A method for accurately measuring the wall thickness of a scots pine tracheid based on a micro-CT image, a scots pine high-carbon fixation strain breeding method and system

Abstract

This invention discloses a method for accurately measuring tracheid wall thickness in *Pinus sylvestris* based on micro-CT images, a method for selecting high carbon-fixing varieties of *Pinus sylvestris*, and a system, belonging to the field of forest tree genetics and breeding technology. Addressing the systematic underestimation of tracheid wall thickness due to partial volume effect in micro-CT measurements, this invention employs the gray-scale centroid method to calculate the sub-pixel position of the wall-cavity boundary, reducing the measurement error from ±1.5 μm to ±0.3 μm. Based on accurate wall thickness data, screening criteria are established: "average wall thickness ≥ 75th percentile of the population, coefficient of variation ≤ 15%, and proportion of thick-walled tracheids ≥ 60%". Combined with carbon storage index verification, this enables precise selection of high carbon-fixing varieties. This method is non-destructive, efficient, and reproducible, and is applicable to the genetic improvement of *Pinus sylvestris* and similar coniferous trees.

Description

Technical Field

[0001] This invention relates to the field of forest tree genetics and breeding technology, specifically to a method for accurately measuring the tracheid wall thickness of Pinus sylvestris based on micro-CT images, a method and system for selecting high carbon-fixing varieties of Pinus sylvestris. Background Technology

[0002] Pinus sylvestris var. mongolica is an important shelterbelt and timber species in the Three-North Region of my country, characterized by its cold resistance, drought resistance, rapid growth, and excellent wood quality. With the implementation of the national "dual carbon" strategy, the carbon sequestration function of forests has received high attention, and breeding Pinus sylvestris varieties with high carbon sequestration capacity has become an important direction for breeding work.

[0003] Current forest tree breeding primarily focuses on growth rate and timber yield, with limited research on systematic breeding for carbon sequestration capacity. Existing technologies suffer from the following shortcomings: ① lack of rapid, non-destructive methods for detecting the microstructure of wood; ② lack of a quantitative relationship between tracheid wall thickness and carbon storage; ③ lack of screening criteria that comprehensively consider both material uniformity and carbon sequestration capacity; ④ traditional breeding methods are time-consuming and inefficient. The carbon sequestration capacity of wood mainly depends on its density and carbon content, and the cell wall thickness of tracheids, as the main constituent cells of coniferous trees, directly determines wood density. Studies have shown that a 10% increase in tracheid wall thickness can increase wood density by 8-12%, correspondingly increasing carbon storage per unit volume.

[0004] In the prior art, the measurement of tracheid wall thickness mainly relies on the following methods:

[0005] (1) Optical microscopy method: Wood slices are prepared and measured directly under an optical microscope. The resolution of this method is limited by the optical diffraction limit (about 0.2 μm), and it is a two-dimensional destructive sampling method, which cannot obtain three-dimensional structural information.

[0006] (2) Scanning electron microscopy: It has high resolution (up to the nm level), but the sample preparation is complicated, requiring dehydration and gold sputtering. It is also a destructive detection method and cannot be used for live selection.

[0007] (3) Micro-CT method: It can realize three-dimensional non-destructive testing, but there is a boundary blurring problem caused by partial volume effect. When the voxel size (1-5μm) is close to the cell wall thickness (2-8μm), the gray value of the wall-cavity boundary region is mixed gray value. The traditional threshold segmentation method will misjudge the boundary voxel as "cavity" or "wall", resulting in a systematic underestimation of the wall thickness by 1-2μm. The spatial resolution is limited, making it difficult to accurately identify the wall-cavity boundary; there is a lack of dedicated image processing algorithms for wood cell structure.

[0008] (4) Traditional threshold segmentation method (global threshold, wall-lumen valley value): In the analysis of wood cell structure based on micro-CT images, the global threshold segmentation method is currently the most commonly used image segmentation method. This method is based on the bimodal feature of the grayscale histogram, and takes the valley value between the high grayscale peak corresponding to the cell wall and the low grayscale peak corresponding to the cell lumen as the global threshold T_seg, and performs binary processing on the image:

[0009] Voxels with grayscale values ≥ T_seg are determined as cell walls (foreground)

[0010] Voxels with grayscale values < T_seg are determined as cell lumens (background)

[0011] However, the traditional threshold segmentation method has the following technical defects:

[0012] ① Limited boundary positioning accuracy: The global threshold method only performs binary division at the integer pixel level and cannot identify the boundary positions at the sub-pixel level. Since the voxel size of micro-CT is usually 1 - 5μm, and the wall thickness of softwood tracheids is usually 2 - 8μm, the integer pixel positioning error can reach more than 50% of the voxel size, seriously affecting the wall thickness measurement accuracy.

[0013] ② Influence of partial volume effect: When the wall-lumen boundary passes through a voxel, the grayscale value of this voxel is the weighted average of the wall material and the lumen material, showing an intermediate gray level. The global threshold method rigidly classifies this intermediate gray level voxel as either the wall or the lumen, resulting in a systematic shift of the boundary position and deviation in the wall thickness measurement result.

[0014] ③ Sensitivity to local grayscale inhomogeneity: The grayscale value of the wood cell wall is affected by factors such as local density, microfibril angle, and chemical composition distribution, showing spatial variation. The global threshold uses a single threshold to process the entire image and cannot adapt to local grayscale changes, easily resulting in over-segmentation or under-segmentation in regions with grayscale inhomogeneity.

