A vein extraction and morphology analysis method and system based on close-up lens imaging

Abstract

The present application relates to the technical field of image processing, and discloses a vein extraction and morphology analysis method and system based on close-up lens imaging. The method comprises the following steps: obtaining a leaf image of a target leaf obtained by a close-up lens imaging system; performing hierarchical vein feature extraction based on the leaf image; representing the target leaf by using a GIS network topology based on the extracted vein features; and performing morphology analysis on the target leaf based on the GIS network topology representation of the target leaf. The present application establishes a complete and highly automated vein morphology analysis system from leaf image acquisition, vein feature extraction, GIS network topology representation to morphology parameter calculation. It not only solves the problem that cost, portability and precision cannot be achieved simultaneously in the prior art, but also provides a powerful analysis tool for botany, ecology and evolutionary biology research through a standardized topology model and quantitative indicators with clear physiological significance.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a method and system for leaf vein extraction and morphological analysis based on close-up imaging. Background Technology

[0002] The branching characteristics of leaf vein networks are an important manifestation of plant adaptation to the environment, and their topology directly affects water transport efficiency and photosynthetic capacity. The number of branches reflects the complexity of the leaf vein system and resource allocation strategies, while the branching angle affects water flow resistance and mechanical support strength. The combination of these three morphological parameters reflects the plant's adaptation strategies to different environmental conditions. In arid environments, plants tend to increase vein density and branching number to improve water transport efficiency; in water-sufficient environments, they may adopt a lower branching number and a larger branching angle to reduce construction costs. By analyzing these parameters, we can assess the plant's adaptability to water stress, predict the plant's maximum photosynthetic rate, understand the plant's resource input strategies, and further conduct research on plant classification and evolution.

[0003] Precise measurement of leaf vein morphology parameters relies on high-quality imaging systems, requiring the capture of the complete structure from the large midrib to the fine terminal veins within a large field of view. Traditional leaf vein imaging methods face numerous limitations: ordinary macro lenses are expensive (a single lens often costs several thousand yuan), limiting the feasibility of large-scale field surveys; microscopes, while offering high resolution, have a small field of view, making it difficult to fully record the macroscopic vein network of the leaf, and their bulky nature makes them inconvenient for fieldwork; scanners, while portable, are ineffective on leathery leaves with poor light transmission, and it is difficult to control the imaging angle and lighting conditions. After obtaining leaf vein images, accurate leaf vein analysis remains a pressing issue. Summary of the Invention

[0004] This invention provides a method and system for leaf vein extraction and morphological analysis based on close-up imaging, in order to overcome the shortcomings of the prior art.

[0005] This invention provides a method for leaf vein extraction and morphological analysis based on close-up imaging, comprising:

[0006] Acquire leaf images of the target leaf using a close-up imaging system; Based on leaf images, hierarchical leaf vein feature extraction is performed; Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology. Based on the GIS network topology representation of the target blade, morphological analysis of the target blade is performed.

[0007] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the close-up imaging system is calibrated using Zhang's Calibration Method.

[0008] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the hierarchical leaf vein feature extraction based on leaf images includes: The leaf image is preprocessed, including any one or any combination of the following: distortion correction (removing lens distortion using calibration parameters), histogram equalization (enhancing the contrast between leaf veins and leaf tissue), color space conversion (converting from RGB to HSV or LAB space, separating luminance and chromaticity information), denoising (using bilateral filtering or nonlocal mean denoising to maintain edge sharpness); and / or, Based on the leaf image, a leaf segmentation algorithm based on Lab space and K-Means clustering is used to segment the leaf region from the background.

[0009] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the hierarchical leaf vein extraction based on leaf images includes: Based on leaf images, primary leaf vein features, secondary leaf vein features, and fine vein features are extracted. The extracted primary vein features, secondary vein features, and fine vein features are fused together, and all vein regions are skeletonized to obtain a centerline with a single pixel width. Morphological pruning is used to remove short branches and repair broken veins. Leaf vein hierarchical classification is performed using topological analysis.

[0010] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the fine vein feature extraction includes: Based on leaf images, semantic segmentation is performed using an improved convolutional neural network to obtain vein features; The improved convolutional neural network includes: an encoder that uses MobileNetV2 as the backbone network to extract features, a decoder that combines shallow details and deep semantic information with multi-scale feature fusion, and an attention mechanism that introduces spatial and channel attention modules to enhance leaf vein features.

