Error correction method for tree-ring width measurement

Abstract

The present invention relates to the research fields of dendrochronology and dendrometry, and in particular to a method for correcting tree ring width measurement errors. The method comprises the following steps: obtaining a tree ring sample image; performing multi-band adaptive enhancement processing on the tree ring sample image to obtain an enhanced tree ring image; identifying tree ring boundaries based on the enhanced tree ring image using a multi-strategy boundary detection method; determining a dynamic measurement path configuration according to the tree ring boundaries, wherein the dynamic measurement path configuration includes multiple measurement paths; obtaining the coordinates of intersections between the tree ring boundaries and the multiple measurement paths; reconstructing a complete tree ring boundary curve through multi-dimensional curve fitting based on the intersection coordinates; calculating a local tilt angle distribution according to the tree ring boundary curve; correcting the tree ring width using a regionalized adaptive correction model based on the local tilt angle distribution; generating corrected tree ring width data and a quality assessment report, and realizing automated and high-precision identification of tree ring boundaries using the multi-strategy boundary detection method.

Description

Technical Field

[0001] The present invention relates to the research fields of dendrochronology and dendrometry, and in particular to a method for correcting errors in tree ring width measurement, for improving the accuracy of tilted tree ring width measurement. Background Art

[0002] Measuring tree ring width is fundamental data for dendrochronology and dendrometry. Currently, the most common method for tree ring detection in China is growth cone sampling. However, growth cone sampling often results in the drill being off-center, unable to accurately determine the pith location. Consequently, the resulting tree rings are often tilted (with a non-zero tangent angle).

[0003] Prior art, such as CN101922916A, discloses a method for correcting errors in tree ring width measurement. This method sets two upper and lower measurement lines on a scanned tree ring image. The upper and lower coordinates of each tree ring's intersection with the measurement lines are measured, and the average is calculated to obtain the intermediate coordinate. The difference in the horizontal coordinates between adjacent intermediate coordinates is the width of the tilted tree ring to be corrected. The difference in the vertical and horizontal coordinates between the upper and lower endpoints can be used to calculate the tilt angle of the tree ring boundary, or the tangent angle. The average of the tilt angles of the inner and outer boundaries of each tree ring is used as the ring's tilt angle. Correction of the tilted tree ring is performed based on trigonometric principles.

[0004] However, these methods have the following shortcomings: 1) They use only two parallel lines with a fixed spacing as the measurement reference, making them incapable of adapting to samples of varying sizes and morphologies; 2) they require manual marking of the intersection of each growth ring with the measurement line, which is labor-intensive and prone to human error; 3) they lack the ability to automatically identify growth ring boundaries, especially for fuzzy or atypical growth rings; 4) they simply average the inclination angles of the two boundaries, ignoring the complexity of growth ring morphology; 5) they require high quality raw images and lack image enhancement and preprocessing mechanisms; and 6) they are unable to effectively identify and handle anomalies such as false and missing growth rings. These issues limit their accuracy and efficiency in practical applications. Summary of the Invention

[0005] The present invention aims to provide a method for correcting tree-ring width measurement errors to solve the problems existing in the prior art.

[0006] The tree-ring width measurement error correction method includes:

[0007] Obtaining tree ring sample images;

[0008] performing multi-band adaptive enhancement processing on the tree ring sample image to obtain an enhanced tree ring image;

[0009] Based on the enhanced tree ring image, a multi-strategy boundary detection method is used to identify the tree ring boundary;

[0010] determining a dynamic measurement path configuration according to the growth ring boundary, wherein the dynamic measurement path configuration includes a plurality of measurement paths;

[0011] Obtaining the coordinates of the intersections of the growth ring boundary and the multiple measurement paths;

[0012] Based on the intersection coordinates, reconstructing the complete annual ring boundary curve by multi-dimensional curve fitting;

[0013] Calculating the local tilt angle distribution according to the growth ring boundary curve;

[0014] Based on the local tilt angle distribution, a regionalized adaptive correction model is used to correct the annual ring width;

[0015] Generate corrected annual ring width data and quality assessment report.

[0016] Preferably, the performing multi-band adaptive enhancement processing on the tree ring sample image specifically includes:

[0017] Decomposing the tree ring sample image into three frequency bands: high frequency, medium frequency and low frequency;

[0018] performing detail enhancement and noise suppression processing on the high frequency band;

[0019] performing edge sharpening and texture enhancement processing on the mid-frequency band;

[0020] performing contrast optimization and illumination balancing processing on the low frequency band;

[0021] Adaptively adjust the enhancement parameters of each frequency band based on the tree species characteristic parameters;

[0022] The three processed frequency bands are fused to obtain the enhanced tree ring image.

[0023] Preferably, the multi-strategy boundary detection method is used to identify the growth ring boundary, specifically including:

[0024] Perform edge detection based on gradient changes to obtain a first boundary detection result;

[0025] Performing region segmentation based on texture features to obtain a second boundary detection result;

[0026] Performing periodicity detection based on morphological features to obtain a third boundary detection result;

[0027] Performing weighted fusion on the first boundary detection result, the second boundary detection result, and the third boundary detection result to obtain a fused boundary;

[0028] Continuity optimization and anomaly detection are performed on the fusion boundary to obtain the growth ring boundary.

[0029] Preferably, the continuity optimization and anomaly detection of the fusion boundary specifically includes:

[0030] detecting breakpoints and isolated points in the fused boundary;

[0031] Based on boundary continuity constraints, repairing the breakpoints and removing the isolated points;

[0032] Identify abnormal growth rings within the fusion boundary, including false rings, missing rings, and traumatic rings;

[0033] Marking or correcting the abnormal growth rings according to the growth law of the growth rings;

[0034] Generate a boundary quality score to identify areas requiring manual intervention.

[0035] Preferably, determining the dynamic measurement path configuration specifically includes:

[0036] Analyze the morphological characteristics of the growth ring boundaries, including overall shape, eccentricity, and complexity;

[0037] Based on the described morphological characteristics and sample size, the optimal number of measurement paths was calculated, ranging from 3 to 9;

[0038] For circular growth rings, configure a radial measurement path layout;

[0039] For elliptical growth rings, the density of measurement paths is increased along the major and minor axes;

[0040] For irregular growth rings, the measurement path distribution is dynamically adjusted according to the morphological complexity;

[0041] Evaluate the coverage uniformity of the measurement path to ensure critical areas are adequately measured.

