Intelligent bulk recognition and analysis system for cross-section cells of gymnosperm wood

Abstract

This invention belongs to the field of computer image processing and computational wood science, specifically relating to an intelligent batch identification and analysis system for cells in cross-sections of gymnosperm wood. The system includes sequentially connected modules for image input, preprocessing, cell instance segmentation, unique ID generation and management, batch calculation of geometric parameters, and result output and interactive filtering. The method involves: after preprocessing the input image, using instance segmentation based on traditional image processing algorithms to batch identify cell outlines; assigning a unique ID to each cell and establishing a mapping; calculating the cell lumen area, diameter, and wall thickness in batches based on the mapping relationship; and finally, performing visualization and interactive filtering. This invention, by integrating batch parallel processing and cell instance-ID mapping management, achieves minute-level fully automated identification and measurement of massive numbers of cells in wood cross-sections, improving efficiency by hundreds of times compared to manual serial operations, and supports fine-grained data management and statistics.

Description

Technical Field

[0001] This invention belongs to the field of computer image processing and computational wood science, specifically relating to an intelligent batch recognition and analysis system for cross-sectional cells of gymnosperm wood. Background Technology

[0002] The microscopic anatomy of wood is crucial for studying its properties, identifying tree species, and reconstructing paleoenvironments. Currently, this field mainly relies on researchers manually outlining and measuring each cell using general-purpose image software (such as ImageJ). This "manual serial" approach has significant drawbacks: (1) it is extremely inefficient, as a single high-resolution image containing thousands of cells often requires hours to days to analyze, making large-scale sample statistics difficult to implement; (2) it is highly subjective, with measurement results dependent on personal experience and poor reproducibility; and (3) it lacks cell-level data management, making it impossible to establish a permanent association between cell entities and data, resulting in low efficiency for subsequent fine screening and population statistics.

[0003] While some studies have attempted to use rule-based traditional image processing algorithms (such as thresholding and watershed segmentation) for semi-automatic assistance, these methods are typically applied in isolation. In natural images like wood cross-sections, which have complex textures and severe cell adhesion, their generalization ability and stability are insufficient, and they also lack the ability to systematically manage massive amounts of cell data. In recent years, deep learning technology has been applied to cell image segmentation (e.g., patent CN120388375B), but model training heavily relies on a large number of precisely labeled samples. However, manual labeling of wood cells is extremely costly, failing to provide a complete system-level solution integrating automatic identification, batch computation, and structured data management. This cannot meet the needs of efficient, traceable, and interactive analysis in wood science research.

[0004] Therefore, there is currently a lack of a complete system that can deeply integrate the reliability of traditional algorithms, the interpretability of domain knowledge, and the interactivity of modern data management to achieve efficient, accurate, and traceable batch analysis of tens of thousands of cells in cross-sectional images of wood. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent batch identification and analysis system for cells in cross-sections of gymnosperm wood, solving the technical problem that "it is difficult to achieve high-precision, batch, and fully automated statistical analysis of massive numbers of cells in cross-sections of wood with traceability".

[0006] To achieve the above objectives, the present invention is implemented using the following technical solution: In a first aspect, the present invention provides an intelligent batch identification and analysis system for cross-sectional cells of gymnosperm wood, comprising: The image processing and analysis unit is used to preprocess and segment the input microscopic image of the cross-section of gymnosperm wood, and output the contour data of all independent cells. The cell data management unit is used to assign a unique identifier to each cell instance and establish a structured mapping relationship between the identifier and cell outline data and geometric parameters. The parameter calculation and interaction unit calculates the geometric parameters of cells in batches according to the structured mapping relationship, and supports users to interactively filter and manage data based on the identifier.

[0007] Furthermore, the image processing and analysis unit includes: Image input module, used for inputting images; The preprocessing module is used to perform color space conversion, noise filtering, contrast enhancement, and normalization on the input image; specifically, it includes: converting the RGB image to a grayscale image and HSV color space; using Gaussian filtering and median filtering for noise reduction; using the Otsu adaptive thresholding algorithm for contrast enhancement; and converting the data type to uint8 format.