[0011] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the fine vein feature extraction includes: Based on leaf images, an improved image segmentation model is used to perform hollow spatial pyramid pooling (ASPP) to capture contextual information at different scales to extract vein features. The improved image segmentation model uses MobileNetV2 as a lightweight backbone, and edge loss is added during the training of the improved image segmentation model to improve the accuracy of the extracted leaf vein boundaries.

[0012] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the target leaf is characterized using GIS network topology based on the extracted leaf vein features, including: Based on the extracted leaf vein features, a graph theory-based leaf vein network model is constructed, which includes undirected graph representation and directed graph representation.

[0013] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the morphological analysis of the target leaf is performed based on the GIS network topology representation of the target leaf, including: Based on the GIS network topology representation of the target blade, morphological analysis is performed on the target blade to obtain the number of branches, length ratio, and branch angle of the target blade. Based on the number of branches, length ratio, and branch angle of the target leaf, perform multi-parameter correlation analysis (including calculating the correlation coefficient between the number of branches and the length ratio, analyzing the relationship between the branch angle and the length ratio, and constructing a multi-dimensional feature vector for species classification or environmental adaptability analysis) and / or calculate the leaf vein complexity index.

[0014] In one embodiment, the number of branches refers to the number of times a first-order vein branches off from its superior vein. For example, the number of second-order veins branching off from a first-order main vein is the number of branches of the second-order veins on the first-order vein. The length ratio refers to the ratio of the length of an i-th order vein segment to the length of its directly connected (i-1)-th order vein segment, reflecting the scalar variation of the vein system. The branching angle refers to the angle formed between the child-order vein and the parent-order vein at the branching point, reflecting the spatial distribution pattern of the veins.

[0015] According to the present invention, a leaf vein extraction and morphological analysis method based on close-up imaging is provided, wherein the number of branches is obtained based on node degree statistics or topological traversal.

[0016] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging includes a method for calculating the length ratio, comprising: Measure the length of leaf vein segments; Establish a father-son relationship; The length ratio is obtained by dividing the length of the parent leaf vein segment by the length of the child leaf vein segment.

[0017] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging includes a method for calculating the branch angle, comprising: Identify the branch point P, the parent leaf vein segment AP, and the child leaf vein segment PB; Calculate the parent direction vector v1 = P – A; Calculate the sub-level direction vector v2 = B – P; Calculate the branch angle θ = arccos((v1·v2) / (|v1||v2|)); Convert to degrees θ_degree = θ×180 / π.

[0018] According to the present invention, a method for leaf vein extraction and morphological analysis based on close-up imaging is provided, wherein the expression for the leaf vein complexity index is: VCI = α·BN_norm + β·LR_norm + γ·BA_norm In the formula, VCI is the vein complexity index, BN, LR, and BA are the normalized number of branches, length ratio, and branch angle, respectively, and α, β, and γ are weighting coefficients.

[0019] This invention also provides a leaf vein extraction and morphological analysis system based on close-up imaging, comprising: The image acquisition module is used to: acquire leaf images of the target leaf obtained through a close-up imaging system; The leaf vein feature extraction module is used to perform hierarchical leaf vein feature extraction based on leaf images. The leaf network topology representation module is used to: represent the target leaf using GIS network topology based on the extracted leaf vein features; The blade morphology analysis module is used to perform morphological analysis on the target blade based on the GIS network topology representation of the target blade.

[0020] The present invention also provides an electronic device, including a processor and a memory storing a computer program, wherein the processor executes the computer program to implement any of the above-described methods for leaf vein extraction and morphological analysis based on close-up imaging.

[0021] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for leaf vein extraction and morphological analysis based on close-up imaging.

[0022] The present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute any of the above-described methods for leaf vein extraction and morphological analysis based on close-up imaging.

[0023] The present invention provides a method and system for leaf vein extraction and morphological analysis based on close-up imaging, which can bring at least the following beneficial effects: This invention employs a close-up imaging system as the image acquisition front-end, whose cost is far lower than that of professional macro lenses costing thousands of yuan, greatly reducing the equipment barrier and making large-scale field surveys possible. Simultaneously, to overcome image distortion that may be caused by close-up lenses, this invention uses the Zhang Zhengyou calibration method to calibrate the close-up imaging system, ensuring the accuracy of the physical measurement values ​​obtained from the images. This allows the method of this invention to achieve accuracy and reliability comparable to expensive equipment while maintaining low cost (field data can be collected using portable devices such as mobile phones) and ease of operation, overcoming the shortcomings of small microscope field of view, poor scanner performance for specific leaves (such as leathery leaves), and difficulty in controlling imaging conditions.