[0042] Preferably, the reconstructing the complete annual ring boundary curve by multi-dimensional curve fitting specifically includes:

[0043] Determining a preliminary outline of the growth ring boundary based on the intersection coordinates;

[0044] According to the complexity of the tree ring boundary, the curve fitting method is adaptively selected, including:

[0045] When the boundary is relatively regular, piecewise polynomial fitting is used;

[0046] When the boundary is moderately complex, spline curve fitting is used;

[0047] When the boundary is highly complex, nonparametric curve fitting is used;

[0048] Optimize the fitting results to ensure that the boundaries are smooth, continuous, and conform to biological growth laws;

[0049] Generate a complete mathematical model of the tree ring boundary and support the precise coordinate calculation of any point.

[0050] Preferably, the calculating the local tilt angle distribution specifically includes:

[0051] uniformly sampling a plurality of points on the growth ring boundary curve;

[0052] Calculate the tangent direction at each sampling point to obtain the local tangent vector;

[0053] Calculating the local tilt angle of each sampling point based on the local tangent vector;

[0054] Establish a local tilt angle distribution map to describe the tilt characteristics of the entire annual ring boundary;

[0055] Identify the area with sudden change of tilt angle and divide the tilt angle area to provide a basis for the selection of correction model.

[0056] Preferably, the method of correcting the growth ring width using a regionalized adaptive correction model specifically includes:

[0057] Dividing the annual ring boundary into a plurality of correction regions according to the local tilt angle distribution;

[0058] For areas with small variations in tilt angle, a uniform tilt model is applied;

[0059] For areas where the tilt angle changes gradually, a gradient tilt model is applied;

[0060] For areas with sudden changes in tilt angle, a segmented tilt model is applied;

[0061] Calculate the correction parameters for each region separately and perform precise correction;

[0062] Calculate the true width between adjacent growth ring boundaries and generate corrected growth ring width data.

[0063] Preferably, generating the corrected annual ring width data and quality assessment report specifically includes:

[0064] Evaluate various sources of uncertainty in the measurement process, including uncertainty caused by image resolution, boundary identification, tilt angle measurement, and correction model selection;

[0065] Calculate the combined uncertainty of each growth ring width;

[0066] The quality of the annual ring width data is graded according to the uncertainty.

[0067] Generates standard format output containing raw data, calibration data, calibration parameters, and quality ratings;

[0068] Provides data visualization results to intuitively display measurement and correction effects.

[0069] Preferably, before the multi-band adaptive enhancement processing, the method further includes:

[0070] Obtain tree species characteristic information;

[0071] extracting corresponding image processing parameters from a pre-established tree species characteristic database according to the tree species characteristic information;

[0072] The multi-band adaptive enhancement processing adopts the image processing parameters for parameter configuration;

[0073] The tree species characteristic information includes tree species type, annual ring density, earlywood and latewood contrast, and boundary clarity.

[0074] The present invention achieves high-precision automated correction of annual ring width measurement by introducing technologies such as multi-band adaptive image enhancement, multi-strategy boundary detection, dynamic measurement path configuration, multi-dimensional curve fitting reconstruction, and regionalized adaptive correction.

[0075] The present invention has the following beneficial effects:

[0076] 1. Through multi-band adaptive enhancement processing, the quality of tree ring images is significantly improved, making the boundary features clearer and laying the foundation for subsequent analysis;

[0077] 2. The multi-strategy boundary detection method is used to achieve automatic and high-precision identification of growth ring boundaries, reducing manual intervention and improving work efficiency;

[0078] 3. Introducing dynamic measurement path configuration, adaptively adjusting the measurement strategy according to the morphological characteristics of the tree rings to adapt to tree ring samples of different morphologies;

[0079] 4. Reconstruct the complete annual ring boundary through multi-dimensional curve fitting, fully capture the morphological characteristics of the annual ring, and improve the integrity of the measurement;

[0080] 5. Based on the local tilt angle distribution, a regional adaptive correction model is adopted to achieve refined correction, significantly improving the accuracy of annual ring width measurement;

[0081] 6. A complete quality assessment system has been established to provide reliable uncertainty assessment for measurement results and enhance data credibility.

[0082] In summary, the tree-ring width measurement error correction method provided by the present invention improves measurement accuracy by 40% to 60% compared with the existing technology, significantly improves the degree of automation, and is more adaptable, providing a more reliable data basis for dendrochronology and paleoclimate research. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 Flowchart of the tree ring width measurement error correction method of the present invention. DETAILED DESCRIPTION

[0084] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0085] Please refer to Figure 1 , Figure 1 Flowchart of the tree ring width measurement error correction method provided by the present invention. Figure 1 As shown, the method includes the following steps:

[0086] Step S101: Acquire a tree ring sample image;

[0087] Step S102: performing multi-band adaptive enhancement processing on the tree ring sample image to obtain an enhanced tree ring image;

[0088] Step S103: Based on the enhanced tree ring image, a multi-strategy boundary detection method is used to identify the tree ring boundary;

[0089] Step S104: determining a dynamic measurement path configuration according to the growth ring boundary, wherein the dynamic measurement path configuration includes multiple measurement paths;

[0090] Step S105: obtaining the coordinates of the intersections of the growth ring boundary and the multiple measurement paths;

[0091] Step S106: reconstructing a complete annual ring boundary curve through multi-dimensional curve fitting based on the intersection coordinates;

[0092] Step S107: calculating the local tilt angle distribution according to the annual ring boundary curve;

[0093] Step S108: Based on the local tilt angle distribution, a regionalized adaptive correction model is used to correct the annual ring width;

[0094] Step S109: Generate corrected annual ring width data and a quality assessment report.

[0095] In one embodiment of the present invention, in step S101, the tree ring sample image can be acquired using a variety of methods, including but not limited to a high-resolution scanner, a microscope camera, or a digital camera. Preferably, a scanner with a resolution of at least 1200 dpi is used to acquire the tree ring sample image to ensure clear and discernible tree ring boundary details. When acquiring the image, standard color chart and scale information can also be collected for subsequent image calibration to ensure measurement accuracy.

[0096] In step S102, multi-band adaptive enhancement processing is performed on the tree ring sample image, specifically including:

[0097] The tree ring sample image is decomposed into three frequency bands: high frequency, medium frequency, and low frequency. In a preferred embodiment of the present invention, wavelet transform is used for image decomposition. Specifically, Daubechies wavelet (such as db4 or db6) can be used for multi-scale decomposition to decompose the image into 3 to 5 scale levels. The mathematical expression of wavelet transform is:

[0098] ,

[0099] in: is the wavelet coefficient, is the scale parameter, is the translation parameter, is the input signal (image in this case), is the conjugate of the wavelet basis function. The high frequency band is subjected to detail enhancement and noise suppression processing. The high frequency band contains detailed information of the annual ring boundary, but also contains noise. The present invention uses an adaptive threshold method to perform processing, enhancing boundary details while suppressing noise. Preferably, the following formula is used to adjust the high frequency coefficient:

[0100] ,

[0101] in: is the adjusted high frequency coefficient, is the original high frequency coefficient, is the pixel coordinate, is the enhancement factor (usually between 0.5 and 2.0), is the noise threshold (the empirical value is the 10-15 percentile of the absolute value of the high-frequency coefficient), is the maximum threshold (usually the 95th percentile of the absolute value of the high-frequency coefficient), It is an indicator function, which takes the value of 1 when the condition is met and 0 otherwise.

