Single tree dynamic monitoring method and system based on multi-period point cloud and deep learning

Abstract

The invention discloses a single tree dynamic monitoring method and system based on multi-stage point cloud and deep learning, and the method comprises the steps: collecting multi-stage multispectral images and laser point cloud data of a research region, and carrying out the preprocessing; performing registration fusion on the preprocessed multi-period multispectral image and the laser point cloud data to generate a multi-period fusion point cloud; based on the multi-period fusion point cloud, performing individual tree segmentation by using a deep learning model; and based on the individual tree segmentation result, inverting the crown breadth and the tree height of the individual tree, and analyzing growth parameters of the individual tree according to the inversion result of the crown breadth and the tree height. By fusing the multi-temporal point cloud and the multi-spectral image data and combining the deep learning model, accurate segmentation and dynamic growth monitoring of the single tree are realized, and the defects of low single tree segmentation accuracy, insufficient multi-source data fusion and incapability of capturing forest tree growth changes in the prior art are overcome. The crown breadth and tree height inversion model is adopted, the precision of growth parameter estimation is effectively improved, and a scientific basis is provided for forest resource management.

Description

Technical Field

[0001] The present invention belongs to the technical field of spectral imaging and laser point cloud, and in particular relates to a single tree dynamic monitoring method and system based on multi-period point cloud and deep learning. Background Art

[0002] Dynamic monitoring of individual trees is an important part of forest resource management and ecological research, and can provide a scientific basis for evaluating forest growth conditions, calculating carbon sinks, and controlling pests and diseases. At present, common methods for monitoring individual trees mainly rely on manual measurement and crown extraction technology based on remote sensing images. However, manual measurement is inefficient, costly, and limited in operability in complex environments. Although remote sensing images can cover large areas of forest land, they are limited by resolution and it is difficult to accurately extract the three-dimensional structural information of trees. In addition, traditional point cloud data processing methods are usually based on rule models or threshold-based segmentation algorithms. These methods show great limitations in dealing with the complexity of tree morphology and environmental noise, especially in the case of dense forest stands and overlapping crowns. It is difficult to achieve accurate segmentation and dynamic growth monitoring of individual trees.

[0003] With the development of deep learning technology, deep learning models based on point clouds have provided new ideas for single tree segmentation and growth monitoring. However, when processing multi-temporal point clouds, existing technologies often lack a mechanism for effectively fusing multi-source data (such as laser point clouds and multispectral images), resulting in large registration errors and insufficient feature extraction. In addition, the monitoring of the growth dynamics of single trees mostly stays at the static feature inversion stage, and cannot accurately capture the growth and change patterns of trees at different time points. Therefore, how to fully integrate multi-temporal point clouds and multispectral image data, and combine deep learning models to achieve accurate dynamic monitoring of single trees and inversion of growth parameters, has become a problem that needs to be solved urgently in the current technical field. Summary of the invention

[0004] The purpose of the present invention is to provide a single tree dynamic monitoring method and system based on multi-period point cloud and deep learning to solve the problem of inaccurate inversion accuracy in the prior art.

[0005] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention provides a single tree dynamic monitoring method based on multi-period point cloud and deep learning, the method comprising:

[0006] Collect multi-period multispectral images and laser point cloud data of the study area and perform pre-processing;

[0007] The pre-processed multi-period multispectral images are registered and fused with the laser point cloud data to generate a multi-period fused point cloud;

[0008] Based on the multi-phase fused point cloud, a deep learning model is used to perform single tree segmentation;

[0009] Based on the single tree segmentation result, the crown width and tree height of the single tree are inverted, and the growth parameters of the single tree are analyzed according to the inversion results of the crown width and tree height.

[0010] As a further improvement of an embodiment of the present invention, the method further includes that the registration and fusion step includes:

[0011] Convert the preprocessed multispectral image into a multispectral image point cloud and standardize the scale of each point cloud;

[0012] Extracting feature lines from the multispectral image point cloud and the laser point cloud respectively; the feature lines include edge feature lines representing tree boundary lines and corner feature lines representing structural key points;

[0013] Based on the extracted feature lines, registration and fusion are performed by calculating the spatial correspondence between the feature lines to generate a multi-phase fused point cloud; the multi-phase fused point cloud is a comprehensive point cloud data set containing multi-phase three-dimensional spatial information and spectral features.

[0014] As a further improvement of an embodiment of the present invention, the method further includes that the single wood splitting step includes:

[0015] Based on multi-phase fusion point clouds, a deep learning model is used to segment individual trees and output a point cloud subset for each individual tree. ,in, is a single point in the point cloud, Indicates the single tree number, Represents the segmentation label of the i-th point;

[0016] The use of a deep learning model to perform single tree segmentation includes:

[0017] Use PointNet++ model to fusion point cloud Segmentation is performed, where each point , is the spatial coordinate, The spectral characteristics.

[0018] As a further improvement of an embodiment of the present invention, the method further includes: the crown width inversion step includes:

[0019] Through the feature extraction function from Extract feature vectors from ;

[0020] Calculate crown width ,in is the neural network model adopted;

[0021] When training the neural network model, optimize the mean square error loss ;

[0022] in, is the total number of single trees, The actual crown width.

[0023] As a further improvement of an embodiment of the present invention, the method further includes that the crown width inversion step further includes:

[0024] The time series smoothing term is defined by regularization to capture the smooth trend of crown amplitude change, which is expressed as:

[0025] ,

[0026] in, is the predicted value of the crown width of the kth tree in the tth period, is the point cloud subset of the kth tree in the tth period The characteristic vector of , T is the time period;

[0027] Use environment variables with Concatenate to form an extended feature vector and calculate the crown width using a neural network model that integrates environmental data ;in, is the average annual rainfall, is the average temperature;

[0028] The dynamic adjustment of regression model parameters is achieved by optimizing the total loss function, which is expressed as:

[0029] ,

[0030] in, is the true crown width, and λ is the smoothing weight coefficient.