[0015] ④ Strong noise sensitivity: Micro-CT images have quantum noise and electronic noise, and the valley position of the grayscale histogram is easily disturbed by noise and is unstable, resulting in uncertainty in threshold selection and reducing measurement repeatability.

[0016] ⑤ Lack of boundary smoothing mechanism: The boundary after threshold segmentation is jagged and contains a large number of isolated noise points. Directly using it for wall thickness calculation will introduce large errors and requires additional post-processing steps.

[0017] This error has a significant impact on breeding screening: Assuming the true wall thickness is 6μm and the measured value is 4.5μm (underestimated by 25%), it may misjudge high carbon sequestration individuals as ordinary individuals, resulting in the loss of genetic resources.

[0018] Therefore, there is an urgent need for a method to measure tracheid wall thickness that can accurately correct for partial volume effects, so as to provide reliable data support for the breeding of high carbon fixation strains. Summary of the Invention

[0019] To overcome the shortcomings of the prior art, one of the objectives of this invention is to provide a method for accurate measurement of tracheid wall thickness based on sub-pixel boundary positioning, thereby solving the technical problem of systematic underestimation of wall thickness.

[0020] The second objective of this invention is to provide a method for breeding high carbon-fixing strains of Pinus sylvestris, which applies the above-mentioned method for accurate measurement of tracheid wall thickness based on sub-pixel boundary positioning.

[0021] The third objective of this invention is to provide a micro-CT image processing system for the precise measurement of tracheid wall thickness in Pinus sylvestris, which implements the aforementioned method for precise measurement of tracheid wall thickness based on sub-pixel boundary positioning in the selection of high carbon-fixing Pinus sylvestris varieties.

[0022] One of the objectives of this invention is achieved through the following technical solution:

[0023] A method for accurately measuring the tracheid wall thickness of Pinus sylvestris based on micro-CT images includes the following steps:

[0024] Step S1: Sample preparation and image acquisition

[0025] A sample of Pinus sylvestris wood was taken, with dimensions of 5 mm radially, 5 mm tangentially, and 20 mm longitudinally. The sample was fixed on the rotating stage of the micro-CT equipment, ensuring that the long axis of the sample was parallel to the rotation axis. The scanning parameters were set as follows: tube voltage 60 kV, tube current 100 μA, exposure time 500 ms, projection number 1600, and voxel size 2 μm. The FDK reconstruction algorithm was used to obtain three-dimensional grayscale image data. The image was 16-bit grayscale with a grayscale value range of 0-65535.

[0026] Step S2: Grayscale Correction and Noise Reduction

[0027] Based on the known CT values ​​of the air region and the aluminum standard part, a gray-to-density conversion relationship is established: ρ = 0.0018 × G - 0.0236, where ρ is the density (g / cm³) and G is the image gray value; a three-dimensional nonlocal mean filter is performed on the three-dimensional gray image, with a search window size of 7 × 7 × 7 voxels, a similarity window size of 3 × 3 × 3 voxels, and a filter parameter h = 0.8 × σ, where σ is the standard deviation of image noise;

[0028] Step S3: Coarse segmentation of tracheid regions

[0029] Calculate the grayscale histogram of the processed image in step S2, identify the high-grayscale peak corresponding to the cell wall and the low-grayscale peak corresponding to the cell cavity; determine the valley grayscale value T_seg between the two peaks. The method for determining T_seg is as follows: find the minimum point between the high-grayscale peak and the low-grayscale peak in the histogram. If there are multiple local minima, take the one closest to the low-grayscale peak; binarize the image with T_seg as the threshold: voxels with grayscale values ≥ T_seg are labeled as the cell wall (value 1), and voxels with grayscale values < T_seg are labeled as the cell cavity (value 0).

[0030] Step S4: Precise localization of the wall-cavity boundary

[0031] Sub-step S41: Identification of boundary voxels

[0032] Traverse the binary image to extract the wall-cavity boundary voxels. The definition of the wall-cavity boundary voxels is as follows: in the 26-neighborhood, there are at least one voxel in the cell wall region and at least one voxel in the cell cavity region simultaneously.

[0033] Sub-step S42: Determination of the normal direction

[0034] For each wall-cavity boundary voxel, calculate its local normal direction. The calculation method is as follows: extract the binary values of this wall-cavity boundary voxel and its 8 nearest neighbor voxels, calculate the grayscale gradient direction through the Sobel operator, and the opposite direction of this direction is the normal direction of the wall-cavity interface, pointing to the cell cavity.

[0035] Sub-step S43: Extraction of grayscale distribution

[0036] Along the normal direction, with the wall-cavity boundary voxel as the center, extend 2 voxels towards the cell wall side and the cell cavity side respectively, and extract the grayscale values of a total of n = 5 voxels to form a grayscale distribution sequence {G1, G2, G3, G4, G5}, where G3 is the grayscale value of the wall-cavity boundary voxel.

[0037] Sub-step S44: Calculation of sub-pixel boundary

[0038] Calculate the cell cavity background grayscale G_bg. The method is as follows: take the arithmetic mean of the grayscale values of all voxels marked as the cell cavity in this sample as G_bg; apply the grayscale centroid method to calculate the sub-pixel-level boundary position:

[0039]

[0040] where x_i is the position coordinate of the i-th voxel along the normal direction relative to the center of the wall-cavity boundary voxel, x_i = (i - 3)*d, d is the voxel size (2μm), and i = 1, 2, 3, 4, 5.

