A sidewalk tree rapid mapping, positioning and statistics method based on handheld slam
By combining handheld SLAM devices with improved point cloud processing technology, the problems of low efficiency and large identification errors in urban street tree mapping have been solved. This has enabled efficient and accurate acquisition of 3D information of street trees and automatic extraction of multi-dimensional parameters, supporting greening management.
Patent Information
- Application Number
- CN202610798006.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-04
- Publication Date
- 2026-08-25
AI Technical Summary
Existing technologies are inefficient in urban street tree mapping, are greatly affected by environmental interference, and cannot quickly obtain three-dimensional information. Furthermore, general SLAM and forestry single tree segmentation algorithms have poor adaptability and are prone to misidentifying street lamp poles as trees.
A handheld SLAM device is used to walk along the sidewalk edge. Combined with voxel grid downsampling and statistical filtering, point cloud processing is performed using the features of the roadside trees. Improved Euclidean clustering with distance threshold and height hierarchies is used to separate ground points from non-tree features, calculate tree trunk coordinates, and generate structured tables.
It enables efficient and accurate street tree surveying and statistics in urban street environments, improves field data collection efficiency by 80%, reduces the impact of dynamic interference, improves recognition accuracy, provides automatic extraction of multi-dimensional parameters, and supports greening management.
Smart Images
Figure CN122637217A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of urban greening maintenance and management technology, specifically a method for rapid mapping, positioning, and statistical analysis of roadside trees based on handheld SLAM. Background Technology
[0002] Traditional methods for locating and counting roadside trees require total stations or RTKs to measure each tree individually. However, due to the significant impact of GNSS signal obstruction in urban areas (high-rise buildings, tree shade), it is difficult to quickly obtain three-dimensional morphology, resulting in low efficiency. Moreover, existing handheld SLAMs are mostly designed for indoor or general terrain. In the complex environment of urban streets (dynamic interference from vehicles and pedestrians), the point cloud noise is high, making it unsuitable for accurate roadside tree counting. General SLAMs are also unsuitable. Furthermore, existing forestry algorithms are incompatible. Single-tree segmentation algorithms are mostly designed for natural forests and do not consider the characteristics of roadside trees such as "linear arrangement, regular spacing, and proximity to municipal facilities." This can easily lead to misidentification of streetlights and traffic signs as trees, resulting in significant errors. Summary of the Invention
[0003] To address the shortcomings of existing technologies, this invention provides a method for rapid mapping, localization, and statistical analysis of roadside trees based on handheld SLAM. The invention aims to solve the problems of low efficiency, high susceptibility to environmental interference, inability to quickly acquire 3D tree information, poor adaptability of general SLAM and forestry single-tree segmentation algorithms in urban street scenes, and susceptibility to misidentification in existing roadside tree mapping methods. The goal is to achieve efficient, accurate, and automated mapping, localization, and statistical analysis of roadside trees.
[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0005] A method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM includes the following steps:
[0006] S1. Data Acquisition: The operator walks along the outer edge of the sidewalk, uses a handheld SLAM scanner and keeps the device facing one side of the road to ensure that the field of view covers the trunks and crowns of the roadside trees, and acquires the trunk ring point cloud to complete the mapping.
[0007] S2. Target Extraction: Voxel grid downsampling and statistical filtering are used to remove outlier noise. An improved Euclidean clustering method combining distance threshold and height stratification is adopted. Taking advantage of the characteristic that the trunks of roadside trees are usually 1.3 meters above the ground and are vertical poles, height constraints are set to separate ground points from non-tree features. Principal component analysis of point cloud clusters is calculated.
[0008] S3. Positioning and Statistics: Extract the projection of individual tree point cloud clusters onto the horizontal plane, calculate their geometric center as the trunk plane coordinates (X, Y), combine the trajectory data to assign them an absolute coordinate system, automatically number the extracted trees according to the SLAM trajectory progression, calculate the diameter at breast height (DBH), tree height, and crown width, and generate a structured table.
[0009] Preferably, in step S1, a small-range closed-loop walk is performed at the start and end of each road segment to constrain SLAM drift.
