Urban road tree information identification method based on streetscape data

Abstract

The invention discloses an urban road tree information identification method based on streetscape data, and the method comprises the steps: obtaining road network data, and carrying out the preprocessing of the road network data, and obtaining a streetscape picture data set in a research region; inputting the streetscape picture data set into the trained deep learning recognition model to obtain an output image containing tree information; constructing a size reference model, and estimating the diameter at breast height and the height of the tree in combination with the size reference model and the output image; and matching and summarizing the tree information with the diameter at breast height and the height of the tree to obtain complete tree information corresponding to the tree. According to the invention, efficient and high-accuracy urban road tree information identification is realized.

Description

Technical Field

[0001] The present invention relates to the field of urban vegetation image recognition, and in particular to a method for identifying urban road tree information based on street view data. Background Art

[0002] With the rapid development of cities, the reduction of vegetation has led to an obvious intensification of the global climate greenhouse effect. In all aspects, it is necessary to carry out model urban greening construction from the perspective of carbon neutrality to ensure that the carbon sink function of urban gardening and greening can play an efficient role. In order to maximize the carbon sink function of urban gardening and greening, it is necessary to adopt advanced technical means to comprehensively monitor and accurately evaluate the ecological benefits of urban vegetation.

[0003] The calculation of carbon sequestration of urban road trees requires the acquisition of information on urban road trees, including the tree's diameter at breast height, tree height, tree type, and tree number.

[0004] Existing tree information identification methods mainly rely on Lidar point cloud data or sample plot inventory methods. These methods are not only time-consuming, but also require the purchase of related equipment, resulting in huge consumption of manpower and material resources. In contrast, street view data has unique advantages. Its acquisition process is simple and can intuitively display the vertical structure of trees. In addition, combined with relevant technologies, street view data can obtain tree species and structural characteristic data. However, the existing technology has not yet realized a method for directly using street view data for tree information identification and carbon sink estimation. Summary of the invention

[0005] In order to solve the technical problems that the existing tree information recognition methods are time-consuming and difficult to implement, the present invention provides a method for identifying tree information on urban roads based on street view data. The technical solution adopted by the present invention is:

[0006] A first aspect of the present invention provides a method for identifying tree information on urban roads based on street view data, the method comprising:

[0007] Obtain road network data and preprocess it to obtain a street view image dataset in the study area;

[0008] Input the street view image dataset into the trained deep learning recognition model to obtain the output image containing tree information;

[0009] Constructing a size reference model, and estimating the diameter at breast height and the height of the tree by combining the size reference model and the output image;

[0010] The tree information is matched and summarized with the tree's diameter at breast height and tree height to obtain complete tree information corresponding to the tree.

[0011] As a preferred solution, the method of obtaining the road network data and preprocessing it to obtain the street view image dataset in the study area includes:

[0012] Download road network data;

[0013] Simplify, vectorize and georeference road network data;

[0014] At the preset image angle, sample the observation points and overlay the road angle value and road number information corresponding to each point. After completion, export and organize the sampling point data into a plain text file;

[0015] Read the imported plain text file and store the sampling point information, generate the uniform resource locator of the application programming interface request to crawl the street view image data;

[0016] The street view image data is screened according to a preset definition, and the screened images are enhanced and corrected, and finally a street view image data set in the study area is obtained.

[0017] As a preferred solution, the method for vectorizing and georeferencing road network data includes:

[0018] Importing the road network data into a preset geospatial platform, performing binary classification with a preset threshold, and then reclassifying, setting the values ​​of non-road categories to null values, thereby simplifying the road network data into a pure road network grid map;

[0019] Use ArcScan tool to extract the road centerline of the road network data and complete the vectorization process;

[0020] Use the road network of the Open Street Map as a reference to adjust the road network data and complete the georeferencing.

[0021] As a preferred solution, the method of filtering the street view image data according to a preset definition, performing image enhancement and correction processing on the filtered images, and finally obtaining a street view image data set in the study area includes:

[0022] Screening the street view image data to obtain a street view image that meets a preset definition;

[0023] Perform brightness and contrast enhancement processing on street view images that meet the preset definition;

[0024] The preset radial distortion model and tangential distortion model are used to correct the enhanced street view images.

