Image processing-based pinniform induction seam identification method

Abstract

The invention belongs to the technical field of image data processing, and relates to an image processing-based pinniform induction seam identification method, which comprises the following steps of S1, obtaining a basic grey-scale map; s2, obtaining an ROI region image; s3, based on the ROI region image, performing image enhancement processing to obtain a crack skeleton image; s4, performing skeleton extraction according to the crack skeleton image to obtain a global skeleton image and an optimal symmetry axis; s5, obtaining a plurality of skeleton point clusters through clustering according to the global skeleton image; based on the optimal symmetry axis, performing sine curve fitting on each skeleton point cluster to obtain an optimal fitting curve of each skeleton point cluster; and S6, obtaining a pinniform induction seam image. According to the method, the optimal channel is selected as the basic grey-scale map, the skeleton is extracted after image enhancement is carried out on the basic grey-scale map, and sine curve fitting is carried out on the skeleton to obtain the accurate pinniform induction seam image, so that the problem of insufficient pinniform induction seam automatic identification precision is solved.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology, and specifically discloses a feather-induced slit recognition method based on image processing. Background Technology

[0002] Against the backdrop of global oil and gas exploration and development continuously deepening towards unconventional reservoirs such as shale and tight sandstone, the formation of a highly conductive artificial fracture network within the reservoir through fracturing directly determines the final production capacity of oil and gas wells due to its geometric complexity and spatial distribution. Within this artificial fracture network, feather-induced fractures, as a secondary microfracture system developing along the main fracture wall and exhibiting quasi-periodic distribution characteristics, can significantly expand the reservoir stimulation volume and effectively connect matrix porosity, making them crucial for comprehensively evaluating fracturing effects and predicting oil and gas production. However, due to the inherent characteristics of feather-induced fractures, such as their high angle, short spacing, incomplete morphology, topological variability, and strong background interference, they often appear in electrical imaging logging images as a series of discrete, short, discontinuous textures arranged approximately like a sine curve. Therefore, accurately identifying feather-induced fractures using electrical imaging logging technology presents a significant challenge.

[0003] With the development of computer vision technology, current electrical imaging logging technology typically uses gradient operators such as Sobel, Canny, and Laplacian to extract fracture edges. However, when dealing with the characteristics of feather-induced fractures, such as short spacing, multiple branches, blurred edges, and severe background noise interference, the automatic identification accuracy of feather-induced fractures is insufficient. Summary of the Invention

[0004] The purpose of this invention is to provide a feather-shaped induced suture recognition method based on image processing, which solves the problem of insufficient accuracy in automatic recognition of feather-shaped induced sutures.

[0005] The specific solution of the present invention is as follows: A feather-induced suture recognition method based on image processing includes the following steps: S1. Preprocess the acquired electrical imaging logging images to obtain a basic grayscale image; S2. Based on the basic grayscale image, obtain the ROI region image through the global configuration system and YOLOv11s model; S3. Based on the ROI region image, perform image enhancement processing to obtain the crack skeleton image; S4. Extract the skeleton from the crack skeleton image to obtain the global skeleton image and the optimal axis of symmetry. S5. Based on the global skeleton image, several skeleton point clusters are obtained through clustering; Based on the optimal axis of symmetry, the optimal fitting curve for each skeleton point cluster is obtained by sinusoidal curve fitting. S6. Based on the optimal fitting curve, redundant curves are merged to obtain the effective fitting curve. The effective fitting curve is then superimposed on the background of the electrical imaging logging image to obtain the feather-shaped induced fracture image.

[0006] Preferably, obtaining the basic grayscale image includes: S11. Acquire electrical imaging logging images; S12. Based on the electrical imaging logging images, multiple channel images are obtained through a channel separation algorithm; S13. Based on the images of each channel and the acquired standard grayscale contrast, obtain the grayscale contrast enhancement rate index for each grayscale contrast. S14. Obtain the edge density enhancement rate index based on the images of each channel and the acquired standard edge density; S15. Based on the gray-level contrast enhancement rate index and the edge density enhancement rate index, select the optimal channel image as the base gray-level image.

[0007] Preferably, the grayscale contrast enhancement rate indicators include: The average gray level of each channel is obtained by calculating the average gray level of each channel image. The standard deviation of each channel image is calculated based on the average gray level of each channel to obtain the gray level contrast of each channel. The gray level contrast enhancement rate index of each channel is obtained based on the gray level contrast of each channel and the obtained standard gray level contrast.

[0008] Preferably, the edge density enhancement rate indicators include: Based on the images of each channel, the edge density of each channel is obtained by calculating the channel edge density of each channel image using the edge pixel indicator function; The edge density enhancement rate index is obtained based on the edge density of each channel and the obtained standard edge density.