[0010] Preferably, in step S1, the trunk of each target tree is briefly stopped on the front and both sides to obtain a complete trunk ring point cloud.
[0011] Preferably, in step S2, the height of the tree trunk above the ground is 1.3 meters, which is the tree diameter at breast height (DBH) measurement height.
[0012] Preferably, in step S2, if the linearity of the first principal component is higher than the threshold and its height matches the characteristics of a street lamp pole, it is filtered out to avoid misclassifying the street lamp pole as a tree.
[0013] Preferably, in step S3, the trees are numbered using the road name plus a number.
[0014] Preferably, in step S3, the tree diameter at breast height is the diameter of the fitted circle of the point cloud cross section at a height of 1.3 meters, the tree height is the Z-axis range of the point cloud, and the crown width is the minimum bounding rectangle of the horizontal projection of the crown.
[0015] Preferably, in step S3, the structured table includes tree number, planar coordinates, diameter at breast height (DBH), tree height, crown width, and road name.
[0016] Compared with the prior art, the present invention has the following beneficial effects:
[0017] (1) Compared with traditional tree-by-tree measurement, continuous data collection can be carried out while walking, and three-dimensional information of all roadside trees can be obtained at once. It does not rely on GNSS signals and can be used smoothly in densely shaded areas. The efficiency of field data collection is increased by 80%, which significantly improves the efficiency of operation.
[0018] (2) A “walking along the edge line + multi-view supplementary scanning” acquisition path for urban streets was proposed, and a scenario-based acquisition SOP was developed to solve the problem of SLAM accumulating errors in long straight-line environments.
[0019] (3) By filtering out dynamic interference and enhancing the features of road trees, the impact of dynamic objects such as vehicles and pedestrians on the quality of point clouds is reduced, adapting to complex street scenes and improving the modeling stability and accuracy in urban environments.
[0020] (4) By utilizing the linear arrangement pattern and morphological characteristics of roadside trees, trees can be effectively distinguished from similar features such as streetlights and traffic signs, resulting in high recognition accuracy and reducing misidentification and omission.
[0021] (5) Cloud processing is integrated with greening maintenance business. While obtaining the location of a single tree, it automatically extracts multi-dimensional parameters such as tree height, diameter at breast height, and crown width, and performs business-oriented data processing to provide refined data support for greening management, tree health monitoring, and green volume calculation. Attached Figure Description
[0022] Figure 1 This is an overall flowchart of the present invention;
[0023] Figure 2 This invention uses a scenario-based acquisition of SOP side views;
[0024] Figure 3 This invention provides a scenario-based SOP top view for data acquisition.
[0025] Figure 4 This is a functional block diagram of the business data processing instruction processing pipeline of the present invention;
[0026] Figure 5 This is a schematic diagram illustrating the output of the business results of this invention.
[0027] In the image: 1. Handheld SLAM device; 2. Worker; 3. Tree trunk; 4. Tree canopy; 5. Curbstone; 6. Recommended data collection route; Detailed Implementation
[0028] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0029] like Figure 1-5 As shown, this invention provides a technical solution: a method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM, comprising the following steps:
[0030] S1. Data Acquisition: The operator walks along the outer edge of the sidewalk. The recommended horizontal distance d between the walking path and the tree trunk column is 2 to 3 meters. The "Qiyu Innovation Lingguang L2" handheld SLAM scanner, which is available on the market, is used. The handheld SLAM (Simultaneous Localization and Mapping) scanner is used and the device is kept facing the side of the road. The vertical distance from the ground to the center of the device is 1.2 to 1.5 meters. The device scanning tilt angle α (the angle between the device's central axis and the horizontal plane) is -10° to +10° to ensure that the field of view covers the trunks and crowns of the roadside trees.
[0031] Small-scale closed-loop walking is performed at the beginning and end of each road segment to constrain SLAM drift. During this process, the front and sides of the trunk of each target tree are briefly stopped to obtain a complete trunk ring point cloud and complete the mapping.