[0041] Sub-step S45: Boundary smoothing and outlier removal

[0042] For all calculated sub-pixel boundary points, a moving average method is used for smoothing, with a window width of 3 voxels. The local mean μ_local and standard deviation σ_local of the wall thickness estimate values ​​of each sub-pixel boundary point and other sub-pixel boundary points in its neighborhood are calculated. If the wall thickness estimate value of a point deviates from μ_local by more than 2σ_local, it is identified as an outlier and removed, and the value of the point is replaced by μ_local.

[0043] Step S5: Calculation of tracheid wall thickness

[0044] The cross-sectional contour of a single intact tracheid is identified as follows: The contour of the closed cavity region is extracted from the binary image, and regions with an area in the range of 50-500 μm² are selected as effective tracheids. For each effective tracheid, four azimuth angles are selected on its cross-section: 0°, 45°, 90°, and 135°. At each azimuth, the sub-pixel coordinates (x_in, y_in) and (x_out, y_out) of the inner and outer sidewall-cavity boundaries are determined. The wall thickness T_w and θ at this azimuth angle are calculated. The arithmetic mean of the wall thicknesses at the four azimuth angles is taken as the wall thickness T_w of the tracheid.

[0045] Step S6: System Error Correction

[0046] Preparation of standard thickness samples: Aluminum foil stacks with thicknesses of 2.0 μm, 4.0 μm, 6.0 μm, 8.0 μm, and 10.0 μm were used, with 10 replicates for each thickness. The standard samples were subjected to micro-CT scanning under the same scanning conditions as in step S1, and the CT wall thickness of each sample was measured using the methods in steps S2-S5. A linear regression relationship was established: T_corrected = a × T_measured + b, where a and b were determined by least squares fitting. This regression relationship was applied to correct the wall thickness measurements of all Pinus sylvestris samples to obtain accurate measurements of the tracheid wall thickness of Pinus sylvestris.

[0047] Furthermore, in step S1, the pine wood samples are taken from the diameter at breast height of 15-20 year old trees, specifically from the 8th-12th growth rings, avoiding the pith and bark areas. The sample surface is sanded to 1000 grit to ensure flatness.

[0048] Furthermore, the micro-CT device in step S1 is a ZEISS Xradia 510 Versa or a device with equivalent performance, with a spatial resolution better than 2μm and a 16-bit sCMOS detector.

[0049] Furthermore, the three-dimensional nonlocal mean filtering in step S2 is replaced by three-dimensional median filtering, with a filtering window size of 3×3×3 voxels, which is suitable for image data with low signal-to-noise ratio.

[0050] Furthermore, in step S3, when the bimodal characteristics of the grayscale histogram are not obvious, the Otsu automatic thresholding method is used to determine T_seg. The discrimination criterion of the Otsu method is to maximize the variance between the two classes, cell wall and cell cavity.

[0051] The second objective of this invention is achieved by the following technical solution:

[0052] A method for breeding high carbon-fixing cultivars of Pinus sylvestris includes the following stages:

[0053] Phase P1: Establishment of the experimental population

[0054] Select 30-50 superior clones or half-sib families of Pinus sylvestris and establish a comparative experimental forest under the same site conditions. The planting density is 2m×3m, and 3-5 replicates are set up, with no less than 9 trees in each replicate.

[0055] Phase P2: Sample Collection and Wall Thickness Measurement

[0056] When the forest is 15-20 years old, 3-5 sample trees are randomly selected from each clone or family. The core is drilled at the diameter at breast height (1.3m height) using a growth cone. The core diameter is 12mm. Samples with a radial diameter of 5mm, a tangential diameter of 5mm, and a longitudinal diameter of 20mm are extracted from the core. The tracheid wall thickness is measured using the precise measurement method of Pinus sylvestris var. mongolica based on micro-CT images as described in any one of claims 1-5. At least 100 intact tracheids are measured for each sample. The following indicators are calculated: mean wall thickness T_w_mean, standard deviation of wall thickness σ_Tw, coefficient of variation of wall thickness CV_Tw=(σ_Tw / T_w_mean)×100%, and proportion of thick-walled tracheids P_thick, that is, the proportion of tracheids with a wall thickness ≥6μm.

[0057] Phase P3: Screening of High Carbon Fixation Candidates

[0058] Based on the measurement results of stage P2, clones or families that simultaneously meet the following three conditions are identified as candidates for high carbon fixation:

[0059] Condition C1: T_w_mean ≥ T_75, where T_75 is the 75th percentile of the wall thickness distribution of all samples in the population being measured.

[0060] Condition C2: CV_Tw ≤ 15%;

[0061] Condition C3: P_thick ≥ 60%;

[0062] Phase P4: Verification of carbon fixation performance

[0063] For the candidate strains selected in stage P3, the following indicators were measured:

[0064] Basic density of wood ρ_basic: The water displacement method was used, in accordance with GB / T1933-2009, and the sample size was 20mm×20mm×20mm;

[0065] Carbon content (C_content): determined using an elemental analyzer, in accordance with GB / T35809-2018. The sample was pulverized and passed through a 100-mesh sieve.

[0066] The carbon storage index (CSI) is calculated as ρ_basic × C_content, with units of kgC / m³.

