An online method and system for analyzing rock fragment morphology based on image recognition

Abstract

The present application relates to the technical field of debris detection, and discloses a kind of based on image recognition's debris morphology online analysis method and system, the method includes: to debris sample is carried out line array scanning, and the original debris strip image is carried out gray scale gradient detection, and the edge response area is obtained;The boundary connectivity repair is carried out to the response area adjacent in position, and the complete closed contour is obtained;The background of original debris strip image is removed, and single particle image is obtained;The edge of contour is carried out corner feature coding, to construct corner degree distribution atlas;Single particle image is carried out long axis direction comparison, and the single particle image close to comparison result is merged into the directional group of debris particle, and the dominant direction vector of directional group is quantified;The dominant direction vector and corner degree distribution atlas are morphologically matched with the standard debris atlas of debris particle, and the debris morphology analysis result of debris sample is obtained;The present application can improve the efficiency of a kind of based on image recognition's debris morphology online analysis.

Description

Technical Field

[0001] This invention relates to the field of rock cuttings detection technology, and in particular to an online method and system for rock cuttings morphology analysis based on image recognition. Background Technology

[0002] Existing rock fragment morphology analysis techniques mostly rely on manual observation or traditional image recognition methods, which have insufficient image acquisition accuracy for rock fragment samples. Furthermore, edge breakpoints and incomplete contours are prone to occur during the rock fragment edge detection stage, making it impossible to accurately capture the true contour features of rock fragments. This leads to a significant reduction in the accuracy of subsequent morphological feature extraction. At the same time, traditional methods lack systematic quantitative means for analyzing the angular and directional distribution features of rock fragments, and can only make qualitative judgments. It is difficult to construct a standardized rock fragment morphology feature atlas, resulting in poor objectivity and comparability of rock fragment morphology analysis results.

[0003] In traditional rock debris morphology analysis workflows, noise removal and single-particle separation in the background culling stage are ineffective, easily leading to the inclusion of irrelevant background pixels or particle adhesion issues, affecting the purity of individual particle images. Furthermore, the lack of scientific clustering algorithms during rock debris particle orientation grouping makes it impossible to accurately quantify the dominant direction vector of rock debris particles, resulting in insufficient feature dimensions in the morphology matching stage, leading to low matching accuracy and efficiency, and failing to meet the real-time and accuracy requirements of online rock debris morphology analysis. Therefore, improving the accuracy of rock debris morphology image recognition and the efficiency of online analysis has become an urgent problem to be solved. Summary of the Invention

[0004] This invention provides an online analysis method and system for rock debris morphology based on image recognition, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, this invention provides an online analysis method for rock debris morphology based on image recognition, comprising: S1. Perform linear array scanning on the collected rock debris samples and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. S2. Perform boundary connectivity repair on adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment particles; S3. Using the complete closed contour as a mask, the background of the original rock debris strip image is removed to obtain a single-particle image of the rock debris particles; S4. Encode the angular features of the contour edges of the single-particle image to construct the angular distribution map of the rock debris particles; S5. Perform long axis direction comparison on the single particle images, group single particle images with similar comparison results into orientation groups of the rock cutting particles, and quantify the dominant direction vector of the orientation group; S6. The dominant direction vector and the angular distribution map are morphologically matched with the standard rock debris map of the rock debris particles to obtain the rock debris morphology analysis results of the rock debris sample.

[0006] In a preferred embodiment, the step of performing linear array scanning on the collected rock debris samples and performing grayscale gradient detection on the original rock debris strip images obtained from the scanning to obtain the edge response regions of the rock debris particles in the rock debris samples includes: The collected rock debris samples were dispersed and spread out, and the resulting sample distribution layer was subjected to linear array push-broom imaging to obtain the original rock debris strip image of the rock debris sample. Gradient magnitude reconstruction is performed on the pixels in the original rock debris strip image to obtain the edge intensity characterization of the pixels; Based on the edge intensity characterization, local extremum screening is performed on the pixels to obtain candidate edge points of rock debris particles in the rock debris sample; Eight-neighbor connectivity analysis is performed on the candidate edge points to merge interconnected candidate edge points into connected regions. The connected regions are then delineated using the minimum bounding rectangle to obtain the edge response region of the rock fragments.