[0102] The mid-frequency band is subjected to edge sharpening and texture enhancement processing. The mid-frequency band contains the main structural information of the annual rings. Enhancing this frequency band helps to highlight the boundaries of the annual rings. The present invention uses a directional enhancement filter for processing, and its mathematical expression is:

[0103] ,

[0104] in: is the adjusted intermediate frequency coefficient, is the original intermediate frequency coefficient, is the enhancement factor (usually 0.3-0.8), is the directional weight, which is related to the consistency of the local gradient direction and is calculated as:

[0105] ,

[0106] in: for point The neighborhood (usually taken as or ), represents the gradient operator, is a small constant (such as 0.0001) to prevent the denominator from being zero. The low-frequency band is subjected to contrast optimization and illumination equalization. The low-frequency band mainly contains the background information and overall brightness distribution of the image. The present invention adopts a method combining histogram equalization and gamma correction, and its processing formula is:

[0107] ,

[0108] in: is the adjusted low-frequency coefficient, is the original low-frequency coefficient, represents the histogram equalization operation, is the gamma correction coefficient (usually between 0.8-1.2, adjusted according to image characteristics), It is a scaling factor used to adjust the overall brightness level. It is usually set so that the average brightness of the processed image is close to that of the original image.

[0109] Based on the tree species characteristic parameters, the enhancement parameters of each frequency band are adaptively adjusted. This invention has established a database containing the characteristic parameters of the annual rings of common tree species, such as the typical parameters of conifers (pine, spruce, etc.) and broad-leaved trees (oak, ash, etc.). For conifers, since the boundary between earlywood and latewood is usually clear, the mid-frequency enhancement coefficient can be appropriately increased. (For example, take 0.6-0.8); for broad-leaved trees, since the boundaries are relatively fuzzy, it is necessary to strengthen the high-frequency details, and the high-frequency enhancement factor can be appropriately increased. (For example, 1.5-2.0 is selected). The three processed frequency bands are fused to obtain the enhanced annual ring image. Preferably, the following formula is used for frequency band fusion:

[0110] ,

[0111] in: is the enhanced image, indicating that after processing at position Pixel value at ; IDWT ) represents the inverse discrete wavelet transform, which is the inverse operation of the wavelet transform and is used to reconstruct the coefficients of the transform domain into an image in the spatial domain; are the low-frequency, mid-frequency and high-frequency coefficients after processing respectively.

[0112] In a preferred embodiment of the present invention, before step S102, it also includes: obtaining tree species characteristic information; extracting corresponding image processing parameters from a pre-established tree species characteristic database based on the tree species characteristic information; the multi-band adaptive enhancement processing uses the image processing parameters for parameter configuration; wherein the tree species characteristic information includes tree species type, annual ring density, earlywood and latewood contrast, and boundary clarity.

[0113] Tree species characteristic information can be obtained through various methods, such as user input or automatic identification. Automatic identification can be achieved using machine learning methods based on image texture features and tree ring morphology. In one embodiment of the present invention, a texture-based random forest classifier is used for tree species identification, achieving an accuracy rate of 85% to 90%. The tree species characteristic database contains image processing parameters for common tree species, such as typical parameter settings for conifers (pine, spruce, etc.) and broadleaf trees (oak, ash, etc.).

[0114] In step S103, a multi-strategy boundary detection method is used to identify growth ring boundaries, specifically including:

[0115] Edge detection is performed based on gradient changes to obtain a first boundary detection result. In a preferred embodiment of the present invention, a method combining the Sobel operator and the Canny edge detector is used. First, the image gradient is calculated:

[0116] ,

[0117] ,

[0118] in: and are the horizontal and vertical gradients, respectively, indicating that the image is at position The rate of change in the horizontal and vertical directions; and is the Sobel operator kernel, which is horizontal and vertical respectively. Convolution kernel; is the input image; is the pixel coordinate; Indicates Double summation of convolution operations within a neighborhood.

[0119] Then, calculate the gradient magnitude and direction:

[0120] ,

[0121] ,

[0122] in: is the gradient amplitude, which indicates the strength of the gradient; is the gradient direction, which indicates the angle of the gradient; is the inverse tangent function, used to calculate angles.

[0123] Next, non-maximum suppression and double thresholding are applied to obtain edge pixels. For tree ring images, the low threshold is usually set to the 20th-30th percentile of the gradient magnitude histogram, and the high threshold is set to the 60th-70th percentile.

[0124] Based on the texture features, the region segmentation is performed to obtain the second boundary detection result. The present invention adopts a method combining the local binary pattern (LBP) texture features and the watershed algorithm. First, the LBP features are calculated:

[0125] ,

[0126] in: Center point The LBP value represents the local texture feature; is the number of sampling points, indicating the radius The number of sampling points evenly distributed on the circumference is usually 8; The sampling radius represents the radius of the sampling circle, which is usually 1 or 2; is the gray value of the center point; For the Gray value of sampling points; is a step function, when 1 if yes, 0 otherwise represents the sum of all sampling points; Indicates the weight used to convert a binary result to a decimal value.

[0127] Then, the watershed algorithm is applied based on the LBP feature map to obtain the region segmentation result. Finally, the region boundary is extracted as the second boundary detection result.

[0128] Based on the morphological features, periodic detection is performed to obtain the third boundary detection result. An important feature of the annual ring is its periodic distribution. The present invention uses this feature for detection. The specific method is: first, radial cumulative projection is performed on the image:

[0129] ,

[0130] in: Angle The projection value of a direction represents the cumulative sum of pixels along that direction; is the radial distance, which represents the distance from the origin (usually the pith position) to the pixel point; and are the minimum and maximum radial distances, respectively; is the image value in the polar coordinate system; represents the summation operation within the radial range. Then, the projection curve Fourier analysis is performed to detect the periodic components. Finally, based on the detected periodicity, the possible location of the annual ring is predicted to generate the third boundary detection result.