[0025] As a preferred solution, the radial distortion model is:

[0026] u′=u(1+k1r2 +k2r 4 +k3r 6 )

[0027] v′=v(1+k1r 2 +k2r 4 +k3r 6 )

[0028] The tangential distortion model is:

[0029] u′=u+[2P1v+P2(r 2 +2u 2 )]

[0030] →′=v+[P1(r 2 +2v 2 )+2P2u]

[0031] Among them, u and v are the horizontal and vertical coordinates of the image point, u′ and v′ are the horizontal and vertical coordinates of the image point after distortion correction, k1, k2, k3 are radial distortion coefficients, P1 and P2 are tangential distortion coefficients, and r is the distance from the center of the image to the coordinate point (u, v).

[0032] As a preferred solution, the training method of the deep learning recognition model includes:

[0033] Label different tree species in the street view image dataset;

[0034] Input the labeled street view image dataset into a preset deep learning recognition model for training to learn the features and patterns of the samples, wherein the deep learning recognition model is a YOLOv7 model;

[0035] The test set is used to evaluate the performance of the model, and the model with the best performance is retained to obtain a trained deep learning recognition model.

[0036] As a preferred solution, the tree information includes:

[0037] The horizontal and vertical coordinates of the trees, the tree ID, the type of trees, and the number of trees.

[0038] As a preferred solution, a size reference model is constructed, and a method for estimating the diameter at breast height and the height of a tree by combining the size reference model and the output image includes:

[0039] Select a reference object, wherein the reference object includes at least one of the following:

[0040] Shared bicycles, bus stops, cars, road fire hydrants and sanitation barrels;

[0041] Determine the average size of the reference object through field research, and construct a corresponding size reference model according to the average size of the reference object;

[0042] The size reference model and the output image are combined, and a reference scale of the size reference model is set to estimate the diameter at breast height and the height of the tree.

[0043] As a preferred embodiment, the method further comprises:

[0044] According to the allometric growth equations of different tree species combined with the corresponding complete tree information, the carbon storage of all trees is calculated and statistically analyzed to obtain the total tree carbon sink.

[0045] As a preferred solution, the carbon storage of all trees is calculated and counted according to the allometric growth equations of different tree species combined with the corresponding complete tree information. The method for obtaining the total carbon storage of trees includes:

[0046] Based on the complete tree information, statistical classification of tree species types is carried out;

[0047] Query the stock calculation formulas for different types of trees, the coefficients of the volume source biomass method, and the carbon storage calculation formula;

[0048] Select the corresponding formula according to the tree species, substitute the diameter at breast height D and tree height H into the volume formula, and calculate the vegetation volume V of the tree, V = aD b H c , where V is the vegetation stock, D is the tree diameter at breast height, and H is the tree height;

[0049] Using the volume source biomass method, the vegetation storage volume V is substituted into the formula to calculate the tree biomass B of the tree;

[0050] B=aV+b

[0051] Among them, a and b are constants, B is biomass, V is stock volume, and the values ​​of parameters a and b are related to tree species;

[0052] Multiply the calculated tree biomass B by the carbon content coefficient to obtain the carbon storage C of the tree;

[0053] C=B×C c

[0054] C is carbon storage, C c is the carbon content coefficient, B is the tree biomass;

[0055] Using the preset script, the complete tree information was combined with the above formula to calculate the carbon sink of road trees in the entire study area.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] The present invention has significant efficiency in data acquisition and processing. By utilizing existing street view image data, the high cost and complex process required for Lidar data acquisition and sample site inventory is avoided. By preprocessing street view images and labeling different tree species, the accuracy and efficiency of subsequent image recognition are significantly improved.

[0058] The present invention uses advanced image recognition technology. Using a deep learning recognition model to automatically identify trees in street view images, it can efficiently and accurately classify different tree species and count the number of trees in the entire area. This technology not only improves recognition efficiency, but also ensures the accuracy of recognition results.

[0059] In terms of size calculation, the present invention provides a standard for calculating tree size by establishing a size reference model, making the results more accurate and consistent. By calculating the diameter at breast height and tree height of each tree, accurate measurement of the basic properties of the tree is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A flow chart of a method for identifying tree information on urban roads based on street view data provided in this embodiment;

[0061] Figure 2 A schematic diagram of the Make Sense tool provided in this embodiment;

[0062] Figure 3 This is a schematic diagram of ImageJ tree height calculation provided in this embodiment. DETAILED DESCRIPTION

[0063] The accompanying drawings are only used for illustrative purposes and are not to be construed as limiting the present invention;

[0064] It should be clear that the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of the present application.