[0008] The instance segmentation module is used to segment the cell wall using HSV adaptive thresholding based on K-means clustering and segment the cell cavity using grayscale thresholding based on the Otsu algorithm, generating an initial cell mask. The post-processing module is used to perform morphological filtering, intelligent hole filling, small object removal, and multi-condition feature filtering (area ≥ 100 pixels, roundness > 0.3, aspect ratio < 5, perimeter > 30 pixels, brightness > dynamic threshold) on the initial mask to obtain accurate independent cell instances.

[0009] Furthermore, the parameter calculation and interaction unit includes: The parameter calculation module is used to calculate the area, equivalent diameter, and cell wall thickness of each cell cavity in batches. The interactive filtering module is used to visually display the outlines and parameter lists of cells with identifiers, and respond to user commands to filter cells with specific identifiers, dynamically updating the statistical analysis results.

[0010] Secondly, this invention provides a method for intelligent batch identification and analysis of cross-sectional cells in gymnosperm wood based on the above-mentioned system, comprising the following steps: (1) Perform color space conversion, noise filtering, contrast enhancement and standardization on the input microscopic images of cross-sections of gymnosperm wood; (2) Perform dual-channel segmentation on the preprocessed image: use HSV adaptive threshold segmentation based on K-means clustering to obtain the cell wall mask, and use grayscale threshold segmentation based on Otsu algorithm to obtain the cell cavity mask; perform morphological filtering, intelligent hole filling and multi-condition feature filtering on the mask to obtain multiple independent cell instances, and extract the precise contour data of each independent cell instance. (3) Assign a globally unique identifier to each independent cell instance obtained above and establish a data structure mapping relationship; based on the data structure mapping relationship, calculate the cell cavity area, equivalent diameter and cell wall thickness of each cell instance in batches, and associate the calculation results with the mapping relationship; (4) Visualize the cell outlines and parameter list with the identifier, and respond to the user's filtering or exclusion instructions for specific cells based on the identifier, and dynamically update the statistical analysis results of the cell population.

[0011] Furthermore, the structured mapping relationship established in the cell data management unit is implemented using pandasDataFrame, with the unique identifier as the index and a structured data object containing cell outline data and geometric parameters as the value, to achieve efficient storage and instant retrieval of cell data.

[0012] Furthermore, the method for calculating cell wall thickness is as follows: calculate the direction vector according to the set ray angle, and extend it from the centroid of the cell cavity along the ray direction to both sides; first detect the cell cavity boundary, and then continue to extend and measure through the cell wall region until an adjacent cell is detected or the maximum search distance is reached; measure the left wall thickness and the right wall thickness respectively, and calculate the average value as the average wall thickness of the cell.

[0013] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the method described above.

[0014] Compared with the prior art, the beneficial effects achieved by the present invention are: (1) This invention uses a batch instance segmentation and parallel computing architecture based on traditional image processing algorithms. Without the need for deep learning model training and high-performance computing equipment, it reduces the analysis time of a single image with more than 2,000 cells from several hours required by manual methods to about 3 minutes, improving efficiency by hundreds of times. This makes batch analysis of large-scale samples possible, and the system has a low deployment threshold and can be run on ordinary computers.

[0015] (2) This invention achieves structured management of massive cell data by assigning a unique and traceable ID to each cell and establishing a mapping database. Users can easily perform number-based review and screening, and generate parameter distribution histograms of the entire cell population and calculate various statistics with one click, providing powerful data tools that traditional methods cannot support for macroscopic research.

[0016] (3) The dual-channel segmentation strategy (HSV color space segmentation of cell wall + Otsu grayscale segmentation of cell cavity) adopted by the invention combines color information and brightness information, and shows good segmentation effect and generalization ability in cross-sectional images of gymnosperm wood. It can maintain a high recognition rate for different tree species (such as Qilian juniper, Pinus tabuliformis, and Cunninghamia lanceolata). The invention realizes the organic combination of fully automatic machine preliminary analysis and manual fine review and screening. Users can quickly locate and exclude invalid cells with incomplete edges, ensure the quality of the final analysis data, and greatly reduce the workload. Attached Figure Description

[0017] Figure 1 This is a flowchart of Embodiment 1 of the present invention.