[0067] Phase P5: Comprehensive Evaluation and Breeding Decisions

[0068] Strains that simultaneously meet the following conditions are selected as superior strains with high carbon fixation:

[0069] CSI≥1.2×CSI_pop, where CSI_pop is the arithmetic mean of the carbon storage index of the experimental population;

[0070] Tree height growth H ≥ 0.8 × H_pop, where H_pop is the arithmetic mean of tree height in the experimental population; no obvious susceptibility to pests and diseases or trunk defects;

[0071] Genetic stability of the selected strains was evaluated, and wall thickness and CSI were repeatedly measured for three consecutive years. Strains with a coefficient of variation of <20% were selected for regional trials.

[0072] Furthermore, the superior clones in stage P1 are derived from the asexual reproduction of superior trees in the National Pinus sylvestris Seed Orchard or superior seed source area, and the half-sib families are derived from the controlled pollination or open pollination of superior trees.

[0073] Furthermore, in stage P2, the core sampling avoids the growth ring boundaries to ensure that the sample is located in a single region of earlywood or latewood, or earlywood and latewood samples are collected separately for independent analysis.

[0074] The third objective of this invention is achieved by the following technical solution:

[0075] A micro-CT image processing system for precise measurement of tracheid wall thickness in Pinus sylvestris, comprising:

[0076] Image acquisition module: Equipped with a micro-CT device, used to acquire three-dimensional grayscale image data of Pinus sylvestris wood samples, with voxel size of 1-5μm;

[0077] Preprocessing module: connected to the image acquisition module, used to perform grayscale correction and noise reduction processing. The grayscale correction is based on establishing a grayscale-density conversion relationship between air and aluminum standard parts, and the noise reduction adopts three-dimensional nonlocal mean filtering.

[0078] Segmentation module: Connected to the preprocessing module, it is used for binary segmentation based on the bimodal valley values ​​of the gray-level histogram to distinguish between cell wall regions and cell cavity regions;

[0079] Boundary localization module: Connected to the segmentation module, it performs sub-pixel level boundary localization, including:

[0080] Boundary voxel recognition unit: identifies boundary voxels in which both wall and cavity voxels exist within a 26-neighborhood;

[0081] Normal calculation unit: Calculates the normal direction of the wall-cavity interface based on the Sobel operator;

[0082] Grayscale extraction unit: Extracts the grayscale distribution of 5 voxels along the normal direction of the wall-cavity interface;

[0083] Centroid Calculation Unit: The gray-scale centroid method is applied to calculate the sub-pixel boundary coordinates;

[0084] Smoothing unit: Performs moving average smoothing and outlier removal on boundary points;

[0085] Wall thickness calculation module: connected to the boundary positioning module, used to identify the outline of a single tracheid, calculate the wall thickness in four azimuth angles and take the average value;

[0086] Calibration module: Connected to the wall thickness calculation module, it stores the regression correction parameters of standard thickness samples and performs systematic error correction on the measured values;

[0087] Output module: Used to output corrected tracheid wall thickness data and statistical indicators.

[0088] Furthermore, it also includes a breeding decision module, which is connected to the output module and is used to perform high carbon fixation candidate screening based on wall thickness data, specifically including:

[0089] Statistical calculation unit: Calculates average wall thickness, coefficient of variation, and proportion of thick-walled tracheids;

[0090] Screening and judgment unit: Candidates are judged based on preset thresholds (average wall thickness ≥ 75th percentile, coefficient of variation ≤ 15%, and proportion of thick-walled tracheids ≥ 60%).

[0091] Data storage unit: Stores wall thickness data and screening results for candidate strains.

[0092] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0093] (1) This invention provides a method for accurately measuring the tracheid wall thickness of Pinus sylvestris based on micro-CT images, a method and system for selecting high carbon-fixing strains of Pinus sylvestris, and a system for this purpose.

[0094] (1) Significantly improved measurement accuracy: By using sub-pixel boundary positioning, the wall thickness measurement error is reduced from ±1.5μm in the traditional method to ±0.3μm (relative error <5%), solving the systematic underestimation problem of thick-walled tracheids (>5μm). (2) Non-destructive and efficient: The single sample scanning time is about 30 minutes, which can obtain three-dimensional data of thousands of tracheids. Compared with the traditional slicing method, the efficiency is improved by more than 100 times, and the samples can be reused. (3) Quantitative screening criteria: The screening threshold (75th percentile) based on population statistics is established to avoid subjective judgment and ensure the repeatability and comparability of the breeding results. (4) Clear genetic gain: The strains bred by this method are expected to have a genetic gain of 12-15% in wood density and 15-18% in carbon storage, without affecting the growth rate.

[0095] Significantly improved measurement accuracy: (1) Subpixel-level positioning: Subpixel-level boundary positioning is achieved by using the gray-scale centroid method, breaking through the voxel size limitation, and the positioning accuracy can reach 0.1-0.2μm, which is far better than the 2μm accuracy of the traditional whole-pixel method. (2) System error correction: A regression correction model is established by using standard thickness samples to effectively eliminate some volume effects and systematic errors caused by instrument drift, improving the measurement accuracy by 15-20%. (3) Local adaptive processing: The normal direction determination and local background correction algorithm adapt to changes in cell wall curvature and local gray-scale non-uniformity, improving the measurement reliability of complex areas.