[0031] As a further improvement of an embodiment of the present invention, the method further includes that the tree height inversion step includes:

[0032] Point cloud subset Perform tree species classification or tree shape recognition, and output tree species through classification models ;

[0033] Based on tree species Select Curve Matching the crown shape; the type of the curve includes a parabola, a high-order polynomial or a spline curve;

[0034] Will Project to the XZ plane and get the point set ;

[0035] The curve was fitted by the least squares method Generate a crown height distribution model with the optimization goal as follows:

[0036] ,

[0037] in, ,express The number of points in is the height of the projection point;

[0038] The calculation of tree height is expressed as:

[0039] ,

[0040] in, is the ground height.

[0041] As a further improvement of an embodiment of the present invention, the method further includes that the tree height inversion step further includes:

[0042] The fitting effect is optimized by embedding the adaptive adjustment mechanism of the growth model parameters, including:

[0043] Introduced in tree height fitting and maximum height potential , adjust the fitting function to:

[0044] ,

[0045] in, For tree age, is the spatial distribution function;

[0046] From point cloud subset Extract feature vectors from , and combined with external environmental data and tree species information, growth parameters are calculated through regression models or rule mapping:

[0047] ,

[0048] ,

[0049] in, For prediction models based on neural networks or empirical formulas, the parameters are adaptively updated according to the current point cloud and environmental conditions;

[0050] Based on the adjusted , optimize the target by the least square method, generate a dynamic crown height distribution model, and calculate the tree height;

[0051] The optimization objective is expressed as:

[0052] .

[0053] As a further improvement of an embodiment of the present invention, the method further includes that analyzing the growth parameters of a single tree according to the inversion results of the crown width and the tree height includes:

[0054] Based on the multi-period single tree segmentation results, time-space matching is performed, where the single tree set in period t is ;

[0055] in, is the total number of trees in period t, and each tree Corresponding point cloud subset ;

[0056] Calculate the center point of a single tree ;

[0057] The time-space matching formula is defined as: ,but , it is determined to be a successful match, otherwise 0; among them, is the single tree number of the t+1 period, is the distance threshold;

[0058] If the match is successful, calculate the tree height growth rate , and crown growth rate ;

[0059] in, is the tree height inversion result, is the crown width inversion result, For the time interval.

[0060] As a further improvement of an embodiment of the present invention, the method further includes that the analysis of the growth parameters of a single tree based on the inversion results of the crown width and the tree height further includes:

[0061] By introducing a closed-loop feedback mechanism to improve detection accuracy, specifically including:

[0062] Record the single wood collection of the previous issues , matching results , tree height and crown width As historical data;

[0063] Calculate the matching success rate based on historical data and historical growth rate Adjust the distance threshold:

[0064] ,

[0065] in, is the adjustment coefficient, is a penalty factor to adapt to changes in tree growth and forest area characteristics;

[0066] Use historical tree height sequences and crown width sequences to dynamically adjust inversion parameters, and recalculate based on the adjusted parameters and ;

[0067] The matching results and inversion parameters of the current period are fed back to the next analysis period to realize a closed-loop process of monitoring, feedback and optimization, thereby improving the success rate of spatiotemporal matching and the accuracy of growth rate calculation.

[0068] To achieve one of the above-mentioned purposes of the invention, an embodiment of the present invention further provides a single tree dynamic monitoring system based on multi-period point cloud and deep learning, the system comprising a data acquisition module, a data fusion module, a segmentation module and an analysis module;

[0069] The data acquisition module is used to collect multi-period multispectral images and laser point cloud data of the research area and perform preprocessing;

[0070] The data fusion module is used to register and fuse the preprocessed multi-period multispectral images with the laser point cloud data to generate a multi-period fused point cloud;

[0071] The segmentation module is used to perform single tree segmentation using a deep learning model based on the multi-phase fused point cloud;

[0072] The analysis module is used to invert the crown width and tree height of the single tree based on the single tree segmentation result, and analyze the growth parameters of the single tree according to the inversion results of the crown width and tree height.

[0073] Compared with the prior art, the present invention provides a method and system for dynamic monitoring of single trees based on multi-period point cloud and deep learning. By integrating multi-phase point cloud and multispectral image data and combining deep learning models, the present invention realizes accurate segmentation and dynamic growth monitoring of single trees, overcoming the defects of low accuracy of single tree segmentation, insufficient fusion of multi-source data, and inability to capture changes in tree growth in the prior art. The inversion model of crown width and tree height is adopted to effectively improve the accuracy of growth parameter estimation, providing a scientific basis for tree growth monitoring and forest resource management. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Figure 1 It is an overall flow chart of the single tree dynamic monitoring method based on multi-period point cloud and deep learning described in the present invention.

[0075] Figure 2 It is a schematic diagram of the architecture of the single tree dynamic monitoring system based on multi-period point cloud and deep learning described in the present invention. DETAILED DESCRIPTION

[0076] The present invention will be described in detail below in conjunction with the specific embodiments shown in the accompanying drawings. However, these embodiments do not limit the present invention, and any structural, methodological, or functional changes made by a person skilled in the art based on these embodiments are all within the scope of protection of the present invention.

[0077] The embodiments of the present invention are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0078] In the first embodiment of the present invention, the present invention provides a single tree dynamic monitoring method based on multi-period point cloud and deep learning, such as Figure 1 As shown, the method includes,

[0079] S1: Collect multi-period multispectral images and laser point cloud data of the study area and perform preprocessing;

[0080] S2: register and fuse the preprocessed multi-period multispectral images with the laser point cloud data to generate a multi-period fused point cloud;

[0081] S3: Based on the multi-phase fused point cloud, a deep learning model is used to perform single tree segmentation;

[0082] S4: Based on the single tree segmentation result, the crown width and tree height of the single tree are inverted, and the growth parameters of the single tree are analyzed according to the inversion results of the crown width and tree height.

