Plant classification labeling method

Abstract

The invention provides a plant classification labeling method. The method comprises the following steps: S1, acquiring an orthoimage of a target area through an unmanned aerial vehicle, and segmenting to generate plant pattern spots; s2, superposing the orthoimage and the base map of the electronic map, and establishing a composite coordinate system; s3, displaying the superposed map on the mobile terminal, and receiving a marking instruction of a user on the pattern spots; when the user cannot identify the plant type, calling a plant identification API to classify the plant image uploaded by the user, and associating the result to the corresponding plant pattern spot; s4, storing the labeled plant data to a database, and supporting statistics and management of the vegetation types, number and distribution of the target area through a geofence function; s5, a visual interface is provided, a user is allowed to check detailed information of the plants by clicking map marks, the detailed information comprises names, photos and statistical results, through the plant classification marking method, the types, the number and the distribution of the vegetation in the statistical area can be systematically analyzed, data are accurate, and presentation is visual.

Description

Technical Field

[0001] The present invention belongs to the technical field of plant classification and labeling, and more specifically relates to a plant classification and labeling method. Background Art

[0002] Plant diversity is the foundation of biodiversity, and commonly used plant diversity assessments are based on plant classification. Currently, however, plant classifications are based solely on plant morphology, physiology, and life form, such as broad-leaved trees, coniferous trees, shrubs, perennial herbs, and annual herbs. This classification is suitable for field research because all plants grow naturally and are not directly influenced by human factors (such as direct intervention such as species screening, planting, maintenance, and management). However, this classification alone is inappropriate for studying plant diversity in urban areas.

[0003] And the existing landscaping managers conduct manual inventory to count the types, quantity and distribution of vegetation in the area.

[0004] Existing technical books have the problem that manual inventory methods are prone to omissions and duplications, making it difficult to obtain accurate data and unable to provide intuitive data presentation. Summary of the Invention

[0005] Therefore, in order to solve the above technical problems, the present invention proposes a plant classification and labeling method, comprising the following steps: S1, obtaining an orthophoto of the target area through a drone, and segmenting and generating plant patches; S2, superimposing the orthophoto with the electronic map base to establish a composite coordinate system; S3, displaying the superimposed map on the mobile terminal, and receiving the user's labeling instructions for the patches; when the user cannot identify the plant species, calling the plant recognition API to classify the plant images uploaded by the user, and associating the results with the corresponding plant patches; S4, storing the labeled plant data in a database, supporting the statistics and management of the vegetation species, quantity and distribution of the target area through the geographic fence function; S5, providing a visual interface, allowing users to view detailed information of the plant by clicking on the map mark, including name, photo and statistical results. Through this plant classification and labeling method, the vegetation species, quantity and distribution of the statistical area can be systematically analyzed, saving manpower, and the data is accurate and intuitive.

[0006] A plant classification and marking method comprises the following steps:

[0007] S1, obtain the orthophoto of the target area through the UAV and segment it to generate plant patches;

[0008] S2, superimposing the orthophoto image with the electronic map base map to establish a composite coordinate system;

[0009] S3: Display the superimposed map on the mobile terminal and receive the user's instructions for labeling the map spots. If the user cannot identify the plant species, call the plant recognition API to classify the plant image uploaded by the user and associate the result with the corresponding plant map spot.

[0010] S4 stores the labeled plant data in the database, supporting the statistics and management of vegetation types, quantities, and distribution in the target area through the geo-fencing function;

[0011] S5,provides a visual interface that allows users to view detailed information of plants, including names, photos, and statistical results,by clicking on map markers.

[0012] Furthermore, in S1, the orthophoto of the target area is obtained by using a multispectral camera mounted on a drone, and the orthophoto stitching and coordinate calibration are achieved through the SLAM algorithm. The obtained orthophoto data is used to distinguish vegetation and non-vegetation areas through NDVI calculation, patch segmentation and vectorization to generate plant patches.