[0018] Figure 2 This is a schematic diagram of the visualization output interface of the analysis results of the system in Embodiment 1 of the present invention. Detailed Implementation

[0019] The technical solutions of the present invention will be clearly and completely described below with reference to the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0020] Configuration and explanation of key system parameters: Based on extensive experiments and analysis of microscopic images of cross-sections of typical gymnosperm wood, the core algorithm parameters used in this invention have been optimized and determined. These parameters collectively constitute the key to achieving high-precision, batch recognition. Specifically: Image preprocessing parameters: To effectively suppress noise while preserving cell boundaries, a cascaded denoising method using a 7×7 pixel Gaussian filter kernel and a 5×5 pixel median filter kernel is used.

[0021] Morphological treatment employed a 3×3 elliptical nucleus structure to conform to the natural shape of the cell outline.

[0022] The coefficients of the Otsu thresholding algorithm can be adjusted between 0.5 and 1.5, with a default value of 1.1, to accommodate images with different contrasts.

[0023] Cell instance segmentation parameters: In the HSV color space, the hue (H) is set to a selection range of 115-180 for typical coloration of gymnosperm wood cell walls.

[0024] The color space is pre-segmented using the K-means clustering algorithm with a cluster number of K=5, which can effectively separate the cell wall, cell cavity and background.

[0025] To distinguish between valid cells and debris, the following filtering thresholds are set: minimum cell area of ​​100 pixels, circularity > 0.3, aspect ratio < 5, and minimum perimeter of 30 pixels.

[0026] Intelligent hole filling only processes holes with an area of ​​less than 80 pixels, avoiding the mistaken filling of real cell cavities.

[0027] Geometric parameter measurement parameters: Cell wall thickness is measured using ray tracing, and the ray direction angle can be set by the user (usually an average of multiple angles such as 0°, 45°, 90°, 135°, etc.) to comprehensively characterize the wall thickness.

[0028] The search step accuracy along the ray direction during measurement is 0.1 pixels, and sub-pixel interpolation is used to improve accuracy.

[0029] The maximum search distance is set to 50 pixels and the minimum effective wall thickness to 4 pixels as physical constraints for measurement. For discontinuous cell walls, wall segments with a maximum gap of 12 pixels are allowed to be merged to deal with situations with low local contrast.

[0030] System performance metrics (under the specified parameter configuration): Processing a typical high-resolution image (containing approximately 2000-3000 cells) takes approximately 135 seconds, enabling minute-level analysis of massive numbers of cells.

[0031] On the validation set, the accuracy rate for cell contour detection exceeded 95%, and the proportion of invalid recognition areas automatically filtered by the system was approximately 5%. Through testing on various gymnosperms such as Pinus tabuliformis, Cunninghamia lanceolata, and Ginkgo biloba, the average recognition rate remained above 90%, demonstrating good universality.

[0032] The above parameter set represents the preferred solution verified under typical conditions. Those skilled in the art should understand that these parameters (especially the hue range H, cluster number K, and morphological filtering thresholds) can be adaptively fine-tuned within the framework of the principles given in this invention based on the microscopic image characteristics (such as cell wall color, contrast, and cell morphology) of specific tree species, and such fine-tuning falls within the protection scope of this invention.

[0033] In the following embodiments, the system software was developed based on the Python language and used open source libraries such as OpenCV, scikit-image, NumPy, and Pandas. All image processing experiments were conducted on a regular desktop computer equipped with an Intel Core i7 processor and 16GB of memory.

[0034] Example 1: Batch analysis of cells in cross-section of Qilian juniper wood Reference Figure 1 This embodiment demonstrates the complete and efficient analysis process and results of the system of the present invention on Qilian juniper.

[0035] S1. Sample and Image Input: Standard cross-sections of Qilian juniper wood were prepared, and high-resolution digital images (12612×1827 pixels, TIFF format) were acquired under a 20x optical microscope. The images were then imported and read using the image input module.

