standard deviation of the regional statistical samples to form a set of carbon storage level evaluation systems that are horizontally comparable and vertically traceable. At the same time, a theoretical carbon storage upper limit model is constructed. Based on the optimal carbon storage density per unit area and the structural weighting factor, the maximum carbon sink potential of the sample plot under the current structure is estimated. By comparing the actual carbon storage with the theoretical upper limit, the relative deviation rate is calculated, and then the adaptation gap between the sample plot structure and carbon storage is identified.

[0082] As Figure 6 shown, as a preferred embodiment of the present invention, the individual tree data module 200 includes: The device installation unit 201 is used to obtain device information, obtain installation requirements, and install the device according to the installation requirements. The installation requirements include installation height and calibration board.

[0083] In this module, the device installation unit 201 obtains device information and obtains the core parameter information of the selected acquisition device, including the functional specifications, interface standards, and fixing methods of the lidar model (such as LiGrip V100), 360° panoramic camera model (such as Insta360 Pro), and differential GPS device (such as SR1 Pro). The device information should include the scanning frequency, field of view angle, and point cloud density of the lidar, the image resolution, synchronization mechanism, etc. of the panoramic camera to ensure that subsequent installation operations meet the technical compatibility requirements. At the same time, it is also necessary to obtain the power supply method and data synchronization interface standard of the device to provide a hardware basis for integrated installation.

[0084] Before installation, clear installation requirements also need to be obtained. The installation height of the device should be uniformly set to 1.3 meters. The bracket structure should have a height locking and anti-offset function and cooperate with a level to complete the installation correction to ensure that the device scanning axis is perpendicular to the ground and the pitch angle is 0°. High-contrast QR code calibration boards are arranged at the four corners of the sample plot. The calibration boards should be made of high-strength anti-reflective materials, and the installation height is about 1.2 - 1.5 meters.

[0085] Taking a 0.1-hectare urban sample plot as an example, a platform-type bracket is set in the center of the sample plot, and the coaxial installation of LiGrip V100 and Insta360 Pro is completed. Calibration boards numbered A1 to A4 are arranged at the four corners of the sample plot, respectively, and the GPS acquisition error does not exceed ±5 cm. After the installation is completed, each device enters the test run. After confirming that the image and point cloud output are synchronized and cover completely, it can enter the formal acquisition process.

[0086] An information acquisition unit 202 is configured to obtain the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the device is installed. The point cloud is obtained by lidar scanning, and the image recognition confidence and the acquisition integrity threshold are acquired. The panoramic image is evaluated for image recognition confidence and acquisition integrity according to the threshold.

[0087] In this module, the information acquisition unit 202 obtains the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the device is installed. After the coaxial installation of the lidar and the 360° panoramic camera is completed, the geometric correspondence between each pixel direction in the panoramic image and the direction of the lidar point cloud center is established through a spatial calibration process. This correspondence is based on the spatial mapping between the calibration board coordinates and the image pixel points, and a projection transformation matrix between the image coordinate system and the point cloud coordinate system is constructed.

[0088] A dual evaluation mechanism of image recognition confidence and image acquisition integrity is introduced. The image recognition confidence refers to the classification probability output given by the species recognition result of each single-tree image region based on the deep learning model, denoted as P class . The acquisition integrity refers to whether the pixel coverage of the crown structure in the image space is complete, and whether there are situations such as occlusion, overexposure, and underexposure that affect the extraction of structural information. It is usually comprehensively evaluated by edge connectivity, image entropy, and brightness histogram equalization, and thresholds T c and T i are respectively used to determine whether the recognition is credible and whether the image is complete.

[0089] For example, in the image region of a single tree, if the confidence given by the species recognition model is 0.92, and the contour region image edge is clear and the pixel distribution is complete, it can be considered that both the recognition confidence and the integrity meet the standards. On the contrary, if the confidence is lower than 0.7, or the image edge is broken due to light, and the number of edge pixels is less than the preset threshold, the image is regarded as an unreliable input and can be marked as "to be re-acquired", and subsequent weight reduction or reconstruction processing is given in carbon storage calculation and species matching.

[0090] A single-tree binding unit 203 is configured to construct a projection envelope relationship model between the image and the point cloud according to the qualified geometric relationship, generate a single-tree data unit according to the relationship model, and bind a unique number to the single-tree data unit. The single-tree data unit includes diameter at breast height, tree height, and crown width.