[0083] In a specific embodiment of the present invention, the registration and fusion step is specifically,

[0084] Convert the preprocessed multispectral image into a multispectral image point cloud and standardize the scale of each point cloud;

[0085] Extracting feature lines from the multispectral image point cloud and the laser point cloud respectively; the feature lines include edge feature lines representing tree boundary lines and corner feature lines representing structural key points;

[0086] Based on the extracted feature lines, registration and fusion are performed by calculating the spatial correspondence between the feature lines to generate a multi-phase fused point cloud; the multi-phase fused point cloud is a comprehensive point cloud data set containing multi-phase three-dimensional spatial information and spectral features.

[0087] It should be noted that in the present invention, multi-period fusion point cloud refers to a comprehensive point cloud data set containing multi-phase three-dimensional spatial information and spectral features generated by spatially registering and fusing multi-period multispectral image point clouds with corresponding laser point cloud data. Specifically, multi-period multispectral images refer to multispectral image data of the study area collected in different time periods (such as every year or every season), while laser point cloud data refers to three-dimensional point cloud data acquired by laser radar technology in the corresponding time period. The goal of registration and fusion is to align these multi-period data in space and integrate their spatial coordinates and spectral information to provide a unified high-precision data basis for subsequent single tree segmentation and dynamic monitoring. In the implementation process, the pre-processed multi-period multispectral images are first converted into multispectral image point clouds, where each point cloud contains spatial coordinate information (x, y, z) and corresponding spectral feature information c. The spatial coordinates (x, y, z) are generated by geometric correction and depth estimation of the image, indicating the position of the point in three-dimensional space; the spectral feature c is derived from the reflectance of each band of the multispectral image (such as visible light and near-infrared band values), which characterizes the spectral characteristics of the forest. To ensure that the point clouds from different data sources maintain consistency in spatial distribution and scale, the generated multispectral image point clouds and laser point clouds are scale-normalized. Specifically, scale normalization calculates the spatial range of the point cloud and normalizes the coordinate values ​​of all points to a unified scale interval (such as [0, 1]), thereby reducing the registration error caused by differences in data acquisition equipment or resolution.

[0088] Furthermore, feature extraction is performed on the multispectral image point cloud and the laser point cloud respectively, and the extracted feature lines include edge feature lines and corner feature lines. "Feature lines" refer to a set of linear features that can characterize the outline of the forest structure extracted from the point cloud data by analyzing the spatial distribution characteristics of the point cloud. Specifically, the edge feature line is the boundary line of the forest determined by calculating the gradient change rate of the point cloud, that is, identifying areas with significant changes in height or density (such as the boundary between the crown and the background) in the point cloud, and connecting the points in these areas into continuous line segments; the corner feature line is formed by identifying key points with sudden changes in curvature in the point cloud (such as the top of the crown or the intersection of branches). These key points usually have high local curvature values ​​and can be extracted and connected into lines through curvature calculation formulas (such as the main curvature estimation based on neighborhood points). This process ensures that the feature lines extracted from the multispectral image point cloud and the laser point cloud have sufficient discrimination and robustness, can accurately reflect the spatial structural characteristics of the forest, and provide a reliable reference for subsequent registration.

[0089] Furthermore, the registration fusion is achieved through a "feature line matching algorithm", which aims to establish a spatial correspondence between the multispectral image point cloud and the feature lines of the laser point cloud, and finally generate a fused point cloud by optimizing the relative positions of the two. In this embodiment, the "feature line matching algorithm" includes the following steps: Feature line extraction and description: Edge feature lines and corner feature lines are extracted from multispectral image point clouds and laser point clouds, and descriptors are generated for each feature line. Descriptors can be vector representations based on line segment length, direction, curvature distribution, or neighborhood point statistical features (such as point density) to quantify the geometric and spatial properties of feature lines.

[0090] Calculation of correspondence between feature lines: By comparing the descriptors of two sets of feature lines, the similarity between them is calculated to determine the matching pairs. "Correspondence between feature lines" means that in multispectral image point clouds and laser point clouds, the feature line pairs that represent the same forest structure (such as the edge or vertex of the same crown) in space are found by calculating the Euclidean distance or cosine similarity as the similarity metric. By traversing all feature line pairs, a preliminary set of matching relationships is generated.

[0091] Registration optimization: Based on the set of matching relationships, the least squares method or iterative closest point (ICP) algorithm is used to optimize the registration results. The least squares method calculates the optimal rigid transformation of the two point clouds by minimizing the sum of the squares of the distances between the matching feature line pairs; the ICP algorithm further refines the registration accuracy by iteratively finding the closest point pairs and updating the transformation parameters. After optimization, the spatial positions of the multispectral image point cloud and the laser point cloud are aligned, and their point data are integrated in chronological order to form a multi-period fused point cloud.

[0092] In a specific embodiment of the present invention, the single wood splitting step is specifically,

[0093] Based on multi-phase fusion point clouds, a deep learning model is used to segment individual trees and output a point cloud subset for each individual tree. ,in, is a single point in the point cloud, Indicates the single tree number, Represents the segmentation label of the i-th point;

[0094] The use of a deep learning model to perform single tree segmentation includes:

[0095] Use PointNet++ model to fusion point cloud Segmentation is performed, where each point , is the spatial coordinate, The spectral characteristics.

[0096] Furthermore, the segmentation loss function is expressed as:

[0097] ,

[0098] in, is the true category distribution, is the predicted probability, The total number of single trees.

[0099] It should be noted that the goal of single tree segmentation is to accurately separate the point cloud data of each tree from the multi-period fused point cloud, laying the foundation for subsequent parameter inversion and dynamic analysis. Specifically, this step uses the multi-period fused point cloud as input, performs single tree segmentation through a deep learning model, and finally outputs a point cloud subset of each single tree. The multi-period fused point cloud is generated by registration and fusion of the multispectral image point cloud and the laser point cloud, which contains the three-dimensional spatial information and spectral properties of the study area in different time periods. Compared with the traditional segmentation method based on single-period data, this step makes full use of the spatiotemporal continuity of multi-period data, and realizes high-precision single tree identification through deep learning technology. It is suitable for dynamic monitoring scenarios and can capture the morphological changes of trees during their growth process.

