A method for constructing a foraminifera morphological feature database

Abstract

The present application relates to the technical field of database construction, and particularly relates to a construction method of a foraminifera morphological feature database, comprising the following steps: S1, collecting foraminifera images from multiple sources, identifying scale drift and shooting deviation based on principal axis normalization and local vorticity field, performing structure-preserving image enhancement, and outputting an enhanced image set consistent in shell substance, shell arrangement, shell opening, suture line and micro-pore features; S2, extracting a multi-scale shell structure atlas, constructing a structure-semantic coupled label nested tree, performing polysemy recognition and ambiguity resolution, and generating a structured annotation set; S3, fusing the image set and the annotation set, and constructing a three-layer database model containing a part layer, an attribute layer and a variation layer according to the label nested tree. Through image structure enhancement, semantic ambiguity resolution and three-layer database construction, the present application effectively improves the recognition accuracy of foraminifera morphological features, the label organization logicality and the information retrieval efficiency.

Description

Technical Field

[0001] This invention relates to the field of database construction technology, and in particular to a method for constructing a database of foraminifera morphological characteristics. Background Technology

[0002] Foraminifera, as important biostratigraphic indicator fossils, are widely used in paleoenvironmental reconstruction, stratigraphic correlation, and marine ecological change research due to their shell morphology. Given their diverse morphological types and complex shell structures, constructing a high-quality morphological database is of great value for improving species identification accuracy and supporting automated classification. In recent years, with the development of digital image acquisition and intelligent recognition technologies, establishing image-based foraminifera morphological databases has gradually become a research hotspot, especially with broad application prospects in ecological geological surveys, marine monitoring, and intelligent identification of microfossils.

[0003] Existing foraminifera databases largely rely on manual naming and static image annotation, lacking systematic modeling of image structural consistency and ambiguous labels. This results in incomplete image feature representation and weak label logic, making it difficult to support automated extraction of complex structural information. Furthermore, during image acquisition, inconsistencies in sample pose, shooting angle, and scale often lead to problems such as local blurring and structural distortion. Traditional image enhancement methods struggle to maintain consistency in microstructural features, resulting in insufficient reliability of sample structures in the database and failing to meet the requirements for high-precision recognition and deep structural semantic modeling. Summary of the Invention

[0004] This invention provides a method for constructing a foraminifera morphological feature database. It builds a database with structural consistency and semantic coupling, proposes a structure-preserving image enhancement method to improve the consistency and recognizability of key structures such as shells, apertures, sutures, and micropores in images, and achieves logical organization of ambiguous labels through multi-scale structural mapping and label ambiguity resolution mechanisms. Finally, it establishes a three-layer database model including a location layer, an attribute layer, and a variation layer, significantly improving the recognizability of foraminifera image data, the logical consistency of label organization, and the database's retrieval and analysis capabilities.

[0005] A method for constructing a database of foraminifera morphological characteristics includes the following steps: S1: Acquire original images of foraminifera from multiple sources. Based on shell principal axis normalization and local curl field calculation, identify potential scale drift and non-ideal shooting angles, perform structure-preserving image enhancement processing, and output a set of morphologically enhanced images that maintain the consistency of local structures such as shell material (porcelain, glass, cement), shell arrangement (single chamber, single row, double row, plano-spiral, spiral), aperture, suture, and shell wall micropores. S2, based on the output morphologically enhanced image set, extracts multi-scale shell structure maps, constructs a structure-semantic coupled label nesting tree, executes a label recommendation and ambiguity resolution process based on a polysemy recognition mechanism, and outputs a structured annotation set with structural logical paths; S3 combines the morphologically enhanced image set with the structured annotation set, and establishes a three-layer database model based on the nested label tree structure, including the part layer (shell, orifice), the attribute layer (morphology, texture, thickness) and the variation layer (evolutionary trend, environmental response characteristics).