[0013] Furthermore, the spectral band range of the multispectral camera is 400-1000nm, and the spatial resolution is not less than 0.5m.

[0014] Furthermore, in S2, the electronic map base map is selected from one of Tiandi Map or Amap, and the coordinate system of the electronic map base map is WGS-84 or CGCS2000.

[0015] Furthermore, in S2, to solve the problem of coordinate offset between the orthophoto and the electronic map, an affine transformation algorithm is used for registration, and the error is controlled within ±0.5 meters. Specifically, the following steps are included:

[0016] S2.1, coordinate system unification: unify the original coordinate system of the orthophoto and the coordinate system of the electronic map base map (selected from Tiandi Map or Amap) to WGS-84 or CGCS2000;

[0017] S2.2, Control Point Selection: Manually or automatically select at least four pairs of control points with the same name on the orthophoto and electronic map base. The control points are locations with stable geographical features, such as road intersections, building vertices, or fixed landmarks.

[0018] S2.3, affine transformation parameter calculation: Based on the coordinate pairs of the control points, the six parameters of the affine transformation matrix (including translation, rotation, scaling, and shearing parameters) are solved by the least squares method to establish a coordinate mapping relationship from the orthophoto to the electronic map;

[0019] S2.4, image resampling and correction: performing geometric correction on the orthophoto according to the affine transformation matrix, resampling the image pixels using bilinear interpolation to eliminate geometric distortion;

[0020] S2.5, Error Verification and Iterative Optimization: Calculate the residual between the corrected image and the electronic map. If the residual exceeds ±0.5 meters, readjust the control points or optimize the transformation parameters until the accuracy requirements are met.

[0021] S2.6, Image overlay output: Overlay the corrected orthophoto with the electronic map base at the pixel level to generate a fused map in a composite coordinate system for subsequent annotation and analysis.

[0022] Furthermore, in S3, the labeling instruction content includes the geographical location of the plant, the species name and the associated plant photos.

[0023] Furthermore, the plant photos include multi-angle photos of distant views, close views and detailed features, wherein the detailed feature photos must clearly show the leaf veins, inflorescences or fruit morphology, and the resolution of a single photo must not be less than 20 million pixels.

[0024] Furthermore, in S3, the annotation instruction content supports manual editing, including adding, modifying or removing plant tags. The editing operation needs to be managed through user authority levels, and administrator authority can overwrite the annotation data of ordinary users.

[0025] Furthermore, in S3, the plant identification API uses a machine learning model to extract and classify plant photos and output plant species determination results. The model is trained based on public plant image datasets (such as PlantCLEF) with a classification accuracy of no less than 90%. It calls the ResNet-50 pre-trained model and fine-tunes it for the local plant library, with an identification accuracy of >95%.

[0026] Furthermore, the annotated data is dynamically updated synchronously with the drone aerial images to ensure real-time data. The synchronization mechanism adopts an incremental update strategy, and the update frequency can be configured to be hourly or daily.

[0027] Furthermore, in S4, the geo-fence function is used to set the boundaries of the management area and automatically count the types, quantities and spatial distribution data of vegetation within the boundaries. The spatial clustering algorithm (DBSCAN) is used in the statistical process to optimize the accuracy of the distribution data.

[0028] Furthermore, in S4, the statistical results include one or more of vegetation coverage, species ratio and distribution heat map, wherein the distribution heat map is generated based on a kernel density estimation algorithm (KDE).

[0029] Furthermore, in S5, the visualization interface supports cross-platform access, including mobile terminals, tablet devices and computer browsers, and the visualization interface uses WebGL technology to achieve three-dimensional map rendering.