[0100] Furthermore, in the initial stage of single tree segmentation, the multi-phase fused point cloud generated by the aforementioned registration and fusion steps is used as input data, and an advanced deep learning model is used for single tree segmentation. The multi-phase fused point cloud is a comprehensive data set obtained by registering and fusion of multi-phase multispectral image point cloud and laser point cloud data, which contains the three-dimensional spatial information (coordinates) and spectral characteristics (reflectivity) of trees in the study area at different time nodes. By analyzing this data set through a deep learning model, the spatial boundaries and category affiliation of each single tree can be effectively identified, and finally a point cloud subset of each single tree is output. This segmentation method is not only applicable to single-phase data, but also can capture the dynamic change characteristics of trees through the spatiotemporal continuity of multi-phase data, laying the foundation for the subsequent analysis of growth parameters.

[0101] Furthermore, the PointNet++ model uses a hierarchical feature extraction and classification process to fused point clouds from multiple periods. The core steps include the following aspects: Local feature extraction: The model first groups the point cloud based on the spatial neighborhood of the point, and uses spherical query or KNN algorithm to extract local features for each point. Determine the neighborhood point set within its local area. Then, the spatial geometric features and spectral features of each local area are extracted through a multi-layer perceptron to form a preliminary feature representation. Multi-scale feature aggregation: PointNet++ adopts a multi-scale sampling and grouping strategy to aggregate local features into global features by gradually reducing the sampling points and expanding the neighborhood range. This hierarchical structure can capture diverse information from crown details to overall morphology, and is particularly suitable for segmentation tasks in overlapping crowns or dense forest areas. Classification and label output: After feature extraction is completed, the model uses a fully connected layer and a softmax function to classify each point. Generate category probability distribution and output segmentation label Finally, the label of each point After confirmation, the point cloud subset Can be filtered The points are automatically generated.

[0102] Furthermore, in order to train the PointNet++ model and optimize its segmentation performance, the cross entropy loss function is used to measure the difference between the predicted results and the true labels. The model uses the back-propagation algorithm and optimizers (such as the Adam optimizer) to adjust the network parameters and gradually improve the segmentation accuracy. The training data usually comes from a manually annotated point cloud dataset, where the point cloud subset and label of each individual tree are predetermined. Through multiple iterations, the model can learn the complex mapping relationship between the spatial structure and spectral characteristics of the point cloud and the individual tree category.

[0103] In a specific embodiment of the present invention, the crown width inversion step is specifically as follows:

[0104] Through the feature extraction function from Extract feature vectors from ;

[0105] Calculate crown width ,in is the neural network model adopted;

[0106] When training the neural network model, optimize the mean square error loss ;

[0107] in, is the total number of single trees, The actual crown width.

[0108] It should be noted that the crown width inversion step aims to accurately estimate the crown width of each tree from the point cloud subset obtained by single tree segmentation. The specific implementation process first uses a machine learning regression model to obtain the crown width of each tree from the point cloud subset obtained by single tree segmentation. The present invention introduces a machine learning regression model to integrate multiple features of point clouds, overcoming the limitations of traditional methods when the crown shape is irregular or the point cloud data is sparse. This data-driven approach significantly improves the accuracy and robustness of the estimation by learning the complex mapping relationship between point clouds and crown widths.

[0109] Furthermore, before regressing the crown width, it is necessary to extract the point cloud subset Specifically, the feature extraction process usually includes the following aspects: first, extracting spatial geometric features, such as the boundary range of the point cloud, height distribution statistics (mean, variance, etc.) and three-dimensional shape descriptors (such as the characteristic axis obtained by principal component analysis); second, combining spectral features, such as the reflectance mean or texture information of each band in the multispectral image point cloud. These features are integrated into a multidimensional vector , fully characterized The diversity and comprehensiveness of feature extraction provide rich input information for the subsequent neural network model, ensuring that crown width estimation can adapt to the differences between different tree species and growth environments.

[0110] Furthermore, based on the extracted feature vector , using a neural network model to calculate the crown width of a single tree The neural network model here A multi-layer perceptron (MLP), convolutional neural network (CNN) or other regression network structures can be used. The core function is to transform the feature vector Convert to crown amplitude The model’s input layer receives the feature vector After hidden layer processing (such as nonlinear transformation of activation function ReLU) and parameter weight adjustment, the predicted value of crown width is finally generated in the output layer. Compared with traditional geometric measurement methods, neural networks can learn the implicit relationship between point cloud features and crown width, especially when dealing with non-circular or irregular crowns, showing higher flexibility and accuracy. In addition, if the point cloud data contains noise or missing parts, the generalization ability of neural networks can also effectively alleviate the impact of these problems.

[0111] It should be noted that in order to ensure the prediction accuracy of the neural network model, the model parameters are adjusted by optimizing the mean square error loss function during the training phase. The goal of the mean square error loss function is to quantify the gap between the predicted value and the true value by minimizing , the model gradually learns the precise mapping relationship between point cloud features and crown width. During the training process, the back propagation algorithm and optimizer (such as Adam or SGD) are used to update the network weights, usually combined with batch training and learning rate scheduling strategies to accelerate convergence and avoid overfitting. In addition, to enhance the generalization ability of the model, regularization techniques (such as L2 regularization) or data enhancement methods (such as point cloud rotation and scaling) can be introduced to improve its performance on different data sets.

[0112] Preferably, in a specific embodiment of the present invention, the inversion framework of the crown amplitude further integrates the time series smoothing term and environmental data feedback to achieve dynamic adaptive adjustment of the regression model parameters.