[0007] In a preferred embodiment, the step of restoring the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment includes: Edge discontinuity detection is performed on the edge response region to obtain the contour fracture endpoints of the rock fragments; The contour fracture endpoint with the smallest included angle between the direction vectors is taken as the endpoint pair to be connected for the rock cutting particles; Edge direction interpolation fitting is performed on the gap region between the endpoints to be connected to obtain the transition edge segment of the endpoints to be connected. The transition edge segment is merged with the edge response area to obtain the complete closed contour of the rock fragment.

[0008] In a preferred embodiment, the step of using the complete closed contour as a mask to remove the background from the original rock debris strip image to obtain a single-particle image of the rock debris grain includes: The complete closed contour is filled with a binarized region to obtain a binarized mask layer of the rock fragments; A morphological closure operation is performed on the small cavities in the binarized mask layer to obtain a repaired mask layer for the rock debris particles; The restorative mask layer is registered pixel-by-pixel with the original rock debris strip image to obtain the mask extraction image of the rock debris particles; Isolated noise points are removed from the mask-extracted image to obtain a purified image of the rock fragments. The cleaned rock debris image is subjected to connected component separation and cropping to obtain single-particle images of the rock debris particles.

[0009] In a preferred embodiment, the step of removing isolated noise points from the mask-extracted image to obtain a purified image of the rock fragments includes: Topological labeling is performed on the connected components in the mask-extracted image to obtain a connected component labeling image of the rock debris particles; The number of pixels in the connected component is measured to obtain the scale representation value of the connected component. Based on the scale characterization value, scale-level stripping is performed on the connected components to obtain the isolated noise connected domains of the rock debris particles. The isolated noise connected components are removed from the connected component labeled image to obtain the purified rock debris image of the rock debris particles.

[0010] In a preferred embodiment, encoding the angular features of the contour edges of the single-particle image to construct an angular distribution map of the rock debris particles includes: The single-particle image is traced to obtain the continuous edge contour lines of the rock debris particles; The continuous edge contour line is resampled with equal arc length contour points to obtain a discrete contour point sequence of the rock fragments. Local orientation analysis is performed on adjacent contour points in the discrete contour point sequence to obtain the contour curvature feature value of the contour points; Based on the contour curvature feature value, the contour points are calibrated for edge angle level to obtain the edge angle feature coding sequence of the rock fragments; Using the circumferential position of the contour points as the horizontal axis of the map and the corner feature encoding value as the vertical axis of the map, the corner feature encoding sequence is mapped to obtain the corner distribution map of the rock debris particles.

[0011] In a preferred embodiment, the step of performing long-axis alignment on the single-particle images, grouping single-particle images with similar alignment results into orientation groups of the rock debris particles, and quantifying the dominant direction vector of the orientation group includes: The principal axis orientation of the single-particle image is extracted to obtain the individual principal direction vector of the rock cutting particle; By mapping the principal direction vector of the individual rock fragments to angle values, the directional angle values ​​of the rock fragments are obtained. Density peak clustering is performed on the directional angle values ​​to obtain the orientation groups of the rock fragments; The dominant direction vector of the directional group is obtained by superimposing the individual principal direction vectors of the rock fragments in the directional group.