[0009] Preferably, obtaining the ROI region image includes: Based on the base grayscale image, the system automatically calls the annotation file corresponding to the base grayscale image through the global configuration system, and parses and extracts the spatial coordinates and category number of the detection box from the annotation file. Based on the spatial coordinates and category number of the detection box, the YOLOv11s model is used to filter and locate candidate feather-shaped induced seam regions in the base grayscale image, and the detection box is superimposed on the base grayscale image according to the candidate feather-shaped induced seam regions to generate ROI region images.

[0010] Preferably, obtaining the crack skeleton image includes: S31. Based on the ROI region image, an enhanced image is obtained through an adaptive histogram equalization algorithm; S32. Obtain the filtered image from the enhanced image using a median filtering algorithm; S33. Obtain a binarized image from the filtered image using an adaptive thresholding algorithm, and then obtain a crack skeleton image from the binarized image using a morphological skeletonization algorithm.

[0011] Preferably, obtaining the enhanced image includes: The cumulative distribution function is obtained from the ROI region image through an adaptive histogram equalization algorithm. The enhanced image is then obtained by truncating the histogram based on the cumulative distribution function using a contrast limiting factor.

[0012] Preferably, obtaining the global skeleton image and the optimal axis of symmetry includes: S41. Based on the crack skeleton image, obtain the left branch skeleton image and the right branch skeleton image through the spatial continuity criterion and multi-scale length threshold filtering mechanism. S42. Based on the left and right limb skeleton images, a global skeleton image is constructed by skeleton integration. The optimal axis of symmetry is obtained through a mirror mapping mechanism based on the global skeleton image.

[0013] Preferably, obtaining the optimal axis of symmetry includes: The left and right limb skeleton images are integrated and processed in the standard coordinate system to construct a global skeleton image by combining the skeleton point set. Meanwhile, multiple sets of symmetry axes are preset, and a mirror mapping mechanism is adopted. Each set of symmetry axes is traversed through a grid search method, and the symmetry index under each set of symmetry axes is calculated. The symmetry axis corresponding to the maximum value of the symmetry index is selected as the optimal symmetry axis.

[0014] Preferably, the optimal fitting curves for each skeleton point cluster include: The set of skeleton pixels in the global skeleton image is clustered according to the criteria of spatial continuity and proximity to obtain several skeleton point clusters. The positional distribution information of each skeleton point cluster relative to the optimal axis of symmetry is used to estimate the initial fitting parameters of the sine curve according to the sine curve fitting formula. Based on the initial fitting parameters of the sine curve, the initial curve fitting of each skeleton point cluster is performed to obtain the initial fitting curve of each skeleton point cluster. Based on the initial fitted curve, the optimal fitting parameters of the sine curve for each skeleton point cluster are solved by the nonlinear least squares algorithm. The optimal fitting curve of each skeleton point cluster is obtained by fitting the optimal fitting parameters of the sine curve to each skeleton point cluster.

[0015] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. This invention divides electrical imaging logging images into multiple channels using a channel separation algorithm. The optimal channel image is selected as the base grayscale image by calculating the grayscale contrast enhancement rate and edge density enhancement rate of each channel image, ensuring maximum preservation of the texture details, grayscale gradient, and edge structure of the feather-shaped induced fractures. Candidate feather-shaped induced fracture regions are selected using the YOLOv11s model, and detection boxes are superimposed to obtain ROI region images, accurately representing the spatial distribution of the feather-shaped induced fractures. Adaptive histogram equalization, value filtering, adaptive thresholding, and morphological skeletonization algorithms are used to enhance the ROI region images, obtaining fracture skeleton images and enhancing the texture features and structural edges of the feather-shaped induced fractures. This method improves the expression quality of fracture features, image coherence, fracture edges, main structure, and branch features. It accurately extracts the skeleton of the fracture skeleton image through spatial continuity criteria, multi-scale length threshold filtering, and mirror mapping to obtain a global skeleton image and optimal symmetry axis, achieving noise suppression and feature preservation. Through clustering and nonlinear least squares algorithms, it accurately obtains quantitative parameters such as amplitude, angular frequency, phase, and translation of the fitted curve, ensuring that the distribution pattern of the optimal fitted curve in coordinate space is highly consistent with the geometric morphology of the feather-induced fracture. By merging the optimal fitted curves, it obtains an accurate feather-induced fracture image, clearly displaying the overall geometric structure of the feather-induced fracture, thereby improving the automatic recognition accuracy of feather-induced fractures. Attached Figure Description

[0016] Figure 1 This is a flowchart of a feather-shaped induced suture recognition method based on image processing in Embodiment 1 of the present invention.