[0030] Beneficial effects of the present invention: The present invention proposes a plant classification and labeling method, comprising the following steps: S1, obtaining an orthophoto of a target area through a drone, and segmenting and generating plant patches; S2, superimposing the orthophoto with an electronic map base map to establish a composite coordinate system; S3, displaying the superimposed map on a mobile terminal, and receiving a user's labeling instruction for the patches; when the user is unable to identify the plant species, calling a plant recognition API to classify the plant image uploaded by the user, and associating the result with the corresponding plant patch; S4, storing the labeled plant data in a database, and supporting statistics and management of the vegetation species, quantity and distribution in the target area through a geographic fence function; S5, providing a visual interface, allowing users to view detailed information of the plant by clicking on a map mark, including name, photo and statistical results. Through this plant classification and labeling method, the vegetation species, quantity and distribution in the statistical area can be systematically analyzed, manpower is saved, the data is accurate, and the presentation is intuitive. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 The present invention provides a flow chart of a plant classification and labeling method.

[0032] The following specific embodiments will further illustrate the present invention in conjunction with the above-mentioned drawings. DETAILED DESCRIPTION

[0033] The following examples are described to assist in understanding the present application, and the examples are not and should not be interpreted in any way as limiting the scope of protection of the present application.

[0034] In the following description, those skilled in the art will recognize that throughout this discussion, components may be described as separate functional units (which may include sub-units), but those skilled in the art will recognize that various components or portions thereof may be divided into separate components or may be integrated together (including within a single system or component).

[0035] At the same time, the connections between components or systems are not intended to be limited to direct connections. Instead, data between these components may be modified, reformatted, or otherwise changed by intermediate components. In addition, additional or fewer connections may be used. It should also be noted that the terms "coupled," "connected," or "input" should be understood to include direct connections, indirect connections through one or more intermediate devices, and wireless connections. Example 1:

[0036] like Figure 1The figure shows a flow chart of a plant classification and labeling method of the present invention.

[0037] A plant classification and marking method comprises the following steps:

[0038] S1, obtain the orthophoto of the target area through the UAV and segment it to generate plant patches;

[0039] S2, superimposing the orthophoto image with the electronic map base map to establish a composite coordinate system;

[0040] S3: Display the superimposed map on the mobile terminal and receive the user's instructions for labeling the map spots. If the user cannot identify the plant species, call the plant recognition API to classify the plant image uploaded by the user and associate the result with the corresponding plant map spot.

[0041] S4 stores the labeled plant data in the database, supporting the statistics and management of vegetation types, quantities, and distribution in the target area through the geo-fencing function;

[0042] S5,provides a visual interface that allows users to view detailed information of plants, including names, photos, and statistical results,by clicking on map markers.

[0043] In S1, a multispectral camera mounted on a drone is used to obtain orthophotos of the target area. The orthophotos are stitched and calibrated using the SLAM algorithm. The obtained orthophoto data is then used to distinguish vegetation from non-vegetation areas through NDVI calculation, patch segmentation, and vectorization to generate plant patches.

[0044] The spectral band range of the multispectral camera is 400-1000nm, and the spatial resolution is not less than 0.5m.

[0045] In S2, the electronic map base map is selected from one of Tiandi Map and Amap, and the coordinate system of the electronic map base map is WGS-84 or CGCS2000.

[0046] In S2, to solve the problem of coordinate offset between the orthophoto and the electronic map, an affine transformation algorithm is used for registration, and the error is controlled within ±0.5 meters. The specific steps include:

[0047] S2.1, coordinate system unification: unify the original coordinate system of the orthophoto and the coordinate system of the electronic map base map (selected from Tiandi Map or Amap) to WGS-84 or CGCS2000;

[0048] S2.2, Control Point Selection: Manually or automatically select at least four pairs of control points with the same name on the orthophoto and electronic map base. The control points are locations with stable geographical features, such as road intersections, building vertices, or fixed landmarks.