[0131] The first boundary detection result, the second boundary detection result and the third boundary detection result are weightedly fused to obtain a fused boundary. The present invention adopts a weighted fusion method based on local consistency, and the fusion formula is:

[0132] ,

[0133] in: is the fusion boundary, indicating that The fusion result at For the Boundary detection results, Corresponding to the results of the three detection methods; is the local weight, indicating the Methods in position Reliability of the department; Represents the weighted sum of the results of the three methods. The calculation formula is:

[0134] ,

[0135] in: is the global confidence coefficient, which is based on the overall reliability of each method and is usually set to 0.3-0.5; is the local consistency influencing factor, which controls the importance of local consistency and is usually set to 0.5-1.0; is a local consistency measure, indicating the The result of this method is at point The degree of consistency with other methods is in the range of [0,1].

[0136] The fusion boundary is subjected to continuity optimization and anomaly detection to obtain the growth ring boundary. Continuity optimization includes breakpoint connection and isolated point removal. In a preferred embodiment of the present invention, the following method is used for breakpoint connection: for the distance less than a preset threshold, (usually set to 10% of the average annual ring width) breakpoints, connected using Bezier curves to maintain the smoothness of the boundaries. For isolated points, if the distance to the nearest boundary is greater than the threshold (usually set to 50% of the average annual ring width), it will be removed.

[0137] In a preferred embodiment of the present invention, the continuity optimization and anomaly detection of the fusion boundary in step S103 specifically include:

[0138] Detect breakpoints and isolated points in the fused boundary. Breakpoints refer to gaps or discontinuous areas in the boundary, while isolated points refer to isolated pixels or small areas away from the main boundary. This detection is performed using a connected component analysis method. The specific steps are: first, binarize the fused boundary image; then, label the connected components; and finally, analyze the size and positional relationships of each connected component to identify breakpoints and isolated points.

[0139] Based on the boundary continuity constraint, the breakpoints are repaired and the isolated points are removed. For the breakpoints, the present invention adopts a curve fitting method to repair them. Preferably, a cubic Bezier curve is used to connect the two ends of the breakpoint to ensure the continuity and smoothness of the curve. The parameterized expression of the Bezier curve is:

[0140] ,

[0141] in: is a point on the curve, representing the parameter The corresponding curve point coordinates; and are the known boundary points at both ends of the breakpoint, indicating the starting and ending points of the curve; and is the control point, determined by maintaining the continuity of the tangent direction; is a parameter with a value range of [0,1]; 、 、 and is the Bessel basis function, which controls the shape of the curve. For isolated points, if their area is less than the threshold (usually set to 5% of the average annual ring area) and the distance to the nearest boundary is greater than the threshold (usually set to 50% of the average annual ring width), it will be removed.

[0142] Identify abnormal growth rings within the fusion boundary, including false growth rings, missing growth rings, and traumatic growth rings. Identify abnormal growth rings based on morphological characteristics and growth patterns. Specific methods include:

[0143] (1) False ring identification: False rings are usually narrow and discontinuous. The present invention adopts the following judgment criteria: if the annual ring width is less than 30% of the average width of a normal annual ring and the circumferential continuity is less than 70%, it is marked as a potential false ring.

[0144] (2) Missing ring identification: By analyzing the spacing between adjacent annual rings, if the spacing at a certain point exceeds 1.8 times the normal spacing, there may be a missing ring.

[0145] (3) Traumatic wheel identification: Traumatic wheels usually manifest as local distortion. The present invention identifies traumatic wheels by analyzing the local curvature anomaly. If the local curvature exceeds three times the average curvature and the affected area is limited, it is likely a traumatic wheel.

[0146] According to the growth law of annual rings, the abnormal annual rings are marked or corrected. For the identified abnormal annual rings, the present invention adopts different processing strategies:

[0147] (1) False wheel: marked as a false wheel and excluded from subsequent measurements;

[0148] (2) Missing ring: Based on the width characteristics of the adjacent annual rings, the possible position and width of the missing ring are estimated and marked as estimated values ​​in the results;

[0149] (3) Traumatic wheel: retain the traumatic wheel, but adjust the measurement path in the affected area to avoid the distorted part.

[0150] Generate a boundary quality score to identify areas requiring human intervention. This paper establishes a boundary quality scoring system based on metrics such as boundary continuity, morphological regularity, and consistency with expected patterns. The quality score is calculated on a 0-100 scale, with areas with scores below 60 being marked as requiring human intervention. In practice, the system generates a boundary quality distribution map that visually displays the quality status of each area, helping researchers quickly locate problem areas.

[0151] In step S104, the dynamic measurement path configuration is determined, specifically including:

[0152] Analyze the morphological characteristics of the growth ring boundary, including overall shape, eccentricity and complexity. In a preferred embodiment of the present invention, the morphological characteristics analysis adopts the following method:

[0153] (1) Overall shape: The overall shape is determined by fitting an ellipse and calculating the direction of the main axis. The center of the fitted ellipse , major axis a, minor axis b and rotation angle θ are calculated by the least squares method.

[0154] (2) Eccentricity: Calculate the eccentricity e using the parameters of the fitted ellipse:

[0155] ,

[0156] Among them: e is the eccentricity, which indicates the flatness of the ellipse, and its value range is [0,1]; a is the length of the major axis; b is the length of the minor axis; Represents square root operation. The closer the eccentricity is to 1, the flatter the growth rings are; the closer it is to 0, the rounder the growth rings are.

[0157] (3) Complexity: quantified by calculating the fractal dimension of the boundary or the ratio of the boundary length to the area. The complexity index C can be expressed as:

[0158] ,

[0159] Where: C is the complexity index, indicating the complexity of the boundary; P is the perimeter of the boundary; A is the area of ​​the enclosed region; π is the circumference of the circle, approximately equal to 3.14159; and 4π is the normalization factor, such that C = 1 for a circle. A larger value of C indicates a more complex boundary.

[0160] According to the morphological characteristics and sample size, the optimal number of measurement paths is calculated, ranging from 3 to 9. The present invention uses the following adaptive formula to determine the number of measurement paths N:

[0161] ,

[0162] in: The number of measurement paths indicates the number of measurement paths that need to be set; Represents a round-up function to ensure an integer result; is the normalized value of the sample size, calculated as the sample diameter divided by the reference value of 10 mm; is the eccentricity; is the normalized value of the complexity index; and γ are weight coefficients that control the influence of each factor. The empirical values ​​are 0.3, 2.0, and 1.5 respectively. Thus, the larger the sample, the flatter the shape, and the more complex the boundary, the more measurement paths are required. For circular growth rings, a radial measurement path layout is configured. For eccentricity The present invention adopts a radial layout with uniform angle distribution. If the number of measurement paths is , then The angle of the path for:

[0163] ,

[0164] in: For the The angle of a measurement path represents the angle between the path and the reference direction; is the initial angle, usually set to 0° (i.e. horizontal direction); is the path index, ranging from 0 to The angle of a complete circle; Indicates the angular interval between adjacent paths.