[0017] Figure 2 This is a flowchart of obtaining the basic grayscale image in Embodiment 1 of the present invention.

[0018] Figure 3 This is a flowchart of obtaining the crack skeleton image in Embodiment 1 of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0020] Example 1 A feather-induced suture recognition method based on image processing, such as Figure 1 As shown, it includes the following steps: S1. Preprocess the acquired electrical imaging logging images to obtain a basic grayscale image; like Figure 2 As shown, obtaining the basic grayscale image includes: S11. Acquire electrical imaging logging images.

[0021] An electrical imaging logging image database is established through a global configuration system, and electrical imaging logging images are obtained from a preset path through a batch processing framework based on the electrical imaging logging image database.

[0022] The global configuration system includes surface instruments, cable transmission, downhole instruments, and interpretation workstations; the acquired electrical imaging logging images are color images and include feather-shaped induced fractures.

[0023] S12. Based on the electrical imaging logging images, multiple channel images are obtained through a channel separation algorithm.

[0024] Channel separation is performed on the electrical imaging logging images to obtain multiple channel images. The electrical imaging logging images are denoted as... , Indicates the row number of the electrical imaging logging image. This indicates the number of columns in the electrical imaging logging image.

[0025] The channel splitting algorithm uses OpenCV's cv2.split() function to split a multi-channel image into independent single-channel images.

[0026] S13. Based on the images of each channel and the obtained standard grayscale contrast ratio, obtain the grayscale contrast ratio index for each grayscale contrast ratio.

[0027] The average gray level of each channel is obtained by calculating the average gray level of each channel image. The gray level contrast of each channel is obtained by calculating the standard deviation of each channel image based on the average gray level of each channel. The gray level contrast enhancement rate of each channel is obtained based on the gray level contrast of each channel and the standard gray level contrast.

[0028] The formula for calculating the average gray level of a channel is: , in, The average gray level of the channel; Indicates the row number of the electrical imaging logging image; Indicates the column number of the electrical imaging logging image; Let the grayscale component of the channel in the i-th row and j-th column be denoted as . , R represents the red channel, G represents the green channel, and B represents the blue channel.

[0029] The formula for calculating channel grayscale contrast is: , in, This refers to the channel's grayscale contrast. Indicates the row number of the electrical imaging logging image. This indicates the column number of the electrical imaging logging image. For the grayscale component of the channel in the i-th row and j-th column, This represents the average gray level of the channel.

[0030] Channel average grayscale and channel grayscale contrast are used to reflect the overall contrast of electrical imaging logging images and can be used to determine the potential for preserving fracture details in electrical imaging logging images.

[0031] The formula for calculating the grayscale contrast ratio improvement rate is: , in, This refers to the grayscale contrast ratio improvement rate. This refers to the channel's grayscale contrast. Standard grayscale contrast.

[0032] The grayscale contrast enhancement rate metric represents the percentage improvement in contrast compared to a standard grayscale image.

[0033] S14. Obtain the edge density enhancement rate index based on the images of each channel and the acquired standard edge density.

[0034] Based on the images of each channel, the edge density of each channel is obtained by calculating the channel edge density of the images of each channel through the edge pixel indicator function; the edge density enhancement rate index is obtained based on the edge density of each channel and the standard edge density.

[0035] The expression for the edge pixel indicator function is: , in, This is a function indicating edge pixels.

[0036] The formula for calculating the channel edge density is: , in, For channel edge density, Indicates the row number of the electrical imaging logging image. This indicates the column number of the electrical imaging logging image. This is a function indicating edge pixels.

[0037] The formula for calculating the edge density improvement rate index is: , in, The edge density improvement rate is the indicator. For channel edge density, Standard edge density.

[0038] The edge density enhancement rate metric represents the percentage improvement in texture preservation relative to a standard grayscale image.

[0039] S15. Based on the gray-level contrast enhancement rate index and the edge density enhancement rate index, select the optimal channel image as the base gray-level image.

[0040] The channel image with the maximum grayscale contrast enhancement rate and edge density enhancement rate in each channel image is used as the base grayscale image.

[0041] The performance of different channels in preserving crack edge features is quantitatively evaluated by using grayscale contrast enhancement rate and edge density enhancement rate metrics. This ensures that the texture details, grayscale gradients, and edge structures of the feather-induced crack are preserved to the greatest extent, providing an accurate image basis for subsequent feather-induced crack recognition.