[0065] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. The singular forms of "a", "said" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0066] When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the attached claims. In the description of the present application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and do not have to be used to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0067] In addition, in the description of this application, unless otherwise specified, "multiple" means two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0068] The present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0069] Example 1

[0070] Please refer to Figure 1 This embodiment provides a method for identifying tree information on urban roads based on street view data, the method comprising:

[0071] S1: Obtain road network data and preprocess it to obtain a street view image dataset in the study area;

[0072] In a specific embodiment, a method for obtaining road network data and preprocessing it to obtain a street view image dataset in a research area includes:

[0073] Download road network data;

[0074] Simplify, vectorize and georeference road network data;

[0075] At the preset image angle, sample the observation points and overlay the road angle value and road number information corresponding to each point. After completion, export and organize the sampling point data into a plain text file (CSV);

[0076] Read the imported plain text file and store the sampling point information, generate the uniform resource locator of the application programming interface request to crawl the street view image data;

[0077] The street view image data is screened according to a preset definition, and the screened images are enhanced and corrected, and finally a street view image data set in the study area is obtained.

[0078] In a specific embodiment, the method for vectorizing and georeferencing road network data includes:

[0079] The road network data is imported into a preset geospatial platform (ArGIS), binarized and classified with a preset threshold, and then reclassified (Reclassify), and the values ​​of non-road classes are set to null values ​​(NoData), thereby simplifying the road network data into a pure road network raster map;

[0080] The ArcScan tool is used to extract the road centerline of the road network data and complete the vectorization processing; wherein the ArcScan tool is an extension module of ArGIS, which is used to convert the image into a vector feature layer.

[0081] Use the road network of the Open Street Map (OSM) as a reference to adjust the road network data and complete the georeferencing.

[0082] In a specific embodiment, the method of filtering the street view image data according to a preset definition, performing image enhancement and correction processing on the filtered images, and finally obtaining a street view image dataset in the research area includes:

[0083] Screening the street view image data to obtain a street view image that meets a preset definition;

[0084] Perform brightness and contrast enhancement processing on street view images that meet the preset definition;

[0085] The preset radial distortion model and tangential distortion model are used to correct the enhanced street view images.

[0086] In a specific embodiment, the radial distortion model is:

[0087] u′=u(1+k1r 2 +k2r 4 +k3r 6 )

[0088] v′=v(1+k1r 2 +k2r 4 +k3r 6 )

[0089] The tangential distortion model is:

[0090] u′=u+[2P1v+P2(R 2 +2u 2)]

[0091] v′=v+[P1(r 2 +2v 2 )+2P2u]

[0092] Among them, u and v are the horizontal and vertical coordinates of the image point, u′ and v′ are the horizontal and vertical coordinates of the image point after distortion correction, k1, k2, k3 are radial distortion coefficients, P1 and P2 are tangential distortion coefficients, and r is the distance from the center of the image to the coordinate point (u, ν).

[0093] S2: Input the street view image dataset into the trained deep learning recognition model to obtain an output image containing tree information;

[0094] In a specific embodiment, the training method of the deep learning recognition model includes:

[0095] Label different tree species in the street view image dataset;

[0096] Input the labeled street view image dataset into a preset deep learning recognition model for training to learn the features and patterns of the samples, wherein the deep learning recognition model is a YOLOv7 model;

[0097] The test set is used to evaluate the performance of the model, and the model with the best performance is retained to obtain a trained deep learning recognition model.

[0098] In a specific embodiment, the tree information includes:

[0099] The horizontal and vertical coordinates of the trees, the tree ID, the type of trees, and the number of trees.

[0100] S3: constructing a size reference model, and estimating the diameter at breast height and the height of the tree by combining the size reference model and the output image;

[0101] In a specific embodiment, a method for constructing a size reference model and estimating the diameter at breast height and the height of a tree by combining the size reference model and the output image includes:

[0102] Select a reference object, wherein the reference object includes at least one of the following:

[0103] Shared bicycles, bus stops, cars, road fire hydrants and sanitation barrels;

[0104] Determine the average size of the reference object through field research, and construct a corresponding size reference model according to the average size of the reference object;

[0105] The size reference model and the output image are combined, and a reference scale of the size reference model is set to estimate the diameter at breast height and the height of the tree.

[0106] S4: Match and summarize the tree information with the diameter at breast height and the tree height of the tree to obtain complete tree information corresponding to the tree.