[0006] Optionally, S1 includes: S11. Obtain original images of foraminifera from field collection, laboratory shooting and historical database. To address the inconsistency in the spatial arrangement of foraminifera shells in the original images, an image alignment algorithm based on principal axis direction estimation is used to normalize the shells and calculate the local curl field of the original images of foraminifera. This is used to capture local rotation features and scale drift anomalies caused by shooting angle deviation or non-standard sample posture, and to identify distorted regions under non-ideal imaging conditions. S12, after completing normalization and local distortion recognition, performs structure-preserving image enhancement processing based on the sensitivity of shell microstructure. With the preservation of shell material type (such as porcelain, glass, cement) and microstructure consistency as the core, it adopts a combination of guided filtering, edge enhancement and texture reconstruction to enhance the local features of shell arrangement (single chamber, single row, double row, planar spiral, spiral), shell aperture morphology, suture texture and shell wall micropore distribution, while avoiding distortion of the original morphological structure, and finally generating a morphological enhancement image set with local structural consistency.

[0007] Optionally, S11 includes: S111, collecting raw images of foraminifera from multiple sources, specifically including: Field image acquisition: Images of foraminifera in sediment samples were captured in the field using portable digital microscopes, and metadata was recorded by combining GPS and sample point numbers; Laboratory images: Images of cleaned foraminifera individuals were captured under an optical microscope, with bright field, polarization mode, and multiple shooting angles. Historical database images: Access public databases and literature atlases, and perform unified image resolution conversion, noise reduction processing, and standard naming; All original foraminifera images were converted to a uniform grayscale format and normalized to a fixed resolution. S112, The shell in the original image of foraminifera is normalized and rotated using the image principal axis direction estimation algorithm; S113, calculate the local horizontal and vertical gradients of the original foraminifera image after normalization and rotation processing, and calculate the two-dimensional local curl field. ,like If the region is deemed to be a local structural anomaly region, it is marked as a distorted region. Threshold for determining the distorted region.

[0008] Optionally, S112 includes: S1121, Extract the shell edge contour from the original image of foraminifera and construct a binary contour image. ; S1122, Calculate the spatial distribution covariance matrix of the shell edge profile. ; S1123, calculate the eigenvectors of the spatial distribution covariance matrix; the direction corresponding to the principal eigenvectors is the direction of the shell's principal axis. Rotate the original foraminifera image to a uniform reference orientation. Generate normalized original images of foraminifera. .

[0009] Optionally, S12 includes: S121, Normalized original image of foraminifera Using oneself as a guide map Guided filtering is then used for global smoothing enhancement to generate a guided-filter enhanced image. ; S122, using the multi-scale Laplacian edge operator to normalize the original foraminifera image. Perform edge enhancement to generate multi-scale edge-enhanced images. ; S123, employing a joint mechanism of nonlocal mean (NLM) and texture fidelity constraints to normalize the original foraminifera image. Achieve detailed restoration and generate texture reconstruction images ; S124, Image enhancement through guided fusion filtering Multi-scale edge enhancement images and texture reconstruction images Obtain enhanced images with consistent local structure And generate a set of morphologically enhanced images with local structural consistency. .

[0010] Optionally, S2 includes: S21. Perform multi-scale structural analysis on the morphological enhancement image set, extract the morphological attribute labels of the shell, including the shell opening shape, shell arrangement, suture direction and shell wall micropore distribution, and construct a structural map by using nodes to represent structural parts and edges to represent hierarchical relationships. S22, bind each node in the structural graph to the corresponding morphological label, and construct a nested label tree with parent-child dependency and logical order according to the composition relationship of the shell structure, so as to realize the coupling organization of structure and semantics; S23. For nodes with structural ambiguity or semantic overlap, identify label ambiguity, recommend labels and resolve conflicts by ranking based on semantic similarity, structural context and confidence, and generate a structured annotation set with logical consistency.

[0011] Optionally, S21 includes: S211 constructs a scale-space pyramid from the morphologically enhanced image set and generates multiple scale images through Gaussian downsampling; S212, on images at various scales, the structural partitioning algorithm is used to extract the shell component regions, and each shell component region is assigned a morphological attribute label, including the shell opening shape, shell arrangement, suture direction and shell wall micropore distribution. S213, assign all detected shell component areas to nodes. This means that each node carries a set of attributes. Directed edges are constructed based on spatial contact relationships and structural hierarchy. To form a structural map ,in, For structural component nodes, This is the set of edges connecting parent and child nodes.