[0042] S2. Based on the basic grayscale image, obtain the ROI region image through the global configuration system and YOLOv11s model; Based on the base grayscale image, the system automatically calls the annotation file corresponding to the base grayscale image through the global configuration system, and parses and extracts the spatial coordinates and category number of the detection box from the annotation file. Based on the spatial coordinates and category number of the detection box, the YOLOv11s model is used to filter and locate candidate feather-shaped induced seam regions in the base grayscale image, and the detection box is superimposed on the base grayscale image according to the candidate feather-shaped induced seam regions to generate ROI region images.

[0043] The ROI region image can accurately depict the spatial distribution characteristics of the feather-shaped induced slits and include category labels, providing a precise input basis for subsequent image enhancement and skeleton extraction analysis.

[0044] S3. Based on the ROI region image, perform image enhancement processing to obtain the crack skeleton image; like Figure 3 As shown, the obtained crack skeleton image includes: S31. Based on the ROI region image, an enhanced image is obtained through an adaptive histogram equalization algorithm.

[0045] Based on the ROI region image, the ROI region image is defined as... ,in Representing pixel coordinates, the cumulative distribution function is obtained through an adaptive histogram equalization algorithm, and then a contrast limiting factor is applied based on the cumulative distribution function. The enhanced image is obtained by histogram truncation.

[0046] The expression for the cumulative distribution function is: , in, The cumulative distribution function; The function is defined as follows: when the grayscale value of the pixel in the i-th row and j-th column is less than or equal to the preset grayscale level. The value is 1 if it is true, and 0 otherwise. For each local region Total number of pixels in the local area Refer to The central segmented area.

[0047] The formula for calculating histogram truncation is: , in, Truncate the histogram of gray level g; This is the contrast limiting factor, which can be set to 2.0; The cumulative distribution function; The total number of gray levels, usually =256, preset grayscale level .

[0048] The formula for calculating image enhancement is: , in, To enhance the image, For ROI region images, To truncate the histogram, This represents the total number of gray levels.

[0049] The adaptive histogram equalization algorithm effectively improves the local grayscale contrast of the ROI region image, enhances the texture features and structural edges of the feathered induced slit, and improves the expression quality of the slit features.

[0050] S32. Obtain the filtered image from the enhanced image using a median filtering algorithm.

[0051] The formula for calculating the filtered image is: , in, To filter images; To enhance the image; Indicates The filtering window area centered on the image noise level is dynamically adjusted and can be selected. .

[0052] By using the median filtering algorithm, the median of the local neighborhood of each pixel is used as the replacement value. The resulting filtered image can effectively smooth high-frequency interference, maintain the continuity of the crack trunk edge, improve the image coherence of the crack trunk and the expression quality of the fracture edge, and further suppress the influence of isolated noise points in the image on the skeleton extraction results.

[0053] S33. Obtain a binarized image from the filtered image using an adaptive thresholding algorithm, and then obtain a crack skeleton image from the binarized image using a morphological skeletonization algorithm.

[0054] Adaptive threshold segmentation is an image processing technique that dynamically calculates thresholds based on the features of local regions in an image. It analyzes the local neighborhood information of each pixel in the image, such as grayscale distribution and brightness changes, to dynamically calculate the optimal threshold for that region, rather than using a globally uniform threshold. This reduces the impact of illumination changes or shadows on the segmentation results and improves segmentation accuracy.

[0055] By using an adaptive thresholding algorithm, the crack area is represented as a high-contrast black area, forming a clear separation from the background.

[0056] The binarized image is transformed into a single-pixel-width centerline structure by using a morphological skeletonization algorithm to extract its main outline and spatial orientation, thereby obtaining a crack skeleton image.

[0057] The morphological skeletonization algorithm effectively removes isolated noise points, maintains the continuity of the main crack structure, significantly reduces background interference, and clearly presents the main structure and branch features of the feather-induced crack.

[0058] S4. Extract the skeleton from the crack skeleton image to obtain the global skeleton image and the optimal axis of symmetry. S41. Based on the crack skeleton image, obtain the left branch skeleton image and the right branch skeleton image through the spatial continuity criterion and multi-scale length threshold filtering mechanism.

[0059] By using the spatial continuity criterion and multi-scale length threshold filtering mechanism, the skeleton structure in the crack skeleton image is quantitatively analyzed and screened, the main branch skeleton is accurately identified and extracted, and the small bifurcation and noise interference are suppressed to obtain the left branch skeleton image and the right branch skeleton image.

[0060] The principle of spatial continuity is one of the core principles of natural regionalization, requiring that the natural regions to be divided remain spatially complete and continuous, and cannot be divided or repeated. This principle, together with principles such as uniformity and relative consistency, constitutes the methodological framework of natural regionalization, and regional division is achieved through sequential division and sequential merging methods.