[0091] In this module, the single-tree binding unit 203 constructs a projection envelope relationship model between the image and the point cloud based on the evaluated qualified geometric relationship. This model is used to describe the corresponding envelope area between the crown profile of a single tree in the image space and its three-dimensional structure in the point cloud space. The construction process depends on the geometric calibration matrix collected coaxially, projects the three-dimensional space information in the point cloud into the image space, and at the same time maps the crown contour extracted in the image back to the point cloud space in reverse to form a spatially consistent closed area of the single-tree contour.

[0092] On this basis, the segmented point cloud clustering blocks are used to perform projection overlap analysis with the image crown profile. For the image-point cloud matching pairs with an overlap degree greater than the set threshold (such as 85%), they are confirmed as single-tree instances, and then single-tree data units are generated. The content of this data unit includes three types of core structure parameters: the diameter at breast height (D) is extracted from the horizontal section of the point cloud, the tree height (H) is obtained from the vertical range of the point cloud, and the crown width (W) is calculated jointly by the image contour and the horizontal projection of the point cloud to ensure spatial integrity and recognition accuracy.

[0093] To achieve the traceability of the data unit and the consistency of subsequent calculations, each single-tree data unit is automatically assigned a unique number. The numbering rule adopts a three-level structure: area number + plot number + tree number. For example, "XZ001-03-T12" represents the 12th tree in plot No. 3 in Xicheng District. All numbers and corresponding structure parameters are automatically written into the structured database, supporting direct indexing and retrieval in subsequent steps such as biomass model invocation, species classification, and spatial structure division.

[0094] As Figure 7 shown, as a preferred embodiment of the present invention, the secondary peeling module 300 includes: A contour feature unit 301 for extracting the contour features of the image area corresponding to the single-tree data unit. The contour features include crown width morphology, texture symmetry, and edge continuity, and identify whether there are contradictions in the contour features. The contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal.

[0095] In this module, the contour feature unit 301 extracts the contour features of the image area corresponding to the single-tree data unit. The contour features mainly include three aspects: crown width morphology, texture symmetry, and edge continuity. These features are extracted by a deep convolutional neural network in cooperation with an edge detection operator and are represented in vector form.

[0096] During the recognition process, the contour features extracted from the image are compared with the results of point cloud clustering for spatial consistency verification. If it is found that a certain clustering block has a complete form in the point cloud, but its corresponding image area shows obvious fragmentation, occlusion, or boundary expansion, or there are obvious single-tree contours in the image while the corresponding structure is not detected in the point cloud, it is determined as a contradictory state where the image display is inconsistent with the point cloud clustering. Such contradictions indicate that there may be tree body mergers in the current point cloud clustering, such as misidentifying two trees as one, or tree body omissions, such as failure to recognize due to occlusion.

[0097] For example, in a dense forest belt, a certain point cloud clustering block is recognized as a single large tree in the LiDAR data, but the image area shows two clear crown boundary areas with low texture continuity in the middle. This contradiction indicates that two adjacent trees may have been mismerged into one. At this time, the system automatically marks this clustering as "abnormal" and enters the subsequent image-assisted stripping process for secondary segmentation to ensure the independence and accuracy of single-tree recognition.

[0098] The secondary stripping unit 302 is used to perform secondary stripping on the abnormal contour. The secondary stripping is to construct an auxiliary cutting plane in the point cloud guided by the trunk direction or the crown edge in the image, and re-divide the clustering block.

[0099] In this module, the secondary stripping unit 302 performs secondary stripping on the abnormal contour. First, it analyzes the structural characteristics of the abnormal clustering area in the image and preferentially extracts its trunk direction, that is, the longitudinal direction from the crown top to the base and the contour trend of the crown edge.

[0100] Next, the extracted main direction information of the image is back-projected into the point cloud space, and an auxiliary cutting plane perpendicular to the trunk direction is constructed in the point cloud. This cutting plane is generally based on the dividing line between the two sub-contours of the crown in the image or the smallest closed area of the edge opening angle to determine its spatial position and the direction of the normal vector. The auxiliary cutting operation realizes the re-division of the original clustering by intercepting the point set in the point cloud clustering block that is farther than the threshold distance from the cutting plane, thereby separating the two mismerged tree bodies into independent units.