[0036] S2. Automated assembly line processing: 1. Image Preprocessing: The preprocessing module within the image processing and analysis unit automatically performs the following operations: Color space conversion: Convert RGB images to grayscale (for cell cavity segmentation) and HSV color space (for cell wall segmentation) through format conversion. Noise filtering: Gaussian filtering (7×7 kernels, σ=0) is applied to the grayscale image, and median filtering (5×5 kernels) is applied to the original image. Data standardization: Convert image data to uint8 format (0-255 range).

[0037] 2. Instance segmentation and post-processing: (1) The preprocessed image is input into the instance segmentation module for dual-channel segmentation: HSV color segmentation (cell wall): K-means clustering algorithm (K=5) is used to cluster the sampled HSV images; clusters with hue H in the range of 115-180 are selected as cell wall regions; the HSV threshold range is determined according to the 2-98 percentile of the clustering results; after generating the cell wall mask, morphological closing operation (2 iterations) is performed using a 3×3 elliptic kernel to connect the broken cell walls; Otsu Thresholding (Cellular): The grayscale image is binarized using the Otsu adaptive thresholding algorithm with a threshold coefficient of 1.1; morphological processing is performed using a 3×3 elliptical kernel, and opening and closing operations are performed; Intelligent Hole Filling: Only small holes (noise) with an area of ​​less than 80 pixels are filled, while cells in large holes are preserved. (2) Connected component labeling: Connected components are labeled based on the cell cavity segmentation results, and each connected component represents an independent cell cavity instance; (3) Multi-condition feature filtering: Filtering is based on regional attributes. The filtering conditions include: area ≥ 100 pixels, circularity > 0.3, aspect ratio < 5, perimeter > 30 pixels; Dynamic brightness filtering: The average brightness of the brightest 50% area × 0.75 is used as the brightness threshold; Ray cell filtering: Long strip ray cells are excluded based on aspect ratio (< 5.0) and eccentricity (< 0.95).

[0038] 3. Cell Identification and Mapping Establishment: For each independent cell instance obtained in the above steps, the cell data management unit assigns a globally unique identifier (ID) according to its spatial order of identification in the image (following the default scanning order of skimage.measure.label, i.e., from top to bottom and from left to right). The ID is in the format of sequential numbers (e.g., 1, 2, 3, ..., N). Subsequently, a pandas DataFrame data structure is created in the system memory, using the ID as an index to store the corresponding cell outline point coordinates, centroid coordinates, and other data.

[0039] 4. Batch Calculation of Geometric Parameters: The parameter calculation module within the interaction unit iterates through the DataFrame and calculates the contour corresponding to each ID. Cell cavity area: The total number of pixels inside the cell cavity is counted based on the cell cavity mask, and then converted into physical area (μm²) according to the scale (4.26 pixels / μm in this embodiment). Equivalent diameter: Based on chordal measurement, calculate the sum of the distances to the left and right boundaries, and multiply it by the pixel size to obtain the equivalent diameter (μm). Cell wall thickness (ray tracing method): Calculate the direction vector (dx, dy) based on the set ray angle; extend from the centroid of the cell cavity along the ray direction to both sides, first detecting the cell cavity boundary; continue to extend the measurement through the cell wall region until an adjacent cell is detected or the maximum search distance (50 pixels) is reached; the wall thickness measurement step is 0.1 pixels, and bilinear interpolation is used to improve sub-pixel accuracy; measure the left wall thickness (upper wall thickness) and the right wall thickness (lower wall thickness) respectively, and calculate the average value as the average wall thickness of the cell.

[0040] S3. Results Interaction and Output: The results output and interactive filtering module displays the original image with overlaid colored outlines and IDs (green outlines represent cell boundaries, yellow numbers represent cell numbers, and red lines represent wall thickness measurement lines), as well as a data list containing all IDs and parameters.

[0041] When the user clicks the "Export" button, the system exports the unique identifiers (IDs) of 2633 valid cells and all their geometric parameters into a structured Excel file containing two worksheets: "Wall Thickness Measurement Results" and "Statistical Summary".

[0042] Table 1 fully displays the data content, format, and verification results output by the system: First, it lists the detailed data of the system's automatic measurements for four representative cells (1702, 1697, 1688, and 1668) selected consecutively by ID; second, it displays the manually verified cell lumen area for each representative cell and calculates the relative deviation, intuitively presenting single-cell-level accuracy; finally, it summarizes the statistical results of the system measurements based on all 2633 cells and the statistical results of the manual measurements based on 200 verification samples.