[0061] Multi-scale length threshold filtering is an image or point cloud filtering method that combines multi-scale analysis with threshold processing. It achieves noise suppression and feature preservation by setting thresholds at different scales.

[0062] S42. Based on the left and right limb skeleton images, a global skeleton image is constructed by skeleton integration. The optimal axis of symmetry is obtained through a mirror mapping mechanism based on the global skeleton image.

[0063] The left and right lateral skeletal images are integrated and processed in a standard coordinate system to construct a global skeleton image. At the same time, multiple sets of symmetry axes are preset, and a mirror mapping mechanism is used to traverse each set of symmetry axes through a grid search method. The symmetry index under each set of symmetry axes is calculated, and the symmetry axis corresponding to the maximum value of the symmetry index is selected as the optimal symmetry axis of the feather-shaped induced suture structure in the global skeleton image.

[0064] The formula for calculating the symmetry index is: , in, S As a symmetry index, This indicates the number of skeletal points that overlap between the mirrored image of the left ramus and the image of the right ramus. This represents the total number of skeleton points in the right lateral branch bone image. This represents the total number of skeleton points in the right branch skeleton image.

[0065] S5. Based on the global skeleton image, several skeleton point clusters are obtained through clustering; based on the optimal axis of symmetry, sine curve fitting is performed on each skeleton point cluster to obtain the optimal fitting curve for each skeleton point cluster. For the skeleton pixel set in the global skeleton image, clustering is performed based on spatial continuity and proximity criteria to obtain several skeleton point clusters. Each skeleton point cluster approximately represents a circumferential segment of the main trunk of the crack. The positional distribution information of each skeleton point cluster relative to the optimal axis of symmetry is used to estimate the initial fitting parameters of the sine curve according to the sine curve fitting formula. Based on the initial fitting parameters of the sine curve, initial curve fitting is performed on each skeleton point cluster to obtain the initial fitting curve of each skeleton point cluster. Based on the initial fitting curve, the optimal fitting parameters of the sine curve are solved for the objective function of each skeleton point cluster using a nonlinear least squares algorithm. The optimal fitting curve of each skeleton point cluster is obtained by fitting the optimal fitting parameters of the sine curve.

[0066] Clustering is the process of dividing a dataset into multiple classes or clusters composed of similar objects, with the aim of maximizing the similarity of objects within the same cluster and maximizing the differences between different clusters.

[0067] The formula for fitting a sine curve is: , in,y It is a sine curve. A Represents amplitude, Represents angular frequency, For phase, b This represents the amount of translation.

[0068] The initial fitting parameters of the sine curve are used for subsequent nonlinear optimization fitting, which can improve convergence efficiency and fitting accuracy.

[0069] The expression for the objective function is: , in, L This is the least squares error function, which is the sum of squared errors between all skeleton points and the fitted sine curve; A Represents amplitude; Indicates angular frequency; For phase; b Represents the translation amount; For the first point cluster in the skeleton The coordinates of n pixels, i = 1, 2, ..., n.

[0070] The optimal fitting curve obtained by fitting the optimal fitting parameters of the sine curve solved by the nonlinear least squares algorithm effectively eliminates local fluctuations, enhances the smoothness and continuity of the fitting curve, initially improves the redundancy of the fitting curve, preserves the spatial distribution characteristics of the skeleton point clusters, and makes the distribution law of the optimal fitting curves of each skeleton point cluster in the coordinate space highly consistent with the geometric shape of the feather-shaped induced slit.

[0071] In solving for the optimal fitting parameters of the sine curve, error convergence constraints are applied to eliminate invalid results that do not converge or have excessively large errors, ensuring the physical rationality and structural interpretability of the fitting expression.

[0072] Error convergence constraints include: setting the maximum number of iterations, the convergence threshold for parameter updates and changes in the objective function during the fitting process. The maximum number of iterations can be 10, and the convergence threshold can be 0.5. S6. Based on the optimal fitting curve, redundant curves are merged to obtain the effective fitting curve. The effective fitting curve is then superimposed on the background of the electrical imaging logging image to obtain the feather-shaped induced fracture image.

[0073] Based on the optimal fitted curve, the geometric consistency between curve pairs is measured by similarity quantification index. Under the premise of satisfying the fitting error constraint, the original curve segments are merged to obtain the effective fitted curve, avoiding the repetitive expression caused by multiple fittings. The similarity quantification index includes parameter difference degree, fitting residual difference, etc. During the fitting process, the fitting parameters, matching index and error statistics of each fitted curve are stored in a standard structure.