[0012] Optionally, S22 includes: S221, for each component node in the structural diagram According to its attribute set In the predefined tag library Label matching is performed using a weighted matching mechanism based on attribute similarity and prior rule reasoning to determine the optimal label. ; S222, based on the existing node connection relationships in the structure graph, construct parent-child dependency edges for the label layer. That is, if node yes If it is a child component, then its corresponding tag As The child tag of; S223, All tag nodes and the father-son dependency between them Organized as a nested tree of tags ,in, A collection of tag nodes. Given the set of directed edges between labels, ultimately, each path from the root to a leaf... This represents a structure-semantic coupling path.

[0013] Optionally, S23 includes: S231, for each node in the nested tag tree If there is a semantic inconsistency or contextual conflict between the current node and its parent tag, or if the node's structural attributes match multiple tags (i.e., candidate tags), then... (If the number of elements in the middle is greater than 1) or the highest confidence score in the label scoring function does not reach the confidence threshold. (Right now This is considered a semantically ambiguous node, and all candidate labels are extracted. Then proceed to the digestion process; S232, for each candidate label Taking into account semantic similarity, structural context consistency, and initial confidence, a multidimensional conflict score is calculated. ; S233, Based on the multidimensional conflict score, the label with the highest score is selected as the final label. ,like If , then it is recorded as a node with an uncertain label, where The candidate label with the highest score. As the second highest-scoring candidate label, To resolve the score gap threshold, all nodes output their label values, forming a logically consistent structured annotation set. .

[0014] Optionally, S3 includes: S31, bind the morphologically enhanced image set with the structured annotation set, and establish a mapping index table through image ID and label ID; S32, nest the tag tree Mapped to a three-tier database structure, specifically including: Level-1 (Partial Layer): Represents major part categories using component type as the primary key; Attribute layer (Level-2): Attached to the part node, it records the structural attribute tags of the shape, texture, and thickness of each type of component; Level-3 (variation layer): records the variable characteristics of this component under historical evolution or environmental influences; It is formalized into a three-level nested record structure. ,in, For component layer nodes (such as shell). For attribute layer tags (such as "flat" or "spiral" arrangement). For the description of the variation layer (e.g., "spiral volatility = 0.23"); S33 maps the three-tier database structure to database record tables, uses primary key and foreign key structures to support efficient retrieval, and provides a condition-based combined query interface.

[0015] The beneficial effects of this invention are: This invention, through principal axis normalization, local curl field analysis, and structure-preserving image enhancement processing of foraminifera images from multiple sources, can effectively improve the clarity of key details such as shell type, shell arrangement, aperture morphology, suture lines, and micropore structure in images, and overcome the problems of scale drift and local distortion caused by non-standard shooting angles and sample postures.

[0016] This invention constructs a structure-semantic coupled nested tag tree and introduces a polysemy recognition and conflict resolution mechanism, which can accurately identify and automatically correct structurally ambiguous or tag-overlapping areas. In the tag recommendation process, semantic similarity, structural compatibility and confidence scores are integrated to achieve the generation of a structured tag set with clear semantic hierarchy and accurate logical relationship, which significantly improves the stability and interpretability of tag assignment.