[0043] Table 1

[0044] like Figure 2 As shown, Figure 2 The original image is shown with color outlines and IDs superimposed (green outlines represent cell boundaries, yellow numbers represent cell numbers, and red line segments represent wall thickness measurement lines), as well as a schematic diagram of a data list containing all IDs and parameters.

[0045] Results: The system completed the automatic identification and measurement of 2633 cells in a single image within 3 minutes. In comparison, experienced researchers used ImageJ to manually measure 200 cells, which took about 50 minutes. If calculated at this manual speed, it would take about 11 hours to complete the measurement of all 2633 cells in the image. As shown in Table 1, the relative deviation of the cell lumen area measured by the system and the manual measurement of 4 randomly selected cells was within ±1%. The average relative deviation of all 200 validation samples was less than 2%. Moreover, the average value and standard deviation of the cell lumen area of ​​the two systems (n=2633 in the system and n=200 in the manual measurement) were very close. Therefore, the system of this invention can achieve fully automatic batch analysis at the minute level while its measurement accuracy is comparable to that of manual measurement. It can also directly output the statistical distribution parameters of all cells that are difficult to obtain quickly manually.

[0046] Example 2: Verification of the universality of the present invention in the analysis of wood from multiple tree species of gymnosperms This embodiment aims to verify the universal applicability of the technical solution of the present invention (batch cell identification based on instance segmentation, unique ID mapping management and automatic parameter calculation) to different gymnosperm tree species, and to prove that it is a universal intelligent analysis solution for wood microstructure.

[0047] S1. Sample and Test Set Preparation Three typical gymnosperm woods different from those in Example 1 (Qilian juniper): Pinus tabuliformis, Cunninghamia lanceolata, and Ginkgo biloba were selected. Standard cross-sections were prepared for each species, and five high-quality digital micrographs of each species were acquired at the same specific magnification (e.g., 20x) to form a test set containing 15 images.

[0048] S2. Standardized Processing Flow The system settings and processing parameters are kept exactly the same as in Example 1, including: image preprocessing flow (color space conversion, Gaussian filtering, median filtering), dual-channel segmentation parameters (K-means clustering K=5, HSV hue range 115-180, Otsu threshold coefficient 1.1), post-processing parameters (morphological kernel 3×3 elliptical kernel, smart hole filling threshold 80 pixels, minimum cell area 100 pixels), and ray tracing parameters (adjustable angle, step 0.1 pixels, maximum search distance 50 pixels).

[0049] S3. Results and Validation Fifteen test images were input into the system for fully automated analysis. The system was able to output clear, uniquely IDed overlay images of cell outlines and structured parameter lists for all tree species images, without any interruptions or failures due to differences in tree species. Table 2 below shows the analysis results of one representative image for each of the two tree species.

[0050] Table 2

[0051] Example 2 demonstrates that, without altering any core algorithms or processes, the system of this invention can be applied to the woods of various gymnosperms with significant differences in anatomical structure, such as Pinus tabuliformis and Cunninghamia lanceolata. When processing Pinus tabuliformis images, the system successfully distinguishes axial tracheids from resin ducts and achieves complete identification of regular cells in Cunninghamia lanceolata. The cell identification rate for all three tree species remains above 90%. This verifies the adaptability of the technical solution of this invention to the differences in microstructure among different tree species, indicating that it is a method and system universally applicable to the quantitative analysis of gymnosperm wood cells.

[0052] Comparative Example 1: A Simplified System for Unique Number Management Module To verify the necessity of the "cell data management unit" and the "structured mapping relationship" established in this invention, a simplified implementation of the system is provided for comparison. The only difference between this simplified system and Embodiment 1 of this invention is that the "cell data management unit" and its related operations in Embodiment 1 are omitted, and therefore the step of establishing the mapping relationship between cell identifiers and contour data through DataFrame is not performed; otherwise, the algorithms and parameters for sample images, preprocessing, instance segmentation and postprocessing, and parameter calculation are completely consistent with those in Embodiment 1.