[0113] It should be noted that the inversion framework of crown width aims to accurately estimate the width of the tree crown from a subset of single tree point clouds, and its basis is to predict the crown width through a machine learning regression model. However, a single static regression model may not be able to fully adapt to the dynamic changes in the growth process of trees or the diversity of environmental conditions. To this end, the time series smoothing term and the environmental data feedback mechanism are further integrated to form a dynamic adaptive inversion framework. This framework not only relies on the spatial and spectral characteristics of the point cloud, but also optimizes the parameter adjustment process of the regression model by introducing time series information and external environmental variables, thereby improving the accuracy and stability of crown width estimation. Such a design can better reflect the crown width change trend of trees at different growth stages and ecological conditions, and provide more reliable data support for dynamic monitoring.

[0114] Furthermore, in traditional crown amplitude inversion, the regression model usually independently predicts the crown amplitude based on single-period point cloud data. However, tree growth has temporal continuity, and the crown width changes between multiple periods of data should show a smooth trend rather than abrupt changes. To capture this temporal characteristic, a temporal smoothing term is introduced into the framework, which is specifically integrated into the loss function through regularization. Assume that the crown width prediction value of the tth period is ,in, Point cloud subset The characteristic vector of the time series smoothing term is defined as the difference penalty of the crown amplitude prediction values ​​in adjacent time periods:

[0115] ,

[0116] Where T is the time period, Smoothness of the model output in the temporal dimension is encouraged to avoid jumps in crown width prediction caused by point cloud noise or short-term anomalies.

[0117] Furthermore, the growth of tree crowns is not only affected by their own structure, but also closely related to external environmental conditions, such as light, rainfall, temperature, and soil moisture. To further improve the adaptability of the regression model, an environmental data feedback mechanism is integrated into the framework. Specifically, environmental variables (such as annual average rainfall R t , average temperature T temp，t ) as additional input, together with the point cloud feature vector Concatenate to form an extended feature vector The regression model is updated to ,in This is a neural network model that integrates environmental data. The introduction of environmental data enables the model to perceive the ecological context in which the trees grow, such as predicting smaller crown expansion in dry years and larger crown expansion in wet years. This feedback mechanism dynamically adjusts the model output through external conditions, enhancing its ability to respond to environmental changes.

[0118] Furthermore, the optimized total loss function combines the original mean square error and the time series smoothing term:

[0119] ,

[0120] in, is the true crown width, and λ is the smoothing weight, which is used to balance accuracy and smoothness. This time series constraint enables the model to learn the long-term variation of the crown width and improves the consistency of multi-period data prediction.

[0121] Furthermore, based on the time series smoothing term and environmental data feedback, the inversion framework realizes dynamic adaptive adjustment of the regression model parameters. In the training phase, the model optimizes the total loss function Learn the initial parameters; in the inference stage, for each period of data, the model and the previous period's forecast value Dynamically update parameters. For example, you can introduce an adaptive learning rate or online learning strategy so that the model weights adjust over time and environmental conditions:

[0122] ,

[0123] in, is the model parameter of the tth period, η is the learning rate, is the loss gradient. This dynamic adjustment mechanism enables the model to adapt to new data in real time, overcoming the performance degradation problem of static models in long-term monitoring. In addition, if environmental data changes significantly (such as abnormal drought), the model can quickly correct the prediction deviation through the feedback mechanism to ensure the continued accuracy of crown width estimation.

[0124] In a specific embodiment of the present invention, the tree height inversion step is specifically,

[0125] Point cloud subset Perform tree species classification or tree shape recognition, and output tree species through classification models ;

[0126] Based on tree species Select Curve Matching the crown shape; the type of the curve includes a parabola, a high-order polynomial or a spline curve;

[0127] Will Project to the XZ plane and get the point set ;

[0128] The curve was fitted by the least squares method Generate a crown height distribution model with the optimization goal as follows:

[0129] ,

[0130] in, ,express The number of points in is the height of the projection point;

[0131] The calculation of tree height is expressed as:

[0132] ,

[0133] in, is the ground height.

[0134] It should be noted that the first step of tree height inversion is to subset the point cloud of a single tree. Perform tree species classification or tree shape identification to determine the biological characteristics of each tree. Specifically, the classification model From point cloud subset Output tree species identification Classification Model A deep learning network (such as PointNet or Random Forest classifier) ​​can be used, whose input is The geometric features (such as height distribution, point density) and spectral features (such as multispectral band values) of the tree are output as tree species categories. The purpose of tree species classification is to provide a basis for subsequent curve selection, because the tree shapes of different tree species (such as cone, round or irregular) have a significant impact on height distribution. By identifying tree species or tree shapes, this step can provide a personalized modeling basis for tree height estimation, which significantly improves adaptability and accuracy compared to the traditional inversion method of uniformly assuming tree shapes.

[0135] Furthermore, in determining the tree species Then, select the appropriate curve type according to its characteristics. , which is used to adapt to the tree species morphology and improve fitting accuracy. Curve types include parabola, high-order polynomial or spline curves, and the specific choice depends on the typical tree shape characteristics of the tree species. For example, for conifers (such as pine trees), their tree shape is usually conical, so parabolas (such as ) fitting; for broad-leaved trees (such as oaks), the crown may be wider and irregular, and higher-order polynomials (such as cubic or quartic polynomials) or spline curves can be used to capture complex vertical contours. Traditional tree height inversion methods often assume fixed geometric models (such as cylinders or simple straight lines) and cannot adapt to diverse tree shapes. This step achieves adaptive tree height modeling through tree species-specific curve selection. This flexibility not only improves fitting accuracy, but also enhances the versatility of the method in different forest types.

[0136] It should be noted that for curve fitting, the point cloud subset Project to the XZ plane to generate a two-dimensional point set The projection process retains the horizontal position x and height z of the point cloud, ignoring the y coordinate to simplify the calculation, forming the side profile of the tree in the vertical direction. This projection method is based on the definition of tree height (i.e. the difference between the top of the crown and the ground height), aiming to extract the height distribution characteristics of the point cloud. The projected point set provides direct two-dimensional data input for subsequent curve fitting, simplifying the conversion process of three-dimensional point cloud to one-dimensional height curve.

