Tree dynamic coding method and system based on high-precision positioning and visual analysis

Abstract

The invention discloses a tree dynamic coding method and system based on high-precision positioning and visual analysis. The method comprises the following steps: acquiring pose data and an image; calculating geographic coordinates of the tree; preprocessing the image; extracting tree ontology features and environmental reference features, and establishing a corresponding relationship between the two features; determining whether the tree is a coded tree or not based on the geographic coordinates of the tree, the tree ontology characteristics and the environmental reference object characteristics; if not, an identity code of the tree is generated, and the identity code of the tree comprises a basic code and a dynamic extension code; the basic codes comprise codes corresponding to geographic coordinates and serial numbers; the dynamic extension code comprises a state code used for representing the current operation and maintenance state of the tree and a version code used for recording the number of updating times of the robust template. By implementing the method provided by the invention, the stable, unique and traceable digital identity of the single tree in the whole life cycle can be established, and the defects of the existing method in the aspects of reliability, uniqueness and long-term continuity are overcome.

Description

TECHNICAL FIELD

[0001] The present application relates to computer vision, and more particularly to a tree dynamic coding method and system based on high-precision positioning and visual analysis. BACKGROUND

[0002] With the rapid advancement of urban greening construction, the number of street trees and landscape green plants has increased significantly. How to achieve long-term stable digital identity recognition and full life cycle management of individual trees has become a key challenge for fine management. Current technical solutions mainly fall into two categories: physical identification methods and remote sensing and image recognition-based methods, but both have obvious limitations.

[0003] Firstly, physical identification methods usually involve hanging or fixing identification plates (such as nameplates, RFID tags, QR codes, etc.) on trees. For example, some areas use QR code tags on tree trunks to facilitate public access to tree information. Although this method is simple and direct at the initial stage of implementation and is convenient for information association, in the complex natural environment outdoors, physical identification is easily affected by weathering, aging, falling off or human damage, and is difficult to be attached to trees stably for a long time. In addition, a large number of hanging identification plates not only affect the city's appearance, but also may be embedded in the tree due to the thickening of the tree trunk, causing damage to the tree health, and cannot adapt to the changes of tree morphology. At the same time, this method has high labor cost and needs regular inspection and maintenance.

[0004] Secondly, remote sensing and image recognition-based methods mainly use aerial images, satellite remote sensing or unmanned aerial vehicle aerial photography for large-scale tree detection and identification, for example, a tree crown recognition method based on pixel-level classifier and template matching. However, the original design of these technologies is for population census and tree species classification, not individual identity management. The core limitation is the lack of individual uniqueness, which can only identify the tree crown or determine the tree species, and cannot distinguish different individuals of the same species; the positioning accuracy is low, which is difficult to meet the precise needs in urban environment; and the visual feature template is static, which is difficult to cope with the changes in tree growth, seasonal replacement, pruning and other factors, and cannot handle the impact of changes in the surrounding environment, so it is difficult to establish a long-term and stable one-to-one identity association.

[0005] In summary, the existing tree identity management technology has not solved the core problem of how to give and maintain a lifelong unique, physical carrier-free and self-adaptive digital identity for dynamically growing individual trees. Specifically, the physical identification method is easy to damage and fall off, lacks long-term reliability, and has high labor maintenance cost; the method relying only on geographic coordinates is difficult to ensure the uniqueness and stability of the identification in complex environment or when trees are transplanted; the method relying only on image features is easily affected by the natural growth of trees and external environmental changes, leading to feature drift, making it difficult to achieve long-term consistency of identity.

[0006] Therefore, it is necessary to design a new method to realize the establishment of a stable, unique and traceable digital identity of a single tree in the whole life cycle, and overcome the shortcomings of the existing method in reliability, uniqueness and long-term continuity. SUMMARY

[0007] The purpose of the present application is to overcome the defects of the prior art and provide a tree dynamic coding method and system based on high-precision positioning and visual analysis.

[0008] To achieve the above-mentioned purpose, the present application adopts the following technical solutions: a tree dynamic coding method based on high-precision positioning and visual analysis, comprising: Obtaining the pose data and images collected by the device installed on the vehicle; Calculating the geographic coordinates of the tree based on the pose data; Preprocessing the images to obtain preprocessed images; Extracting tree body features and environmental reference features based on the preprocessed images, and establishing a correspondence between the tree body features and the environmental reference features; Determining whether the tree is an already coded tree based on the geographic coordinates of the tree, the tree body features and the environmental reference features; If the tree is not an already coded tree, generating an identity code of the tree, wherein the identity code of the tree includes a basic code and a dynamic expansion code; the basic code includes an encoding corresponding to the geographic coordinates and a serial number; the dynamic expansion code includes a state code for indicating the current operation and maintenance state of the tree and a version code for recording the update number of the robust template.

[0009] Further technical solutions thereof are as follows: If the tree is an already coded tree, dynamically adjusting the feature weight based on the matching degree of the tree body features and historical tree body features and the matching degree of the environmental reference features and historical environmental reference features, and fusing with the historical template to generate a new feature template, and recording in the identity code of the tree.

[0010] Further technical solutions thereof are as follows: Determining a candidate tree set based on the geographic coordinates of the tree through global similarity comparison; When there is a candidate tree set, fine comparison is made based on the tree body features and the environment reference features to determine whether the tree is an already coded tree.

[0011] A further technical solution is that the geographic coordinates of the tree are calculated based on the pose data, comprising: The relative distance and angle of the tree base relative to the vehicle are calculated by using a visual algorithm combined with the pose data and the image to obtain the relative distance and angle; The relative distance and angle are converted into increments in the geodetic coordinate system using a rotation matrix, and the absolute coordinates of the tree are updated to the vehicle position to determine the geographic coordinates of the tree. The geographic coordinates of the tree are coded to obtain the code corresponding to the geographic coordinates.

[0012] A further technical solution is that the tree body features and the environment reference features are extracted based on the preprocessed image, and the correspondence between the tree body features and the environment reference features is established, comprising: The preprocessed image is analyzed by a convolutional neural network to extract high-dimensional visual features including the trunk, bark texture, and key points, and a picture enhancement algorithm under multi-view and lighting conditions is used for robustness; The three-dimensional coordinates of the fixed reference in the environment involved in the preprocessed image are identified and calculated, the relative distance and angle between the tree and the fixed reference are determined combined with the heading angle to form the environment reference features; The tree body features and the environment reference features are associated by using a spatial relationship vector to construct the relative position relationship between them.

[0013] A further technical solution is that when there is a candidate tree set, fine comparison is made based on the tree body features and the environment reference features to determine whether the tree is an already coded tree, comprising: When there is a candidate tree set, the matching degree of the tree body features and the environment reference features with the tree body features and the environment reference features corresponding to the candidate tree set is compared to determine whether the tree is an already coded tree.

[0014] A further technical solution is that when there is a candidate tree set, the matching degree of the tree body features and the environment reference features with the tree body features and the environment reference features corresponding to the candidate tree set is compared to determine whether the tree is an already coded tree, comprising: When there is a candidate tree set, the matching degree of the tree body features and the environment reference features with the tree body features and the environment reference features corresponding to the candidate tree set is calculated to obtain the tree feature matching degree. determining whether the tree feature matching degree meets the requirement; if the tree feature matching degree meets the requirement, determining that the tree is a tree with an existing code; if the tree feature matching degree does not meet the requirement, calculating a matching degree of the environmental reference feature and the environmental reference feature corresponding to the candidate tree set to obtain an environmental reference feature matching degree; if the environmental reference feature matching degree meets the requirement, determining that the tree identity is replaced or significantly changed; if the environmental reference feature matching degree does not meet the requirement, freezing the environmental reference feature matching degree.