[0074] The fitting error constraint refers to the following: before merging the segments of the same origin curves, the optimal fitting curve is checked for errors. The normalized mean square error must be lower than a first error threshold, which can be 0.15, and the coefficient of determination R0 must be within a certain range. 2 The error exceeds the second error threshold, which can be 0.6. This occurs when the normalized mean square error is below the first error threshold and the coefficient of determination R is also satisfied. 2 Only curves exceeding the second error threshold are considered valid and included in the merging process; otherwise, they are discarded to ensure the accuracy and reliability of the final results.

[0075] The formula for calculating the coefficient of determination is: , in, The coefficient of determination; The i-th actual observation value, i.e., the y-coordinate of the crack skeleton point; The fitted curve in the corresponding The predicted value below; The mean of the observed values.

[0076] By structurally screening and merging redundant or overlapping curve segments in the optimal fitting curve, not only is redundant expression reduced, but the complete expression of key geometric features is also maintained. This effectively reduces the redundancy of the fitting curve, improves the accuracy and uniqueness of the expression of crack geometry, and clearly shows the overall geometric structure of the feather-shaped induced crack.

[0077] The electrical imaging logging image is divided into multiple channels using a channel separation algorithm. The optimal channel image is selected as the base grayscale image by calculating the grayscale contrast enhancement rate and edge density enhancement rate of each channel image, ensuring maximum preservation of the texture details, grayscale gradient, and edge structure of the feather-shaped induced fractures. Candidate feather-shaped induced fracture regions are selected using the YOLOv11s model, and detection boxes are superimposed to obtain ROI region images, accurately representing the spatial distribution of the feather-shaped induced fractures. Image enhancement of the ROI region images is performed using adaptive histogram equalization, value filtering, adaptive thresholding, and morphological skeletonization algorithms to obtain fracture skeleton images, enhancing the texture features and structural edges of the feather-shaped induced fractures. The system enhances the representation quality of fracture features, image coherence, fracture edges, main structure, and branch features. It accurately extracts the skeleton of the fracture skeleton image using spatial continuity criteria, multi-scale length threshold filtering, and mirror mapping mechanisms to obtain a global skeleton image and optimal symmetry axis, achieving noise suppression and feature preservation. Through clustering and nonlinear least squares algorithms, it accurately obtains quantitative parameters such as amplitude, angular frequency, phase, and translation of the fitted curve, ensuring that the distribution pattern of the optimal fitted curve in coordinate space is highly consistent with the geometric morphology of the feather-induced fracture. By merging the optimal fitted curves, it obtains an accurate feather-induced fracture image, clearly displaying the overall geometric structure of the feather-induced fracture, thereby improving the automatic recognition accuracy of feather-induced fractures.

[0078] Example 2 A method for identifying feather-shaped induced fractures in Block X of an oilfield includes the following steps: S1. Obtain electrical imaging logging images of an oilfield in area X.

[0079] The global configuration system allows for the preset of key parameters, including image storage path, processing mode, and performance statistics control. An electrical imaging logging image database for an oilfield's X area is constructed using the global configuration system. Based on this database, a batch processing framework is used to retrieve electrical imaging logging images containing feather-shaped induced fractures from preset paths.

[0080] S2. Based on the electrical imaging logging images, multiple channel images are obtained through the channel separation algorithm. Based on each channel image, the gray-scale contrast enhancement rate and edge density enhancement rate of each channel image are calculated, and the optimal channel image is selected as the base gray-scale image. Based on the electrical imaging logging images, red, green, and blue channel images were obtained using a separation algorithm. The grayscale contrast enhancement rate and edge density enhancement rate of the red, green, and blue channel images were calculated respectively. The results showed that the grayscale contrast enhancement rate of the red channel image was 0.9% and the edge density enhancement rate was -2.7%; the grayscale contrast enhancement rate of the green channel image was 4.1% and the edge density enhancement rate was 4.4%; and the grayscale contrast enhancement rate of the blue channel image was -2.8% and the edge density enhancement rate was 4.2%. It can be seen that the green channel image has the maximum grayscale contrast enhancement rate and edge density enhancement rate, indicating that the green channel is the optimal channel. Therefore, the green channel image was used as the base grayscale image.

[0081] S3. Based on the basic grayscale image, obtain the ROI region image through the global configuration system and YOLOv11s model; Based on the base grayscale image, the annotation file corresponding to the green channel image is automatically called through the global configuration system. The spatial coordinates and category numbers of the detection boxes are parsed and extracted from the annotation file. Based on the spatial coordinates and category numbers of the detection boxes, the YOLOv11s model is used to filter and locate candidate feather-shaped induced seam regions in the green channel image. Based on the candidate feather-shaped induced seam regions, detection boxes are superimposed on the green channel image to generate ROI region images.

[0082] The ROI region image can accurately depict the spatial distribution characteristics of the feather-shaped induced slits and include category labels, providing a precise input basis for subsequent image enhancement and skeleton extraction analysis.