[0024] To address the multi-scale characteristics of leaf vein networks, ranging from coarse to fine, this invention designs a hierarchical leaf vein feature extraction process. By extracting and fusing features from first-level, second-level, and fine veins separately, the topological structure of the leaf vein network is fully preserved. Particularly for the most challenging fine vein features, this invention employs an improved deep learning model: by introducing a lightweight backbone network based on MobileNetV2 and attention mechanisms (spatial and channel attention modules), it ensures both extraction accuracy and computational efficiency; by combining Spatial Pyramid Pooling with Hollows (ASPP) and an edge loss function, it effectively captures contextual information at different scales and significantly improves the extraction accuracy of fine vein boundaries. This combination of techniques enables this invention to extract the complete hierarchical structure of leaf veins with high accuracy and automation, laying a solid data foundation for subsequent morphological analysis.

[0025] This invention transforms the extracted leaf vein network into a graph-based (directed / undirected graph) GIS network topology model. This standardized representation framework not only accurately describes the connectivity of leaf veins, but more importantly, it enables seamless integration of leaf vein data into existing, powerful GIS analysis tools and ecological models. This strong scalability provides unlimited possibilities for subsequent more complex spatial analyses, network structure comparisons, and correlation analyses with other geographic environmental data, significantly enhancing the application value of this invention.

[0026] This invention, based on GIS network topology representation, accurately calculates three core morphological parameters: the number of branches, the length ratio, and the branch angle. By performing multi-parameter correlation analysis on these three parameters (such as calculating correlation coefficients and constructing multidimensional feature vectors) and constructing the vein complexity index (VCI), this invention can quantitatively assess a plant's adaptability to water stress, predict photosynthetic rates, and conduct species classification and environmental adaptability studies, elevating the observation of vein morphology to a scientific level of quantitative analysis and functional prediction.

[0027] In summary, this invention establishes a complete and highly automated leaf vein morphology analysis system, encompassing leaf image acquisition, vein feature extraction, GIS network topology representation, and morphological parameter calculation. It not only solves the problem of balancing cost, portability, and accuracy in existing technologies but also provides a powerful analytical tool for botany, ecology, and evolutionary biology research through standardized topological models and quantitative indicators with clear physiological significance. Attached Figure Description

[0028] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0029] Figure 1 This is a flowchart illustrating a leaf vein extraction and morphological analysis method based on close-up imaging provided by the present invention.

[0030] Figure 2 This illustrates the principles of macro imaging and close-up imaging.

[0031] Figure 3 This illustrates the difference in imaging between the use of a close-up lens and the use of a close-up lens.

[0032] Figure 4 The results of leaf vein extraction and analysis of sweet cherry leaves photographed with a close-up lens are shown; the leaf vein extraction results and parameter definitions are shown in the figure.

[0033] Figure 5 Showing based on Figure 4 The analysis results of the number of bifurcations in the case shown.

[0034] Figure 6 Showing based on Figure 4 The analysis results of the length ratio in the case shown.

[0035] Figure 7 Showing based on Figure 4 The analysis results of the branch angle in the case shown. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, embodiments of this invention, and should not be construed as limiting the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. In the description of this invention, it should be understood that the terminology used is for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0037] Figure 1 This is a flowchart illustrating a method for leaf vein extraction and morphological analysis based on close-up imaging provided by the present invention. The executing entity of this method can be any applicable terminal-side device or network-side device, such as a leaf vein extraction and morphological analysis device based on close-up imaging.

[0038] See Figure 1 The present invention provides a method for leaf vein extraction and morphological analysis based on close-up imaging, which may include: S110, Acquiring an image through a close-up lens system ( Figure 2 The image of the target blade obtained by illustrating the imaging principle is shown.

[0039] In one embodiment, Zhang's Calibration Method is used to calibrate the close-up imaging system. The purpose of camera calibration is to determine the camera's intrinsic parameters (focal length, principal point position, distortion coefficient) and extrinsic parameters (camera position and orientation), establishing a mapping relationship between image coordinates and world coordinates. In leaf vein morphology measurement, accurate calibration is crucial for obtaining reliable parameters such as vein length and branch angles.