[0015] Further technical solutions thereof are as follows: labeling the state code to generate a new tree identity code and associating it with the original tree identity code.

[0016] Further technical solutions thereof are as follows: when the matching degree of the tree body feature and the historical tree body feature and the matching degree of the environmental reference feature and the historical environmental reference feature both exceed a first set threshold, keeping the weight corresponding to the tree body feature and the weight corresponding to the environmental reference feature in a proportion close to each other; if the matching degree of the environmental reference feature and the historical environmental reference feature is higher than the first set threshold, and the matching degree of the tree body feature and the historical tree body feature is lower than a second set threshold, then increasing the weight corresponding to the environmental reference feature and decreasing the weight corresponding to the tree body feature to highlight the environmental feature; if the matching degree of the tree body feature and the historical tree body feature is higher than the first set threshold, and the matching degree of the environmental reference feature and the historical environmental reference feature is lower than the second set threshold, then increasing the weight corresponding to the tree body feature and decreasing the weight corresponding to the environmental reference feature to emphasize the tree feature; in the case that the matching degree of the tree body feature and the historical tree body feature and the matching degree of the environmental reference feature and the historical environmental reference feature are both lower than the second set threshold, freezing the weight corresponding to the environmental reference feature to avoid false update.

[0017] The application also provides a tree dynamic coding system based on high-precision positioning and visual analysis, comprising: An acquisition unit is configured to acquire pose data and images collected by a device installed on a vehicle. A calculation unit is configured to calculate geographical coordinates of a tree based on the pose data. A preprocessing unit is configured to preprocess the images to obtain preprocessed images. An extraction unit is configured to extract tree body features and environmental reference features based on the preprocessed images and establish a correspondence between the tree body features and the environmental reference features. A comparison unit is configured to determine whether the tree is a tree with an existing code based on the geographical coordinates of the tree, the tree body features and the environmental reference features. A code generation unit is configured to generate an identity code of the tree if the tree is not a tree with an existing code, wherein the identity code of the tree comprises a basic code and a dynamic extension code; the basic code comprises an encoding corresponding to the geographical coordinates and a serial number; and the dynamic extension code comprises a state code for indicating a current operation and maintenance state of the tree and a version code for recording an update number of a robust template.

[0018] Compared with the prior art, the application has the following beneficial effects: the device installed on the vehicle is used to collect pose data and images, and the geographical coordinates of the tree are calculated by using the information; the images are preprocessed to extract tree body features and environmental reference features and establish a correspondence therebetween; then, whether the tree is a tree recorded in the existing record is determined according to the geographical coordinates of the tree, the tree body features and the environmental features; if the tree is a newly discovered tree, an identity code of the tree is generated, which comprises a basic code composed of an encoding corresponding to the geographical coordinates and a serial number and a dynamic extension code containing a state code for indicating an operation and maintenance state of the tree and a version code for recording an update number of a robust template. This method realizes a stable, unique and traceable digital identity of a single tree in its whole life cycle, solves the deficiencies of the prior art in reliability, uniqueness and long-term continuity, and thus ensures the efficiency and accuracy of tree management.

[0019] The application will be further described below in combination with the drawings and specific embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0021] Figure 1 A flowchart of a tree dynamic coding method based on high-precision positioning and visual analysis provided by an embodiment of the present application is shown in FIG. 1. Figure 2 An identification diagram of a tree provided by an embodiment of the present application is shown in FIG. 2. Figure 3 A top view of an identification diagram of a tree provided by an embodiment of the present application is shown in FIG. 3. Figure 4 An identification effect diagram of a tree provided by an embodiment of the present application is shown in FIG. 4. Figure 1 An identification effect diagram of a tree provided by an embodiment of the present application is shown in FIG. 5. Figure 5 Figure 2 An identification effect diagram of a tree provided by an embodiment of the present application is shown in FIG. 6. Figure 6 A diagram of a fine comparison provided by an embodiment of the present application is shown in FIG. 7. Figure 7 A schematic block diagram of a tree dynamic coding system based on high-precision positioning and visual analysis provided by an embodiment of the present application is shown in FIG. 8. Figure 8 A schematic block diagram of a computer device provided by an embodiment of the present application is shown in FIG. 9. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0023] It should be understood that, when used in the specification and the appended claims, the terms "comprise" and "include" indicate the presence of the described features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0024] It should also be understood that the terms used in the present application specification are only for the purpose of describing particular embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, the singular forms "a", "an" and "the" are intended to include the plural forms, unless the context clearly indicates otherwise.

[0025] It should be further understood that the term "and / or" used in the present application specification and the appended claims means one or more of the associated listed items as well as all possible combinations of the items.

[0026] Please refer to​Figure 1 , Figure 1 The schematic flowchart of the tree dynamic coding method based on high-precision positioning and visual analysis provided by the embodiments of the present application. The tree dynamic coding method based on high-precision positioning and visual analysis is applied in a server, calculates the geographical coordinates of the trees through the pose data and images collected by the devices on the vehicle, pre-processes the images to extract the tree body features and environmental reference features, and establishes the feature correspondence relationship. This method uses global similarity comparison and fine comparison to determine whether the tree is an existing coded tree, and generates an identity code including a basic code and a dynamic extension code for a newly identified tree, realizing a stable, unique and traceable digital identity of a single tree throughout its life cycle. In addition, by dynamically adjusting the feature weights and fusing historical templates to update the tree feature templates, the reliability in long-term continuity is ensured, and the deficiencies of the prior art in terms of uniqueness and reliability are overcome. This process allows accurate monitoring of the state changes of the trees, and even in the case of tree replacement or major changes, the accuracy and consistency of the digital identity of the trees can be maintained.

[0027] Figure 1 The flowchart of the tree dynamic coding method based on high-precision positioning and visual analysis provided by the embodiments of the present application. As shown in Figure 1 , the method comprises the following steps S110 to S150.

[0028] S110, acquiring the pose data and images collected by the devices installed on the vehicle.

[0029] In this embodiment, the pose data refers to the information about the position and attitude of the vehicle collected by the devices installed on the vehicle. Specifically, these devices include but are not limited to global navigation satellite system (GNSS) receivers (supporting RTK differential positioning), inertial measurement units (IMU), etc. high-precision acquisition modules and computing units. The GNSS module is used to acquire the centimeter-level absolute geographical coordinates P vehicle (X v , Y v ) and the accurate heading angle θ (the angle between the vehicle head direction and the north) of the collection vehicle in real time, forming the accurate pose data of the vehicle. This pose data not only provides the position information of the vehicle, but also includes its motion state (such as speed, acceleration) and heading angle, which is crucial for accurately calculating the geographical coordinates of the trees relative to the vehicle.

[0030] The images refer to the image acquisition results of the trees on both sides of the road through the high-definition vehicle-mounted camera. These images have undergone a preprocessing process, including denoising, illumination normalization, etc., to obtain standardized images, so as to ensure the accuracy of subsequent feature extraction. In order to ensure that the trees are located in the center of the field of view, a target detection algorithm is usually applied to the collected images to crop the tree area, generating a cropped image containing only the main part of the tree. This step helps to improve the efficiency and accuracy of subsequent visual analysis, especially when advanced technologies such as convolutional neural networks (CNN) are used to extract tree body features (such as tree trunk texture, bark texture, trunk / branch key points, etc.). In addition, by analyzing tree images taken at different time points, different angles of view, and different lighting conditions, a more robust time series feature can be constructed, further enhancing the reliability of tree recognition and identity confirmation. The images not only contain rich visual information of the trees themselves, but also include fixed reference information in the surrounding environment, which is crucial for establishing the association between the trees and their geographical locations.