[0165] For elliptical growth rings, the density of measurement paths is increased along the major and minor axes. For elliptical growth rings, the present invention uses a non-uniform angular distribution to increase the path density in the major and minor axis directions. Specifically, the angular distribution is adjusted so that the angular density is inversely proportional to the ellipse radius, ensuring a denser measurement path in areas with greater curvature (usually along the minor axis).

[0166] For irregular growth rings, the measurement path distribution is dynamically adjusted according to the morphological complexity. or complexity index The present invention adopts an adaptive distribution strategy based on curvature for irregular growth rings. First, the curvature distribution of the boundary is calculated; then, the density of measurement paths is increased in areas with larger curvature. The specific algorithm is: divide the boundary into equal parts. part( Much greater than , usually take ), calculate the average curvature of each segment , then determine based on the cumulative distribution of curvature values The positions of the measurement paths ensure that the path distribution is adapted to the curvature distribution.

[0167] Evaluate the coverage uniformity of the measurement path to ensure that key areas are fully measured. This invention introduces the coverage uniformity index U, which is calculated as follows:

[0168] ,

[0169] in: is the coverage uniformity index, which indicates the uniformity of the measurement path distribution, and its value range is [0,1]; is the standard deviation of the distance between adjacent measurement paths, indicating the degree of dispersion of the distance distribution; is the average distance; is the coefficient of variation, which indicates the relative degree of dispersion. The closer the value is to 1, the more uniform the coverage. In practice, if , the measurement path distribution needs to be adjusted to improve uniformity. In addition, the present invention also checks whether there is sufficient measurement path coverage in key areas (such as high-curvature areas). If the measurement path density in a certain key area is lower than the threshold, the system will automatically add additional measurement paths.

[0170] In step S105, the intersection coordinates of the annual ring boundary and the multiple measurement paths are obtained. This step realizes the accurate intersection positioning of the annual ring boundary and the measurement paths. In a preferred embodiment of the present invention, the acquisition of the intersection coordinates adopts sub-pixel-level precise positioning technology, including gray-scale interpolation method, gradient maximum positioning, and edge fitting method.

[0171] The gray-scale interpolation method constructs a high-precision gray-scale curve near the intersection, and then determines the true intersection position at a finer scale. The gradient maximum positioning improves the intersection positioning accuracy by finding local gradient peak points. The edge fitting method performs local curve fitting on the edge near the intersection to further improve the positioning accuracy. By comprehensively using these technologies, the present invention can achieve sub-pixel-level intersection positioning accuracy, significantly improving the accuracy of subsequent measurements.

[0172] In step S106, based on the intersection coordinates, the complete annual ring boundary curve is reconstructed by multi-dimensional curve fitting, specifically including:

[0173] Based on the intersection coordinates, the preliminary contour of the annual ring boundary is determined. In a preferred embodiment of the present invention, the determination of the preliminary contour adopts a method combining intersection connection method and morphological feature analysis. First, the intersections on the same annual ring boundary are connected in sequence to form a preliminary contour; then, it is corrected by combining morphological features (such as the expected annual ring shape) to remove obviously unreasonable parts.

[0174] According to the complexity of the annual ring boundary, the curve fitting method is adaptively selected. The present invention provides three curve fitting methods with different complexities and automatically selects the most suitable method according to the complexity of the annual ring boundary:

[0175] When the boundary is relatively regular, piecewise polynomial fitting is adopted. For a regular boundary with a complexity index C < 1.2, piecewise polynomial fitting can ensure both fitting accuracy and computational efficiency. In a preferred embodiment of the present invention, a cubic polynomial is used for piecewise fitting, and each segment contains 3 - 5 intersections. First-order derivative continuity is maintained between adjacent segments to ensure the smoothness of the overall curve.

[0176] When the boundary is moderately complex, spline curve fitting is adopted. For a moderately complex boundary with a complexity index 1.2 < C < 1.5, the spline curve can better adapt to local changes. The present invention adopts a cubic B-spline curve, and its parametric expression is:

[0177] When the boundary is moderately complex, spline curve fitting is used. For medium-complex boundaries, spline curves can better adapt to local changes. The present invention adopts a cubic B-spline curve, whose parameterized expression is:

[0178] ,

[0179] in: is a point on the curve, representing the parameter The corresponding curve point coordinates; is the cubic B-spline basis function, which controls the shape of the curve; is the control point, which is determined by the intersection coordinates; is the number of control points; and is the parameter range; Represents the weighted sum of all control points. When the boundary is highly complex, non-parametric curve fitting is used. For complexity index For highly complex boundaries, non-parametric curve fitting can better capture the local details of the boundary. The present invention adopts a non-parametric fitting method based on radial basis function (RBF), and its expression is:

[0180] ,

[0181] in: is the fitting function, indicating the point The function value at ; are the known intersection coordinates; is the radial basis function, usually a Gaussian function or multiple quadratic functions ,in For distance, and is the shape parameter; is the weight coefficient, which is determined by solving the linear equations; is a low-order polynomial term, usually a first-order or second-order polynomial; is the number of intersection points; represents the weighted sum of all intersection points; Indicates a point To the intersection The Euclidean distance.

[0182] The fitting results are optimized to ensure smooth and continuous boundaries and conform to biological growth laws. The present invention uses the following methods to optimize the fitting results:

[0183] (1) Smoothness optimization: By adding a smoothness penalty term, fluctuations caused by overfitting are avoided. The optimization objective function is:

[0184] ,

[0185] in: is the optimization objective function, which represents the overall optimization goal; is the data fitting error, which represents the deviation between the fitting curve and the actual intersection point, usually using the mean square error; As a measure of smoothness, the square integral of the second derivative of the curve is usually used; The smoothing coefficient controls the importance of smoothness and is adaptively adjusted according to data quality. It is usually in the range of 0.01-0.1.

[0186] (2) Biological rationality constraints: Based on the growth patterns of trees, morphological constraints are added. For example, tree rings are usually concentric circles or ellipses, and adjacent tree rings usually maintain similar morphological characteristics. During the fitting process, if these patterns are violated, the system will automatically adjust the fitting parameters to ensure the biological rationality of the results.

[0187] A complete mathematical model of the tree ring boundary is generated, supporting precise coordinate calculations for any point. The final mathematical model of the tree ring boundary uses a parametric representation, enabling calculations at any point under any parameter value. For piecewise polynomials and spline curves, their parametric expressions are directly used; for non-parametric curves, calculations at any point are achieved through interpolation. Furthermore, the present invention provides a discrete representation of the boundary to facilitate subsequent analysis and visualization.