[0032] S2. Target Extraction: Voxel grid downsampling and statistical filtering are used to remove outliers and noise. An improved Euclidean clustering method combining distance threshold and height stratification is adopted. Taking advantage of the characteristic that the trunks of roadside trees are usually 1.3 meters above the ground and are vertical poles, a height constraint is set, where the 1.3-meter height of the tree trunk above the ground is the tree diameter at breast height (DBH). Ground points are separated from non-tree features. Principal component analysis (PCA) is calculated for the point cloud clusters. If the linearity of the first principal component is higher than the threshold and the height matches the characteristics of a street lamp pole, it is filtered out to avoid misclassifying street lamp poles as trees.
[0033] S3. Positioning and Statistics: Extract the projection of individual tree point cloud clusters onto the horizontal plane, calculate their geometric center as the trunk plane coordinates (X, Y), and assign them an absolute coordinate system based on the trajectory data. According to the SLAM trajectory's progression, automatically number the extracted trees, using the road name plus a number, such as "Road Name_001". Calculate the diameter at breast height (DBH), tree height, and crown width, and generate a structured table. The DBH is the diameter of the fitted circle of the point cloud cross-section at a height of 1.3 meters, the tree height is the Z-axis range of the point cloud, and the crown width is the minimum bounding rectangle of the horizontal projection of the crown. The structured table includes the tree number, plane coordinates, DBH, tree height, crown width, and road name.
[0034] The statistical breakdown of business results is shown in the table below:
[0035] Tree number X-coordinate (m) Y coordinate (m) Tree height (m) Diameter at breast height (m) Crown width - East-West (m) Crown width - North-South (m) Front plant spacing (m) state 001 0.0 2.0 5.8 0.35 4.2 4.0 —— normal 002 5.0 2.0 8.7 0.36 4.3 4.1 5.0 normal 003 10.0 2.0 8.2 0.33 4.0 3.9 5.0 normal 004 15.0 2.0 8.9 0.37 4.4 4.2 5.0 normal 005 20.0 2.0 8.4 0.34 4.1 4.0 5.0 normal
[0036] The road runs east-west. The X coordinate is along the road direction (east is positive), and the Y coordinate is the distance from the center of the tree to the curb. The roadside trees are planted at equal intervals of 5m, and the Y coordinates are the same. Tree number 001 has no predecessor and is marked as "—".
[0037] Based on a handheld SLAM-based method for rapid mapping, localization, and statistical analysis of street trees, a set of scenario-based standard operating procedures (SOPs) for handheld SLAM data acquisition in urban street environments is defined. The SOPs include: path planning rules before acquisition (requiring walking along one side of the street tree column while maintaining a relatively constant distance and speed), equipment operation specifications during acquisition (equipment posture and scanning angle requirements to optimize point cloud acquisition in the tree trunk area), and real-time data quality checks and re-acquisition rules (if multiple consecutive trees are found to have missing or distorted point clouds, fragments need to be re-acquired on-site according to the rules).
[0038] Implementing scenario-based standard operating procedures is crucial to ensuring data quality consistency. The specific implementation is as follows:
[0039] 1. Pre-collection planning: Mark the road section to be collected on the electronic map. The planning principle is to move along the sidewalk on one side of the row of trees to ensure that the scanning fan of the equipment can cover the entire row of trees.
[0040] 2. On-site data collection operation:
[0041] 1) The operator 2 holds a handheld SLAM device 1 that integrates LiDAR, IMU and visual sensors to scan the lower part of the tree trunk and the canopy in a balanced manner;
[0042] 2) Proceed along the recommended collection path 6, which is parallel to the curb 5 and maintains a horizontal distance d of 2 to 3 meters from the center line of the tree trunks. Walk at a constant speed and smoothly, avoiding violent shaking.
[0043] 3) The main scanning plane of the SLAM device 1 should be basically perpendicular to the street direction, and the tilt angle α (the angle between the device and the horizontal plane) should be finely adjusted within -10° to +10° to ensure that the laser can effectively scan the trunk of the roadside tree 3 from the root to the main branch;
[0044] 4) The device software displays the coverage of the collected point cloud in real time. If a single tree's point cloud is severely missing due to a brief obstruction, or if multiple trees are missing, the segment must be re-collected on-site immediately by retracing the original path or slowly to ensure the integrity of the point cloud for each target tree trunk 3.