[0137] Furthermore, based on the projection point set , fit the selected curve type by the least squares method , generate a height distribution model and provide accurate tree height estimation. The core of the least squares method is to minimize the sum of square errors between the fitting curve and the observation points by adjusting the curve parameters (such as a, b, c of the parabola). This process can effectively smooth the noise or discreteness in the point cloud data and generate a continuous tree height contour curve. Compared with the traditional method of directly taking the highest point as a simple estimate of the tree height, this step captures the overall height trend of the crown through curve fitting, avoids the interference of local outliers on the results, and improves the reliability of tree height estimation.

[0138] It should be noted that the tree height The calculation of the curve accurately reflects the actual height of a single tree by the difference between the vertex of the curve and the ground height. This curve-based tree height definition not only takes into account the overall shape of the crown, but also The correction eliminates the influence of terrain undulation. Compared with the rough estimation of the traditional method that only relies on a single highest point, this method calculates the tree height by the maximum value of the curve, which can better deal with the situation where the top point of the crown is missing or unevenly distributed.

[0139] Preferably, in a specific embodiment of the present invention, an adaptive adjustment mechanism of growth model parameters is further embedded in the tree height inversion to optimize the fitting effect.

[0140] It should be noted that the goal of tree height inversion is to accurately estimate the height of trees from a subset of single tree point clouds, which is usually achieved by fitting curves (such as parabolas or high-order polynomials). However, traditional static fitting methods often assume that tree height distribution follows a fixed model, ignoring the dynamic characteristics of tree growth and the differences between different tree species and environmental conditions. In order to improve the fitting effect and enhance the adaptability of the inversion method, an adaptive adjustment mechanism of growth model parameters is further embedded in tree height inversion. This mechanism introduces the parameters of the tree growth model and dynamically adjusts these parameters in combination with point cloud data and external information to optimize the accuracy and robustness of curve fitting. Such a design not only improves the accuracy of tree height estimation, but also provides a more reliable basis for growth trend analysis for dynamic monitoring of multi-period data.

[0141] Furthermore, in tree height inversion, growth model parameters are embedded in the curve fitting process to describe the law of tree height changes over time or environment. Projecting point sets onto point clouds Optimization is performed, and this mechanism further introduces growth model parameters such as growth rate and maximum height potential These parameters are derived from the biological characteristics of tree growth and are in the following basic form:

[0142] ,

[0143] in, For the tree to be tall, For tree age, is the initial time, is the growth rate. To adapt to the point cloud data, the growth model is combined with the geometric curve, and the adjusted fitting function can be expressed as:

[0144] ,

[0145] in, is a spatial distribution function (such as a parabola), and the growth parameters dynamically control the overall scale of tree height.

[0146] Furthermore, the adaptive adjustment mechanism of the growth model parameters is dynamically optimized through point cloud features and external data. , Specifically, firstly, from the point cloud subset Extract feature vectors from , including height statistics (such as maximum height, distribution variance) and spectral characteristics (such as canopy density indicators); at the same time, combined with external inputs such as environmental data and tree species information This information is used to predict initial growth parameters using regression models or rule mapping, such as:

[0147] ,

[0148] ,

[0149] in, It can be an artificial neural network model based on deep learning, or a mathematical expression based on domain knowledge and data. The adaptive adjustment mechanism updates the parameters according to the current point cloud and environmental conditions at each fitting to ensure It can reflect the actual growth status of trees. For example, under drought conditions, May decrease, reflecting a slowing growth rate.

[0150] Furthermore, based on the adaptively adjusted growth parameters, the optimization objective still uses the least squares method, but the fitting function is updated to a dynamic form:

[0151] ,

[0152] in, The growth model parameters are integrated. The optimization process is iteratively adjusted and The parameters of make the curve fit the point cloud distribution better. The final tree height is calculated as:

[0153] ,

[0154] The adaptability of growth parameters enables the fitting curve to dynamically adapt to the growth stage of trees and environmental changes, such as the rapid growth of young trees or the high saturation of mature trees, which significantly improves the accuracy of the fitting effect.

[0155] In a specific embodiment of the present invention, the growth parameters of a single tree are analyzed based on the inversion results of the crown width and the tree height, specifically,

[0156] Based on the multi-period single tree segmentation results, time-space matching is performed, where the single tree set in period t is ;

[0157] in, is the total number of trees in period t, and each tree Corresponding point cloud subset ;

[0158] Calculate the center point of a single tree ;

[0159] The time-space matching formula is defined as: ,but , it is determined to be a successful match, otherwise 0; among them, is the single tree number of the t+1 period, is the distance threshold;

[0160] If the match is successful, calculate the tree height growth rate , and crown growth rate ;

[0161] in, is the tree height inversion result, is the crown width inversion result, For the time interval.

[0162] It should be noted that the core goal of dynamic analysis is to monitor the growth changes of individual trees through multi-period data. The first step is to perform spatiotemporal matching based on the multi-period individual tree segmentation results to identify the corresponding relationship of the same tree in different time periods. The process of spatiotemporal matching aims to determine which individual trees belong to the same tree in the time series from multi-period data, providing a basis for subsequent growth parameter calculations. Traditional methods may rely on manual marking or simple spatial proximity judgments, while this step achieves automated and efficient individual tree tracking through a systematic center point matching algorithm, which is particularly suitable for large-scale forest monitoring scenarios.

[0163] Furthermore, to achieve spatiotemporal matching, the center point of each tree is first calculated as a representative of its spatial position. The calculation of the center point integrates the spatial distribution of the point cloud and can effectively characterize the location characteristics of a single tree, providing a stable reference point even if the crown shape is irregular or the point cloud density is uneven.