[0096] Measurement efficiency is greatly improved: (1) Non-destructive testing: No need for slicing and sample preparation, it only takes 2-3 hours to obtain wall thickness data from scanning a single sample, while the traditional optical microscopy method takes 1-2 days. (2) Three-dimensional information acquisition: A single scan can obtain the three-dimensional structural information of thousands of tracheids, which has strong statistical representativeness, while traditional methods can only observe a limited cross section. (3) Automated processing: The image processing process is standardized and automated, reducing manual intervention and making it suitable for large-scale sample analysis.

[0097] The selection efficiency has been significantly improved: (1) Early selection: Effective selection can be carried out when the forest age is 15-20 years, which is 10-15 years earlier than the traditional selection based on the properties of mature wood. (2) Non-destructive screening: Sampling is carried out using growth cones, which does not affect the normal growth of plants and can retain superior genotypes for subsequent propagation. (3) Balance of multiple traits: The average wall thickness, uniformity and the proportion of thick-walled cells are comprehensively considered, taking into account both carbon fixation capacity and material stability. Detailed Implementation

[0098] The present invention will be further described below with reference to specific embodiments. It should be noted that, without conflict, the embodiments or technical features described below can be arbitrarily combined to form new embodiments.

[0099] Example 1: Precise Measurement of Tracheid Wall Thickness in Pinus sylvestris var. mongolica

[0100] Sample preparation

[0101] Core samples were extracted from an 18-year-old Pinus sylvestris plantation in Qiqihar City, Heilongjiang Province, using a growth cone at a diameter at breast height (DBH) of 1.3m. Three earlywood samples were taken from the 10th growth ring, each measuring 5mm radially, 5mm tangentially, and 20mm longitudinally, using a sharp blade. The sample surfaces were progressively sanded to 1000 grit, ultrasonically cleaned for 5 minutes, and dried at 60℃ to constant weight.

[0102] Micro-CT scan

[0103] Scanning was performed using a ZEISS Xradia 510 Versa micro-CT scanner. The sample was fixed on a rotating stage, and its position was adjusted so that its long axis was parallel to the rotation axis. Scanning parameters were set as follows: tube voltage 60kV, tube current 100μA, exposure time 500ms, projection number 1600, voxel size 2μm. The scanning field of view was set to 10.24mm × 10.24mm, and the reconstructed image size was 5120 × 5120 pixels.

[0104] Image processing and wall thickness measurement

[0105] Image processing and wall thickness measurement are performed according to steps S2-S5 of the present invention:

[0106] (1) Gray-scale correction and noise reduction: Based on the air region (gray-scale value ≈ 1500) and the aluminum standard part (gray-scale value ≈ 45000), a conversion relationship ρ = 0.0018 × G - 0.0236 is established. Three-dimensional nonlocal mean filtering is applied, with a search window of 7 × 7 × 7 voxels and a similarity window of 3 × 3 × 3 voxels, h = 0.8 × σ.

[0107] (2) Region segmentation: Calculate the gray-level histogram, identify the cell wall peak (gray value ≈ 28000) and the cell cavity peak (gray value ≈ 8000), determine the threshold T_seg=16500, and perform binary segmentation.

[0108] (3) Boundary localization: Identify boundary voxels, calculate normal directions, extract grayscale distribution sequences (n=5), and calculate sub-pixel boundary positions using the grayscale centroid method. After moving average smoothing and outlier removal, the boundary coordinates are obtained.

[0109] (4) Wall thickness calculation: Identify effective tracheids with an area of ​​50-500 μm², and measure the wall thickness at four angles of 0°, 45°, 90° and 135° for each tracheid, and take the average value.

[0110] System error correction

[0111] Standard aluminum foil samples were prepared with thicknesses of 2.0, 4.0, 6.0, 8.0, and 10.0 μm, ten of each type, and scanned and measured under the same conditions. A regression equation was established: T_corrected = 1.08 × T_measured - 0.15 (R² = 0.994). This correction equation was applied to correct the measured values.

[0112] Results Analysis

[0113] After correction, the mean wall thicknesses of the three samples were 3.82 μm, 3.91 μm, and 3.75 μm, respectively, with standard deviations of 0.42 μm, 0.38 μm, and 0.45 μm, and coefficients of variation of 11.0%, 9.7%, and 12.0%, respectively. The number of tracheids measured per sample was 156, 142, and 168, respectively, with coefficients of variation (CV_n) of 7.8%, 8.2%, and 7.5%, respectively, all <10%, indicating sufficient sample size.

[0114] Compared with the traditional optical microscopy method, the measurement results of the method of the present invention have a correlation coefficient of 0.96 with the microscopy method when the slice thickness is 15μm, but the measurement time is shortened by 80% and the slice damage is avoided.

[0115] Example 2: Breeding of high carbon-fixing strains of Pinus sylvestris var. mongolica

[0116] Establishment of experimental groups

[0117] (1) 35 superior clones from the Pinus sylvestris seed orchard in Heilongjiang Province were selected and a comparative experimental forest was established under the same site conditions: sandy soil, average annual precipitation of 350 mm, planting density of 2 m × 3 m, randomized block design, 4 replicates, 9 trees per replicate.

[0118] Sample collection and wall thickness measurement

[0119] When the forest was 18 years old, four specimens were randomly selected from each clone, and core samples were drilled at the diameter at breast height (DBH). Tracheid wall thickness was measured according to the method in Example 1, with 120 tracheids measured per sample.