[0012] In a preferred embodiment, the step of performing density peak clustering on the orientation angle values ​​to obtain the orientation groups of the rock fragments includes: The distribution range of the direction angle value is analyzed to obtain the neighborhood cutoff distance of the direction angle value; The frequency of neighborhood points is statistically analyzed for the direction angle value to obtain the local density value of the direction angle value; wherein, the calculation formula for the local density value is as follows: ; in, Indicates the first Local density values ​​for each directional angle. Indicates the first Each direction angle value, Indicates the first Each direction angle value, Indicates the neighborhood cutoff distance, This represents the summation function. Indicates an indicator function, when hour, ,otherwise ; Based on the local density value, minimum distance addressing is performed on the direction angle value to obtain the relative distance value of the direction angle value; Based on the local density value and the relative distance value, cluster centers are identified for the direction angle value to obtain the cluster center point of the direction angle value; Using the cluster center point as the core, the directional angle values ​​are clustered and assigned to obtain the orientation groups of the rock fragments.

[0013] In a preferred embodiment, the step of morphological matching the dominant direction vector and the angular distribution map with the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphology analysis results of the rock fragment sample includes: The dominant direction vector and the edge angle distribution map are fused to obtain the feature vector to be matched for the rock fragments; Based on the feature vector to be matched, a similarity search is performed in the standard rock debris map library of the rock debris particles to obtain the candidate matching map of the rock debris particles; The candidate matching map and the feature vector to be matched are compared with local similarity to obtain the local similarity sequence of the rock fragments; Based on the local fit sequence, the candidate matching maps are evaluated for comprehensive confidence, and the rock fragment morphology analysis results of the rock fragment samples are obtained.

[0014] To address the aforementioned problems, the present invention also provides an online rock debris morphology analysis system based on image recognition, the system comprising: The image acquisition and edge response extraction module is used to perform linear array scanning on the acquired rock debris samples and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. The contour connectivity repair module is used to repair the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment particles. The single-particle image extraction module is used to remove the background from the original rock debris strip image by using the complete closed contour as a mask, so as to obtain the single-particle image of the rock debris particle. An edge feature encoding module is used to encode the edge features of the contour edges of the single particle image in order to construct an edge angle distribution map of the rock debris particles; The orientation group identification and dominant direction quantification module is used to compare the single particle images along their long axis, group single particle images with similar comparison results into orientation groups of the rock cutting particles, and quantify the dominant direction vector of the orientation group. The morphological matching analysis module is used to perform morphological matching between the dominant direction vector and the angular distribution map and the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphological analysis results of the rock fragment sample. Compared with the prior art, the present invention has the following beneficial effects: 1. This technology combines linear scanning with grayscale gradient detection to accurately extract the edge response regions of rock debris particles. Boundary connectivity restoration then forms complete closed contours. Masking technology is used to remove backgrounds and separate individual particles, achieving high-precision extraction of rock debris images. Simultaneously, edge angle distribution maps are constructed through edge feature encoding, and density peak clustering is used to divide rock debris particles into directional groups and quantify the dominant direction vector. This enables systematic and quantitative extraction of rock debris morphological features, significantly improving the comprehensiveness and accuracy of rock debris morphological feature analysis and laying a high-quality data foundation for subsequent morphology matching.

[0015] 2. This technology fuses the dominant direction vector of rock fragments with the angular distribution map to form a multi-dimensional feature vector to be matched. Similarity retrieval and local fit comparison are then performed in a standard rock fragment image library. After comprehensive confidence evaluation, the rock fragment morphology analysis results are output, effectively improving the accuracy of morphology matching. The entire technology relies on a modular system architecture to achieve efficient connection and automated processing of each step, significantly improving the overall efficiency of rock fragment morphology analysis. It enables online analysis of rock fragment morphology, and the analysis results are standardized and quantifiable, enhancing the objectivity and practical application value of the rock fragment morphology analysis results. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an online rock debris morphology analysis method based on image recognition, provided in an embodiment of the present invention. Figure 2 A functional module diagram of an online rock debris morphology analysis system based on image recognition, provided in an embodiment of the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0017] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0018] This application provides an online analysis method for rock debris morphology based on image recognition. The execution entity of this online analysis method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the online analysis method for rock debris morphology based on image recognition can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0019] Reference Figure 1 The diagram shown is a flowchart illustrating an online rock debris morphology analysis method based on image recognition, according to an embodiment of the present invention. In this embodiment, the online rock debris morphology analysis method based on image recognition includes: S1. Perform linear array scanning on the collected rock debris samples and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. In this embodiment of the invention, the step of performing linear array scanning on the collected rock debris samples and performing grayscale gradient detection on the original rock debris strip image obtained from the scanning to obtain the edge response region of the rock debris particles in the rock debris sample includes: The collected rock debris samples were dispersed and spread out, and the resulting sample distribution layer was subjected to linear array push-broom imaging to obtain the original rock debris strip image of the rock debris sample. Gradient magnitude reconstruction is performed on the pixels in the original rock debris strip image to obtain the edge intensity characterization of the pixels; Based on the edge intensity characterization, local extremum screening is performed on the pixels to obtain candidate edge points of rock debris particles in the rock debris sample; Eight-neighbor connectivity analysis is performed on the candidate edge points to merge interconnected candidate edge points into connected regions. The connected regions are then delineated using the minimum bounding rectangle to obtain the edge response region of the rock fragments.