[0188] In step S107, the local tilt angle distribution is calculated according to the growth ring boundary curve, which specifically includes:

[0189] Multiple points are uniformly sampled on the growth ring boundary curve. To ensure the uniformity of sampling, the present invention adopts an equal arc length sampling strategy rather than a simple equal parameter sampling. The specific method is: first calculate the total arc length of the boundary curve , then determine the number of sampling points (usually 50-100), and finally uniformly sample in the arc length parameter space to obtain a sampling point sequence. , No. The arc length parameter of each sampling point is:

[0190] ,

[0191] in: For the The arc length parameter of each sampling point represents the distance along the curve from the starting point; is the total arc length of the boundary curve; is the number of sampling points; The sampling point index ranges from 0 to Represents the arc length interval between adjacent sampling points. Calculate the tangent direction at each sampling point and get the local tangent vector. For parameterized curves , the tangent vector can be calculated by taking the derivative:

[0192] ,

[0193] in: is the tangent vector, indicating that the curve is at the parameter The tangent direction at ; and The curve parameter equations are and Parameters The derivative of and The rate of change of direction. For discrete curves, the tangent direction can be approximately calculated by the difference between adjacent points.

[0194] Based on the local tangent vector, the local tilt angle of each sampling point is calculated. Local tilt angle It is defined as the angle between the tangent direction and the reference direction (usually the horizontal direction), and the calculation formula is:

[0195] ,

[0196] in: is the local tilt angle, which represents the angle between the tangent direction and the horizontal direction; and The tangent vectors are and Quantity; To ensure the continuity of the angle, the present invention uses the four-quadrant inverse tangent function arctan2 ,get Angle value within the range.

[0197] A local tilt angle distribution diagram is created to describe the tilt characteristics of the entire annual ring boundary. This local tilt angle distribution diagram is a functional relationship between sampling position and tilt angle, and can be used to visually display the tilt of different parts of the annual ring boundary. In a preferred embodiment of the present invention, a polar coordinate display is used, where the angular coordinate represents the circumferential position of the sampling point, the radial coordinate represents the magnitude of the local tilt angle, and the color code represents the rate of change of the tilt angle.

[0198] Identify the tilt angle mutation area and divide the tilt angle area to provide a basis for the correction model selection. The present invention identifies the tilt angle mutation area by analyzing the change rate of the local tilt angle. The specific method is: calculate the change rate of the tilt angle between adjacent sampling points , if the rate of change exceeds the threshold (usually set to 5° / mm), it is marked as a mutation area. is the tilt angle difference between adjacent sampling points, in degrees (°); is the physical arc length between adjacent sampling points, in millimeters (mm); is the rate of change threshold, measured in degrees per millimeter (° / mm). Based on the distribution of mutation regions, the annual ring boundary is divided into multiple tilt angle zones, each with a relatively gentle tilt angle change. This regional division provides an important basis for the subsequent selection of adaptive correction models.

[0199] In step S108, based on the local tilt angle distribution, a regionalized adaptive correction model is used to correct the annual ring width, specifically including: dividing the annual ring boundary into multiple correction areas according to the local tilt angle distribution. The regional division is based on the changing characteristics of the tilt angle to ensure that the tilt angle changes in each area are relatively consistent. The present invention adopts an adaptive segmentation algorithm based on the rate of change, and the specific steps are: first, calculate the first-order derivative (rate of change) of the tilt angle; then, identify the point where the rate of change exceeds the threshold as the segmentation point; finally, obtain multiple correction areas. Preferably, according to the complexity of the annual ring morphology, the number of regions is usually between 3 and 8. For areas with small changes in tilt angle, a uniform tilt model is applied. For the standard deviation of the tilt angle The region can be considered as uniformly tilted, using a simple cosine correction model:

[0200] ,

[0201] in: is the corrected true width, indicating the actual growth width of the annual ring; is the measured tilt width, which means the width obtained by direct measurement; is the average tilt angle in the region; is the cosine correction factor, projecting the tilt measurement onto a plane perpendicular to the growth direction.

[0202] For areas where the tilt angle changes gradually, the gradient tilt model is applied. In the region, the present invention adopts a gradient tilt model:

[0203] ,

[0204] in: is the true width after correction; is the arc length parameter, indicating the position along the boundary; and is the starting and ending position of the area; is the function of the inclination angle with respect to the arc length, indicating the position The tilt angle at For location Correction factor at ; Indicates the integral operation within the region, corresponding to continuous correction. Usually a linear model is used ,parameter and Determined by linear regression, where is the intercept, indicating the initial tilt angle, is the slope, which represents the rate of change of the inclination angle.

[0205] For areas with sudden changes in tilt angle, a segmented tilt model is applied. For regions with a uniform tilt angle, the present invention uses a segmented tilt model. Specifically, the region is further subdivided into multiple subregions, each of which is then subjected to a uniform tilt model, and the results are then combined. The subregions are divided based on the local extreme values ​​of the tilt angle, ensuring that the tilt angle within each subregion varies smoothly.

[0206] Calculate the correction parameters for each area and perform accurate correction. According to the selected correction model, calculate the corresponding correction parameters for each area. For the uniform tilt model, the correction parameter is the average tilt angle ; For the gradual tilt model, the correction parameter is the linear coefficient and For the segmented tilt model, the correction parameters are the average tilt angles of each sub-region. These parameters are determined based on the local tilt angle distribution data using the least squares method or other optimization methods.

[0207] Calculate the true width between adjacent growth ring boundaries and generate corrected growth ring width data. The present invention obtains the true width by calculating the shortest distance between adjacent growth ring boundaries. The specific method is: for each point on the growth ring boundary, calculate its shortest distance to the adjacent growth ring boundary, and then correct it according to the local inclination angle. The final growth ring width is taken as the weighted average of the correction results of multiple measurement points, and the weight is related to the quality score of the measurement point. This method takes into account the complex morphology of the growth ring boundary and can more accurately reflect the actual growth situation.

[0208] In step S109, the corrected annual ring width data and quality assessment report are generated, specifically including:

[0209] Evaluate various sources of uncertainty in the measurement process, including uncertainty caused by image resolution, boundary identification, tilt angle measurement, and correction model selection. This paper adopts an uncertainty evaluation method based on error propagation theory, specifically considering the following sources of uncertainty:

[0210] Evaluate various sources of uncertainty in the measurement process, including uncertainty caused by image resolution, boundary identification, tilt angle measurement, and correction model selection. This paper adopts an uncertainty evaluation method based on error propagation theory, specifically considering the following sources of uncertainty:

[0211] (1) Image resolution uncertainty : Related to the spatial resolution of the image, usually expressed as half the pixel size. (pixels / mm) of the image, mm, where is the number of pixels per millimeter, and 0.5 represents an uncertainty of half a pixel.