[0164] Furthermore, based on the center point of a single tree, the time-space matching formula is used to determine the single tree in period t. The single tree in period t+1 Whether they correspond to the same tree. is the preset distance threshold. By traversing all single trees in the t+1 period , find The closest distance and less than of , to determine the matching relationship. If the distance is less than , then the match is determined to be successful (match=1), indicating and are the same tree; otherwise, the match fails (match=0). The choice of should be determined according to the tree density and growth rate of the study area. For example, it can be set to a smaller value (such as 1-2 meters) in sparse forest areas, and appropriately relaxed (such as 3-5 meters) in dense forest areas to balance the accuracy and fault tolerance of matching. This matching algorithm achieves efficient identification of single tree identity by quantifying spatial distance and combining the spatiotemporal continuity of multi-period data, avoiding the complex morphological comparison or manual correction in traditional methods.

[0165] Preferably, in a specific embodiment of the present invention, a closed-loop feedback mechanism is further introduced in the dynamic analysis to adjust the spatiotemporal matching and inversion parameters in real time according to historical monitoring data to improve the overall monitoring accuracy.

[0166] It should be noted that in order to improve the overall monitoring accuracy, a closed-loop feedback mechanism is further introduced to use historical monitoring data to adjust the spatiotemporal matching and inversion parameters in real time. This mechanism constructs a dynamic feedback loop, takes the previous monitoring results as input, and feeds them back to the current analysis process, thereby optimizing the matching threshold and inversion model parameter settings. This design not only improves the accuracy of single tree tracking, but also enhances the reliability of tree height, crown width and other parameter estimates, providing more accurate technical support for long-term forest dynamic monitoring.

[0167] Furthermore, the core of the closed-loop feedback mechanism is to use historical monitoring data (such as the individual tree matching results and inversion parameters of previous periods) as the basis for real-time adjustment, forming a "monitoring-feedback-optimization" cycle. Specifically, assuming that the individual tree set in period t is , tree height is obtained by time-space matching and inversion calculation , Crown Width And matching relationship These results are stored as historical data and are stored together with the current input point cloud. Together, the feedback model is input. The feedback model evaluates the effectiveness of the current parameters based on the regularity of historical data (such as growth rate trend, matching success rate), and dynamically adjusts the distance threshold of spatiotemporal matching. and the weights of the inversion model The closed-loop feedback enables the system to adaptively respond to tree growth and environmental changes by continuously learning historical information, improving the continuity and consistency of monitoring.

[0168] Furthermore, in the traditional method of spatiotemporal matching relying on the calculation of the center point distance, As a fixed value, it is difficult to adapt to the differences in tree density or growth rate in different forest areas. Through a closed-loop feedback mechanism, historical matching data is used to dynamically adjust , define the matching success rate , and based on the historical growth rate Estimate the spatial movement range of trees. The adjustment formula can be expressed as:

[0169] ,

[0170] in, is the adjustment coefficient, is the penalty factor. If the historical matching success rate Lower, Appropriate relaxation to improve the matching rate; if the growth rate Larger, Increase to adapt to the center point offset caused by changes in tree height. This real-time adjustment mechanism enables spatiotemporal matching to dynamically adapt to the actual situation in the monitoring area and reduce mismatching or missed matching.

[0171] Furthermore, the inversion parameters (such as the tree height fitting curve , or crown width regression model ) is also optimized through a closed-loop feedback mechanism. Historical monitoring data provide a time series of inversion results (e.g. ), the feedback model analyzes its trends and errors and dynamically updates the current parameters. For example, for tree height inversion, the growth parameters can be adjusted according to the historical tree height:

[0172] ,

[0173] in, is the predicted tree height for the tth period, is the learning rate. If the historical prediction is too low, Similarly, the weights of the crown width regression model Updates via online learning:

[0174] ,

[0175] This parameter optimization uses historical errors to correct the model in real time, ensuring that the inversion results are consistent with the actual growth trend and improving the accuracy of long-term monitoring.

[0176] Furthermore, the closed-loop feedback mechanism forms a dynamically optimized monitoring system through the coordinated adjustment of time and space matching and inversion parameters. After each monitoring is completed, the results are recorded and fed back to the next analysis, and the cycle is iterated and optimized. , , The overall accuracy improvement is reflected in: the success rate of spatiotemporal matching is improved, avoiding the interruption of single tree tracking; the adaptability of inversion parameters is enhanced, reducing the systematic error of tree height and crown width estimation. In multi-period monitoring, this mechanism can smooth the change curve of growth parameters, reduce the impact of noise, and provide more reliable growth rates (such as , ) calculation results.

[0177] In the second embodiment of the present invention, the present invention provides a single tree dynamic monitoring system based on multi-period point cloud and deep learning, such as Figure 2 As shown, the system includes a data acquisition module 1, a data fusion module 2, a segmentation module 3 and an analysis module 4;

[0178] The data acquisition module 1 is used to collect multi-period multispectral images and laser point cloud data of the research area and perform preprocessing;

[0179] The data fusion module 2 is used to register and fuse the pre-processed multi-period multispectral images with the laser point cloud data to generate a multi-period fused point cloud;

[0180] The segmentation module 3 is used to perform single tree segmentation based on the multi-phase fused point cloud using a deep learning model;

[0181] The analysis module 4 is used to invert the crown width and tree height of the single tree based on the single tree segmentation result, and analyze the growth parameters of the single tree according to the inversion results of the crown width and tree height.

[0182] In summary, the present invention provides a method and system for dynamic monitoring of single trees based on multi-period point cloud and deep learning. By integrating multi-phase point cloud and multispectral image data and combining deep learning models, the accurate segmentation and dynamic growth monitoring of single trees are achieved, overcoming the defects of low accuracy of single tree segmentation, insufficient fusion of multi-source data, and inability to capture changes in tree growth in the prior art. The inversion model of crown width and tree height is adopted to effectively improve the accuracy of growth parameter estimation, providing a scientific basis for tree growth monitoring and forest resource management.

[0183] It should be understood that although this specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each implementation mode may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

[0184] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the modules described above can refer to the corresponding process in the aforementioned method implementation, and will not be repeated here.

[0185] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present implementation scheme.

[0186] In addition, each functional module in each embodiment of the present application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0187] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for a computer system (which can be a personal computer, a server, or a network system, etc.) or a processor to perform some steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0188] Finally, it should be noted that the above implementation modes are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned implementation modes, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned implementation modes, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various implementation modes of the present application.