[0031] S120, calculating the geographical coordinates of the tree based on the pose data.

[0032] In this embodiment, the geographical coordinates of the tree refer to the specific position of the tree relative to the earth's surface determined through the combination of the vehicle's high-precision positioning information (including centimeter-level absolute geographical coordinates and accurate heading angle) and the image data collected by the vehicle-mounted camera, and processed by visual algorithms. This geographical location is usually represented by latitude and longitude, and in order to balance the needs of uniqueness and data volume, Geohash encoding is also performed.

[0033] In an embodiment, the above step S120 can include steps S121-S123.

[0034] S121, using a visual algorithm to calculate the distance and azimuth angle of the tree base relative to the vehicle based on the pose data and the image, to obtain the relative distance and angle.

[0035] In this embodiment, the relative distance and angle refer to the position parameters of the tree base relative to the vehicle determined by analyzing the images taken by the vehicle-mounted camera through visual algorithms. Specifically: Relative distance: refers to the straight-line distance d from the center of the vehicle to the base of the tree.

[0036] Azimuth angle: refers to the included angle a between the line connecting the tree and the vehicle center and the vehicle heading line.

[0037] These parameters are obtained by detecting and measuring the trees and their surrounding environmental references (such as streetlights, manholes, etc.) in the images, ensuring accurate identification of the position of the trees even under different lighting conditions or angles of view.

[0038] S122, convert the relative distance and angle into increments in the geodetic coordinate system using a rotation matrix, and update to the vehicle position to determine the absolute coordinates of the tree to obtain the geographic coordinates of the tree.

[0039] In this embodiment, this step involves the process of converting the observation value in the vehicle coordinate system to the geodetic coordinate system. The specific operation is as follows: According to the heading angle θ of the vehicle, a rotation matrix R(θ) is constructed.

[0040] Using the formula P tree =P vehicle +R(θ)∗[d⋅cosα,d⋅sinα] T , where P vehicle is the absolute geographic coordinates of the vehicle, the relative distance and angle are converted into increments in the geodetic coordinate system.

[0041] Finally, the absolute coordinates of the tree (i.e. its geographic coordinates) are obtained, which have a latitude and longitude accuracy of 9 decimal places, and generally take the last 6 decimal places for Geohash encoding, ensuring a plane accuracy of about 0.1 meters.

[0042] S123, encode the geographic coordinates of the tree to obtain the encoding corresponding to the geographic coordinates.

[0043] In this embodiment, in this step, the geographic coordinates of the tree are converted into a form that is convenient for storage and query - encoding. Specifically: First, the geographic coordinates of the tree are encoded using the Geohash encoding method. Geohash is a geographic encoding method that can encode two-dimensional latitude and longitude data into one-dimensional string form, retaining a certain spatial resolution and facilitating database indexing and retrieval.

[0044] The encoded result not only reflects the specific position of the tree, but also can be used as part of the basic encoding to construct a unique identity for the tree. This identity remains unchanged throughout the tree's life cycle, achieving unique and traceable management of the tree.

[0045] Through the above steps, the system can effectively determine the precise position of each tree and generate a unique identity code for it, thereby achieving effective management and long-term tracking of individual trees.

[0046] S130, pre-process the image to obtain a pre-processed image.

[0047] In this embodiment, the pre-processed image refers to the standardized image generated through a series of processing operations on the original collected image, such as removing noise, adjusting lighting, cropping the region of interest, etc. These preprocessing steps are crucial for improving image quality, enhancing the accuracy and robustness of feature recognition. Specifically, S130 can include the following key steps: Filtering techniques such as Gaussian filtering, median filtering, etc. are used to smooth the image and reduce the impact of noise. Random noise in the image is reduced or eliminated, which may come from the sensor itself or external environmental factors (such as weather conditions).

[0048] Histogram equalization, adaptive contrast enhancement, etc. are used to adjust the image brightness distribution, so that the image can maintain good visual effects under various lighting conditions. The problem of inconsistent image brightness caused by changes in lighting conditions is solved, ensuring that tree images taken at different time points are comparable.

[0049] First, use target detection algorithms (such as YOLO, SSD, etc.) to locate the position of trees in the image; Then, according to the detection results, enlarge the boundary box range according to a certain proportion, ensure the complete inclusion of trees and their nearby fixed reference objects (such as street lamps, manhole covers, etc.), and form a cropped image.

[0050] The part containing the target tree is accurately cropped from the whole image, excluding irrelevant background information, focusing on the tree itself and its surrounding environment reference objects.

[0051] The cropped image is uniformly scaled to a specified size, and its pixel values are normalized to a specific interval (such as [0, 1] or [-1, 1]) to better adapt to the requirements of deep learning models such as convolutional neural networks.

[0052] Make all images input into the subsequent model have consistent size and format, facilitate batch processing and accelerate the calculation process.

[0053] Through the above series of preprocessing operations, the original image is converted into a high-quality standard format, not only improving the efficiency and accuracy of the subsequent feature extraction stage, but also providing a reliable data foundation for realizing the identity identification of single trees. In addition, good preprocessing can help the system to run stably in complex and variable actual environment, effectively cope with problems such as lighting changes, angle differences, etc., enhancing the robustness and practicality of the whole system.

[0054] S140, based on the pre-processed image, extracting tree body features and environmental reference object features, and establishing the corresponding relationship between the tree body features and the environmental reference object features.

[0055] In this embodiment, the tree body features refer to information extracted from the tree image to describe its unique physical and visual attributes. These features include but are not limited to tree trunk texture, bark details, key point positions of the trunk and branches, etc., which together form the basis for identifying and distinguishing different trees.

[0056] The environmental reference features refer to the positional information of fixed objects in the tree's surrounding environment (such as street lamps, manhole covers, etc.) and their relative relationships, which provide assistance to improve the accuracy and stability of tree identity identification.

[0057] In an embodiment, the above step S140 can include steps S141-S143.

[0058] S141, analyze the pre-processed image through a convolutional neural network, extract high-dimensional visual features including tree trunk, bark texture, and key points, and use a multi-view and light condition picture enhancement algorithm for robustness.

[0059] In this embodiment, a pre-trained convolutional neural network (CNN) is used to perform feature extraction on the input standardized image.

[0060] The extracted features include but are not limited to the texture, structural information of the tree trunk and bark, key point positions of the trunk and main branches, etc.

[0061] To enhance the robustness of the model, images taken under various viewing angles and lighting conditions are used for training or testing to ensure that the system can work stably under different environmental conditions.

[0062] Specifically, a deep learning model suitable for image feature extraction is selected, and these models have performed well in many computer vision tasks. The intermediate layers of the CNN are designed with specific convolutional layers and pooling layers to capture local detail information of the tree, such as tree trunk texture and bark structure. At the same time, high-level feature maps are used to identify the key point positions of the trunk and main branches. Transfer learning technology is used, and a model pre-trained on a large-scale natural image dataset is used as a basis, and then fine-tuned to adapt to the extraction requirements of tree features.