[0040] S120. Based on the leaf image, perform hierarchical leaf vein feature extraction.

[0041] In one embodiment, S120 can preprocess the leaf image, wherein the preprocessing includes any one or any combination of the following: distortion correction (removing lens distortion using calibration parameters), histogram equalization (enhancing the contrast between leaf veins and leaf tissue), color space conversion (converting from RGB to HSV or LAB space, separating luminance and chromaticity information), and denoising (using bilateral filtering or non-local mean denoising to maintain clear edges).

[0042] In one embodiment, S120 can segment the leaf region from the background based on the leaf image using a leaf segmentation algorithm based on Lab space and K-Means clustering. The core idea is: The Excess Green (ExG) index is calculated to enhance green areas and initially distinguish leaves from the background; the ExG image is adaptively binarized to obtain the initial foreground (leaves) and background masks; GrabCut is initialized using a fuzzy Gaussian mixture model (FGMM), and the foreground / background mask of GrabCut is automatically generated using ExG and Otsu results; Initialize with automatic masking and iteratively optimize the segmentation results; Morphological processing removes minor noise and fills in small holes to obtain the final blade mask.

[0043] In one embodiment, S120 may include: S1201. Based on the leaf image, extract the features of primary veins, secondary veins, and fine veins. Specifically, primary veins refer to the main veins that extend from the leaf base to the leaf tip, which are the thickest and most prominent. Secondary veins are secondary veins that periodically branch off from the primary veins and extend towards the leaf margin. Tertiary veins are the smallest and densest reticulate veins that connect with the secondary veins to form a fine network.

[0044] S1202. The extracted primary vein features, secondary vein features, and fine vein features are fused together, and all vein regions are skeletonized to obtain a centerline with a single pixel width. Short branches are removed using morphological pruning operations to repair broken veins.

[0045] In one embodiment, skeletonization can be used to extract the centerline of leaf veins, aiming to refine objects in an image into a single-pixel-wide "skeleton," typically used to extract the structural centerline or simplified form of objects in an image. Skeletonization operations are usually based on the repeated application of morphological operators (such as erosion, dilation, opening, and closing operations) until the target region in the image is compressed into a thin centerline, preserving the object's topological structure and connectivity features. The skeletonization process: (1) Erosion: The erosion operation shrinks the target area in the image and erodes the edges. After multiple erosions, the outline of the object in the image will gradually shrink until it reaches a single pixel width.

[0046] (2) Iterative operation: By repeatedly applying the erosion operation, the width of the leaf vein region is gradually reduced while retaining its main structure. After each erosion operation, some small and unnecessary branches are checked and removed, and finally a leaf vein skeleton with a width of one pixel is obtained.

[0047] (3) Repairing broken veins: During the skeletonization process, veins may break due to noise, image quality, or excessively thin branches. Morphological repair is typically used to repair these broken veins. This operation combines local dilation and erosion to connect broken areas in the skeleton, restoring the continuity of the vein structure.

[0048] (4) Removal of short branches: During the process of skeletalizing leaf veins, some branches that are too small or unimportant may be retained. These branches may not have a substantial impact on the main structure of the leaf veins, so they need to be removed through morphological pruning. The pruning operation usually determines whether to retain certain branches by setting a minimum length threshold. Branches smaller than this threshold will be deleted to ensure that the final extracted skeleton is more concise and highlights the main structure of the leaf veins.

[0049] S1203. Use topological analysis to classify leaf vein hierarchies.

[0050] In one embodiment, the process of extracting primary leaf vein features may include: (1) According to the definition, primary veins have a certain width. Therefore, we first perform multi-color space analysis on them, including HSV: H (hue), S (saturation), V (brightness). The veins are darker in the V channel; LAB: L (brightness), A (green-red axis), B (blue-yellow axis). The veins are darker in the L channel. We use the values ​​of the L and V channels to enhance the veins. Next, we use the CLANE parameter to enhance the contrast. Then, we perform sato filtering and binarization to obtain the vein area with width, which may still contain some secondary veins. (2) The extracted leaf vein results are coordinated, and then principal component analysis is performed to obtain the position and direction of the principal axis, i.e., the first-order leaf vein; combined with the extracted leaf vein skeleton, the two endpoints of the principal axis can be obtained. (3) Perform morphological preprocessing on the total leaf vein extraction results, including closure, dilation, and erosion; then perform distance transformation and skeletonization to obtain the leaf vein skeleton; use the leaf vein skeleton to create a network graph, including node creation, node connection, and connectivity enhancement; use the network graph of nodes, combined with the endpoint position, first find the node closest to the endpoint (because there may be offset), then use Dijkstra's algorithm to calculate the shortest path, that is, obtain the skeleton of the first-level leaf vein; use the skeleton, use the extracted leaf vein results to dilate, and the first-level leaf vein can be obtained.