[0107] In a specific embodiment, the method for identifying urban road tree information based on street view data includes:

[0108] According to the allometric growth equations of different tree species combined with the corresponding complete tree information, the carbon storage of all trees is calculated and statistically analyzed to obtain the total tree carbon sink.

[0109] In a specific embodiment, the carbon storage of all trees is calculated and counted according to the allometric growth equations of different tree species combined with the corresponding complete tree information to obtain the total carbon storage of trees, including:

[0110] Based on the complete tree information, statistical classification of tree species types is carried out;

[0111] Query the stock calculation formulas for different types of trees, the coefficients of the volume source biomass method, and the carbon storage calculation formula;

[0112] Select the corresponding formula according to the tree species, substitute the diameter at breast height D and tree height H into the volume formula, and calculate the vegetation volume V of the tree, V = aD b H c , where V is the vegetation stock, D is the tree diameter at breast height, and H is the tree height;

[0113] Using the volume source biomass method, the vegetation storage volume V is substituted into the formula to calculate the tree biomass B of the tree;

[0114] B=aV+b

[0115] Among them, a and b are constants, B is biomass, V is stock volume, and the values ​​of parameters a and b are related to tree species;

[0116] Multiply the calculated tree biomass B by the carbon content coefficient to obtain the carbon storage C of the tree;

[0117] C=B×C c

[0118] C is carbon storage, C c is the carbon content coefficient, B is the tree biomass;

[0119] Using the preset script, the complete tree information was combined with the above formula to calculate the carbon sink of road trees in the entire study area.

[0120] The present invention combines tree information with growth equations of different tree species to accurately calculate the carbon sink of each tree, providing scientific and reliable carbon sink estimation results. Through comprehensive calculation, the total carbon sink of the entire region can be obtained, providing important data support for environmental assessment and decision-making.

[0121] Example 2

[0122] Please refer to Figure 1 This embodiment provides a method for identifying tree information on urban roads based on street view data, the method comprising:

[0123] S1: Obtain road network data and preprocess it to obtain a street view image dataset in the study area;

[0124] In a specific embodiment, a method for obtaining road network data and preprocessing it to obtain a street view image dataset in a research area includes:

[0125] Download road network data;

[0126] Simplify, vectorize and georeference road network data;

[0127] At the preset image angle, sample the observation points and overlay the road angle value and road number information corresponding to each point. After completion, export and organize the sampling point data into a plain text file (CSV);

[0128] Specifically, the appropriate angle of the street view picture of the study area is selected, fov = 360° for a panoramic view;

[0129] Read the imported plain text file and store the sampling point information, generate the uniform resource locator of the application programming interface request to crawl the street view image data;

[0130] Specifically, Python is used to read and store the sampling point information from the imported CSV, and generate the uniform resource locator of the application programming interface request to crawl the street view image data;

[0131] The street view image data is screened according to a preset definition, and the screened images are enhanced and corrected, and finally a street view image data set in the study area is obtained.

[0132] In a specific embodiment, the method for vectorizing and georeferencing road network data includes:

[0133] The road network data is imported into a preset geospatial platform (ArGIS), binarized and classified with a preset threshold, and then reclassified (Reclassify), and the values ​​of non-road classes are set to null values ​​(NoData), thereby simplifying the road network data into a pure road network raster map;

[0134] The ArcScan tool is used to extract the road centerline of the road network data and complete the vectorization processing; wherein the ArcScan tool is an extension module of ArGIS, which is used to convert the image into a vector feature layer.

[0135] Use the road network of the Open Street Map (OSM) as a reference to adjust the road network data and complete the georeferencing.

[0136] In a specific embodiment, the method of filtering the street view image data according to a preset definition, performing image enhancement and correction processing on the filtered images, and finally obtaining a street view image dataset in the research area includes:

[0137] Screening the street view image data to obtain a street view image that meets a preset definition;

[0138] Perform brightness and contrast enhancement processing on street view images that meet the preset definition;

[0139] The preset radial distortion model and tangential distortion model are used to correct the enhanced street view images.