[0063] Low-level features such as edges and color changes are extracted through convolutional layers; then more complex patterns such as the unique texture of the bark are captured through multiple convolutional layers and nonlinear activation function combinations. Heat map regression or direct regression methods are used to predict the key point coordinates of the trunk and main branches. This step helps subsequent three-dimensional modeling and growth monitoring of trees. Combining feature maps from different levels, a multi-level, multi-scale feature representation is formed to comprehensively describe the morphological features of the tree.

[0064] S142, identify and calculate the three-dimensional coordinates of the fixed reference in the environment involved in the pre-processed image, determine the relative distance and angle between the tree and the fixed reference combined with the heading angle, and form the environment reference feature.

[0065] In this embodiment, the target detection algorithm is used to identify the fixed reference in the image (such as street lamps, manhole covers, etc.), and the three-dimensional coordinates are estimated by stereo vision or other depth perception technology.

[0066] According to the heading angle and the current position of the vehicle, the relative distance and angle between the tree and each reference are calculated.

[0067] The above information is integrated into the environment reference feature, which is used as supplementary data for subsequent feature comparison and identity confirmation process.

[0068] Specifically, the target detection framework is used to identify the fixed reference in the image (such as street lamps, manhole covers, etc.). These algorithms can realize real-time detection while ensuring accuracy. A unique class label is set for each type of reference, and sufficient sample size is included in the training set to cover all possible situations.

[0069] According to the heading angle θ of the vehicle, the relative direction between the tree and the reference is adjusted. Specifically, the observation value in the vehicle coordinate system is converted to the geodetic coordinate system.

[0070] The straight-line distance of the tree base relative to the camera center is calculated; the angle of the tree and the vehicle connection relative to the vehicle heading is determined.

[0071] Based on the above calculation results, combined with the current position P vehicle of the vehicle, the absolute latitude and longitude coordinates P tree of the tree are obtained through spatial coordinate transformation.

[0072] All the information calculated above (such as the three-dimensional coordinates of the reference, the relative distance and angle between the tree and it, etc.) are encoded into a unified format for subsequent processing. The newly generated environment reference feature is stored in the database together with the corresponding tree body feature for subsequent identity confirmation and feature comparison.

[0073] Through the above detailed steps, the system not only accurately extracts the high-dimensional visual features of the tree itself, but also effectively utilizes the fixed reference information in the surrounding environment to construct a stable and reliable tree identity identification system. This method significantly improves the accuracy and reliability of tree identification, especially suitable for long-term monitoring and management of urban green resources application scenarios.

[0074] S143, use the spatial relationship vector to associate the tree body feature and the environment reference feature, and construct the relative position relationship between them.

[0075] In this embodiment, a spatial relationship vector R te is defined, which represents the relative positional relationship between the tree body feature F t and the surrounding environment reference feature F e .

[0076] Although no accurate pairing of each feature point is performed in this process, this relative relationship can help the system to still accurately locate and identify specific tree individuals when facing tree growth, damage, or environmental changes.

[0077] The comprehensive feature template formed finally not only contains rich body information, but also integrates reliable environmental reference data, greatly improving the adaptability and reliability of the entire system.

[0078] Through these three steps, the system can effectively extract the unique features of the tree and closely combine them with the surrounding environment, thereby realizing long-term effective identity identification and tracking management of individual trees. This method overcomes the limitations of traditional single identification methods and improves the accuracy and stability of tree identity identification.

[0079] S150, determining whether the tree is an already coded tree based on the geographical coordinates of the tree, the tree body feature, and the environmental reference feature.

[0080] In an embodiment, the above step S150 can include steps S151-S152.

[0081] S151, determining a candidate tree set based on the geographical coordinates of the tree through global similarity comparison.

[0082] In this embodiment, the candidate tree set refers to the set of all existing coded tree records that are close to the current tree position (e.g., within a 5-meter radius) filtered from the database based on global similarity comparison of tree geographical coordinates.

[0083] Using the geographical coordinates of the tree (obtained through a high-precision positioning system), a set of historical records that may match the current tree, i.e., the candidate tree set, is quickly filtered out.

[0084] In the data collection stage, the GNSS module (supporting RTK differential positioning) is used to obtain the centimeter-level absolute geographical coordinates P vehicle and the accurate heading angle θ of the vehicle.

[0085] The relative distance d and the included angle α of the tree relative to the vehicle are calculated through the visual algorithm combined with the vehicle position and heading angle, and then converted into the absolute latitude and longitude coordinates P tree of the tree.

[0086] The system will query the tree identity database for all existing records within a 5-meter radius based on the calculated geographic coordinates of the tree.

[0087] This step relies on efficient spatial indexing techniques (such as R-tree or Quadtree) to speed up the search process and reduce computational load.

[0088] If no records are found within the specified range, the current tree is considered a newly discovered tree, triggering the new creation process. If multiple potential matches exist, further analysis of the detailed features of these candidates is required for confirmation.

[0089] S152、When there is a candidate tree set, fine comparison is made based on the tree body features and the environmental reference features to determine whether the tree is an existing coded tree.

[0090] Specifically, when there is a candidate tree set, the matching degree of the tree body features and the environmental reference features is compared with the tree body features and the environmental reference features corresponding to the candidate tree set to determine whether the tree is an existing coded tree.

[0091] For each tree in the candidate tree set, fine comparison is made using its body features (texture, structure, etc.) and environmental reference features (position relationship of fixed facilities) to ensure that the final confirmed tree identity is accurate.

[0092] First, the current collected tree body features (including but not limited to trunk texture, bark texture, key point position, etc.) are matched with the historical feature templates of the corresponding trees in the candidate set.

[0093] The similarity score between the two is calculated, and if the score is higher than the set threshold (e.g. 90%), it is considered that the two are highly consistent, and it is inclined to believe that they are the same tree.

[0094] When the matching degree of the body features is insufficient, the environmental reference features are examined instead. This includes the three-dimensional coordinates of permanent facilities (such as streetlights, manhole covers, etc.) around the tree and their relative position relationships.

[0095] The consistency of the distance, angle, etc. of these environmental reference features between the current tree and the candidate tree is compared to assess whether they are located at the same geographic location.

[0096] If both the tree body features and the environmental features show high matching degrees, the tree is finally confirmed as an existing coded tree, and a feature fusion update operation is performed.

[0097] If only the environmental features match well but the body features differ greatly, it is possible that the tree has been replaced or changed significantly, in which case a new identity record needs to be created and associated with the old record, while the status code is marked (e.g., "replaced / transplanted").

[0098] If neither of them can achieve a satisfactory matching degree, it is considered as a suspected case of transplanting and environmental change, and the comparison result is temporarily frozen for subsequent verification.

[0099] After successful matching, the system dynamically fuses the new and old features according to a certain weight ratio to generate an updated feature template, which is stored in the database as one of the historical versions.

[0100] At the same time, the status code and version code of the corresponding tree are adjusted to ensure the integrity and traceability of the data.

[0101] Through the above steps, S150 not only realizes the effective identification and tracking of tree identity, but also establishes a perfect life cycle management system, greatly improving the efficiency and accuracy of urban greening resource management and protection work.

[0102] In an embodiment, the above step S152 can include steps S1521-S1527.

[0103] S1521, when there is a candidate tree set, calculate the matching degree of the tree body features and the tree body features corresponding to the candidate tree set to obtain the tree feature matching degree; S1522, determine whether the tree feature matching degree meets the requirements; S1523, if the tree feature matching degree meets the requirements, determine that the tree is a tree with an existing code; S1524, if the tree feature matching degree does not meet the requirements, calculate the matching degree of the environmental reference features and the environmental reference features corresponding to the candidate tree set to obtain the environmental reference feature matching degree; S1525, if the environmental reference feature matching degree meets the requirements, determine that the tree identity is replaced or changed significantly; S1526, mark the status code to generate a new tree identity code and associate it with the original tree identity code.