[0083] S4. Based on the ROI region image, an enhanced image is obtained through an adaptive histogram equalization algorithm. A filtered image is obtained through a median filtering algorithm based on the enhanced image. A binarized image is obtained through an adaptive threshold segmentation algorithm based on the filtered image. A crack skeleton image is obtained through a morphological skeletonization algorithm based on the binarized image. The left and right branch skeleton images are obtained through a spatial continuity criterion and a multi-scale length threshold filtering mechanism based on the crack skeleton image. The images of the left and right branch skeletons show that the main path of the crack and its bifurcation relationship are significantly restored, and some short branches and unstructured noise lines are included. The stray branches are effectively removed, the main structure remains continuous, the morphology is complete, and it has a good geometric representation basis.

[0084] S5. Based on the left and right limb skeleton images, a global skeleton image is constructed by skeleton integration. The optimal axis of symmetry is obtained through a mirror mapping mechanism based on the global skeleton image. The left and right lateral skeletal images are integrated and processed in a standard coordinate system to construct a global skeleton image. At the same time, multiple sets of symmetry axes are preset, and a mirror mapping mechanism is used to traverse each set of symmetry axes through a grid search method. The symmetry index and symmetry score under each set of symmetry axes are calculated, as shown in Table 1. The symmetry axis corresponding to the maximum value of both the symmetry index and the symmetry score is selected as the optimal symmetry axis of the feather-shaped induced suture structure.

[0085] The formula for calculating the symmetry score is: , Among them, W i Score the symmetry of the i-th group of symmetry axes; i S is the symmetry index of the i-th group of symmetry axes, i = 1, 2, ..., n; i This represents the maximum value of the symmetry index among all symmetry indices.

[0086] Table 1 Symmetry Indices and Symmetry Scores for Each Axis of Symmetry

[0087] As can be seen from Table 1, when the symmetry axis x=188.0, the corresponding symmetry index and symmetry score are both at their maximum values, indicating that the crack structures on both sides of the global skeleton image have the strongest geometric symmetry under this symmetry axis. Therefore, x=188.0 is the optimal symmetry axis.

[0088] S6. Based on the global skeleton image, several skeleton point clusters are obtained through clustering; based on the optimal axis of symmetry, sine curve fitting is performed on each skeleton point cluster to obtain the optimal fitting curve for each skeleton point cluster. For the skeleton pixel set in the global skeleton image, clustering is performed based on spatial continuity and proximity criteria to obtain several skeleton point clusters. Each skeleton point cluster approximately represents a circumferential segment of the main trunk of the crack. The positional distribution information of each skeleton point cluster relative to the optimal axis of symmetry is used to estimate the initial fitting parameters of the sine curve according to the sine curve fitting formula. Based on the initial fitting parameters of the sine curve, initial curve fitting is performed on each skeleton point cluster to obtain the initial fitting curve of each skeleton point cluster. Based on the initial fitting curve, the optimal fitting parameters of the sine curve are solved for the objective function of each skeleton point cluster using a nonlinear least squares algorithm. The optimal fitting curve of each skeleton point cluster is obtained by fitting the optimal fitting parameters of the sine curve.

[0089] S7. Based on the optimal fitting curve, redundant curves are merged to obtain the effective fitting curve. The effective fitting curve is then superimposed on the background of the electrical imaging logging image to obtain the feather-shaped induced fracture image.

[0090] Based on the optimal fitted curve, the geometric consistency between curve pairs is measured by the similarity quantification index. Under the premise of satisfying the fitting error constraint, the original curve segments are merged to obtain the effective fitted curve.

[0091] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A feather-shaped induced slit recognition method based on image processing, characterized in that, Includes the following steps: S1. Preprocess the acquired electrical imaging logging images to obtain a basic grayscale image; S2. Based on the basic grayscale image, obtain the ROI region image through the global configuration system and YOLOv11s model; S3. Based on the ROI region image, perform image enhancement processing to obtain the crack skeleton image; S4. Extract the skeleton from the crack skeleton image to obtain the global skeleton image and the optimal axis of symmetry. S5. Based on the global skeleton image, several skeleton point clusters are obtained through clustering; based on the optimal axis of symmetry, sine curve fitting is performed on each skeleton point cluster to obtain the optimal fitting curve for each skeleton point cluster. S6. Based on the optimal fitting curve, redundant curves are merged to obtain the effective fitting curve. The effective fitting curve is then superimposed on the background of the electrical imaging logging image to obtain the feather-shaped induced fracture image.