[0140] In a specific embodiment, the radial distortion model is:

[0141] u′=u(1+k1r 2 +k2r 4 +k3r 6 )

[0142] v′=v(1+k1r 2 +k2r 4 +k3r 6 )

[0143] The tangential distortion model is:

[0144] u′=u+[2P1v+P2(r 2 +2u 2 )]

[0145] v′=v+[P1(r 2 +2v 2 )+2P2u]

[0146] Among them, u and v are the horizontal and vertical coordinates of the image point, u′ and v′ are the horizontal and vertical coordinates of the image point after distortion correction, k1, k2, k3 are radial distortion coefficients, P1 and P2 are tangential distortion coefficients, and r is the distance from the center of the image to the coordinate point (u, v);

[0147] Combining the above two formulas, the pincushion or barrel distortion caused by the 360° panoramic camera of the street view car is corrected, and the influence of the distortion on the measurement of the tree diameter at breast height and tree height is reduced.

[0148] S2: Input the street view image dataset into the trained deep learning recognition model to obtain an output image containing tree information;

[0149] In a specific embodiment, the training method of the deep learning recognition model includes:

[0150] Label different tree species in the street view image dataset;

[0151] Input the labeled street view image dataset into a preset deep learning recognition model for training to learn the features and patterns of the samples, wherein the deep learning recognition model is a YOLOv7 model;

[0152] Using the test set to evaluate the performance of the model, retaining the model with the best performance, and obtaining a trained deep learning recognition model;

[0153] Specifically, the method of inputting the labeled street view image dataset into a preset deep learning recognition model for training to learn the features and patterns of the samples, wherein the deep learning recognition model is a YOLOv7 model, includes:

[0154] Configure the relevant deep learning environment, install CUDA, download the Pytorch deep learning environment, download YOLOv7 and build the dependencies required by the YOLOv7 model;

[0155] Please refer to Figure 2 , use the Make Sense tool to label the different tree species that need to be studied in the street view image dataset;

[0156] Create a new VOCdata folder, which includes the Annotations folder for storing xml label files, the dataSet_path folder is the street view image dataset after xml conversion to YOLO format, the images folder is used to store the dataset, and the labels is the YOLO format dataset;

[0157] Create a split_train_val.py file in the VOCdata folder to divide the data set;

[0158] Create a new xml_to_yolo.py in the VOCdata folder to convert XML into txt format;

[0159] Create a new myvoc.yaml file in the data folder to configure the street view image dataset.

[0160] Train the model on the selected images.

[0161] In a specific embodiment, the tree information includes:

[0162] The horizontal and vertical coordinates of the trees, the tree ID, the type of trees, and the number of trees.

[0163] S3: Please refer to Figure 3 , constructing a size reference model, and estimating the diameter at breast height and the height of the tree by combining the size reference model and the output image;

[0164] In a specific embodiment, a method for constructing a size reference model and estimating the diameter at breast height and the height of a tree by combining the size reference model and the output image includes:

[0165] Select a reference object, wherein the reference object includes at least one of the following:

[0166] Shared bicycles, bus stops, cars, road fire hydrants and sanitation barrels;

[0167] Determine the average size of the reference object through field research, and construct a corresponding size reference model according to the average size of the reference object;

[0168] Combining the size reference model and the output image, and setting a reference scale of the size reference model, to estimate the diameter at breast height and the height of the tree;

[0169] Specifically, the reference standards of the selected reference objects are as follows: the average height of bicycle tire size is 65 cm, the average height of bus platform is 270 cm, the average height of moving cars is 160 cm, the height of roadside power boxes is 140 cm, the height of plastic sanitation barrels is 100 cm, the height of stainless steel sanitation barrels is 90 cm, and the height of tree whitewashing is 110 cm;

[0170] More specifically, ImageJ software was used to calculate the diameter at breast height and tree height of the trees.

[0171] S4: matching and summarizing the tree information with the diameter at breast height and the tree height of the tree to obtain complete tree information corresponding to the tree;

[0172] Specifically, the trees identified by YOLOv7 have unique horizontal and vertical coordinates in the image, and the name of each image contains the unique longitude and latitude information of the location where it was taken;

[0173] Match the longitude, latitude and vertical coordinates of the trees identified by YOLOv7 with the tree diameter at breast height and tree height calculated by ImageJ;

[0174] Import the matched information into the same CSV file, which at least includes the longitude and latitude of the image, the horizontal and vertical coordinates of the trees in the image, the tree ID, the tree's diameter at breast height, the tree height, the tree type, and the number of trees.

[0175] In a specific embodiment, the method for identifying urban road tree information based on street view data includes:

[0176] According to the allometric growth equations of different tree species combined with the corresponding complete tree information, the carbon storage of all trees is calculated and statistically analyzed to obtain the total tree carbon sink.