[0104] S1527, if the environmental reference feature matching degree does not meet the requirements, freeze the environmental reference feature matching degree.

[0105] In this embodiment, when there is a candidate tree set, first, the matching degree of the ontology features (such as trunk texture, bark texture, trunk / branch key points, and breast height form information, etc.) of the current tree and the corresponding ontology features of each tree in the candidate tree set are calculated to obtain the tree feature matching degree.

[0106] Based on a pre-set threshold (for example, 90%), it is judged whether the above calculated tree feature matching degree reaches or exceeds the threshold. If it does, it is considered that the matching degree meets the requirements; otherwise, it does not.

[0107] If the tree feature matching degree meets the requirements, it means that the currently identified tree is highly similar to a tree in the candidate set, and the system determines the identity of the tree as a tree that already exists in the database.

[0108] If the tree feature matching degree does not meet the requirements, the matching degree of the environmental reference features (such as the relative position and angle with fixed reference objects such as street lamps and guardrails, etc.) of the current tree and the corresponding environmental reference features of the candidate set are calculated to obtain the environmental reference feature matching degree.

[0109] If the environmental reference feature matching degree meets the requirements, it indicates that although the tree itself has changed (possibly due to growth, pruning, or partial damage), the surrounding environmental features have not changed significantly, and the system determines that the tree identity is replaced or has undergone major changes.

[0110] For the tree that is determined to be replaced or to have undergone major changes, the system will update its status code (such as marked as "10 = replaced / transplanted"), generate a new tree identity code, and ensure association with the original tree identity code to ensure complete traceability of the life cycle.

[0111] If the environmental reference feature matching degree also does not meet the requirements, i.e., both the tree and its surrounding environment have undergone significant changes, the system will freeze this matching result, wait for further data collection or manual confirmation, to avoid false updates or confusion of the tree identity.

[0112] This series of steps ensures the accuracy and reliability of tree identity recognition, while also allowing the system to dynamically adapt to the challenges brought by tree growth and environmental changes.

[0113] S160, if the tree is not an existing coded tree, an identity code of the tree is generated, wherein the identity code of the tree includes a basic code and a dynamic expansion code; the basic code includes an encoding corresponding to the geographic coordinates and a serial number; the dynamic expansion code includes a status code for indicating the current operation and maintenance state of the tree and a version code for recording the number of updates of the robust template.

[0114] In this embodiment, first, the absolute geographic coordinates of the new tree are obtained using a high-precision GNSS system (P tree ). To ensure uniqueness in space, the latitude and longitude are converted into a string encoding using Geohash encoding or other similar methods. Typically, the last 6 digits of the latitude and longitude are encoded to balance uniqueness and data volume. This encoding method takes into account the specific geographic location of the tree, ensuring accurate spatial positioning of each tree.

[0115] To further ensure the uniqueness of the basic code, a self-incrementing serial number is added. This can be a way of continuously numbering all trees within a specific geographic area, ensuring that each tree has a unique basic code.

[0116] Once the basic code is generated by the system, it is permanently bound to the specific tree and its geographic location. Whether subsequent template updates, environmental changes, or operational management operations occur, the basic code will not change. This part of the code is equivalent to the "identity card number" of the tree, with global uniqueness, stability, and non-modifiability, and is the core identifier.

[0117] The status code is used to represent the current operational status of the tree. In the initial state, if the tree is healthy and does not require special attention, the status code is set to "00" (normal). The status code can reflect the real-time condition of the tree and be adjusted according to actual needs. Other status codes include but are not limited to: "01": Need to detect (need to call external information for verification, and need human intervention if necessary); "10": Replaced / transplanted (information stored in the database and needs to be regenerated); "11": Damaged / faulty (information stored in the database and needs to be regenerated); Version code: records the number of updates of the robust template, reflecting the evolution history of the tree features in its life cycle. For new trees, the initial version code is set to "0", and the version code increases with subsequent identification, feature fusion, and template updates.

[0118] The status code and version number can be dynamically adjusted according to the identification results and management needs, similar to the file information attached to an identity card, which can be continuously updated to adapt to new situations without affecting the uniqueness and stability of the basic code itself.

[0119] Each basic code is not just a number, but corresponds to a complete data set containing the ontological features of the tree (such as tree trunk texture, structure, etc.), environmental features (such as the relative position to fixed reference objects such as streetlights, guardrails, etc.), geographic coordinates, and historical versions. When a new tree is identified, the system creates a new data set and associates all the relevant information mentioned above with it.

[0120] Over time, as new identifications or features are updated, this information is added to the dataset for the corresponding tree, and the status code and version code are adjusted as needed. This not only maintains the uniqueness of the tree identity, but also provides complete traceability throughout the life cycle.

[0121] In this way, even in the case of tree growth, damage, or environmental changes, the identity of each tree can be effectively managed and tracked, ensuring its identifiability and traceability throughout its life cycle. In addition, this coding mechanism provides flexibility, allowing dynamic adjustment of the state and feature templates of the tree according to actual needs, better adapting to various complex scenarios.

[0122] In this embodiment, the tree identity code is uniquely bound to the individual tree and its geographical location at the time of first generation, following the following principles: Uniqueness: Each tree corresponds to only one base code throughout its life cycle, which cannot be repeated and will not be changed.

[0123] Stability: Once the base code is established, it will remain unchanged forever, just like the "identity card number" of the tree, ensuring long-term effective identification.

[0124] Extensibility: Although the base code is fixed and unchangeable, the corresponding "profile information" (such as feature templates and operation and maintenance status) can be dynamically updated over time, and these updates do not affect the uniqueness and stability of the base code.

[0125] S170, if the tree is an existing coded tree, dynamically adjust the feature weights based on the matching degree of the tree body features and historical tree body features, and the matching degree of the environmental reference features and historical environmental reference features, and fuse with the historical template to generate a new feature template, and record it in the tree identity code.

[0126] In an embodiment, the above step S170 can include steps S171-S174.

[0127] S171, when the matching degree of the tree body features and historical tree body features, and the matching degree of the environmental reference features and historical environmental reference features, both exceed the first set threshold, the weight corresponding to the tree body features and the weight corresponding to the environmental reference features are kept close in proportion.

[0128] In this embodiment, when the matching degree of the tree body features (such as the texture and structure of the trunk) and the tree body features in the historical record and the matching degree of the environmental reference features (such as the relative position with the fixed reference such as street lamps, guardrails, etc.) and the environmental reference features in the historical record both exceed the first set threshold value.

[0129] Operation: In this case, it means that the current collected data is highly consistent with the historical data, so the proportion of the weights W t and the weights W e corresponding to the tree body features and the environmental reference features is close, ensuring the stability and accuracy of the template.

[0130] S172, if the matching degree of the environmental reference features and the historical environmental reference features is higher than the first set threshold value, and the matching degree of the tree body features and the historical tree body features is lower than the second set threshold value, then the weight corresponding to the environmental reference features is increased, and the weight corresponding to the tree body features is reduced to highlight the environmental features.