[0017] This invention combines image data with structured label information and maps it into a three-layer database model of part layer – attribute layer – variation layer, establishing a morphological feature database with high scalability and multi-dimensional query capabilities. It supports refined recording of component attributes, expression of the evolutionary trajectory of environmental response, and multi-level image, label, and trend retrieval, meeting the application needs of multiple scenarios such as ecological research, sedimentary environment evolution analysis, and intelligent annotation training. Attached Figure Description

[0018] 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 only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the construction method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the construction of a morphological enhancement image set according to an embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. Those skilled in the art may employ other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0021] like Figures 1-2 As shown, a method for constructing a database of foraminifera morphological characteristics includes the following steps: S1: Acquire original images of foraminifera from multiple sources. Based on shell principal axis normalization and local curl field calculation, identify potential scale drift and non-ideal shooting angles, perform structure-preserving image enhancement processing, and output a set of morphologically enhanced images that maintain the consistency of local structures such as shell material (porcelain, glass, cement), shell arrangement (single chamber, single row, double row, plano-spiral, spiral), aperture, suture, and shell wall micropores. S2, based on the output morphologically enhanced image set, extracts multi-scale shell structure maps, constructs a structure-semantic coupled label nesting tree, executes a label recommendation and ambiguity resolution process based on a polysemy recognition mechanism, and outputs a structured annotation set with structural logical paths; S3 combines the morphologically enhanced image set with the structured annotation set, and establishes a three-layer database model based on the nested label tree structure, including the part layer (shell, orifice), the attribute layer (morphology, texture, thickness) and the variation layer (evolutionary trend, environmental response characteristics).

[0022] S1 includes: S11. Obtain original images of foraminifera from field collection, laboratory shooting, and historical databases. To address the inconsistency in the spatial arrangement of foraminifera shells in the original images, an image alignment algorithm based on principal axis direction estimation is used to normalize the shells, ensuring that their major axis direction is aligned with a unified reference direction. The local curl field of the original images of foraminifera is also calculated to capture local rotation features and scale drift anomalies caused by shooting angle deviation or non-standard sample posture, and to identify distorted regions under non-ideal imaging conditions. S12, after completing normalization and local distortion recognition, performs structure-preserving image enhancement processing based on the sensitivity of shell microstructure. With the preservation of shell material type (such as porcelain, glass, cement) and microstructure consistency as the core, it adopts a combination of guided filtering, edge enhancement and texture reconstruction to enhance the local features of shell arrangement (single chamber, single row, double row, planar spiral, spiral), shell aperture morphology, suture texture and shell wall micropore distribution, while avoiding distortion of the original morphological structure, and finally generating a morphological enhancement image set with local structural consistency.

[0023] S11 includes: S111, collecting raw images of foraminifera from multiple sources, specifically including: Field image acquisition: Images of foraminifera in sediment samples were captured in the field using portable digital microscopes, and metadata was recorded by combining GPS and sample point numbers; Laboratory images: Images of cleaned foraminifera individuals were captured under an optical microscope, with bright field, polarization mode, and multiple shooting angles. Historical database images: Access public databases and literature atlases, and perform unified image resolution conversion, noise reduction processing, and standard naming; All original images of foraminifera were converted to a uniform grayscale format and normalized to a fixed resolution. , Pixel; S112, The shell in the original image of foraminifera is normalized and rotated using the image principal axis direction estimation algorithm; S113, calculate the local horizontal and vertical gradients of the original foraminifera image after normalization and rotation processing, and calculate the two-dimensional local curl field. ,like If the region is deemed to be a local structural anomaly region, it is marked as a distorted region. The threshold for determining the distorted region is expressed as: ; ; in, , The image gradients are respectively in , The directional component.

[0024] S112 includes: S1121, Extract the shell edge contour from the original image of foraminifera and construct a binary contour image. ; S1122, Calculate the spatial distribution covariance matrix of the shell edge profile. , represented as: ; ; ; ; in, For the number of edge points, For edge coordinates, The coordinates of the center point; S1123, calculate the eigenvectors of the spatial distribution covariance matrix; the direction corresponding to the principal eigenvectors is the direction of the shell's principal axis. Rotate the original foraminifera image to a uniform reference orientation. Generate normalized original images of foraminifera. , represented as: .