[0101] For example, in the image, an abnormal area is found to show a symmetric arrangement of double crowns, and the included angle between the crown edge contours is about 60°. After back-projection, two local density centers are located in the point cloud, and the cutting plane is established with the perpendicular bisector of the line connecting these two centers as the normal vector. Intercepting the point set around this profile line in the point cloud can complete the image-driven re-segmentation of the clustering block, enabling the two actually independent tree bodies to form new single-tree data units respectively, and re-binding the numbers and structural parameters.

[0102] The species classification unit 303 is used to recalculate the diameter at breast height, tree height, and crown width of each sub-block after division to obtain the final individual tree data unit. Species classification is performed based on the final individual tree data unit, and specific tree species are identified through species classification.

[0103] In this module, after the species classification unit 303 is divided, independent calculations of structural parameters are performed on each newly divided sub-block to generate the final valid individual tree data unit. A horizontal section is extracted at a height of 1.3 meters above the ground, and the point cloud slice density profile is used to fit a circular or elliptical cross-section to calculate the diameter. The Z-axis range is calculated between the highest point on the ground and the highest point of the crown. The sub-block is projected onto the horizontal plane, and the maximum envelope boundary of the point cloud on the X-Y plane is extracted to calculate the average diameter of the major axis and the minor axis. After the parameter calculation is completed, combined with the contour, texture, and color characteristics of the corresponding area of the sub-block in the image, further species classification and identification are carried out. This identification is based on a pre-trained convolutional neural network model, which generates feature vectors using the texture directionality, color distribution, and leaf morphological characteristics of the image, and compares the similarity with the standard species image vector library in the database. The output category is the species name, such as Camphor tree, Ginkgo biloba, Celtis sinensis, etc., and at the same time, an identification confidence score is given to judge the reliability of the identification result.

[0104] For example, for the two sub-blocks obtained by one peeling, the diameter at breast height is calculated to be 24.3 cm and 17.5 cm respectively, the tree height is 12.7 m and 9.6 m respectively, and the crown width is 5.1 m and 3.8 m respectively. Through image recognition, the confidence level of the first sub-block is 0.94, matching Camphor tree, and the confidence level of the second sub-block is 0.89, matching Celtis sinensis. Both exceed the threshold and are confirmed as two valid individual trees, finally forming two independent data units, which are written into the database and called by the subsequent carbon storage model.

[0105] As Figure 8 shown, as a preferred embodiment of the present invention, the grade determination module 400 includes: The structural area division unit 401 is used to obtain the sample area, divide the sample plot into different grade structural areas according to the final individual tree data unit. The structural areas include a structurally homogeneous area, a structurally transitional area, and a structurally heterogeneous area, and calculate the average individual tree carbon storage and the total number of trees in the structural area.

[0106] In this module, the structural area division unit 401 obtains the sample area and divides the spatial structure area of the sample plot based on the distribution of the single tree structure characteristics. First, obtain the spatial boundary range of the current sample plot, determine its area, shape, and geographical coordinate distribution. Subsequently, based on the structural parameters (diameter at breast height D, tree height H, crown width W) and spatial distribution coordinates in the individual tree data unit, scan the structural characteristics of the entire sample plot in a local statistical manner (such as a 3×3m or 5×5m sliding window); The sample plots are divided into three types of structural regions: structurally homogeneous regions: characterized by low H0, low M, and low W, with tree species and sizes being highly uniform, commonly found in plantations or neat green belts; structurally transitional regions: having medium H0 and M, commonly found in partially regenerated or small-scale mixed regions; structurally heterogeneous regions: characterized by high H0, high M, and high W, with complex structures and mixed tree species, which are typical features of natural secondary forests or areas with many years of natural growth; For each type of structural region, the total carbon storage and the total number of individual trees within it are respectively counted, and the average carbon storage per individual tree is calculated. The carbon storage calculation is based on the individual carbon storage values generated by the dynamic biomass matching mechanism (such as B j =a(D 2 H) b ×CF, where CF is the carbon factor).

[0107] For example, in a sample plot of an urban green space, the structurally heterogeneous region contains 35 trees, with a total carbon storage of 6.3 t and an average carbon storage per individual tree of 0.18 t; there are 46 trees in the structurally homogeneous region, with a total carbon storage of 5.2 t and an average of 0.11 t; the structurally transitional region is between the two. The above data provides a basis for subsequent construction of regional weights and overall effectiveness evaluation.

[0108] The grade determination unit 402 is used to construct a structural weighting factor in combination with the area of the sample plot, integrate the total carbon storage value of the entire sample plot according to the structural weighting factor, obtain the carbon storage grade evaluation rules for the sample plot, and determine the corresponding grade according to the total carbon storage value of the entire sample plot.