[0131] In this embodiment, if the matching degree of the environmental reference features and the historical environmental reference features is higher than the first set threshold value, and the matching degree of the tree body features and the historical tree body features is lower than the second set threshold value. This means that although the tree may have undergone some changes (such as growth or damage), the environmental reference around it remains unchanged or changes little. At this time, the weight W e of the environmental reference features should be increased, and the weight W t of the tree body features should be correspondingly reduced to highlight the importance of environmental features and ensure accurate identification of the tree in the case of little environmental change.

[0132] S173, if the matching degree of the tree body features and the historical tree body features is higher than the first set threshold value, and the matching degree of the environmental reference features and the historical environmental reference features is lower than the second set threshold value, then the weight corresponding to the tree body features is increased, and the weight corresponding to the environmental reference features is reduced to emphasize the tree features.

[0133] In this embodiment, if the matching degree of the tree body features and the historical tree body features is higher than the first set threshold value, and the matching degree of the environmental reference features and the historical environmental reference features is lower than the second set threshold value.

[0134] This indicates that although the surrounding environment may have changed (such as the addition or removal of reference objects), the characteristics of the tree itself (such as the texture and structure of the trunk) remain highly consistent. At this time, the weight W t of the tree body features should be increased, and the weight W eThis emphasizes the importance of tree characteristics in the identification process.

[0135] S174. If the matching degree between the tree body feature and the historical tree body feature, and the matching degree between the environmental reference feature and the historical environmental reference feature are both lower than the second set threshold, the weight corresponding to the environmental reference feature is frozen to avoid erroneous updates.

[0136] When the matching degree between the tree's intrinsic features and historical tree intrinsic features, and the matching degree between environmental reference features and historical environmental reference features are both lower than the second set threshold.

[0137] In this scenario, there may be significant uncertainty and error risk, such as severe damage to trees or major environmental changes. To avoid erroneous template updates, the system will freeze the weights W of environmental reference features. e Only historical information is retained, pending further confirmation or manual intervention.

[0138] Once an appropriate weight allocation strategy has been determined based on the above conditions, the feature fusion formula F will then be used. m =W t ⋅F t +W e ⋅F e The fused feature template is calculated and smoothly integrated with historical templates to generate a new feature template. This new template not only reflects the currently collected information but also takes into account historical data, ensuring long-term tracking and updating of tree features.

[0139] The results of the merged update are written into the dataset corresponding to the tree's identity code, with the version code incrementing with each update. Furthermore, each collected raw data entry is stored as an independent historical record, ensuring data traceability and integrity. In this way, even when trees grow, are damaged, or their environment changes, the identity of each tree can be effectively managed and tracked, achieving uniqueness and traceability throughout its entire lifecycle.

[0140] For example, please see Figures 2-3 The in-vehicle system can provide centimeter-level precise positioning for the vehicle, obtaining the absolute geographic coordinates P of the vehicle's center. vehicle (X v Y v) and the heading angle of the vehicle (i.e. the angle between the vehicle heading and the true north). After obtaining these information, the system will pre-process the image, marking the street trees in a conspicuous color and other fixed references (such as manhole covers, street signs, power boxes, street lamps, etc. about 20 kinds) in different colors. Through the visual algorithm, the straight-line distance d of the base of the street tree relative to the camera and the relative angle a between the line connecting the tree and the center of the vehicle and the heading line of the vehicle are calculated in the image.

[0141] Next, according to the principle of spatial coordinate transformation, the position of the tree is converted into absolute latitude and longitude coordinates. This process ensures that the accurate and unique absolute position of the tree can be obtained regardless of the vehicle's orientation. Usually, the converted latitude and longitude coordinates have 9 decimal places, but in order to balance between uniqueness and data volume, the last 6 decimal places are usually taken for Geohash encoding, which can guarantee a plane accuracy of about 0.1 meters.

[0142] In addition, the model also extracts the characteristics of the tree itself and the environmental characteristics from multiple sources, and dynamically allocates weights for feature fusion. When the vehicle drives near the tree, the absolute coordinates of the tree are calculated according to the angle between the vehicle and the tree, and the tree body features including the trunk texture and key branch points are extracted, while the environmental feature information such as the relative position and horizontal angle of the tree to the surrounding fixed references (such as manhole covers, street lamps, etc.) is recorded.

[0143] Subsequently, as Figures 4-6 , the system will find all the encoded trees within a 5-meter range of the tree coordinates in the database to form a candidate set, and compare the collected tree body features with the candidate trees in the candidate set. If the similarity exceeds the set threshold (usually set at about 90% to avoid misidentification), it is marked as a candidate. Then, further comparison of the collected environmental feature information and the environmental feature information of the candidate trees, such as the distance, angle and type of the reference objects, is made. If the environmental feature information of the two is also highly matched, it is considered that the newly collected information corresponds to a certain tree in the database.

[0144] Finally, the model dynamically fuses the newly collected tree feature information and environmental feature information with the original data in the database according to the weight ratio. For example, Figure 3 The example shows the information collected by the vehicle in the same area at different time periods, showing the cases of trees T2 and T3. The model will compare T2 and T3 with the existing data in the database such as T1 through the above comparison process, and conclude that T2 and T3 are actually the tree T1 in the database. Therefore, only the existing data needs to be updated, without the need to create a new record. This method improves the consistency and accuracy of the data, while also reducing the generation of redundant data.

[0145] The method of the embodiment combines high-precision positioning information and visual features of trees to provide a brand-new single-tree identity identification method. This method not only avoids the limitations brought by relying on a single physical marker, coordinates or images, but also enhances the stability and reliability of the identification.

[0146] In each identification process, the currently collected tree features are automatically fused with historical data, and the weight is dynamically adjusted according to environmental changes and matching degrees. When encountering feature drift (such as changes caused by tree growth or environmental changes), the system will update the template with the latest features, thereby ensuring the accuracy and consistency of long-term tracking.

[0147] A multi-dimensional feature template is adopted, including the texture, structure and other features of the tree itself, as well as the relative positions of the surrounding permanent facilities. This not only improves the accuracy of identification in complex environments, but also ensures the distinguishability between different trees, increasing the stability of the entire system.

[0148] Based on the unique identity code generated by the position, tree body features and environmental reference features, each tree has a unique digital identity throughout its life cycle. Once the code is generated, it remains unchanged, providing a solid foundation for tree management and research, and achieving full traceability from planting to the final state.

[0149] In summary, the method of the embodiment establishes a stable and dynamically adaptive tree identity identification system by fusing positioning information and visual analysis, solving the problems of easy loss and invalidation over time in traditional methods, and ensuring the uniqueness and traceability of each tree throughout its life cycle through unique coding rules.

[0150] The above-mentioned tree dynamic coding method based on high-precision positioning and visual analysis collects pose data and images through devices installed on vehicles, calculates the geographical coordinates of the tree using these information, and pre-processes the images to extract tree body features and environmental reference features and establish their corresponding relationship. Then, according to the geographical coordinates of the tree, the body features and environmental features, it is determined whether the tree is in the existing record. If it is a newly discovered tree, an identity code is generated for it, including a basic code composed of a code corresponding to the geographical coordinates and a serial number, and a dynamic expansion code containing a state code representing the operation and maintenance state and a version code recording the number of template update times. This method realizes the stable, unique and traceable digital identity of a single tree throughout its life cycle, solves the shortcomings of existing technology in reliability, uniqueness and long-term continuity, and ensures the efficiency and accuracy of tree management work.