[0045] The handheld SLAM-based method for rapid mapping, localization, and statistics of roadside trees also defines an automated processing flow from raw SLAM data to greening management business data. This flow includes: scene-based preprocessing (based on prior knowledge of street scenes, such as linear arrangement and height / reflectivity of common features, for targeted filtering and enhancement), business object identification and segmentation (using the business logic of "roadside trees", such as relative position with curbs and regular spacing constraints, to distinguish trees from streetlights, signs, etc.), business parameter extraction and mapping (according to greening management business requirements, automatically calculating and extracting "tree number, coordinate X, coordinate Y, diameter at breast height, tree height, crown width, and road name" information from single tree point clouds), and one-click generation of business results (automatically associating spatial location and attribute data to generate statistical reports and two-dimensional result maps that conform to industry standards).
[0046] The specific implementation of the business data processing workflow is as follows:
[0047] The collected raw point cloud data 1 is imported into the built-in processing software and enters the automated instruction processing pipeline;
[0048] Data preprocessing and scene enhancement module:
[0049] 1) Dynamic interference filtering: Using spatiotemporal consistency analysis of point clouds in multiple consecutive frames, the point clouds of moving vehicles, pedestrians and other dynamic objects are identified and removed. For temporarily stationary vehicles, their point cloud features (size, shape and typical road tree model) are compared and filtered out.
[0050] 2) Ground segmentation: Improved cloth simulation filtering (CSF) and other algorithms are used to quickly separate ground point clouds from non-ground point clouds. The extracted ground point cloud is used for subsequent calculation of curb stones and definition of reference planes.
[0051] 3) Enhanced separation of pole-shaped features (305): Based on the ground segmentation results and point cloud features, all vertical pole-shaped features (including trees, street lamp poles, and sign poles) are extracted. The point cloud reflection intensity distribution, cross-sectional geometric features (such as roundness and radius variation) and top structure of each pole-shaped target are calculated. The trunks of roadside trees usually have a specific range of reflection intensity and irregular cross-sectional contours, while the reflection intensity features of metal pole-shaped features are different and the shapes are regular. Based on this, preliminary separation and marking are carried out.
[0052] Business object identification and segmentation module:
[0053] 1) Initial screening based on spatial distribution rules: Using the business rule that roadside trees are arranged linearly and approximately equally spaced along the road, Hough transform line detection or cluster analysis is performed on the pole-shaped targets retained after preprocessing on the top plane to identify the linear target sequence arranged along the curb with a spacing within a certain reasonable threshold (such as 4-8 meters). Targets that deviate too far from the sequence or have abnormal spacing (such as a single utility pole) are initially excluded.
[0054] 2) Fine segmentation based on morphology and reflection features: For the candidate targets initially screened in the previous step, fine segmentation is performed using the results of enhanced separation of rod-shaped ground features, with a focus on comparing tree trunks and streetlight poles.
[0055] a) Geometric shape: The cylindrical model that fits the point cloud of the tree trunk may have a radius that varies along the height and has an uneven surface; while the street lamp pole is usually a regular cylinder with a constant radius.
[0056] b) Top structure: The top of the tree trunk point cloud connects to the point cloud clusters of the tree crown (presenting an irregular clump), while the top of the street lamp pole connects to the regular geometric point cloud of the lamp.
[0057] c) Spatial relationships: Combining point cloud intensity information, each independent street tree point cloud is accurately segmented using preset rules or a lightweight classification model.
[0058] Business parameter extraction module:
[0059] 1) Calculation of geometric parameters of a single tree:
[0060] a) Tree height: Calculate the vertical height difference between the highest point in the single tree point cloud and the ground point (obtained by interpolation of the ground point cloud near the tree trunk);
[0061] b) Diameter at breast height (DBH): Extract a point cloud slice of the trunk cross section at a height of 1.3 meters above the ground from the point cloud, fit a circle using the least squares method, and its diameter is the DBH.