[0025] S12 includes: S121, Normalized original image of foraminifera Using oneself as a guide map Guided filtering is then used for global smoothing enhancement to generate a guided-filter enhanced image. , represented as: ; in, , These are the corresponding filter coefficients; Filter coefficients , By minimizing the loss function The loss function is expressed as: ; in, For filtering window, For regularization terms; S122, for structural regions with complex edge features such as the aperture contour and suture line, a multi-scale Laplacian edge operator is used to normalize the original foraminifera image. Perform edge enhancement to generate multi-scale edge-enhanced images. , represented as: ; ; in, For edge responses at different scales For scale The Laplace operator under Gauss, These are the weighting coefficients for edge responses at different scales. A multi-scale set for edge extraction; S123, employing a joint mechanism of nonlocal mean (NLM) and texture fidelity constraints to normalize the original foraminifera image. Achieve detailed restoration and generate texture reconstruction images Specifically, it includes: (1) For each pixel to be enhanced Search for structurally similar reference blocks in the image. Based on structural similarity Weighted fusion, represented as: ; ; in, For A non-local search window centered on the user. The location of the target pixel in the current image to be enhanced. To and The positions of adjacent reference pixels, To normalize the original image of foraminifera in pixels grayscale value at that location , These are the normalized original images of foraminifera. , A vector composed of all gray levels within the central image patch. For similarity sensitivity parameters, Normalization factor; (2) For the micropore region, a texture prior constraint term is added, which is expressed as: ; in, These are the texture prior constraint weight parameters. For texture descriptors; S124, Image enhancement through guided fusion filtering Multi-scale edge enhancement images and texture reconstruction images Obtain enhanced images with consistent local structure And generate a set of morphologically enhanced images with local structural consistency. , represented as: ; ; in, This is an edge structure mask; a value of 1 indicates the opening / suture area. This is a mask for the microporous structure; a value of 1 indicates a microporous region in the shell wall. This is the mask for the remaining non-sensitive areas.

[0026] S2 includes: S21. Perform multi-scale structural analysis on the morphological enhancement image set, extract the morphological attribute labels of the shell, including the shell opening shape, shell arrangement, suture direction and shell wall micropore distribution, and construct a structural map by using nodes to represent structural parts and edges to represent hierarchical relationships. S22, bind each node in the structural graph to the corresponding morphological label, and construct a nested label tree with parent-child dependency and logical order according to the composition relationship of the shell structure, so as to realize the coupling organization of structure and semantics; S23. For nodes with structural ambiguity or semantic overlap, identify label ambiguity, recommend labels and resolve conflicts by ranking based on semantic similarity, structural context and confidence, and generate a structured annotation set with logical consistency.

[0027] S21 includes: S211, constructing the morphologically enhanced image set into a scale-space pyramid, and generating multiple scale images through Gaussian downsampling, represented as: ; in, For the first Images at various scales For scale Gaussian kernel, This represents the total number of scale levels; S212, on images at various scales, a structural partitioning algorithm is used to extract shell component regions, and each shell component region is assigned a morphological attribute label, including the shell opening shape, shell arrangement, suture direction, and distribution of shell wall micropores, where; The aperture morphology is calculated based on the local shape convexity and closure, and is expressed as follows: ; ; in, This represents the actual contour area of ​​the candidate region for the shell opening. Let be the area of ​​the convex hull of the contour region. The convexity of a local shape indicates the area is more concave or has more complex boundaries, and the more distinct the aperture outline is. The length of the inner closed contour. The total boundary profile length, The degree of closure is 1, which indicates that the region is approximately closed and satisfies the characteristics of the shell opening structure. The shell arrangement is identified based on the change in curvature of the centroid sequence, and is represented as follows: ; in, For the first The curvature of the centroid sequence at each point, if If the value is approximately 0 and the direction is stable, it is judged as a single-column type. If there is a periodic positive and negative change, it is a spiral type. If the center of gravity is horizontally symmetrical, it is a double-column type. The suture line direction is extracted based on directional gradient flow tracing and is represented as follows: ; ; in, The gradient direction; Directional consistency is used to verify whether it conforms to the direction of the suture, as shown below: ; in, A score is given for directional consistency, with a value close to 1 indicating that the main direction is stable and conforms to the direction of the suture line. Shell wall micropore distribution based on texture entropy And density threshold judgment, expressed as: ; in, For the first pixel block The probability of gray levels; S213, assign all detected shell component areas to nodes. This means that each node carries a set of attributes. Directed edges are constructed based on spatial contact relationships and structural hierarchy. To form a structural map ,in, For structural component nodes, This is the set of edges connecting parent and child nodes; For example: establishing an edge between a shell node and a shell opening node. , indicates a "containment" relationship.