[0120] Filtering results

[0121] (1) The average wall thickness T_w_mean of the 35 clones ranged from 3.2 to 5.8 μm, and the 75th percentile T_75 = 4.9 μm. Eight clones simultaneously met the conditions C1 (T_w_mean ≥ 4.9 μm), C2 (CV_Tw ≤ 15%), and C3 (P_thick ≥ 60%), with a selection rate of 22.9%.

[0122] Carbon fixation performance verification

[0123] The basic density and carbon content of the wood were determined for eight candidate clones:

[0124] Basic density of wood ρ_basic: 0.42-0.51 g / cm³;

[0125] Carbon content (C_content): 48.5%-50.2%;

[0126] Carbon Storage Index (CSI): 203.7-256.0 kgC / m³;

[0127] The population average CSI_pop = 198.5 kgC / m³. There are 4 clones that simultaneously satisfy CSI ≥ 1.2 × CSI_pop (i.e. ≥ 238.2 kgC / m³) and tree height growth ≥ 0.8 × H_pop.

[0128] Genetic stability assessment

[0129] (1) Repeated measurements were performed on the four selected clones for three consecutive years (18-20 years old). The coefficients of variation of wall thickness and CSI were 8.5%-12.3% and 6.8%-11.5%, respectively, both <20%, indicating good genetic stability and suitable for regional trials.

[0130] Breeding effect

[0131] Compared with traditional breeding methods based on growth rate, the carbon storage of high carbon-fixing strains bred using this method is increased by 25-30%, while the growth rate remains above 85% of the population average, achieving a balanced improvement in carbon fixation capacity and growth performance.

[0132] Example 3: Comparison of different noise reduction methods

[0133] To verify the effectiveness of the noise reduction method of the present invention, three-dimensional nonlocal mean filtering (the method of the present invention) and three-dimensional median filtering (the comparison method) were respectively applied to image data with the same low signal-to-noise ratio.

[0134] result:

[0135] Non-local mean filtering: signal-to-noise ratio improved by 12.5 dB, wall thickness measurement repeatability error reduced by 3.2%;

[0136] Median filtering: signal-to-noise ratio improved by 8.3 dB, wall thickness measurement repeatability error reduced by 5.8%;

[0137] This indicates that nonlocal mean filtering has a better noise reduction effect while preserving edge information, but for images with low signal-to-noise ratio, median filtering can be used as an alternative.

[0138] Example 4: Comparison of different threshold segmentation methods

[0139] For samples with indistinct bimodal features in the grayscale histogram (latewood portion), the valley method and Otsu method of this invention are used to determine the threshold, respectively.

[0140] result:

[0141] Valley method: T_seg=17200, poor cell wall continuity after segmentation.

[0142] Otsu's method: T_seg=18500, with better segmentation results and better cell wall integrity, indicating that when the bimodal feature is not obvious, Otsu's method can provide a more reliable threshold.

[0143] Example 5: Verification of the sub-pixel boundary localization method

[0144] Experimental materials: Standard samples with known wall thicknesses were selected, aluminum foil stacks with thicknesses of 2.0 μm, 4.0 μm, 6.0 μm, 8.0 μm and 10.0 μm, 10 replicates of each, and pine wood samples, 3 clones, 5 replicates of each.

[0145] Experimental equipment: ZEISS Xradia 510 Versa micro-CT, voxel size 2μm.

[0146] Experimental Groups:

[0147] Control group: Traditional threshold segmentation method (global threshold, wall-cavity valley value);

[0148] Experimental group: Subpixel grayscale centroid method of the present invention (n=5).

[0149] result:

[0150] Table 1

[0151]

[0152] The results in Table 1 show that the subpixel method reduces the systematic error from 18-40% to less than 5%, meeting the accuracy requirements for breeding screening.

[0153] The above embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of protection of the present invention. Any non-substantial changes and substitutions made by those skilled in the art based on the present invention shall fall within the scope of protection claimed by the present invention.