[0020] The collected rock cuttings samples were physically dispersed to separate the rock cutting particles from each other without adhesion. The dispersed rock cutting particles were then evenly spread on a designated bearing surface to form a sample distribution layer with uniform thickness and particle distribution. A linear array scanning imaging device was used to continuously push and scan the sample distribution layer along a fixed horizontal direction on the bearing surface. During the pushing and scanning process, optical image information of the sample distribution layer was collected line by line. All continuously collected line image information was seamlessly stitched and integrated in the order of collection to finally generate the original rock cutting strip image of the rock cutting sample.

[0021] Gray values ​​are extracted from each pixel in the original rock debris strip image. The gray value differences between each pixel and its adjacent pixels in the horizontal and vertical directions are calculated. The gray value differences in the two directions are fused to reconstruct the gradient amplitude. The edge intensity representation corresponding to each pixel is determined based on the constructed gradient amplitude. This edge intensity representation directly reflects the degree to which the edge features of rock debris particles are present at the pixel.

[0022] Taking each pixel in the original rock debris strip image as the center, all neighboring pixels around the pixel are selected to form a local neighborhood. The edge intensity representation of the center pixel is compared with the edge intensity representation of all pixels in the local neighborhood one by one. The center pixel with the maximum edge intensity representation value in the local neighborhood is selected. All pixels selected in this way constitute the candidate edge points of rock debris particles in the rock debris sample.

[0023] For each candidate edge point, the system detects whether there are other candidate edge points in the eight directions of the pixel: up, down, left, right, and four diagonals. If there are candidate edge points in the corresponding directions, the two candidate edge points are considered to be connected. After completing the full detection of all candidate edge points in this way, all candidate edge points that are connected to each other are merged into multiple independent connected regions. Geometric contour detection is performed on each independent connected region to determine the position of the outermost pixel of each connected region. A rectangular border that can completely enclose the connected region is drawn based on the outermost pixel. This rectangular border is the minimum bounding rectangle of the connected region. The area covered by each minimum bounding rectangle is the edge response region of the rock debris particles.

[0024] The beneficial effects are that spreading the rock debris samples before linear array push-broom imaging ensures the clarity and independence of rock debris particles in the original rock debris strip images, laying a high-quality image foundation for subsequent image analysis. The edge intensity characterization obtained through gradient amplitude reconstruction can accurately quantify the edge features of each pixel. Based on this, local extremum screening can efficiently and accurately locate candidate edge points of rock debris particles. Eight-neighbor connectivity analysis can accurately merge connected candidate edge points to form connected regions. Combined with the delineation of the minimum bounding rectangle, the edge response region of rock debris particles can be quickly identified. The entire process achieves accurate and efficient extraction of the edge region of rock debris particles, providing a clear and accurate analysis range for subsequent rock debris particle contour repair and analysis, and effectively improving the accuracy and targeting of preprocessing in rock debris morphology image analysis.