[0028] S22 includes: S221, for each component node in the structural diagram According to its attribute set In the predefined tag library Label matching is performed using a weighted matching mechanism based on attribute similarity and prior rule reasoning to determine the optimal label. , represented as: ; in, For nodes With tags Attribute similarity between them For compatibility scoring based on a tag-based logical rule base, , These are the corresponding weight coefficients; ; ; in, For nodes Attribute vectors (such as morphological features and scale levels). For tags Attribute template vector, For type consistency, Scoring is based on logical consistency with the structural context. , These represent the weightings of type matching and structural consistency in the control logic score; S222, based on the existing node connection relationships in the structure graph, construct parent-child dependency edges for the label layer. That is, if node yes If it is a child component, then its corresponding tag As The child tag of is represented as: ; in, In the tag library The set of allowed sub-tags; S223, All tag nodes and the father-son dependency between them Organized as a nested tree of tags ,in, A collection of tag nodes. Given the set of directed edges between labels, ultimately, each path from the root to a leaf... This represents a structure-semantic coupling path.

[0029] S23 includes: S231, for each node in the nested tag tree If there is a semantic inconsistency or contextual conflict between the current node and its parent tag, or if the node's structural attributes match multiple tags (i.e., candidate tags), then... (If the number of elements in the middle is greater than 1) or the highest confidence score in the label scoring function does not reach the confidence threshold. (Set to 0.85) (i.e.) This is considered a semantically ambiguous node, and all candidate labels are extracted. Then proceed to the digestion process; S232, for each candidate label Taking into account semantic similarity, structural context consistency, and initial confidence, a multidimensional conflict score is calculated. , represented as: ; ; ; ; in, The semantic similarity between candidate tags and context tags. Scoring is given for the compatibility between candidate labels and node structural attributes. This represents the confidence score during initial label matching. , , These are the corresponding weight coefficients. , These are the semantic embedding vectors of the candidate label and the context label, respectively. For nodes Attribute vectors, For tags Attribute template vector, This is a parameter for adjusting attribute compatibility. Candidate tags The original matching score; S233, Based on the multidimensional conflict score, the label with the highest score is selected as the final label. ,like If , then it is recorded as a node with an uncertain label, where The candidate label with the highest score. As the second highest-scoring candidate label, To resolve the score gap threshold, all nodes output their label values, forming a logically consistent structured annotation set. , represented as: ; .

[0030] S3 includes: S31, bind the morphologically enhanced image set with the structured annotation set, and establish a mapping index table through image ID and label ID, represented as: ; in, For the first Enhanced image, A set of structured labels associated with an image; S32, nest the tag tree Mapped to a three-tier database structure, specifically including: Level-1 (Part Level): Using component category as the primary key, it represents major part categories, such as "shell", "orifice", "stitching line", etc. Attribute layer (Level-2): Attached to the part node, it records the structural attribute tags of the shape, texture, and thickness of each type of component; Level-3 (variation layer): records the variable characteristics of the component under historical evolution or environmental influences, such as "angle change trend" and "micropore density response to salinity change". It is formalized into a three-level nested record structure. ,in, For component layer nodes (such as shell). For attribute layer tags (such as "flat" or "spiral" arrangement). For the description of the variation layer (e.g., "spiral volatility = 0.23"); S33 maps a three-tier database structure to database record tables, uses primary and foreign key structures to support efficient retrieval, and provides a condition-based combined query interface, including: Image-level retrieval: Extract all parts, attributes, and variation information of an image by its ID; Tag-level reverse lookup: Retrieve all image and attribute instances by tag or component type; Trend query: Supports clustering analysis and trend callback on the mutation layer.