Claims

1. A single tree dynamic monitoring method based on multi-period point cloud and deep learning, characterized by: include, Collect multi-period multispectral images and laser point cloud data of the study area and perform pre-processing; The pre-processed multi-period multispectral images are registered and fused with the laser point cloud data to generate a multi-period fused point cloud; Based on the multi-phase fused point cloud, a deep learning model is used to perform single tree segmentation; Based on the single tree segmentation result, the crown width and tree height of the single tree are inverted, and the growth parameters of the single tree are analyzed according to the inversion results of the crown width and tree height.

2. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 1 is characterized by: The registration and fusion step comprises: Convert the preprocessed multispectral image into a multispectral image point cloud and standardize the scale of each point cloud; Extracting feature lines from the multispectral image point cloud and the laser point cloud respectively; the feature lines include edge feature lines representing tree boundary lines and corner feature lines representing structural key points; Based on the extracted feature lines, registration and fusion are performed by calculating the spatial correspondence between the feature lines to generate a multi-phase fused point cloud; the multi-phase fused point cloud is a comprehensive point cloud data set containing multi-phase three-dimensional spatial information and spectral features.

3. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 2 is characterized by: The single wood splitting step comprises: Based on multi-phase fusion point clouds, a deep learning model is used to segment individual trees and output a point cloud subset for each individual tree. ,in, is a single point in the point cloud, Indicates the single tree number, Represents the segmentation label of the i-th point; The use of a deep learning model to perform single tree segmentation includes: Use PointNet++ model to fusion point cloud Segmentation is performed, where each point , is the spatial coordinate, The spectral characteristics.

4. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 3 is characterized by: The crown width inversion step includes: Through the feature extraction function from Extract feature vectors from ; Calculate crown width ,in is the neural network model adopted; When training the neural network model, optimize the mean square error loss ; in, is the total number of single trees, The actual crown width.

5. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 4 is characterized by: The crown width inversion step further comprises: The time series smoothing term is defined by regularization to capture the smooth trend of crown amplitude change, which is expressed as: ， in, is the predicted value of the crown width of the kth tree in the tth period, is the point cloud subset of the kth tree in the tth period The characteristic vector of , T is the time period; Use environment variables with Concatenate to form an extended feature vector and calculate the crown width using a neural network model that integrates environmental data ;in, is the average annual rainfall, is the average temperature; The dynamic adjustment of regression model parameters is achieved by optimizing the total loss function, which is expressed as: ， in, is the true crown width, and λ is the smoothing weight coefficient.

6. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 3 is characterized by: The tree height inversion step includes: Point cloud subset Perform tree species classification or tree shape recognition, and output tree species through classification models ; Based on tree species Select Curve Matching the crown shape; the type of the curve includes a parabola, a high-order polynomial or a spline curve; Will Project to the XZ plane and get the point set ; The curve was fitted by the least squares method Generate a crown height distribution model with the optimization goal as follows: ， in, ,express The number of points in is the height of the projection point; The calculation of tree height is expressed as: ， in, is the ground height.

7. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 6 is characterized by: The tree height inversion step further comprises, The fitting effect is optimized by embedding the adaptive adjustment mechanism of the growth model parameters, including: Introduced in tree height fitting and maximum height potential , adjust the fitting function to: ， in, For tree age, is the spatial distribution function; From point cloud subset Extract feature vectors from , and combined with external environmental data and tree species information, growth parameters are calculated through regression models or rule mapping: ， ， in, For prediction models based on neural networks or empirical formulas, the parameters are adaptively updated according to the current point cloud and environmental conditions; Based on the adjusted , optimize the target by the least square method, generate a dynamic crown height distribution model, and calculate the tree height; The optimization objective is expressed as: 。 8. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 5 or 7, characterized in that: The growth parameters of a single tree analyzed based on the inversion results of crown width and tree height include: Based on the multi-period single tree segmentation results, time-space matching is performed, where the single tree set in period t is ; in, is the total number of individual trees in period t, and each individual tree Corresponding point cloud subset ; Calculate the center point of a single tree ; The time-space matching formula is defined as: ,but , it is determined to be a successful match, otherwise 0; among them, is the single tree number of the t+1 period, is the distance threshold; If the match is successful, calculate the tree height growth rate , and crown growth rate ; in, is the tree height inversion result, is the crown width inversion result, For the time interval.

9. The single tree dynamic monitoring method based on multi-period point cloud and deep learning according to claim 8, characterized in that: The analysis of the growth parameters of a single tree based on the inversion results of the crown width and the tree height further includes: By introducing a closed-loop feedback mechanism to improve detection accuracy, specifically including: Record the single wood collection of the previous issues , matching results , tree height and crown width As historical data; Calculate the matching success rate based on historical data and historical growth rate Adjust the distance threshold: ， in, is the adjustment coefficient, is a penalty factor to adapt to changes in tree growth and forest area characteristics; Use historical tree height sequences and crown width sequences to dynamically adjust inversion parameters, and recalculate based on the adjusted parameters and ; The matching results and inversion parameters of the current period are fed back to the next analysis period to realize a closed-loop process of monitoring, feedback and optimization, thereby improving the success rate of spatiotemporal matching and the accuracy of growth rate calculation.

10. A single tree dynamic monitoring system based on multi-period point cloud and deep learning, characterized by: It includes data acquisition module, data fusion module, segmentation module and analysis module; The data acquisition module is used to collect multi-period multispectral images and laser point cloud data of the research area and perform preprocessing; The data fusion module is used to register and fuse the preprocessed multi-period multispectral images with the laser point cloud data to generate a multi-period fused point cloud; The segmentation module is used to perform single tree segmentation using a deep learning model based on the multi-phase fused point cloud; The analysis module is used to invert the crown width and tree height of the single tree based on the single tree segmentation result, and analyze the growth parameters of the single tree according to the inversion results of the crown width and tree height.

Images