2. The feather-shaped induced slit recognition method based on image processing according to claim 1, characterized in that, The process of obtaining the base grayscale image includes: S11. Acquire electrical imaging logging images; S12. Based on the electrical imaging logging images, multiple channel images are obtained through a channel separation algorithm; S13. Based on the images of each channel and the acquired standard grayscale contrast, obtain the grayscale contrast enhancement rate index for each grayscale contrast. S14. Obtain the edge density enhancement rate index based on the images of each channel and the acquired standard edge density; S15. Based on the gray-level contrast enhancement rate index and the edge density enhancement rate index, select the optimal channel image as the base gray-level image.

3. The feather-shaped induced slit recognition method based on image processing according to claim 2, characterized in that, The obtained grayscale contrast enhancement rate indicators include: The average gray level of each channel is obtained by calculating the average gray level of each channel image. The standard deviation of each channel image is calculated based on the average gray level of each channel to obtain the gray level contrast of each channel. The gray level contrast enhancement rate index of each channel is obtained based on the gray level contrast of each channel and the obtained standard gray level contrast.

4. The feather-shaped induced slit recognition method based on image processing according to claim 2, characterized in that, The obtained edge density enhancement rate indicators include: Based on the images of each channel, the edge density of each channel is obtained by calculating the channel edge density of the images of each channel through the edge pixel indicator function; and the edge density enhancement rate index is obtained based on the edge density of each channel and the obtained standard edge density.

5. The feather-shaped induced slit recognition method based on image processing according to claim 1, characterized in that, The obtained ROI region image includes: Based on the base grayscale image, the system automatically calls the annotation file corresponding to the base grayscale image through the global configuration system, and parses and extracts the spatial coordinates and category number of the detection box from the annotation file. Based on the spatial coordinates and category number of the detection box, the YOLOv11s model is used to filter and locate candidate feather-shaped induced seam regions in the base grayscale image, and the detection box is superimposed on the base grayscale image according to the candidate feather-shaped induced seam regions to generate ROI region images.

6. The feather-shaped induced slit recognition method based on image processing according to claim 1, characterized in that, The obtained crack skeleton image includes: S31. Based on the ROI region image, an enhanced image is obtained through an adaptive histogram equalization algorithm; S32. Obtain the filtered image from the enhanced image using a median filtering algorithm; S33. Obtain a binarized image from the filtered image using an adaptive thresholding algorithm, and then obtain a crack skeleton image from the binarized image using a morphological skeletonization algorithm.

7. The feather-shaped induced slit recognition method based on image processing according to claim 6, characterized in that, The enhanced image includes: The cumulative distribution function is obtained from the ROI region image through an adaptive histogram equalization algorithm. The enhanced image is then obtained by truncating the histogram based on the cumulative distribution function using a contrast limiting factor.

8. The feather-shaped induced slit recognition method based on image processing according to claim 1, characterized in that, The process of obtaining the global skeleton image and the optimal symmetry axis includes: S41. Based on the crack skeleton image, obtain the left branch skeleton image and the right branch skeleton image through the spatial continuity criterion and multi-scale length threshold filtering mechanism. S42. Based on the left and right limb skeleton images, a global skeleton image is constructed by skeleton integration. The optimal axis of symmetry is obtained through a mirror mapping mechanism based on the global skeleton image.

9. The feather-shaped induced slit recognition method based on image processing according to claim 8, characterized in that, The process of obtaining the optimal axis of symmetry includes: The left and right limb skeleton images are integrated and processed in the standard coordinate system to construct a global skeleton image by combining the skeleton point set. Meanwhile, multiple sets of symmetry axes are preset, and a mirror mapping mechanism is adopted. Each set of symmetry axes is traversed through a grid search method, and the symmetry index under each set of symmetry axes is calculated. The symmetry axis corresponding to the maximum value of the symmetry index is selected as the optimal symmetry axis.

10. The feather-shaped induced slit recognition method based on image processing according to claim 1, characterized in that, The optimal fitting curves for each skeleton point cluster include: The set of skeleton pixels in the global skeleton image is clustered according to the criteria of spatial continuity and proximity to obtain several skeleton point clusters. The positional distribution information of each skeleton point cluster relative to the optimal axis of symmetry is used to estimate the initial fitting parameters of the sine curve according to the sine curve fitting formula. Based on the initial fitting parameters of the sine curve, the initial curve fitting of each skeleton point cluster is performed to obtain the initial fitting curve of each skeleton point cluster. Based on the initial fitted curve, the optimal fitting parameters of the sine curve for each skeleton point cluster are solved by the nonlinear least squares algorithm. The optimal fitting curve of each skeleton point cluster is obtained by fitting the optimal fitting parameters of the sine curve to each skeleton point cluster.

Images