[0031] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0032] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method of constructing a foraminiferal morphological feature database, characterized by, Includes the following steps: S1. Acquire original images of foraminifera from multiple sources. Based on shell principal axis normalization and local curl field calculation, identify potential scale drift and non-ideal shooting angles, perform structure-preserving image enhancement processing, and output a set of morphologically enhanced images that maintain the consistency of local structures such as shell material, shell arrangement, pores, sutures, and shell wall micropores. S2, based on the output morphologically enhanced image set, extracts multi-scale shell structure maps, constructs a structure-semantic coupled label nesting tree, executes a label recommendation and ambiguity resolution process based on a polysemy recognition mechanism, and outputs a structured annotation set with structural logical paths; S3 combines the morphologically enhanced image set with the structured annotation set, and establishes a three-layer database model based on the nested label tree structure, including the part layer, attribute layer, and variation layer.

2. The method for constructing a foraminifera morphological characteristic database according to claim 1, characterized in that, S1 includes: S11. Obtain original images of foraminifera from field collection, laboratory shooting and historical database. To address the inconsistency in the spatial arrangement of foraminifera shells in the original images, an image alignment algorithm based on principal axis direction estimation is used to normalize the shells and calculate the local curl field of the original images of foraminifera. This is used to capture local rotation features and scale drift anomalies caused by shooting angle deviation or non-standard sample posture, and to identify distorted regions under non-ideal imaging conditions. S12, after completing the normalization process and local distortion recognition, performs structure-preserving image enhancement processing based on the sensitivity of the shell microstructure. With the preservation of shell material type and microstructure consistency as the core, it uses a combination of guided filtering, edge enhancement and texture reconstruction to enhance the local features of shell arrangement, shell aperture morphology, suture texture and shell wall micropore distribution, while avoiding distortion of the original morphological structure, and finally generates a set of morphologically enhanced images with local structural consistency.

3. The method for constructing a foraminifera morphological characteristic database according to claim 2, characterized in that, S11 includes: S111, collecting raw images of foraminifera from multiple sources, specifically including: Field image acquisition: Images of foraminifera in sediment samples were captured in the field using portable digital microscopes, and metadata was recorded by combining GPS and sample point numbers; Laboratory images: Images of cleaned foraminifera individuals were captured under an optical microscope, with bright field, polarization mode, and multiple shooting angles. Historical database images: Access public databases and literature atlases, and perform unified image resolution conversion, noise reduction processing, and standard naming; All original foraminifera images were converted to a uniform grayscale format and normalized to a fixed resolution. S112, The shell in the original image of foraminifera is normalized and rotated using the image principal axis direction estimation algorithm; S113, calculate the local horizontal and vertical gradients of the original foraminifera image after normalization and rotation processing, and calculate the two-dimensional local curl field. ,like If the region is deemed to be a local structural anomaly region, it is marked as a distorted region. Threshold for determining the distorted region.

4. The method for constructing a foraminifera morphological characteristic database according to claim 3, characterized in that, S112 includes: S1121, Extract the shell edge contour from the original image of foraminifera and construct a binary contour image. ; S1122, Calculate the spatial distribution covariance matrix of the shell edge profile. ; S1123, calculate the eigenvectors of the spatial distribution covariance matrix; the direction corresponding to the principal eigenvectors is the direction of the shell's principal axis. Rotate the original foraminifera image to a uniform reference orientation. Generate normalized original images of foraminifera. .

5. The method for constructing a foraminifera morphological characteristic database according to claim 4, characterized in that, S12 includes: S121, Normalized original image of foraminifera Using oneself as a guide map Guided filtering is then used for global smoothing enhancement to generate a guided-filter enhanced image. ; S122, using the multi-scale Laplacian edge operator to normalize the original foraminifera image. Perform edge enhancement to generate multi-scale edge-enhanced images. ; S123, employing a joint mechanism of nonlocal mean and texture fidelity constraint to normalize the original foraminifera image. Achieve detailed restoration and generate texture reconstruction images ; S124, Image enhancement through guided fusion filtering Multi-scale edge enhancement images and texture reconstruction images Obtain enhanced images with consistent local structure And generate a set of morphologically enhanced images with local structural consistency. .