[0051] In one embodiment, the process of extracting secondary leaf vein features may include: The results obtained based on Sobel detection and HSI detection mainly include primary and secondary leaf veins. Subtracting the primary leaf vein result from this result yields a rough estimate of the secondary leaf veins. After skeletonization, the secondary veins have little impact on the measurement of leaf vein density and can be used to obtain preliminary results. To enhance the extraction results, the following extraction method can be used: B-COSFIRE filtering based on biological visual mechanisms: Due to the high similarity between leaf veins and blood vessel morphology, this method has been proven effective for extracting leaf veins in complex backgrounds.

[0052] Principle: B-COSFIRE (Bar-Selective Combination of Shifted FilterResponses) is a configurable filter. It simulates neurons in the V1 region of the cerebral cortex, using a combination of responses from a set of Difference of Gaussians (DoG) functions to specifically detect lines of specific thickness and orientation.

[0053] Implementation: Reproduce using Python's COSFIRE package.

[0054] In one embodiment, the process of vein feature extraction may include: Based on leaf images, semantic segmentation is performed using an improved convolutional neural network to obtain vein features; The improved convolutional neural network includes: an encoder that uses MobileNetV2 as the backbone network to extract features, a decoder that combines shallow details and deep semantic information with multi-scale feature fusion, and an attention mechanism that introduces spatial and channel attention modules to enhance leaf vein features.

[0055] In one embodiment, the process of vein feature extraction may include: Based on leaf images, an improved image segmentation model is used to perform hollow spatial pyramid pooling (ASPP) to capture contextual information at different scales to extract vein features. The improved image segmentation model uses MobileNetV2 as a lightweight backbone, and edge loss is added during the training of the improved image segmentation model to improve the accuracy of the extracted leaf vein boundaries.

[0056] In one embodiment, the DeepLabV3 architecture can be used when extracting vein features. DeepLabV3 is an advanced semantic image segmentation model designed to improve image segmentation accuracy through efficient multi-scale feature fusion. A core feature of DeepLabV3 is its dilated convolution, a method that captures a wider range of contextual information by expanding the receptive field, making it particularly suitable for processing detailed images such as leaf veins. The decoder uses dilated spatial pyramid pooling (ASPP) to capture contextual information from different scales through multi-scale pooling operations, effectively improving segmentation accuracy, especially when processing complex structures like veins. Furthermore, ASPP significantly enhances the feature representation of vein regions by increasing multi-scale contextual information while maintaining computational efficiency. Combined with spatial and channel attention modules and edge loss, DeepLabV3 can further optimize vein feature extraction, ensuring accurate segmentation results and detailed representation.

[0057] The first method is based on a convolutional neural network (CNN) and uses the DeepLabV3 architecture for image segmentation. In this invention, DeepLabV3 uses dilated convolution to expand the receptive field, enabling the network to capture fine vein details while maintaining high computational performance. In the decoder part, DeepLabV3 uses dilated spatial pyramid pooling (ASPP) to fuse features at different scales. Simultaneously, to improve the representation of fine vein features, this embodiment adds a spatial attention module and a channel attention module to the network to further enhance the feature representation of the fine vein region. The spatial attention module helps the network focus on important regions in the image, while the channel attention module improves the feature representation of the fine vein region by dynamically adjusting the weights of each channel. In this way, the DeepLabV3 architecture can effectively extract fine vein features and enhance detail representation.

[0058] The second method uses the DeepLabV3 architecture and incorporates edge loss to improve the accuracy of vein boundary identification. In this method, DeepLabV3's ASPP module is also used to capture multi-scale contextual information and extract image features within a larger receptive field through dilated convolution. To further improve the segmentation accuracy of vein boundaries, edge loss is introduced during training. Edge loss helps the model more accurately identify vein contours and avoids blurry vein boundaries. The addition of edge loss makes the network's predictions near vein boundaries more accurate, ensuring that the extraction of vein regions is not only precise but also truly reflects details.