[0053] The simplified system's processing flow is as follows: S1. Sample and Image Input: Import the same cross-sectional microscopic image of Qilian juniper wood as in Example 1 through the image input module.

[0054] S2. Automated assembly line processing: 1. Preprocessing: The process is exactly the same as in Example 1, performing color space conversion, Gaussian filtering, and median filtering.

[0055] 2. Instance segmentation and post-processing: exactly the same as in Example 1, using the same dual-channel segmentation and post-processing parameters to segment cells and obtain independent cell outlines.

[0056] 3. Batch calculation of geometric parameters: The system directly calculates the geometric parameters of each cell in batches based on the above contour data and using the same algorithm as in Example 1.

[0057] S3. Output of Results: The system ultimately outputs a list containing all cell geometric parameters. However, there is no unique identifier associated with each parameter in this list, and the order of the parameters is independent of the spatial location of the cells in the image.

[0058] The simplified system was applied to analyze the images above, and the system successfully completed the batch calculation, outputting a list containing 2633 cell parameters.

[0059] However, compared with Example 1, the simplified system has the following defects: (1) When statistical outliers appear in the output parameter list, the user cannot quickly locate the corresponding cell on the image for verification based on the ID as in Example 1; (2) When it is necessary to exclude invalid cells located at the edge of the image, due to the lack of ID mapping, the ID-based marking and exclusion operation described in Example 1 cannot be implemented, and a tedious full list manual comparison must be performed; (3) All parameters exist in the form of a flat list, which cannot support the dynamic updating of statistical charts or the management of data subsets based on valid / invalid status as described in Example 1.

[0060] This comparative example demonstrates that, while keeping all other processing conditions identical, omitting the cell data management unit and structured mapping relationships will cause the system to lose the core interactive and management functions shown in Example 1. This result directly proves that this unit and mapping relationships are irreplaceable and necessary for achieving the technical effect of this invention—from batch parameter calculation to interactive and traceable intelligent batch analysis.

[0061] Comparative Example 2; Comparison based on single-channel segmentation To compare and verify the superiority of the "dual-channel segmentation strategy" of this invention over single-channel segmentation, in this comparative example, Group A and Group B are identical to Example 1 in all processing steps and parameters except for the segmentation strategy. This comparative example processes the same cross-sectional microscopic images of Qilian juniper wood as Example 1.

[0062] Group A: Segmentation using only HSV color space (cell wall channels) For the preprocessed image, only K-means clustering is used for HSV color space segmentation to obtain the cell wall mask; Subsequently, the mask was processed using the same post-processing steps as in Example 1 (including morphological filtering, smart hole filling, small object removal, and multi-condition feature filtering) in an attempt to obtain cell instances. The lack of cell cavity segmentation information makes it impossible to accurately determine cell boundaries, resulting in a significant decrease in cell recognition accuracy.

[0063] Group B: Otsu grayscale segmentation (cell cavity channel) only. For the preprocessed image, only Otsu adaptive thresholding is used for grayscale segmentation to obtain the cell cavity mask; Subsequently, the mask was processed using the same post-processing steps as in Example 1 (including morphological filtering, smart hole filling, small object removal, and multi-condition feature filtering) to obtain cell instances; Due to the lack of cell wall information, accurate cell wall thickness measurement is not possible, and the segmentation effect for low-contrast images is poor.

[0064] The differences in recognition performance and functional integrity between the dual-channel segmentation strategy and the single-channel segmentation strategy were compared, and the results are shown in Table 3 below: Table 3

[0065] This comparative example shows that single-channel segmentation methods all have lower cell identification accuracy than the dual-channel segmentation strategy of this invention, and neither can accurately measure cell wall thickness. The dual-channel segmentation strategy of this invention combines color information from the HSV color space and brightness information from the grayscale space, enabling more accurate segmentation of the cell wall and cell cavity, providing the necessary foundation for subsequent cell wall thickness measurement. This demonstrates the irreplaceable necessity of the dual-channel segmentation strategy for realizing the full functionality of this invention.

[0066] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No markings in the claims should be construed as limiting the scope of the claims.