[0109] In this module, the grade determination unit 402 constructs a structural weighting factor in combination with the area of the sample plot, and calculates the actual occupied area A i and the total carbon storage C i within each type of structural region (such as structurally homogeneous region, structurally transitional region, structurally heterogeneous region) in the sample plot. Based on the proportion w i =A i / A total of the area in the total area of the entire sample plot, a structural weighting factor w i is constructed, and the carbon storage values of each region are weighted and integrated with this: ; Among them, Ctotal is the weighted total carbon storage value of the sample plot, reflecting the differential contributions of different structural regions to the carbon storage capacity. After integrating the carbon storage values, it is necessary to introduce the carbon storage grade evaluation rules. These rules are based on the historical sample plot statistical data of the same type of regions (such as urban green spaces, ecological isolation belts, secondary forests, etc.), and set the theoretical carbon storage upper limit C max , median C mid and bottom line C min . The C total of the current sample plot is compared with these reference values to divide the carbon storage grade. For example: Level 1 (excellent): C total ≥0.8×C max ; Level 2 (good): 0.6×C max ≤C total <0.8×C max ; Level 3 (medium): 0.4×C max ≤C total <0.6×C max ; Level 4 (poor): C total <0.4×C max ; For example, the area of a sample plot in a certain city is 1 hectare, among which the structurally heterogeneous area accounts for 40%, the structurally transitional area accounts for 35%, and the structurally homogeneous area accounts for 25%. The total carbon storage of the three is 8.2t, 5.4t, and 3.1t respectively. Through weighted calculation, the weighted total carbon storage of the sample plot is 6.18t. If the historical carbon storage upper limit of this area is 8.5t, then 6.18t is about 73% of it, and it is judged as "good" level.

[0110] The deviation rate unit 403 is used to calculate the theoretical carbon storage upper limit, calculate the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot, obtain the deviation rate threshold. If the relative deviation rate is greater than the deviation rate threshold, then obtain the list of improvement suggestions and send the list to the terminal.

[0111] In this module, the deviation rate unit 403 calculates the theoretical carbon storage upper limit. According to the structural area division result, introduce the theoretical optimal single-tree carbon storage amount for each type of area respectively, and then multiply it by the corresponding total number of plants to obtain the theoretical carbon storage upper limit C max . For example, the theoretical single-tree carbon storage in the structurally heterogeneous area is 0.28t, and the actual number of plants is 40, then the upper limit of this area is 11.2t.

[0112] Conduct deviation analysis on the carbon storage performance of the current sample plot and calculate its relative deviation rate: ; This deviation rate reflects the distance between the carbon storage of the sample plot and its potential. In order to distinguish whether intervention is needed, a deviation rate threshold (such as 25%) is introduced as the management trigger standard. If the deviation rate R 偏离 >R 阈值 , then the system determines that there is room for optimization in this sample plot.

[0113] On this basis, reference plots with similar structures to the current plot but excellent carbon storage performance are retrieved from the database, and their characteristics such as species composition, density, mixed planting ratio, and age structure are analyzed to automatically generate a list of improvement suggestions, such as increasing the mixed planting ratio, introducing dominant species with high carbon storage, such as Celtis sinensis and Pistacia chinensis, and optimizing plant spacing. After being formatted, this list is automatically pushed to the forestry work terminal for managers to conduct feasibility assessments and restoration decisions.

[0114] For example, the theoretical upper limit of a certain plot is 9.4t, the actual value is 6.3t, and the deviation rate is 32.9%, which is higher than the set threshold of 25%. The system determines it as "medium-low adaptation" and pushes the suggestion: replant 30 Cinnamomum camphora trees in the structural transition area to increase the mixing index, with an expected carbon storage gain of 0.9t.

[0115] In one embodiment, a computer device is proposed. The computer device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the following steps are implemented: Obtain research area information, where the research area information includes the research area, research time, and research requirements. The research time includes the data collection time and the duration of plot setting; Install the device according to the installation requirements. Through the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center obtained after the device installation, generate single-tree data units according to the relationship model, and bind unique numbers to the single-tree data units; Extract the contour features of the image area corresponding to the single-tree data units, identify whether there are contradictions in the contour features, perform secondary peeling on abnormal contours to obtain the final single-tree data units, and classify the species according to the final single-tree data units; Construct a structure weighting factor in combination with the area of the plot, determine the corresponding grade according to the total carbon storage value of the entire plot, calculate the theoretical upper limit of carbon storage, and calculate the relative deviation rate of the current carbon storage according to the theoretical upper limit of carbon storage and the total carbon storage value of the plot.