[0049] S2.3, affine transformation parameter calculation: Based on the coordinate pairs of the control points, the six parameters of the affine transformation matrix (including translation, rotation, scaling, and shearing parameters) are solved by the least squares method to establish a coordinate mapping relationship from the orthophoto to the electronic map;

[0050] S2.4, image resampling and correction: performing geometric correction on the orthophoto according to the affine transformation matrix, resampling the image pixels using bilinear interpolation to eliminate geometric distortion;

[0051] S2.5, Error Verification and Iterative Optimization: Calculate the residual between the corrected image and the electronic map. If the residual exceeds ±0.5 meters, readjust the control points or optimize the transformation parameters until the accuracy requirements are met.

[0052] S2.6, Image overlay output: Overlay the corrected orthophoto with the electronic map base at the pixel level to generate a fused map in a composite coordinate system for subsequent annotation and analysis.

[0053] In S3, the labeling instruction content includes the geographical location of the plant, the species name and the associated plant photos.

[0054] The plant photos include multi-angle photos of distant views, close views and detailed features, among which the detailed feature photos must clearly show the leaf veins, inflorescences or fruit morphology, and the resolution of a single photo must not be less than 20 million pixels.

[0055] In S3, the annotation instruction content supports manual editing, including adding, modifying or removing plant tags. Editing operations must be managed through user permission levels, and administrator permissions can overwrite ordinary user annotation data.

[0056] In S3, the plant identification API uses a machine learning model to extract and classify plant photos and output plant species determination results. The model is trained based on public plant image datasets (such as PlantCLEF) with a classification accuracy of no less than 90%. It calls the ResNet-50 pre-trained model and fine-tunes it for the local plant library, with an identification accuracy of >95%.

[0057] The labeled data is dynamically updated synchronously with the drone aerial images to ensure real-time data. The synchronization mechanism adopts an incremental update strategy, and the update frequency can be configured to be hourly or daily.

[0058] In S4, the geo-fence function is used to set the boundaries of the management area and automatically count the types, quantities and spatial distribution data of vegetation within the boundaries. The spatial clustering algorithm (DBSCAN) is used in the statistical process to optimize the accuracy of the distribution data.

[0059] In S4, the statistical results include one or more of vegetation coverage, species ratio, and distribution heat map, wherein the distribution heat map is generated based on a kernel density estimation algorithm (KDE).

[0060] In S5, the visualization interface supports cross-platform access, including mobile terminals, tablet devices and computer browsers, and the visualization interface uses WebGL technology to achieve three-dimensional map rendering.

[0061] Beneficial effects of the present invention: The present invention proposes a plant classification and labeling method, comprising the following steps: S1, obtaining an orthophoto of a target area through a drone, and segmenting and generating plant patches; S2, superimposing the orthophoto with an electronic map base map to establish a composite coordinate system; S3, displaying the superimposed map on a mobile terminal, and receiving a user's labeling instruction for the patches; when the user is unable to identify the plant species, calling a plant recognition API to classify the plant image uploaded by the user, and associating the result with the corresponding plant patch; S4, storing the labeled plant data in a database, and supporting statistics and management of the vegetation species, quantity and distribution in the target area through a geographic fence function; S5, providing a visual interface, allowing users to view detailed information of the plant by clicking on a map mark, including name, photo and statistical results. Through this plant classification and labeling method, the vegetation species, quantity and distribution in the statistical area can be systematically analyzed, manpower is saved, the data is accurate, and the presentation is intuitive.

[0062] The above-described embodiments merely illustrate several implementations of the present invention, and while their descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art would be able to make numerous variations and improvements without departing from the spirit of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of the present invention shall be determined by the appended claims.

Claims

1. A plant classification and labeling method, characterized by: The following steps are included: S1, obtain the orthophoto of the target area through the UAV and segment it to generate plant patches; S2, superimposing the orthophoto image with the electronic map base map to establish a composite coordinate system; S3: Display the superimposed map on the mobile terminal and receive the user's instructions for labeling the map spots. If the user cannot identify the plant species, call the plant recognition API to classify the plant image uploaded by the user and associate the result with the corresponding plant map spot. S4 stores the labeled plant data in the database, supporting the statistics and management of vegetation types, quantities, and distribution in the target area through the geo-fencing function; S5,provides a visual interface that allows users to view detailed information of plants, including names, photos, and statistical results,by clicking on map markers.