[0059] In implementation, a pre-trained DeepLabV3 model can be provided via PyTorch, which can be directly loaded and fine-tuned. DeepLabV3 itself integrates backbone networks such as MobileNetV2 and Xception, which can be selected according to needs. See [link to actual extraction results] for details. Figure 4-7 .

[0060] In one embodiment, the process of classifying leaf vein hierarchy using topological analysis may include: (1) Identify the main vein; (2) Identify branch points: detect the intersection points of the leaf vein skeleton; (3) Hierarchical marking: Starting from the main vein, the branches connected to the main vein are marked as secondary veins, the branches connected to the secondary veins are marked as tertiary veins, and so on; (4) Verification and adjustment: Combine the information on leaf vein thickness (width) to correct the erroneous markings.

[0061] S130. Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology.

[0062] Network topology includes the following basic elements: Node: The intersection point in the network, corresponding to the branch point or endpoint of a leaf vein; Edge: A line segment connecting two nodes, corresponding to a leaf vein segment; Connectivity: describes the connection relationships between edges and nodes; Directionality: The flow direction of the edge, corresponding to the hierarchical direction from the main vein to the fine veins in the leaf veins.

[0063] In one embodiment, S130 may include: Based on the extracted leaf vein features, the leaf vein network is represented as an undirected graph G=(V, E): (1) Vertex set V: contains all branch points and endpoints; (2) Edge set E: contains all leaf vein segments; (3) Vertex attributes: position coordinates (x, y), connectivity (degree); (4) Edge attributes: length, width, level, curvature.

[0064] Based on the extracted leaf vein features, the leaf vein network is represented as a directed graph: (1) Source point: petiole junction; (2) Confluence point: the terminal point of a leaf vein; (3) Side direction: from the main vein (lower level) to the fine vein (higher level).

[0065] Undirected and directed graph representations constitute a leaf vein network model based on graph theory.

[0066] S140. Based on the GIS network topology representation of the target blade, perform morphological analysis on the target blade.

[0067] In one embodiment, S140 may include: Based on the GIS network topology representation of the target blade, morphological analysis is performed on the target blade to obtain the number of branches, length ratio, and branch angle of the target blade. Based on the number of branches, length ratio, and branch angle of the target leaf, perform multi-parameter correlation analysis (including calculating the correlation coefficient between the number of branches and the length ratio, analyzing the relationship between the branch angle and the length ratio, and constructing a multi-dimensional feature vector for species classification or environmental adaptability analysis) and / or calculate the leaf vein complexity index.

[0068] In one embodiment, the number of branches refers to the number of times a first-order vein branches off from its superior vein. For example, the number of second-order veins branching off from a first-order main vein is the number of branches of the second-order veins on the first-order vein. The length ratio refers to the ratio of the length of an i-th order vein segment to the length of its directly connected (i-1)-th order vein segment, reflecting the scalar variation of the vein system. The branching angle refers to the angle formed between the child-order vein and the parent-order vein at the branching point, reflecting the spatial distribution pattern of the veins.

[0069] In one embodiment, the method for calculating the number of bifurcations includes: Method 1: Based on node degree statistics (1) Traverse all i-th order leaf vein segments; (2) For each i-th level leaf vein, identify its starting node (the node connected to the i-1 level leaf vein). (3) Count the number of i-th order leaf veins branching off from this node; (4) Calculate the average number of branches = total number of branches / number of leaf vein segments of order i-1; Method 2: Based on topological traversal (1) Starting from the main vein, use depth-first search to traverse the network; (2) For each i-1 level leaf vein segment, count the number of i-level leaf veins directly connected to it; (3) Record the number of branches at each branch point and construct a histogram of the number of branches; In one embodiment, the method for calculating the length ratio includes: Step 1: Measure the length of the leaf vein segment (1) For the skeletalized leaf veins, calculate the Euclidean distance length of each side (leaf vein segment); (2) If the leaf vein is a curve, calculate its arc length: accumulate the distance between pixels along the skeleton; (3) Use calibration parameters to convert pixel length to physical length (mm); Step 2: Establish the father-son relationship (1) For each i-th level leaf vein segment, identify its starting node; (2) The i-1 level leaf vein segments connected by the nodes serve as the parent level; (3) Record the father-son relationship pairs; Step 3: Calculate the length ratio Length ratio R = L_child / L_parent Where L_child is the length of the child-level leaf vein segment, and L_parent is the length of the parent-level leaf vein segment.