[0116] In one embodiment, a computer-readable storage medium is provided. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, the processor is caused to execute the following steps: Obtain research area information, where the research area information includes the research area, research time, and research requirements. The research time includes the data collection time and the duration of plot setting; Install the device according to the installation requirements. Through the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center obtained after the device installation, generate single-tree data units according to the relationship model, and bind unique numbers to the single-tree data units; Extract the contour features of the image region corresponding to the individual tree data unit, identify whether there are contradictions in the contour features, perform secondary peeling on the abnormal contours to obtain the final individual tree data unit, and classify the species according to the final individual tree data unit; Construct a structure weighting factor in combination with the area of the plot, determine the corresponding grade according to the total carbon storage value of the entire plot, calculate the theoretical upper limit of carbon storage, and calculate the relative deviation rate of the current carbon storage according to the theoretical upper limit of carbon storage and the total carbon storage value of the plot.

[0117] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the arrows, these steps do not necessarily have to be executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily have to be executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages does not necessarily have to be sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps.

[0118] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0119] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of concise description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as within the scope described in this specification.

[0120] The above-described embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

[0121] The foregoing is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the carbon storage of urban forest communities based on multi-source data, characterized in that, The method includes: Obtaining research area information, where the research area information includes the research area, research time, and research requirements. The research time includes the data collection time and the duration of plot setting; Installing the equipment according to the installation requirements, obtaining the geometric relationship between the orientation in the panoramic image and the point cloud center direction after the equipment is installed, generating single-tree data units according to the relationship model, and binding unique numbers to the single-tree data units; Extracting the contour features of the image area corresponding to the single-tree data units, identifying whether there are contradictions in the contour features, performing secondary peeling on abnormal contours to obtain the final single-tree data units, and classifying species according to the final single-tree data units; Specifically, it includes extracting the contour features of the image area corresponding to the single-tree data units. The contour features include crown shape, texture symmetry, and edge continuity. Identifying whether there are contradictions in the contour features. The contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal. Performing secondary peeling on the abnormal contour. The secondary peeling is to construct an auxiliary cutting plane in the point cloud guided by the trunk direction or crown edge in the image, and re-divide the clustering block. After the division is completed, recalculate the diameter at breast height, tree height, and crown width of each sub-block to obtain the final single-tree data units, classify species according to the final single-tree data units, and identify specific tree species through species classification; Constructing a structure weighting factor in combination with the area of the plot, determining the corresponding grade according to the total carbon storage value of the entire plot, calculating the theoretical carbon storage upper limit, and calculating the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the plot.

2. The method for evaluating the carbon storage of urban forest communities based on multi-source data according to claim 1, wherein The step of installing the equipment according to the installation requirements, obtaining the geometric relationship between the orientation in the panoramic image and the point cloud center direction after the equipment is installed, generating single-tree data units according to the relationship model, and binding unique numbers to the single-tree data units specifically includes: Obtaining equipment information, obtaining installation requirements, and installing the equipment according to the installation requirements. The installation requirements include the installation height and the calibration board; Obtaining the geometric relationship between the orientation in the panoramic image and the point cloud center direction after the equipment is installed. The point cloud is obtained by lidar scanning, obtaining the image recognition confidence and the acquisition integrity threshold, and evaluating the image recognition confidence and acquisition integrity of the panoramic image according to the threshold; Constructing a projection envelope relationship model between the image and the point cloud according to the qualified geometric relationship, generating single-tree data units according to the relationship model, and binding unique numbers to the single-tree data units. The single-tree data units include the diameter at breast height, tree height, and crown width.

3. The method for evaluating the carbon storage of urban forest communities based on multi-source data according to claim 1, wherein The step of constructing a structure weighting factor in combination with the area of the plot, determining the corresponding grade according to the total carbon storage value of the entire plot, calculating the theoretical carbon storage upper limit, and calculating the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the plot specifically includes: Obtaining the plot area, dividing the plot into different grade structure areas according to the final single-tree data units. The structure areas include structure homogeneous areas, structure transition areas, and structure heterogeneous areas, and calculating the average single-tree carbon storage and the total number of trees in the structure area; Construct a structure weighting factor in combination with the area of the sample plot, integrate the total carbon storage value of the entire sample plot according to the structure weighting factor, obtain the evaluation rules for the carbon storage level of the sample plot, and determine the corresponding level according to the total carbon storage value of the entire sample plot; Calculate the theoretical upper limit of carbon storage, calculate the relative deviation rate of the current carbon storage according to the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot, obtain the deviation rate threshold, if the relative deviation rate is greater than the deviation rate threshold, then obtain a list of improvement suggestions and send the list to the terminal.