2. The plant classification labeling method according to claim 1, characterized in that: In S1, a multispectral camera mounted on a drone is used to obtain orthophotos of the target area. The orthophotos are stitched and calibrated using the SLAM algorithm. The obtained orthophoto data is then used to distinguish vegetation from non-vegetation areas through NDVI calculation, patch segmentation, and vectorization to generate plant patches.

3. The plant classification labeling method according to claim 1, characterized in that: In S2, to solve the problem of coordinate offset between the orthophoto and the electronic map, an affine transformation algorithm is used for registration, and the error is controlled within ±0.5 meters. The specific steps include: S2.1, coordinate system unification: unify the original coordinate system of the orthophoto and the coordinate system of the electronic map base map (selected from Tiandi Map or Amap) to WGS-84 or CGCS2000; S2.2, Control Point Selection: Manually or automatically select at least four pairs of control points with the same name on the orthophoto and electronic map base. The control points are locations with stable geographical features, such as road intersections, building vertices, or fixed landmarks. S2.3, affine transformation parameter calculation: Based on the coordinate pairs of the control points, the six parameters of the affine transformation matrix (including translation, rotation, scaling, and shearing parameters) are solved by the least squares method to establish a coordinate mapping relationship from the orthophoto to the electronic map; S2.4, image resampling and correction: performing geometric correction on the orthophoto according to the affine transformation matrix, resampling the image pixels using bilinear interpolation to eliminate geometric distortion; S2.5, Error Verification and Iterative Optimization: Calculate the residual between the corrected image and the electronic map. If the residual exceeds ±0.5 meters, readjust the control points or optimize the transformation parameters until the accuracy requirements are met. S2.6, Image overlay output: Overlay the corrected orthophoto with the electronic map base at the pixel level to generate a fused map in a composite coordinate system for subsequent annotation and analysis.

4. The plant classification labeling method according to claim 1, characterized in that: In S3, the labeling instruction content includes the geographical location of the plant, the species name and the associated plant photos.

5. The plant classification labeling method according to claim 4, characterized in that: The plant photos include multi-angle photos of distant views, close views and detailed features, among which the detailed feature photos must clearly show the leaf veins, inflorescences or fruit morphology, and the resolution of a single photo must not be less than 20 million pixels.

6. The plant classification labeling method according to claim 1, characterized in that: In S3, the annotation instruction content supports manual editing, including adding, modifying or removing plant tags. Editing operations must be managed through user permission levels, and administrator permissions can overwrite ordinary user annotation data.

7. The plant classification labeling method according to claim 1, characterized in that: In S3, the plant identification API uses a machine learning model to extract and classify plant photos and output plant species determination results. The model is trained based on public plant image datasets (such as PlantCLEF), calls the ResNet-50 pre-trained model, and fine-tunes it for the local plant library.

8. The plant classification labeling method according to claim 6, characterized in that: The labeled data is dynamically updated synchronously with the drone aerial images to ensure real-time data. The synchronization mechanism adopts an incremental update strategy, and the update frequency can be configured to be hourly or daily.

9. The plant classification labeling method according to claim 1, characterized in that: In S4, the geo-fence function is used to set the boundaries of the management area and automatically count the types, quantities and spatial distribution data of vegetation within the boundaries. The spatial clustering algorithm (DBSCAN) is used in the statistical process to optimize the accuracy of the distribution data.

10. The plant classification labeling method according to claim 1, characterized in that: In S4, the statistical results include one or more of vegetation coverage, species ratio, and distribution heat map, wherein the distribution heat map is generated based on a kernel density estimation algorithm (KDE).

Images