Claims

1. A method for accurately measuring the tracheid wall thickness of *Pinus sylvestris* based on micro-CT images, characterized in that, It includes the following steps: Step S1: Sample preparation and image acquisition Take a Pinus sylvestris var. mongolica wood sample with a size of 5 mm in the radial direction, 5 mm in the tangential direction, and 20 mm in the longitudinal direction; fix the sample on the rotating table of the micro-CT device to ensure that the long axis of the sample is parallel to the rotation axis; set the scanning parameters: tube voltage 60 kV, tube current 1 mA, exposure time 500 ms, number of projections 1600, voxel size 2 μm; use the FDK reconstruction algorithm to obtain three-dimensional gray image data, the image is 16-bit gray, and the gray value range is 0 - 65535; Step S2: Gray correction and noise reduction processing Based on the known CT values of the air region and the aluminum standard part, establish a gray-density conversion relationship: ρ = 0.0018×G - 0.0236, where ρ is the density in g / cm³ and G is the image gray value; perform three-dimensional non-local mean filtering on the three-dimensional gray image, the search window size is 7×7×7 voxels, the similarity window size is 3×3×3 voxels, and the filtering parameter h = 0.8×σ, where σ is the standard deviation of the image noise; Step S3: Coarse segmentation of the tracheid region Calculate the gray histogram of the image processed in Step S2, and identify the high-gray peak corresponding to the cell wall and the low-gray peak corresponding to the cell lumen; determine the valley gray value T_seg between the two peaks, and the determination method of T_seg is: find the minimum point between the high-gray peak and the low-gray peak in the histogram, if there are multiple local minima, take the one closest to the low-gray peak; perform binarization on the image with T_seg as the threshold: voxels with a gray value ≥ T_seg are marked as the cell wall with a value of 1, and voxels with a gray value < T_seg are marked as the cell lumen with a value of 0; Step S4: Precise positioning of the wall-lumen boundary Sub-step S41: Identification of boundary voxels Traverse the binary image to extract the wall-lumen boundary voxels, and the definition of the wall-lumen boundary voxels is: in the 26-neighborhood, there are at least one cell wall region voxel and at least one cell lumen region voxel; Sub-step S42: Determination of the normal direction For each wall-lumen boundary voxel, calculate its local normal direction, and the calculation method is: extract the binary values of this wall-lumen boundary voxel and its 8 nearest neighbor voxels, calculate the gray gradient direction through the Sobel operator, and the opposite direction of this direction is the normal direction of the wall-lumen interface, pointing to the cell lumen; Sub-step S43: Extraction of gray distribution Along the normal direction, with the wall-lumen boundary voxel as the center, extend 2 voxels to both the cell wall side and the cell lumen side, and extract the gray values of a total of n = 5 voxels to form a gray distribution sequence {G1, G2, G3, G4, G5}, where G3 is the gray value of the wall-lumen boundary voxel; Sub-step S44: Calculation of the sub-pixel boundary Calculate the cell lumen background gray value G_bg, and the method is: take the arithmetic mean of the gray values of all voxels marked as the cell lumen in this sample as G_bg; apply the gray centroid method to calculate the sub-pixel level boundary position: where, x_i is the position coordinate of the i-th voxel along the normal direction relative to the center of the wall-lumen boundary voxel, x_i = (i - 3)*d, d is the voxel size, which is 2 μm, and i = 1, 2, 3, 4, 5; Sub-step S45: Boundary smoothing and outlier removal For all calculated sub-pixel boundary points, a moving average method is used for smoothing, with a window width of 3 voxels. The local mean μ_local and standard deviation σ_local of the wall thickness estimate values ​​of each sub-pixel boundary point and other sub-pixel boundary points in its neighborhood are calculated. If the wall thickness estimate value of a point deviates from μ_local by more than 2σ_local, it is identified as an outlier and removed, and the value of the point is replaced by μ_local. Step S5: Calculation of tracheid wall thickness The cross-sectional contour of a single intact tracheid is identified as follows: The contour of the closed cavity region is extracted from the binary image, and regions with an area in the range of 50-500 μm² are selected as valid tracheids. For each valid tracheid, four azimuth angles are selected on its cross-section: 0°, 45°, 90°, and 135°. At each azimuth, the sub-pixel coordinates (x_in, y_in) and (x_out, y_out) of the inner and outer sidewall-cavity boundaries are determined. The wall thickness at that azimuth angle is then calculated. Take the arithmetic mean of the wall thickness in four directions as the wall thickness T_w of the tracheid; Step S6: System Error Correction Preparation of standard thickness samples: Aluminum foil stacks with thicknesses of 2.0 μm, 4.0 μm, 6.0 μm, 8.0 μm, and 10.0 μm were used, with 10 replicates for each thickness. The standard samples were subjected to micro-CT scanning under the same scanning conditions as in step S1, and the CT wall thickness of each sample was measured using the methods in steps S2-S5. A linear regression relationship was established: T_corrected = a × T_measured + b, where a and b were determined by least squares fitting. This regression relationship was applied to correct the wall thickness measurements of all Pinus sylvestris samples to obtain accurate measurements of the tracheid wall thickness of Pinus sylvestris.

2. The method for accurately measuring the tracheid wall thickness of *Pinus sylvestris* based on micro-CT images as described in claim 1, characterized in that, In step S1, the pine wood samples are taken from the diameter at breast height of 15-20 year old trees, specifically from the 8th-12th growth rings, avoiding the pith and bark areas. The sample surface is sanded to 1000 grit to ensure flatness.

3. The method for accurately measuring the tracheid wall thickness of *Pinus sylvestris* based on microscopic CT images as described in claim 1, characterized in that... The micro-CT device in step S1 is a ZEISS Xradia 510 Versa or a device with equivalent performance, with a spatial resolution better than 2μm and a 16-bit sCMOS detector.

4. The method for accurately measuring the tracheid wall thickness of *Pinus sylvestris* based on micro-CT images as described in claim 1, characterized in that, The three-dimensional nonlocal mean filtering in step S2 is replaced by three-dimensional median filtering, with a filtering window size of 3×3×3 voxels, which is suitable for image data with low signal-to-noise ratio.

5. The method for accurately measuring the tracheid wall thickness of *Pinus sylvestris* based on micro-CT images as described in claim 1, characterized in that, In step S3, when the bimodal characteristics of the grayscale histogram are not obvious, the Otsu automatic thresholding method is used to determine T_seg. The discrimination criterion of the Otsu method is to maximize the variance between the two classes, cell wall and cell cavity.