Claims

1. A smart batch identification and analysis system for cross-sectional cells of gymnosperm wood, characterized in that, include: The image processing and analysis unit is used to preprocess and segment the input microscopic image of the cross-section of gymnosperm wood, and output the contour data of all independent cells. The instance segmentation adopts a dual-channel segmentation strategy based on traditional image processing algorithms, including HSV color space segmentation of cell walls and Otsu adaptive threshold segmentation of cell cavities, and obtains cell instances through connected component labeling. The cell data management unit is used to assign a unique identifier to each cell instance and establish a structured mapping relationship between the identifier and cell outline data and geometric parameters. The parameter calculation and interaction unit calculates the geometric parameters of cells in batches according to the structured mapping relationship, and supports users to interactively filter and manage data based on the identifier.

2. The system according to claim 1, characterized in that, The image processing and analysis unit includes: Image input module, used for inputting images; The preprocessing module is used to perform color space conversion, noise filtering, contrast enhancement, and normalization on the input image; The instance segmentation module is used to segment the cell wall using HSV adaptive thresholding based on K-means clustering and segment the cell cavity using grayscale thresholding based on the Otsu algorithm, generating an initial cell mask. The post-processing module is used to perform morphological filtering, intelligent hole filling, small object removal, and multi-condition feature filtering on the initial mask to obtain accurate independent cell instances.

3. The system according to claim 2, characterized in that, The HSV adaptive threshold segmentation of K-means clustering includes: sampling the HSV color space and performing clustering using the K-means algorithm; filtering hues and designating clusters that match the hues as cell wall regions; and determining the HSV threshold range based on the percentiles of the clustering results.

4. The system according to claim 1, characterized in that, The multi-condition feature filtering includes: comprehensive filtering based on area, roundness, aspect ratio, perimeter, and brightness.

5. The system according to claim 1, characterized in that, The parameter calculation and interaction unit includes: The parameter calculation module is used to calculate the area, equivalent diameter, and cell wall thickness of each cell cavity in batches. The cell wall thickness is measured using a ray tracing-based method, which measures the distance from the centroid of the cell cavity to the adjacent cell or the maximum search distance along a set angle direction. The interactive filtering module is used to visually display the outlines and parameter lists of cells with identifiers, and respond to user commands to filter cells with specific identifiers, dynamically updating the statistical analysis results.

6. A method for executing the system as described in claim 1, characterized in that, Includes the following steps: (1) Perform color space conversion, noise filtering, contrast enhancement and standardization on the input microscopic images of cross-sections of gymnosperm wood; (2) Perform dual-channel segmentation on the preprocessed image: use HSV adaptive threshold segmentation based on K-means clustering to obtain the cell wall mask, and use grayscale threshold segmentation based on Otsu algorithm to obtain the cell cavity mask; perform morphological filtering, intelligent hole filling and multi-condition feature filtering on the mask to obtain multiple independent cell instances, and extract the precise contour data of each independent cell instance. (3) Assign a globally unique identifier to each independent cell instance obtained above and establish a data structure mapping relationship; based on the data structure mapping relationship, calculate the cell wall thickness, equivalent diameter and cell cavity area of ​​each cell instance in batches, and associate the calculation results with the mapping relationship; (4) Visualize the cell outlines and parameter list with the identifier, and respond to the user's filtering or exclusion instructions for specific cells based on the identifier, and dynamically update the statistical analysis results of the cell population.

7. The method according to claim 6, characterized in that, The structured mapping relationship established in the cell data management unit is implemented using pandas DataFrame. The unique identifier is used as the index, and the structured data object containing cell outline data and geometric parameters is used as the value to achieve efficient storage and instant retrieval of cell data.

8. The method according to claim 6, characterized in that, The method for calculating the cell wall thickness of the cell instance in step (3) is as follows: calculate the direction vector according to the set ray angle, and extend it from the centroid of the cell cavity along the ray direction to both sides; first detect the cell cavity boundary, and then continue to extend and measure through the cell wall area until an adjacent cell is detected or the maximum search distance is reached; measure the left wall thickness and the right wall thickness respectively, and calculate the average value as the average wall thickness of the cell.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 6-8.

Images