[0070] For each level transition (e.g., level 1 to level 2, level 2 to level 3), calculate: (1) Mean length ratio (2) Standard deviation of length ratio (3) Median of length ratio (4) Quartiles of length ratio (25%, 75%) (2) Length proportion distribution curve.

[0071] In one embodiment, the method for calculating the branch angle includes: Identify the branch point P, the parent leaf vein segment AP, and the child leaf vein segment PB; Calculate the parent direction vector v1 = P – A; Calculate the sub-level direction vector v2 = B – P; Calculate the branch angle θ = arccos((v1·v2) / (|v1||v2|)); Convert to degrees θ_degree = θ×180 / π.

[0072] In one embodiment, the expression for the vein complexity index is: VCI = α·BN_norm + β·LR_norm + γ·BA_norm In the formula, VCI is the vein complexity index, BN, LR, and BA are the normalized number of branches, length ratio, and branch angle, respectively, and α, β, and γ are weighting coefficients.

[0073] This invention establishes a complete methodology for leaf vein extraction and morphological analysis based on close-up imaging, proposes a calibration method for close-up imaging systems to ensure accurate physical measurement values, combines traditional image processing and deep learning methods to achieve high-precision extraction of leaf veins at different levels, introduces a GIS network topology model to establish a representation framework for the leaf vein network, elaborates on the calculation methods for three core parameters—the number of branches, the length ratio, and the branch angle—and reveals the physiological and ecological significance of leaf vein morphological parameters.

[0074] This invention establishes a complete and highly automated leaf vein morphology analysis system, encompassing leaf image acquisition, vein feature extraction, GIS network topology representation, and morphological parameter calculation. It not only solves the problem of balancing cost, portability, and accuracy in existing technologies, but also provides a powerful analytical tool for botany, ecology, and evolutionary biology research through standardized topological models and quantitative indicators with clear physiological significance.

[0075] The leaf vein extraction and morphology analysis system based on close-up imaging provided by this invention will be described below. The leaf vein extraction and morphology analysis system based on close-up imaging described below can be referred to in correspondence with the leaf vein extraction and morphology analysis method based on close-up imaging described above.

[0076] This invention provides a leaf vein extraction and morphological analysis system based on close-up imaging, which may include: The image acquisition module is used to: acquire leaf images of the target leaf obtained through a close-up imaging system; The leaf vein feature extraction module is used to perform hierarchical leaf vein feature extraction based on leaf images. The leaf network topology representation module is used to: represent the target leaf using GIS network topology based on the extracted leaf vein features; The blade morphology analysis module is used to perform morphological analysis on the target blade based on the GIS network topology representation of the target blade.

[0077] The present invention provides an electronic device that may include: a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other via the communication bus 840. The processor 810 can call logical instructions in the memory 830 to execute the following steps: Acquire leaf images of the target leaf using a close-up imaging system; Based on leaf images, hierarchical leaf vein feature extraction is performed; Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology. Based on the GIS network topology representation of the target blade, morphological analysis of the target blade is performed.

[0078] Furthermore, the logical instructions in the aforementioned memory 830 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0079] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, and the computer program being executed by a processor, enabling the computer to perform the following steps: Acquire leaf images of the target leaf using a close-up imaging system; Based on leaf images, hierarchical leaf vein feature extraction is performed; Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology. Based on the GIS network topology representation of the target blade, morphological analysis of the target blade is performed.

[0080] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps: Acquire leaf images of the target leaf using a close-up imaging system; Based on leaf images, hierarchical leaf vein feature extraction is performed; Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology. Based on the GIS network topology representation of the target blade, morphological analysis of the target blade is performed.

[0081] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0082] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0083] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for leaf vein extraction and morphological analysis based on close-up imaging, characterized in that, include: Acquire leaf images of the target leaf using a close-up imaging system; Based on leaf images, hierarchical leaf vein feature extraction is performed; Based on the extracted leaf vein features, the target leaf is characterized using GIS network topology. Based on the GIS network topology representation of the target blade, morphological analysis of the target blade is performed.