[0177] In a specific embodiment, the carbon storage of all trees is calculated and counted according to the allometric growth equations of different tree species combined with the corresponding complete tree information to obtain the total carbon storage of trees, including:

[0178] Based on the complete tree information, statistical classification of tree species types is carried out;

[0179] Query the stock calculation formulas for different types of trees, the coefficients of the volume source biomass method, and the carbon storage calculation formula;

[0180] Select the corresponding formula according to the tree species, substitute the diameter at breast height D and tree height H into the volume formula, and calculate the vegetation volume V of the tree, V = aD b H c , where V is the vegetation stock, D is the tree diameter at breast height, and H is the tree height;

[0181] Using the volume source biomass method, the vegetation storage volume V is substituted into the formula to calculate the tree biomass B of the tree;

[0182] B=aV+b

[0183] Among them, a and b are constants, B is biomass, V is stock volume, and the values ​​of parameters a and b are related to tree species;

[0184] Multiply the calculated tree biomass B by the carbon content coefficient to obtain the carbon storage C of the tree;

[0185] C=B×C c

[0186] C is carbon storage, C c is the carbon content coefficient, B is the tree biomass;

[0187] Using the preset script, the complete tree information was combined with the above formula to calculate the carbon sink of road trees in the entire study area.

[0188] Specifically, the tree species types are statistically classified based on the obtained tree specific information CSV file;

[0189] Query the stock calculation formulas for different types of trees, the coefficients of the volume source biomass method, and the carbon storage calculation formula;

[0190] Select the corresponding formula according to the tree species, substitute the diameter at breast height D and tree height H into the stock volume formula, and calculate the vegetation stock volume V of the tree;

[0191] Soft broadleaved forest V=0.0000674286D 1.87657 H 0.92888

[0192] Hard broad-leaved forest V = 0.0000601228D 1.87750 H 0.98496

[0193] Where V is the vegetation storage (m 3 ), D is the tree diameter at breast height (cm), H is the tree height (m);

[0194] Use the volume source biomass method to bring the vegetation storage volume V into the formula to calculate the tree biomass B of the tree;

[0195] B=aV+b

[0196] Among them, the parameter a of soft broadleaf is 0.4754, parameter b is 30.6034; the parameter a of hard broadleaf is 0.7564, parameter b is 8.3103; a and b are both constants, B is biomass, V is stock volume, and the values ​​of parameters a and b are related to tree species;

[0197] The calculated tree biomass B is multiplied by the carbon content coefficient to obtain the tree's carbon storage C. The carbon content coefficient mainly has two parameters, 0.45 and 0.5. Domestic scholars often choose 0.5. In order to improve the calculation accuracy of carbon storage, these two coefficients can be combined for cross calculation.

[0198] C=B×C c

[0199] Where, C is carbon storage (t), C c is the carbon content, B is the tree biomass;

[0200] Use the relevant python script to import the obtained tree information CSV file, and combine it with the above formula to calculate the carbon sink of road trees in the entire study area.

[0201] The present invention accurately calculates the carbon sink of each tree through the growth equation of different tree species, providing scientific and reliable carbon sink estimation results. Through comprehensive calculation, the carbon sink of the entire region can be obtained, providing important data support for environmental assessment and decision-making.

[0202] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A method for identifying tree information on urban roads based on street view data, characterized in that: The method comprises: Obtain road network data and preprocess it to obtain a street view image dataset in the study area; Input the street view image dataset into the trained deep learning recognition model to obtain the output image containing tree information; Constructing a size reference model, and estimating the diameter at breast height and the height of the tree by combining the size reference model and the output image; The tree information is matched and summarized with the tree's diameter at breast height and tree height to obtain complete tree information corresponding to the tree.

2. The method for identifying urban road tree information based on street view data according to claim 1, characterized in that: The method of obtaining the road network data and preprocessing it to obtain the street view image dataset in the study area includes: Download road network data; Simplify, vectorize and georeference road network data; At the preset image angle, sample the observation points and overlay the road angle value and road number information corresponding to each point. After completion, export and organize the sampling point data into a plain text file; Read the imported plain text file and store the sampling point information, generate the uniform resource locator of the application programming interface request to crawl the street view image data; The street view image data is screened according to a preset definition, and the screened images are enhanced and corrected, and finally a street view image data set in the study area is obtained.