6. A method for breeding high carbon-fixing strains of Pinus sylvestris, characterized in that, Includes the following stages: Phase P1: Establishment of the experimental population Select 30-50 superior clones or half-sib families of Pinus sylvestris and establish a comparative experimental forest under the same site conditions. The planting density is 2m×3m, and 3-5 replicates are set up, with no less than 9 trees in each replicate. Phase P2: Sample Collection and Wall Thickness Measurement When the forest is 15-20 years old, 3-5 sample trees are randomly selected from each clone or family. At the diameter at breast height (DBH), i.e., 1.3m in height, a core is drilled using a growth cone. The core diameter is 12mm. From the core, samples with a radial diameter of 5mm, a tangential diameter of 5mm, and a longitudinal diameter of 20mm are extracted. The tracheid wall thickness of Pinus sylvestris is measured using the precise measurement method of Pinus sylvestris var. mongolica based on micro-CT images as described in any one of claims 1-5. At least 100 intact tracheids are measured for each sample. The following indicators are calculated: mean wall thickness T_w_mean, standard deviation of wall thickness σ_Tw, coefficient of variation of wall thickness CV_Tw=(σ_Tw / T_w_mean)×100%, and proportion of thick-walled tracheids P_thick, i.e., the proportion of tracheids with a wall thickness ≥6μm. Phase P3: Screening of High Carbon Fixation Candidates Based on the measurement results of stage P2, clones or families that simultaneously meet the following three conditions are identified as candidates for high carbon fixation: Condition C1: T_w_mean ≥ T_75, where T_75 is the 75th percentile of the wall thickness distribution of all samples in the population being measured. Condition C2: CV_Tw ≤ 15%; Condition C3: P_thick ≥ 60%; Phase P4: Verification of carbon fixation performance For the candidate strains selected in stage P3, the following indicators were measured: Basic density of wood ρ_basic: The water displacement method was used, in accordance with GB / T1933-2009, and the sample size was 20mm×20mm×20mm; Carbon content (C_content): determined using an elemental analyzer, in accordance with GB / T35809-2018. The sample was pulverized and passed through a 100-mesh sieve. The carbon storage index (CSI) is calculated as ρ_basic × C_content, with units of kgC / m³. Phase P5: Comprehensive Evaluation and Breeding Decisions Strains that simultaneously meet the following conditions are selected as superior strains with high carbon fixation: CSI≥1.2×CSI_pop, where CSI_pop is the arithmetic mean of the carbon storage index of the experimental population; Tree height growth H ≥ 0.8 × H_pop, where H_pop is the arithmetic mean of tree height in the experimental population; no obvious susceptibility to pests and diseases or trunk defects; Genetic stability of the selected strains was evaluated, and wall thickness and CSI were repeatedly measured for three consecutive years. Strains with a coefficient of variation of <20% were selected for regional trials.

7. The method for breeding high carbon-fixing strains of Pinus sylvestris as described in claim 6, characterized in that, The superior clones in stage P1 are derived from the asexual reproduction of superior trees in the National Pinus sylvestris Seed Orchard or superior seed source area, and the half-sib families are derived from the controlled pollination or open pollination of superior trees.

8. The method for breeding high carbon-fixing strains of Pinus sylvestris as described in claim 6, characterized in that, In stage P2, core sampling avoids growth ring boundaries to ensure that the sample is located in a single region of earlywood or latewood, or earlywood and latewood samples are collected separately for independent analysis.

9. A micro-CT image processing system for precise measurement of tracheid wall thickness in Pinus sylvestris, characterized in that, include: Image acquisition module: Equipped with a micro-CT device, used to acquire three-dimensional grayscale image data of Pinus sylvestris wood samples, with voxel size of 1-5μm; Preprocessing module: connected to the image acquisition module, used to perform grayscale correction and noise reduction processing. The grayscale correction is based on establishing a grayscale-density conversion relationship between air and aluminum standard parts, and the noise reduction adopts three-dimensional nonlocal mean filtering. Segmentation module: Connected to the preprocessing module, it is used for binary segmentation based on the bimodal valley values ​​of the gray-level histogram to distinguish between cell wall regions and cell cavity regions; Boundary localization module: Connected to the segmentation module, it performs sub-pixel level boundary localization, including: Boundary voxel recognition unit: identifies boundary voxels in which both wall and cavity voxels exist within a 26-neighborhood; Normal calculation unit: Calculates the normal direction of the wall-cavity interface based on the Sobel operator; Grayscale extraction unit: Extracts the grayscale distribution of 5 voxels along the normal direction of the wall-cavity interface; Centroid Calculation Unit: The gray-scale centroid method is applied to calculate the sub-pixel boundary coordinates; Smoothing unit: Performs moving average smoothing and outlier removal on boundary points; Wall thickness calculation module: connected to the boundary positioning module, used to identify the outline of a single tracheid, calculate the wall thickness in four azimuth angles and take the average value; Calibration module: Connected to the wall thickness calculation module, it stores the regression correction parameters of standard thickness samples and performs systematic error correction on the measured values; Output module: Used to output corrected tracheid wall thickness data and statistical indicators.

10. A micro-CT image processing system for precise measurement of tracheid wall thickness in Pinus sylvestris as described in claim 9, characterized in that, It also includes a breeding decision module, which is connected to the output module and is used to perform high carbon sequestration candidate screening based on wall thickness data, specifically including: Statistical calculation unit: Calculates average wall thickness, coefficient of variation, and proportion of thick-walled tracheids; Screening and judgment unit: Candidates are judged based on preset thresholds, which are: average wall thickness ≥ 75th percentile, coefficient of variation ≤ 15%, and proportion of thick-walled tracheids ≥ 60%; Data storage unit: Stores wall thickness data and screening results for candidate strains.

Images