[0025] S2. Perform boundary connectivity repair on adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment particles; In this embodiment of the invention, the step of repairing the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment includes: Edge discontinuity detection is performed on the edge response region to obtain the contour fracture endpoints of the rock fragments; The contour fracture endpoint with the smallest included angle between the direction vectors is taken as the endpoint pair to be connected for the rock cutting particles; Edge direction interpolation fitting is performed on the gap region between the endpoints to be connected to obtain the transition edge segment of the endpoints to be connected. The transition edge segment is merged with the edge response area to obtain the complete closed contour of the rock fragment.

[0026] The contour lines of the edge response area of ​​the rock fragments are traversed point by point for detection. The connection status of each contour pixel with the adjacent contour pixels is checked. The position where there are missing pixels and no continuous connection pixels in the contour lines is determined as the edge discontinuity point. The contour pixel at the edge discontinuity point is the contour fracture endpoint of the rock fragment.

[0027] For each contour fracture endpoint, extract the contour extension direction vector of its location. After sorting out the direction vector information of all contour fracture endpoints, calculate the angle between the direction vectors of any two contour fracture endpoints. After completing the angle calculation of all pairwise combinations of contour fracture endpoints, select the pair of contour fracture endpoints with the smallest angle value and determine the pair of endpoints to be connected for the rock debris particles.

[0028] Extract the contour trend of each of the two endpoints in the endpoint pair to be connected. Select continuous pixels in the gap area between the endpoint pairs to be connected according to the contour extension law along the trend. Connect the selected pixels in sequence according to the natural extension order of the contour to form a continuous line segment that is consistent with the original contour trend. This continuous line segment is the transition edge line segment of the endpoint pair to be connected.

[0029] The two endpoints of the transition edge segment are precisely connected to the corresponding contour break endpoints at the pixel level. All pixel information of the transition edge segment is integrated into the contour pixel information of the edge response area, so that the originally broken contour lines are seamlessly connected through the transition edge segment, and the contour of the rock fragments forms an uninterrupted closed ring structure. This closed ring structure is the complete closed contour of the rock fragments.

[0030] The beneficial effects are that the edge discontinuity detection through point-by-point traversal can accurately locate all fracture endpoints of the rock debris particle contour, providing accurate point location basis for contour repair. The selection of endpoint pairs to be connected based on the minimum angle between direction vectors can ensure the consistency of the endpoints to be connected in the contour direction, avoiding the deviation of the contour direction after repair. The transition edge segments formed by interpolation fitting based on the contour direction can be naturally connected with the original contour, improving the fit of the contour repair. The pixel-level edge fusion of the transition edge segments and the edge response area can achieve complete repair of the fracture contour and form a standard complete closed contour. This provides accurate and complete contour basis for subsequent background removal of rock debris particles and single particle image extraction, effectively improving the integrity and accuracy of rock debris particle contour extraction.

[0031] S3. Using the complete closed contour as a mask, the background of the original rock debris strip image is removed to obtain a single-particle image of the rock debris particles; In this embodiment of the invention, the step of using the complete closed contour as a mask to remove the background from the original rock debris strip image to obtain a single-particle image of the rock debris grain includes: The complete closed contour is filled with a binarized region to obtain a binarized mask layer of the rock fragments; A morphological closure operation is performed on the small cavities in the binarized mask layer to obtain a repaired mask layer for the rock debris particles; The restorative mask layer is registered pixel-by-pixel with the original rock debris strip image to obtain the mask extraction image of the rock debris particles; Isolated noise points are removed from the mask-extracted image to obtain a purified image of the rock fragments. The cleaned rock debris image is subjected to connected component separation and cropping to obtain single-particle images of the rock debris particles.

[0032] The step of removing isolated noise points from the mask-extracted image to obtain a purified image of the rock fragments includes: Topological labeling is performed on the connected components in the mask-extracted image to obtain a connected component labeling image of the rock debris particles; The number of pixels in the connected component is measured to obtain the scale representation value of the connected component. Based on the scale characterization value, scale-level stripping is performed on the connected components to obtain the isolated noise connected domains of the rock debris particles. The isolated noise connected components are removed from the connected component labeled image to obtain the purified rock debris image of the rock debris particles.