[0212] (2) Boundary identification uncertainty :It is related to the accuracy of the boundary detection algorithm and the image quality. The present invention has been determined through repeated tests and is generally 0.5-1.5 pixels, which is converted to an actual length of mm, where represents the pixel uncertainty range of the boundary positioning, is the number of pixels per millimeter.

[0213] (3) Tilt angle measurement uncertainty : Related to the accuracy of the tangent direction calculation. Through error propagation analysis, the standard uncertainty of the tilt angle measurement is usually 1-3°, which represents the angular error range of the tilt angle measurement.

[0214] (4) Uncertainty in calibration model selection :Related to the applicability of the calibration model. This paper evaluates the difference in calibration results of different models, which is usually 3% to 8% of the calibration amount, indicating the relative error range caused by model selection. The comprehensive uncertainty of each annual ring width is calculated. Based on the error propagation theory, the comprehensive standard uncertainty The calculation formula is:

[0215] ,

[0216] in: is the comprehensive standard uncertainty, in millimeters (mm) is the uncertainty caused by image resolution, in millimeters (mm) is the boundary identification uncertainty, in millimeters (mm) is the measured tilt width in millimeters (mm); is the tilt angle in degrees (°) The uncertainty of the tilt angle measurement is expressed in radians, which is determined by the angular uncertainty Converted to: , in radians ( ) The relative uncertainty selected for the calibration model is a dimensionless quantity, usually taken as 0.03-0.08 ( ) is a sine function; Indicates square root operation, expanded uncertainty Usually take , corresponding to a confidence level of approximately 95%, where 2 is the coverage factor.

[0217] The quality of the annual ring width data is graded according to the uncertainty. The present invention establishes a quality grading standard based on relative uncertainty (the ratio of uncertainty to the measured value):

[0218] Level A: Relative uncertainty <2%, high-precision data, can be used for detailed research;

[0219] Level B: Relative uncertainty 2% to 5%, medium to high precision data, suitable for most research needs;

[0220] Level C: Relative uncertainty 5% to 10%, medium-precision data, can be used for trend analysis;

[0221] Level D: Relative uncertainty > 10%, low-precision data, for reference only, not recommended for quantitative research.

[0222] Generates a standard format output containing raw data, correction data, correction parameters and quality ratings. The present invention supports multiple data output formats, including:

[0223] (1) Standard formats for tree-ring research: such as Tucson and Belfast formats, which are easy to integrate with existing tree-ring research software;

[0224] (2) Common table formats: such as CSV, Excel, etc., to facilitate data sharing and further analysis;

[0225] (3) Structured data formats: such as JSON, XML, etc., which facilitate programmatic processing and data exchange.

[0226] The output content includes complete information such as original measurement data, corrected data, correction parameters, quality ratings, abnormal annual ring records, etc., ensuring the traceability and repeatability of the data.

[0227] Provide data visualization results to intuitively display measurement and correction effects. The present invention provides a variety of visualization functions, including:

[0228] (1) Visualization of tree ring boundaries: Displays the identified tree ring boundaries, measurement paths, and intersection locations;

[0229] (2) Comparison of correction effects: intuitively display the difference in annual ring width before and after correction;

[0230] (3) Quality distribution map: Displays the measurement quality of different areas through color coding;

[0231] (4) Tree ring sequence diagram: displays the corrected tree ring width sequence and its temporal variation trend.

[0232] These visualization results help researchers intuitively understand the measurement process and results, improving research efficiency.

[0233] In a preferred embodiment of the present invention, a complete implementation process is provided:

[0234] A high-resolution scan (1200 dpi) of a pine tree ring sample was obtained. The system automatically identified it as a pine tree and applied multi-band adaptive enhancement using preprocessing parameters specific to conifers. The enhanced image showed clearer growth ring boundaries and improved the contrast between earlywood and latewood by approximately 40%.

[0235] Based on the enhanced images, the system employed a multi-strategy boundary detection approach to identify tree ring boundaries. By fusing the results of gradient detection, texture segmentation, and periodicity analysis, the system successfully identified the boundaries of 38 complete tree rings. The system automatically detected and processed two false rings and one damaged ring, achieving an average boundary quality score of 85 out of 100.

[0236] Based on morphological analysis of the growth ring boundaries, the system determined a configuration of five radial measurement paths. Due to the slightly elliptical shape of the sample (eccentricity e=0.35), the system appropriately increased the density of measurement paths along the major and minor axes to ensure uniform coverage (uniformity index U=0.86).

[0237] The system accurately obtains the coordinates of the intersection of the annual ring boundary and the measurement path, and uses sub-pixel positioning technology with a positioning accuracy of 0.2 pixels (about 0.004mm).

[0238] Based on the obtained intersection coordinates, the system reconstructs the complete growth ring boundary through spline fitting. Due to the medium complexity of the growth ring boundary (C=1.3), the system automatically selects a cubic B-spline curve for fitting, with a fitting error of less than 0.5 pixels.

[0239] Based on the reconstructed growth ring boundary curves, the system calculated the local tilt angle distribution. By uniformly sampling 60 points on each growth ring boundary, the calculated local tilt angle distribution ranged from 12° to 38°, with an average tilt angle of 25°.

[0240] Based on the local tilt angle distribution, the system divides the tree ring boundaries into four correction zones and applies different correction models: a uniform tilt model is used in zone 1 (0°-90°); a gradient tilt model is used in zones 2 (90°-180°) and 3 (180°-270°); and a uniform tilt model is used in zone 4 (270°-360°). The corrected tree ring widths are, on average, approximately 9% smaller than the uncorrected widths.

[0241] The system generates corrected tree-ring width data and a quality assessment report. Comprehensive uncertainty analysis shows that the corrected tree-ring width data are primarily of Grade B (85%) and Grade C (15%) quality, with no Grade D data. The quality assessment report details the contributions of each uncertainty source, providing researchers with an important reference for data reliability.

[0242] Another embodiment of the present invention is directed to the processing of complex and abnormal growth rings:

[0243] When inputting an image (1600 dpi) of an oak tree specimen containing multiple false and wound rings, the system automatically identified it as a broadleaf tree and applied multi-band adaptive enhancement using parameters specific to broadleaf trees. After enhancement, even the fuzzy growth ring boundaries became relatively clear.

[0244] The system successfully identified the main growth ring boundaries using a multi-strategy boundary detection method. Morphological analysis identified three false growth rings and one traumatic growth ring. False growth rings are characterized by a narrow width (only 25% of the normal growth ring width) and low circumferential continuity (approximately 65%). Traumatic growth rings exhibit locally high curvature (3.5 times the average curvature).

[0245] For this complex sample, the system automatically planned seven measurement paths, especially increasing the measurement density near the trauma area, and adjusting the path direction to avoid severely distorted parts.