[0062] c) Crown width: Calculate the projection of the canopy point cloud onto the horizontal plane, fit its minimum bounding rectangle or convex hull, and take its maximum diameter or east-west or north-south width as the crown width;
[0063] 2) Spatial location and relationship calculation;
[0064] a) Location (coordinates and numbering): The three-dimensional coordinates of the center point of the bottom of the tree trunk point cloud are used as the location of the tree, and it can be automatically numbered in sequence according to the acquisition direction;
[0065] b) Continuous tree spacing: Calculate the horizontal distance between the positions of two adjacent trees;
[0066] c) Missing Tree Location: In the identified tree sequence, if no tree point cloud is detected at a location where a tree is expected to be present (inferred from the average tree spacing), and the location meets the tree planting conditions (such as not being on a manhole cover or entrance), then the location is marked as a "missing tree" point and its coordinates are recorded.
[0067] Results packaging module:
[0068] This module integrates all the extracted information in a structured way and generates a business deliverable package with one click.
[0069] 1) Generate a structured business data table: The table fields should include at least: tree number, coordinates (X, Y), tree height, diameter at breast height, crown width (east-west / north-south), spacing between trees, status (normal / missing trees / suspected disease), etc. This table can be directly imported into the greening management system.
[0070] 2) Generate a locatable 3D result map: In the overall point cloud scene map, mark the position of each identified tree with eye-catching symbols (such as dots) and numbers, and display the tree height and diameter at breast height as labels. Missing trees are also marked with special symbols.
[0071] 3) Output business report: Automatically generate a brief report containing the total number of trees, a list of missing tree locations, and key parameter statistics (such as average tree height and diameter at breast height).
[0072] This invention begins with "pre-collection planning," which plans the collection path based on task segment information. The core execution stage is "scenario-based standard collection," where operators strictly follow specifications to collect field data. The collected raw data then enters the "business-oriented data processing" stage, where it is automatically processed by the instruction processing pipeline, ultimately outputting the system's business results. Figure 5 The data package shown can be directly used for green space maintenance and management.
[0073] It should be noted that, in this document, terms such as “comprising,” “including,” or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0074] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM, characterized in that: Includes the following steps: S1. Data Acquisition: The operator walks along the outer edge of the sidewalk, uses a handheld SLAM scanner and keeps the device facing one side of the road to ensure that the field of view covers the trunks and crowns of the roadside trees, and acquires the trunk ring point cloud to complete the mapping. S2. Target Extraction: Voxel grid downsampling and statistical filtering are used to remove outlier noise. An improved Euclidean clustering method combining distance threshold and height stratification is adopted. Taking advantage of the characteristic that the trunks of roadside trees are usually 1.3 meters above the ground and are vertical poles, height constraints are set to separate ground points from non-tree features. Principal component analysis of point cloud clusters is calculated. S3. Positioning and Statistics: Extract the projection of individual tree point cloud clusters onto the horizontal plane, calculate their geometric center as the trunk plane coordinates (X, Y), combine the trajectory data to assign them an absolute coordinate system, automatically number the extracted trees according to the SLAM trajectory progression, calculate the diameter at breast height (DBH), tree height, and crown width, and generate a structured table.
2. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S1, a small-scale closed-loop walk is performed at the start and end of each road segment to constrain SLAM drift.
3. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S1, briefly pause on the front and sides of the trunk of each target tree to obtain a complete trunk ring point cloud.
4. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S2, the height of the tree trunk above the ground, which is 1.3 meters, is the tree diameter at breast height (DBH) measurement height.
5. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S2, if the linearity of the first principal component is higher than the threshold and its height matches the characteristics of a street lamp pole, it is filtered out to avoid misclassifying the street lamp pole as a tree.
6. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S3, the trees are numbered using the road name plus a number.
7. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S3, the tree diameter at breast height is the diameter of the fitted circle of the point cloud section at a height of 1.3 meters, the tree height is the Z-axis range of the point cloud, and the crown width is the minimum bounding rectangle of the horizontal projection of the crown.
8. The method for rapid mapping, localization, and statistical analysis of road trees based on handheld SLAM according to claim 1, characterized in that: In step S3, the structured table contains tree number, planar coordinates, diameter at breast height (DBH), tree height, crown width, and road name.