6. The method for constructing a foraminifera morphological characteristic database according to claim 5, characterized in that, S2 includes: S21. Perform multi-scale structural analysis on the morphological enhancement image set, extract the morphological attribute labels of the shell, including the shell opening shape, shell arrangement, suture direction and shell wall micropore distribution, and construct a structural map by using nodes to represent structural parts and edges to represent hierarchical relationships. S22, bind each node in the structural graph to the corresponding morphological label, and construct a nested label tree with parent-child dependency and logical order according to the composition relationship of the shell structure, so as to realize the coupling organization of structure and semantics; S23. For nodes with structural ambiguity or semantic overlap, identify label ambiguity, recommend labels and resolve conflicts by ranking based on semantic similarity, structural context and confidence, and generate a structured annotation set with logical consistency.

7. The method for constructing a foraminifera morphological characteristic database according to claim 6, characterized in that, S21 includes: S211 constructs a scale-space pyramid from the morphologically enhanced image set and generates multiple scale images through Gaussian downsampling; S212, on images at various scales, the structural partitioning algorithm is used to extract the shell component regions, and each shell component region is assigned a morphological attribute label, including the shell opening shape, shell arrangement, suture direction and shell wall micropore distribution. S213, assign all detected shell component areas to nodes. This means that each node carries a set of attributes. Directed edges are constructed based on spatial contact relationships and structural hierarchy. To form a structural map ,in, For structural component nodes, This is the set of edges connecting parent and child nodes.

8. The method for constructing a foraminifera morphological characteristic database according to claim 7, characterized in that, S22 includes: S221, for each component node in the structural diagram According to its attribute set In the predefined tag library Label matching is performed using a weighted matching mechanism based on attribute similarity and prior rule reasoning to determine the optimal label. ; S222, based on the existing node connection relationships in the structure graph, construct parent-child dependency edges for the label layer. That is, if node yes If it is a child component, then its corresponding tag As The child tag of; S223, All tag nodes and the father-son dependency between them Organized as a nested tree of tags ,in, A collection of tag nodes. Given the set of directed edges between labels, ultimately, each path from the root to a leaf... This represents a structure-semantic coupling path.

9. The method for constructing a foraminifera morphological characteristic database according to claim 8, characterized in that, S23 includes: S231, for each node in the nested tag tree If there is a semantic inconsistency or contextual conflict between the current node and its parent label, the node's structural attributes match multiple labels, or the highest confidence score in the label scoring function does not reach the confidence threshold, the following conditions will be considered: This is considered a semantically ambiguous node, and all candidate labels are extracted. Then proceed to the digestion process; S232, for each candidate label Taking into account semantic similarity, structural context consistency, and initial confidence, a multidimensional conflict score is calculated. ; S233, Based on the multidimensional conflict score, the label with the highest score is selected as the final label. ,like If , then it is recorded as a node with an uncertain label, where The candidate label with the highest score. As the second highest-scoring candidate label, To resolve the score gap threshold, all nodes output their label values, forming a logically consistent structured annotation set. .

10. The method for constructing a foraminifera morphological characteristic database according to claim 9, characterized in that, S3 includes: S31, bind the morphologically enhanced image set with the structured annotation set, and establish a mapping index table through image ID and label ID; S32, nest the tag tree Mapped to a three-tier database structure, specifically including: Part layer: using component category as the primary key to represent major part categories; Attribute layer: Attached to the part node, it records the structural attribute tags of each type of component, such as shape, texture, and thickness; Variation layer: records the variable characteristics of the component under historical evolution or environmental influences; It is formalized into a three-level nested record structure. ,in, For location layer nodes, For attribute layer tags, Description of the variant layer; S33 maps the three-tier database structure to database record tables, uses primary key and foreign key structures to support efficient retrieval, and provides a condition-based combined query interface.

Images