[0246] The system marks spurious wheels and excludes them from the measurement sequence, and adopts a local adjustment strategy for traumatic wheels, using specific correction parameters in the affected area.

[0247] Due to the complex morphology of the tree rings (complexity index C=1.7), the system automatically selects the non-parametric curve fitting method to reconstruct the tree ring boundaries, and uses biological rationality constraints to ensure that the fitting results conform to the growth laws of trees.

[0248] Tilt angle analysis revealed a wide distribution of tilt angles (8°-45°) for this sample, with distinct regions of sudden change. Based on this, the system divided the growth ring boundaries into six calibration zones, each applying a different calibration model.

[0249] The final quality assessment report showed that the main growth ring data reached Grade B, and the local trauma area was Grade C. The system also provided detailed exception handling records to ensure data traceability.

[0250] The above examples clearly demonstrate the effectiveness of the present invention in various scenarios. Compared to existing technologies, this invention not only significantly improves measurement accuracy (by 40% to 60%) but also significantly enhances automation, reduces manual intervention, and effectively handles complex and unusual situations, providing a more reliable data foundation for dendrochronological research.

[0251] The foregoing is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for correcting tree ring width measurement errors, characterized in that: include: Obtaining tree ring sample images; performing multi-band adaptive enhancement processing on the tree ring sample image to obtain an enhanced tree ring image; Based on the enhanced tree ring image, a multi-strategy boundary detection method is used to identify the tree ring boundary; determining a dynamic measurement path configuration according to the growth ring boundary, wherein the dynamic measurement path configuration includes a plurality of measurement paths; Obtaining the coordinates of the intersections of the growth ring boundary and the multiple measurement paths; Based on the intersection coordinates, reconstructing the complete annual ring boundary curve by multi-dimensional curve fitting; Calculating the local tilt angle distribution according to the growth ring boundary curve; Based on the local tilt angle distribution, a regionalized adaptive correction model is used to correct the annual ring width; Generate corrected annual ring width data and quality assessment report; The determining of the dynamic measurement path configuration specifically includes: Analyze the morphological characteristics of the growth ring boundaries, including overall shape, eccentricity, and complexity; Based on the described morphological characteristics and sample size, the optimal number of measurement paths was calculated, ranging from 3 to 9; For circular growth rings, configure a radial measurement path layout; For elliptical growth rings, the density of measurement paths is increased along the major and minor axes; For irregular growth rings, the measurement path distribution is dynamically adjusted according to the morphological complexity; Evaluate the coverage uniformity of the measurement path to ensure critical areas are adequately measured.

2. The method according to claim 1, characterized in that The performing multi-band adaptive enhancement processing on the tree ring sample image specifically includes: Decomposing the tree ring sample image into three frequency bands: high frequency, medium frequency and low frequency; performing detail enhancement and noise suppression processing on the high frequency band; performing edge sharpening and texture enhancement processing on the mid-frequency band; performing contrast optimization and illumination balancing processing on the low frequency band; Adaptively adjust the enhancement parameters of each frequency band based on the tree species characteristic parameters; The three processed frequency bands are fused to obtain the enhanced tree ring image.

3. The method according to claim 1, characterized in that The multi-strategy boundary detection method is used to identify the growth ring boundary, specifically including: Perform edge detection based on gradient changes to obtain a first boundary detection result; Performing region segmentation based on texture features to obtain a second boundary detection result; Performing periodicity detection based on morphological features to obtain a third boundary detection result; Performing weighted fusion on the first boundary detection result, the second boundary detection result, and the third boundary detection result to obtain a fused boundary; Continuity optimization and anomaly detection are performed on the fusion boundary to obtain the growth ring boundary.

4. The method according to claim 3, characterized in that The continuous optimization and anomaly detection of the fusion boundary specifically includes: detecting breakpoints and isolated points in the fused boundary; Based on boundary continuity constraints, repairing the breakpoints and removing the isolated points; Identify abnormal growth rings within the fusion boundary, including false rings, missing rings, and traumatic rings; Marking or correcting the abnormal growth rings according to the growth law of the growth rings; Generate a boundary quality score to identify areas requiring manual intervention.

5. The method according to claim 1, characterized in that The method of reconstructing a complete annual ring boundary curve by multi-dimensional curve fitting specifically includes: Determining a preliminary outline of the growth ring boundary based on the intersection coordinates; According to the complexity of the tree ring boundary, the curve fitting method is adaptively selected, including: When the boundary is relatively regular, piecewise polynomial fitting is used; When the boundary is moderately complex, spline curve fitting is used; When the boundary is highly complex, nonparametric curve fitting is used; Optimize the fitting results to ensure that the boundaries are smooth, continuous, and conform to biological growth laws; Generate a complete mathematical model of the tree ring boundary and support the precise coordinate calculation of any point.

6. The method according to claim 1, characterized in that The calculating of the local tilt angle distribution specifically includes: uniformly sampling a plurality of points on the growth ring boundary curve; Calculate the tangent direction at each sampling point to obtain the local tangent vector; Calculating the local tilt angle of each sampling point based on the local tangent vector; Establish a local tilt angle distribution map to describe the tilt characteristics of the entire annual ring boundary; Identify the area with sudden change of tilt angle and divide the tilt angle area to provide a basis for the selection of correction model.

7. The method according to claim 1, characterized in that The method of correcting the growth ring width using the regionalized adaptive correction model specifically includes: Dividing the annual ring boundary into a plurality of correction regions according to the local tilt angle distribution; For areas with small variations in tilt angle, a uniform tilt model is applied; For areas where the tilt angle changes gradually, a gradient tilt model is applied; For areas with sudden changes in tilt angle, a segmented tilt model is applied; Calculate the correction parameters for each region separately and perform precise correction; Calculate the true width between adjacent growth ring boundaries and generate corrected growth ring width data.

8. The method according to claim 1, characterized in that The generating of the corrected annual ring width data and quality assessment report specifically includes: Evaluate various sources of uncertainty in the measurement process, including uncertainty caused by image resolution, boundary identification, tilt angle measurement, and correction model selection; Calculate the combined uncertainty of each growth ring width; The quality of the annual ring width data is graded according to the uncertainty. Generates standard format output containing raw data, calibration data, calibration parameters, and quality ratings; Provides data visualization results to intuitively display measurement and correction effects.

9. The method according to claim 1, wherein Before the multi-band adaptive enhancement processing, the method further includes: Obtain tree species characteristic information; extracting corresponding image processing parameters from a pre-established tree species characteristic database according to the tree species characteristic information; The multi-band adaptive enhancement processing adopts the image processing parameters for parameter configuration; The tree species characteristic information includes tree species type, annual ring density, earlywood and latewood contrast, and boundary clarity.

Images