[0151] Figure 7This is a schematic block diagram of a tree dynamic coding system 300 based on high-precision positioning and visual analysis provided in an embodiment of the present invention. Figure 7 As shown, corresponding to the above-described tree dynamic coding method based on high-precision positioning and visual analysis, the present invention also provides a tree dynamic coding system 300 based on high-precision positioning and visual analysis. This tree dynamic coding system 300 includes a unit for executing the above-described tree dynamic coding method based on high-precision positioning and visual analysis, and the system can be configured in a server. Specifically, please refer to... Figure 7 The tree dynamic coding system 300 based on high-precision positioning and visual analysis includes an acquisition unit 301, a calculation unit 302, a preprocessing unit 303, an extraction unit 304, a comparison unit 305, and a coding generation unit 306.

[0152] The system comprises: an acquisition unit 301 for acquiring pose data and images collected by a device installed on a vehicle; a calculation unit 302 for calculating the geographic coordinates of the tree based on the pose data; a preprocessing unit 303 for preprocessing the image to obtain a preprocessed image; an extraction unit 304 for extracting tree body features and environmental reference features based on the preprocessed image, and establishing a correspondence between the tree body features and the environmental reference features; a comparison unit 305 for determining whether the tree is an already coded tree based on its geographic coordinates, tree body features, and environmental reference features; and a coding generation unit 306 for generating an identity code for the tree if it is not an already coded tree, wherein the tree identity code includes a basic code and a dynamic extension code; wherein the basic code includes the code corresponding to the geographic coordinates and a sequence number; and the dynamic extension code includes a status code indicating the current operation and maintenance status of the tree and a version code recording the number of times the robust template has been updated.

[0153] In one embodiment, the tree dynamic coding system 300 based on high-precision positioning and visual analysis further includes: The fusion update unit 307 is used to dynamically adjust the feature weights based on the matching degree between the tree's ontological features and historical tree ontological features, and the matching degree between the environmental reference features and historical environmental reference features, if the tree is an already coded tree, and to fuse it with the historical template to generate a new feature template, which is then recorded in the tree's identity code.

[0154] In one embodiment, the comparison unit 305 includes: a global comparison subunit configured to determine a candidate tree set based on the geographical coordinates of the tree by global similarity comparison; and a refined comparison subunit configured to, when the candidate tree set exists, perform refined comparison based on the tree body features and the environment reference features to determine whether the tree is an existing coded tree.

[0155] In an embodiment, the computing unit 302 comprises: a distance-angle computing subunit configured to calculate the distance and azimuth angle of the tree base relative to the vehicle by using a visual algorithm combined with the pose data and the image to obtain relative distance and angle; a coordinate computing subunit configured to convert the relative distance and angle into an increment in the geodetic coordinate system using a rotation matrix and update to the vehicle position to determine the absolute coordinates of the tree to obtain the geographical coordinates of the tree; and an encoding subunit configured to encode the geographical coordinates of the tree to obtain the code corresponding to the geographical coordinates.

[0156] In an embodiment, the extracting unit 304 comprises: a body feature extracting subunit configured to analyze the preprocessed image by a convolutional neural network to extract high-dimensional visual features including the trunk, bark texture and key points, and use a picture enhancement algorithm under multi-view and light conditions to improve robustness; an environment feature extracting subunit configured to identify and calculate the three-dimensional coordinates of fixed reference objects in the environment involved in the preprocessed image, combine with the heading angle to determine the relative distance and angle between the tree and the fixed reference objects to form the environment reference features; and a relationship constructing subunit configured to use a spatial relationship vector to associate the tree body features and the environment reference features to construct the relative position relationship therebetween.

[0157] In an embodiment, the refined comparison subunit is configured to, when the candidate tree set exists, compare the matching degree of the tree body features and the environment reference features with the tree body features and the environment reference features corresponding to the candidate tree set to determine whether the tree is an existing coded tree.

[0158] In an embodiment, the refined comparison subunit comprises: The body feature matching degree calculation module is configured to calculate a matching degree between the tree body feature and tree body features corresponding to the candidate tree set to obtain a tree feature matching degree when the candidate tree set exists.

[0159] In an embodiment, the fine-grained comparison subunit further comprises: The association module is configured to mark the status code to generate a new tree identity code and associate the new tree identity code with the original tree identity code.

[0160] In an embodiment, the fusion update unit 307 comprises: The first adjustment subunit is configured to keep the weight corresponding to the tree body feature and the weight corresponding to the environment reference feature in a close proportion when the matching degree between the tree body feature and the historical tree body feature and the matching degree between the environment reference feature and the historical environment reference feature both exceed a first set threshold value. The third adjustment subunit is configured to increase the weight corresponding to the tree body feature and reduce the weight corresponding to the environment reference feature to emphasize the tree feature when the matching degree between the tree body feature and the historical tree body feature exceeds the first set threshold value and the matching degree between the environment reference feature and the historical environment reference feature is lower than the second set threshold value. The fourth adjustment subunit is configured to freeze the weight corresponding to the environment reference feature to avoid false update when the matching degree between the tree body feature and the historical tree body feature and the matching degree between the environment reference feature and the historical environment reference feature are both lower than the second set threshold value.

[0161] It should be noted that the specific implementation process of the tree dynamic coding system 300 and each unit based on high-precision positioning and visual analysis described above can be clearly understood by those skilled in the art, which can be referred to the corresponding description in the foregoing method embodiments. For the convenience and brevity of description, it will not be repeated here.

[0162] The tree dynamic coding system 300 based on high-precision positioning and visual analysis described above can be implemented in the form of a computer program, which can run on a computer device as shown in the figure. Figure 8

[0163] Please refer to Figure 8 , Figure 8 is a schematic block diagram of a computer device provided by an embodiment of the present application. The computer device 500 can be a server, wherein the server can be a stand-alone server or a server cluster composed of multiple servers.

[0164] Refer to Figure 8 , the computer device 500 includes a processor 502, a memory and a network interface 505 connected through a system bus 501, wherein the memory can include a non-volatile storage medium 503 and an internal memory 504.

[0165] The non-volatile storage medium 503 can store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions which, when executed, can cause the processor 502 to perform a tree dynamic coding method based on high-precision positioning and visual analysis.

[0166] The processor 502 is configured to provide computing and control capabilities to support the operation of the entire computer device 500.

[0167] The internal memory 504 provides an environment for the running of the computer program 5032 in the non-volatile storage medium 503, which, when executed by the processor 502, can cause the processor 502 to perform a tree dynamic coding method based on high-precision positioning and visual analysis.

[0168] The network interface 505 is configured to perform network communication with other devices. Those skilled in the art can understand that Figure 8 the structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device 500 to which the scheme of the present application is applied. The specific computer device 500 can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0169] ​The processor 502 is configured to run the computer program 5032 stored in the memory to implement all the steps of the tree dynamic coding method based on high-precision positioning and visual analysis.

[0170] It should be understood that, in the embodiments of the present application, the processor 502 can be a central processing unit (CPU), and the processor 502 can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0171] It can be understood by those skilled in the art that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program includes program instructions, and the computer program can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the above-mentioned embodiments.

[0172] Therefore, the present application also provides a storage medium. The storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein the computer program is executed by a processor to make the processor execute all the steps of the tree dynamic coding method based on high-precision positioning and visual analysis.

[0173] The storage medium can be a U disk, a mobile hard disk, a read-only memory (ROM), a magnetic disk or an optical disk, etc. Various computer-readable storage media that can store program codes.