3. The method for identifying urban road tree information based on street view data according to claim 2, characterized in that: Methods for vectorizing and georeferencing road network data include: Importing the road network data into a preset geospatial platform, performing binary classification with a preset threshold, and then reclassifying, setting the values ​​of non-road categories to null values, thereby simplifying the road network data into a pure road network grid map; Use ArcScan tool to extract the road centerline of the road network data and complete the vectorization process; Use the road network of the Open Street Map as a reference to adjust the road network data and complete the georeferencing.

4. The method for identifying urban road tree information based on street view data according to claim 3, characterized in that: The method of filtering the street view image data according to a preset definition, performing image enhancement and correction processing on the filtered images, and finally obtaining a street view image data set in the research area includes: Screening the street view image data to obtain a street view image that meets a preset definition; Perform brightness and contrast enhancement processing on street view images that meet the preset definition; The preset radial distortion model and tangential distortion model are used to correct the enhanced street view images.

5. The method for identifying urban road tree information based on street view data according to claim 4, characterized in that: The radial distortion model is: u′=u(1+k1r 2 +k2r 4 +k3r 6 ) v′=v(1+k1r 2 +k2r 4 +k3r 6 ) The tangential distortion model is: u′=u+[2P1v+P2(r 2 +2u 2 )] v′=ν+[P1(r 2 +2v 2 )+2P2u] Among them, u and v are the horizontal and vertical coordinates of the image point, u′ and v′ are the horizontal and vertical coordinates of the image point after distortion correction, k1, k2 and k3 are radial distortion coefficients, P1 and P2 are tangential distortion coefficients, and r is the distance from the center of the image to the coordinate point (u, v).

6. The method for identifying urban road tree information based on street view data according to claim 1, characterized in that: The training method of the deep learning recognition model includes: Label different tree species in the street view image dataset; Input the labeled street view image dataset into a preset deep learning recognition model for training to learn the features and patterns of the samples, wherein the deep learning recognition model is a YOLOv7 model; The test set is used to evaluate the performance of the model, and the model with the best performance is retained to obtain a trained deep learning recognition model.

7. The method for identifying urban road tree information based on street view data according to claim 1, characterized in that: The tree information includes: The horizontal and vertical coordinates of the trees, the tree ID, the type of trees, and the number of trees.

8. The method for identifying urban road tree information based on street view data according to claim 1, characterized in that: The method of constructing a size reference model and estimating the diameter at breast height and the height of a tree by combining the size reference model and the output image comprises: Select a reference object, wherein the reference object includes at least one of the following: Shared bicycles, bus stops, cars, road fire hydrants and sanitation barrels; Determine the average size of the reference object through field research, and construct a corresponding size reference model according to the average size of the reference object; The size reference model and the output image are combined, and a reference scale of the size reference model is set to estimate the diameter at breast height and the height of the tree.

9. The method for identifying urban road tree information based on street view data according to claim 1, characterized in that: The method further comprises: According to the allometric growth equations of different tree species combined with the corresponding complete tree information, the carbon storage of all trees is calculated and statistically analyzed to obtain the total tree carbon sink.

10. The method for identifying urban road tree information based on street view data according to any one of claims 1 to 9, characterized in that: We can also calculate and count the carbon storage of all trees based on the allometric growth equations of different tree species combined with the corresponding complete tree information to obtain the total carbon storage of trees. The specific method is as follows: Based on the complete tree information, statistical classification of tree species types is carried out; Query the stock calculation formulas for different types of trees, the coefficients of the volume source biomass method, and the carbon storage calculation formula; Select the corresponding formula according to the tree species, substitute the diameter at breast height D and tree height H into the volume formula, and calculate the vegetation volume V of the tree, V = aD b H c , where V is the vegetation stock, D is the tree diameter at breast height, and H is the tree height; Using the volume source biomass method, the vegetation storage volume V is substituted into the formula to calculate the tree biomass B of the tree; B=aV+b Among them, a and b are constants, B is biomass, V is stock volume, and the values ​​of parameters a and b are related to tree species; Multiply the calculated tree biomass B by the carbon content coefficient to obtain the carbon storage C of the tree; C=B×C c C is carbon storage, C c is the carbon content coefficient, B is the tree biomass; Using the preset script, the complete tree information was combined with the above formula to calculate the carbon sink of road trees in the entire study area.

Images