2. The leaf vein extraction and morphological analysis method based on close-up imaging according to claim 1, characterized in that, The hierarchical leaf vein feature extraction based on leaf images includes: The leaf image is preprocessed, wherein the preprocessing includes any one or any combination of the following: distortion correction, histogram equalization, color space conversion, and denoising; and / or, Based on the leaf image, a leaf segmentation algorithm based on Lab space and K-Means clustering is used to segment the leaf region from the background.

3. The leaf vein extraction and morphological analysis method based on close-up imaging according to claim 2, characterized in that, The hierarchical leaf vein extraction based on leaf images includes: Based on leaf images, primary leaf vein features, secondary leaf vein features, and fine vein features are extracted. The extracted primary vein features, secondary vein features, and fine vein features are fused together, and all vein regions are skeletonized to obtain a centerline with a single pixel width. Morphological pruning is used to remove short branches and repair broken veins. Leaf vein hierarchical classification is performed using topological analysis.

4. The leaf vein extraction and morphological analysis method based on close-up imaging according to claim 3, characterized in that, The vein feature extraction includes: Based on leaf images, semantic segmentation is performed using an improved convolutional neural network to obtain vein features; The improved convolutional neural network includes: an encoder that uses MobileNetV2 as the backbone network to extract features; a decoder that employs multi-scale feature fusion to combine shallow details and deep semantic information; and an attention mechanism that introduces spatial and channel attention modules to enhance leaf vein features; or, The vein feature extraction includes: Based on leaf images, an improved image segmentation model is used to perform hollow spatial pyramid pooling to capture contextual information at different scales to extract vein features. The improved image segmentation model uses MobileNetV2 as a lightweight backbone, and edge loss is added during the training of the improved image segmentation model to improve the accuracy of extracted leaf vein boundaries.

5. The leaf vein extraction and morphological analysis method based on close-up imaging according to claim 3, characterized in that, The method of characterizing the target leaf using GIS network topology based on the extracted leaf vein features includes: Based on the extracted leaf vein features, a graph theory-based leaf vein network model is constructed, which includes undirected graph representation and directed graph representation.

6. The method for leaf vein extraction and morphological analysis based on close-up imaging according to any one of claims 1-5, characterized in that, The GIS network topology representation based on the target blade includes morphological analysis of the target blade, comprising: Based on the GIS network topology representation of the target blade, morphological analysis is performed on the target blade to obtain the number of branches, length ratio, and branch angle of the target blade. Based on the number of branches, length ratio, and branch angle of the target leaf, multi-parameter correlation analysis and / or leaf vein complexity index calculation are performed.

7. The leaf vein extraction and morphological analysis method based on close-up imaging according to claim 6, characterized in that, The number of branches is obtained based on node degree statistics or topological traversal. And / or, Methods for calculating length ratios include: Measure the length of leaf vein segments; Establish a father-son relationship; The length ratio is obtained by dividing the length of the parent vein segment by the length of the child vein segment; and / or, The methods for calculating the branch angle include: Identify the branch point P, the parent leaf vein segment AP, and the child leaf vein segment PB; Calculate the parent direction vector v1 = P – A; Calculate the sub-level direction vector v2 = B – P; Calculate the branch angle θ = arccos((v1·v2) / (|v1||v2|)); Convert to degrees θ_degree = θ × 180 / π; and / or, The expression for the vein complexity index is: VCI = α·BN_norm + β·LR_norm + γ·BA_norm In the formula, VCI is the vein complexity index, BN, LR, and BA are the normalized number of branches, length ratio, and branch angle, respectively, and α, β, and γ are weighting coefficients.

8. A system for leaf vein extraction and morphological analysis based on close-up imaging, characterized in that, include: The image acquisition module is used to: acquire leaf images of the target leaf obtained through a close-up imaging system; The leaf vein feature extraction module is used to perform hierarchical leaf vein feature extraction based on leaf images. The leaf network topology representation module is used to: represent the target leaf using GIS network topology based on the extracted leaf vein features; The blade morphology analysis module is used to perform morphological analysis on the target blade based on the GIS network topology representation of the target blade.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the leaf vein extraction and morphological analysis method based on close-up imaging as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the leaf vein extraction and morphological analysis method based on close-up imaging as described in any one of claims 1 to 7.

Images