[0174] Those skilled in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized in electronic hardware, computer software or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been described in general terms in the above description. Whether the functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. A person skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0175] In several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other manners. For example, the division of the system embodiments described above is merely a logical division, and there can be other division manners in actual implementation. For example, two or more units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0176] The steps in the method embodiments of the present application can be adjusted, combined and deleted in sequence according to actual needs. The units in the system embodiments of the present application can be combined, divided and deleted according to actual needs. In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit.

[0177] The integrated unit, if realized in the form of a software functional unit and sold or used as an independent product, can be stored in a storage medium. Based on such understanding, the technical solutions of the present application essentially or say the part of the prior art that makes a contribution, or the whole or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0178] The above description is merely a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of various equivalent modifications or replacements within the technical range disclosed by the present application, and these modifications or replacements should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A tree dynamic coding method based on high-precision positioning and visual analysis, characterized in that, include: Acquire pose data and images collected by devices installed on the vehicle; Calculate the geographic coordinates of the tree based on the pose data; The image is preprocessed to obtain a preprocessed image; Based on the preprocessed image, tree body features and environmental reference features are extracted, and the correspondence between the tree body features and the environmental reference features is established. Based on the tree's geographical coordinates, the tree's physical characteristics, and the characteristics of the environmental reference objects, determine whether the tree is a tree that already has a code; If the tree is not an existing coded tree, an identity code for the tree is generated. The identity code for the tree includes a basic code and a dynamic extension code. The basic code includes the code corresponding to the geographic coordinates and a sequence number. The dynamic extension code includes a status code for indicating the current operation and maintenance status of the tree and a version code for recording the number of times the robust template has been updated.

2. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 1, characterized in that, After determining whether a tree is an already coded tree based on its geographical coordinates, its intrinsic features, and the features of the environmental reference points, the process further includes: If the tree is an already coded tree, the feature weights are dynamically adjusted based on the matching degree between the tree's ontological features and historical tree ontological features, as well as the matching degree between the environmental reference features and historical environmental reference features. These features are then fused with historical templates to generate new feature templates, which are recorded in the tree's identity code.

3. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 1, characterized in that, The process of determining whether a tree is an already coded tree based on its geographical coordinates, its intrinsic features, and environmental reference features includes: Candidate tree sets are determined based on the geographic coordinates of the trees through global similarity comparison; When a candidate tree set exists, a refined comparison is performed based on the tree's intrinsic features and the environmental reference features to determine whether the tree is an already coded tree.

4. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 1, characterized in that, The calculation of the tree's geographic coordinates based on the pose data includes: A visual algorithm is used to combine the pose data and the image to calculate the distance and azimuth angle of the tree base relative to the vehicle, so as to obtain the relative distance and angle. The relative distance and angle are converted into increments in the geodetic coordinate system using a rotation matrix and updated to the vehicle position to determine the absolute coordinates of the tree, thus obtaining the tree's geographic coordinates. The geographic coordinates of the trees are encoded to obtain the corresponding codes.

5. The tree dynamic coding method based on high-precision positioning and visual analysis as described in claim 1, characterized in that, The step of extracting tree features and environmental reference features from the preprocessed image and establishing the correspondence between the tree features and the environmental reference features includes: The preprocessed image is analyzed by a convolutional neural network to extract high-dimensional visual features, including tree trunk, bark texture and key points, and the robustness of the image enhancement algorithm under multiple viewpoints and lighting conditions is utilized. Identify and calculate the three-dimensional coordinates of fixed reference objects in the environment involved in the preprocessed image, and determine the relative distance and angle between the trees and the fixed reference objects by combining the heading angle, thus forming environmental reference object features; The spatial relationship vector is used to associate the tree's physical features with the environmental reference features to construct the relative positional relationship between the two.

6. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 5, characterized in that, When a candidate tree set exists, a refined comparison is performed based on the tree's intrinsic features and the environmental reference features to determine whether the tree is an already coded tree, including: When a candidate tree set exists, the matching degree of the tree ontology features and the environmental reference features with the tree ontology features and environmental reference features corresponding to the candidate tree set is compared to determine whether the tree is an already coded tree.

7. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 6, characterized in that, When a candidate tree set exists, comparing the matching degree between the tree ontological features and the environmental reference features and the tree ontological features and environmental reference features corresponding to the candidate tree set to determine whether the tree is an already coded tree includes: When a candidate tree set exists, the matching degree between the tree ontology features and the tree ontology features corresponding to the candidate tree set is calculated to obtain the tree feature matching degree. Determine whether the tree feature matching degree meets the requirements; If the tree feature matching degree meets the requirements, then the tree is determined to be a tree that has already been coded; If the tree feature matching degree does not meet the requirements, the matching degree between the environmental reference feature and the environmental reference feature corresponding to the candidate tree set is calculated to obtain the environmental reference feature matching degree. If the environmental reference features match the requirements, then it is determined that the tree's identity has been replaced or significantly altered. If the environmental reference feature matching degree does not meet the requirements, then the environmental reference feature matching degree is frozen.

8. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 7, characterized in that, After determining whether the tree's identity has been replaced or significantly altered, the process further includes: The status code is marked to generate a new tree identification code, which is then associated with the original tree identification code.

9. The tree dynamic coding method based on high-precision positioning and visual analysis according to claim 1, characterized in that, The process of dynamically adjusting feature weights based on the matching degree between the tree's intrinsic features and historical tree intrinsic features, and the matching degree between the environmental reference features and historical environmental reference features, and fusing them with historical templates to generate new feature templates, which are then recorded in the tree's identity encoding, includes: When the matching degree between the tree body feature and the historical tree body feature, and the matching degree between the environmental reference feature and the historical environmental reference feature both exceed the first set threshold, the weight corresponding to the tree body feature and the weight corresponding to the environmental reference feature are kept close in proportion. If the matching degree between the environmental reference feature and the historical environmental reference feature is higher than a first set threshold, and the matching degree between the tree body feature and the historical tree body feature is lower than a second set threshold, then the weight corresponding to the environmental reference feature is increased, and the weight corresponding to the tree body feature is decreased to highlight the environmental feature. If the matching degree between the tree feature and the historical tree feature is higher than a first set threshold, and the matching degree between the environmental reference feature and the historical environmental reference feature is lower than a second set threshold, then the weight corresponding to the tree feature is increased, and the weight corresponding to the environmental reference feature is decreased to emphasize the tree feature. If the matching degree between the tree body feature and the historical tree body feature, and the matching degree between the environmental reference feature and the historical environmental reference feature are both lower than the second set threshold, the weight corresponding to the environmental reference feature is frozen to avoid erroneous updates.

10. A tree dynamic coding system based on high-precision positioning and visual analysis, characterized in that, include: The acquisition unit is used to acquire pose data and images collected by the device installed on the vehicle; A calculation unit is used to calculate the geographic coordinates of the tree based on the pose data; A preprocessing unit is used to preprocess the image to obtain a preprocessed image; The extraction unit is used to extract tree body features and environmental reference features based on the preprocessed image, and to establish the correspondence between the tree body features and the environmental reference features. The comparison unit is used to determine whether the tree is a tree that has been coded based on the tree's geographical coordinates, the tree's physical characteristics, and the characteristics of the environmental reference objects. The encoding generation unit is used to generate an identity code for the tree if the tree is not a tree with an existing code. The identity code of the tree includes a basic code and a dynamic extension code. The basic code includes the code corresponding to the geographic coordinates and a sequence number. The dynamic extension code includes a status code for indicating the current operation and maintenance status of the tree and a version code for recording the number of times the robust template has been updated.

Images