4. A method for evaluating the carbon storage of urban forest communities based on multi-source data according to claim 1, characterized in that, The device information includes lidar and panoramic camera.

5. An urban forest community carbon storage assessment system based on multi-source data, characterized in that, The system includes: A research preparation module that obtains research area information, where the research area information includes the research area, research time, and research requirements, and the research time includes the data collection time and the duration of sample plot setting; A single-tree data module that installs devices according to the installation requirements, obtains the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the devices are installed, generates single-tree data units according to the relationship model, and binds unique numbers to the single-tree data units; A secondary stripping module that extracts the contour features of the image area corresponding to the single-tree data unit, identifies whether there are contradictions in the contour features, performs secondary stripping on abnormal contours to obtain the final single-tree data unit, and classifies species according to the final single-tree data unit; A level determination module that constructs a structure weighting factor in combination with the area of the sample plot, determines the corresponding level according to the total carbon storage value of the entire sample plot, calculates the theoretical upper limit of carbon storage, and calculates the relative deviation rate of the current carbon storage according to the theoretical upper limit of carbon storage and the total carbon storage value of the sample plot.

6. The urban forest community carbon storage assessment system based on multi-source data according to claim 5, characterized in that The single-tree data module includes: A device installation unit that obtains device information, obtains installation requirements, and installs devices according to the installation requirements, where the installation requirements include the installation height and calibration board; An information collection unit that obtains the geometric relationship between the orientation in the panoramic image and the direction of the point cloud center after the devices are installed, where the point cloud is obtained by lidar scanning, obtains the image recognition confidence and the acquisition integrity threshold, and evaluates the image recognition confidence and acquisition integrity of the panoramic image according to the threshold; A single-tree binding unit that constructs a projection envelope relationship model between the image and the point cloud according to the qualified geometric relationship, generates single-tree data units according to the relationship model, and binds unique numbers to the single-tree data units, where the single-tree data units include diameter at breast height, tree height, and crown width.

7. The urban forest community carbon storage assessment system based on multi-source data according to claim 6, characterized in that, The secondary stripping module includes: A contour feature unit that extracts the contour features of the image area corresponding to the single-tree data unit, where the contour features include crown shape, texture symmetry, and edge continuity, identifies whether there are contradictions in the contour features, and the contradiction is that the image display is inconsistent with the point cloud clustering. If it is identified that there is a contradiction, it is determined that the current clustering is abnormal; A secondary stripping unit that performs secondary stripping on abnormal contours, where the secondary stripping is to construct an auxiliary cutting plane in the point cloud guided by the main trunk direction or the crown edge in the image and re-divide the clustering block; A species classification unit that, after the division is completed, recalculates the diameter at breast height, tree height, and crown width of each sub-block to obtain the final single-tree data unit, classifies species according to the final single-tree data unit, and identifies specific tree species through species classification.

8. A system for evaluating the carbon storage of urban forest communities based on multi-source data according to claim 7, characterized in that, The level determination module includes: A structure area division unit obtains a sample plot area and divides the sample plot into different hierarchical structure areas according to the final individual tree data unit. The structure areas include a structure homogeneous area, a structure transition area, and a structure heterogeneous area, and calculates the average individual tree carbon storage and the total number of trees in the structure area; A grade determination unit constructs a structure weighting factor in combination with the area of the sample plot, integrates the total carbon storage value of the entire sample plot according to the structure weighting factor, obtains the sample plot carbon storage grade evaluation rule, and determines the corresponding grade according to the total carbon storage value of the entire sample plot; A deviation rate unit calculates the theoretical carbon storage upper limit, calculates the relative deviation rate of the current carbon storage according to the theoretical carbon storage upper limit and the total carbon storage value of the sample plot, obtains the deviation rate threshold. If the relative deviation rate is greater than the deviation rate threshold, a list of improvement suggestions is obtained and sent to the terminal.

9. The urban forest community carbon storage assessment system based on multi-source data according to claim 8, characterized in that The device information includes a lidar and a panoramic camera.

Images