[0033] The complete closed outline of the rock fragments is binarized into pixel regions. All pixels in the inner region surrounded by the complete closed outline are set to the same gray value, while all pixels outside the complete closed outline are set to another gray value, thus achieving a clear distinction between the two gray values. Then, the inner region of the complete closed outline is filled with pixels across the entire range. After the filling is completed, a binarized mask layer of rock fragments is formed.

[0034] A morphological closing operation is performed on the binarized mask layer. First, pixel dilation is applied to the layer to extend the pixels at the edges of small holes towards the center of the holes, thus covering the pixels of the small holes. Then, pixel erosion is applied to the dilated layer to restore the outline pixels of the layer to their original positions before dilation. By performing dilation followed by erosion, all the small holes in the layer are filled, while maintaining the overall outline and size of the mask, resulting in a repair mask layer for rock debris particles.

[0035] The restoration mask layer and the original rock debris strip image are precisely aligned in pixel coordinates to form a one-to-one correspondence between the coordinate positions of all pixels in the two layers. All pixel information in the original rock debris strip image corresponding to the gray values ​​of the rock debris particle area in the restoration mask layer is extracted. The pixel information corresponding to the gray values ​​of the background area in the restoration mask layer is masked. The extracted pixel information is rearranged according to its original coordinate position to obtain the mask extraction image of rock debris particles.

[0036] Topological structure labeling is performed on all interconnected pixel regions (connected components) in the mask-extracted image. A unique label is assigned to each connected component, and this label is associated with all pixels of the corresponding connected component. This allows each connected component to be accurately identified through its unique label. After labeling all connected components, a connected domain labeling image of rock debris particles is obtained.

[0037] The number of pixels in each connected component with unique label information in the connected component labeled image is counted. All pixels contained in each connected component are counted point by point, and the counted number of pixels is used as the scale representation value of the corresponding connected component. This completes the scale measurement of all connected components in the mask extraction image.

[0038] The scale representation values ​​of all connected components are sorted and numerically sorted. Based on the sorting results, all connected components are divided into scale levels. Connected components with scale representation values ​​at the lowest level are selected from all connected components. These selected connected components are the isolated noise connected domains of rock debris particles.

[0039] Based on the unique labeling information of isolated noise connected regions in the connected region labeled image, the pixel positions of all isolated noise connected regions are accurately located. All pixel information at these positions is removed from the connected region labeled image. The pixel coordinate positions of the image after noise removal are calibrated to maintain the original arrangement structure of the remaining pixels, thus obtaining the purified rock debris image.

[0040] All connected components in the purified rock debris image are identified one by one, and an independent contour boundary range is defined for each connected component. The boundary pixel coordinates of each connected component are determined. The purified rock debris image is precisely cropped according to the defined boundary range, and each connected component is completely separated from the overall image to form an independent image corresponding to each connected component. This independent image is the single-particle image of the rock debris particle.

[0041] The beneficial effects are as follows: by filling the binarized region of the complete closed contour and performing morphological closing operations, a repair mask layer without small holes was constructed, providing a precise and complete mask basis for background removal. Pixel-by-pixel registration extraction achieved accurate separation of the rock debris grain region from the background region, effectively removing background pixels from the original image. Through a series of operations such as topological marking and scale measurement, isolated noise points were accurately identified and removed, thoroughly purifying the rock debris image and removing irrelevant interfering pixels. Finally, connected component separation and cropping achieved accurate extraction of individual rock debris grains. The entire process achieved efficient and thorough background removal, ensuring the purity, integrity, and independence of individual grain images. This provides a high-quality image foundation for subsequent morphological feature studies such as the angular feature encoding and directional group analysis of rock debris grains, effectively improving the accuracy and effectiveness of subsequent rock debris morphology analysis.

[0042] S4. Encode the angular features of the contour edges of the single-particle image to construct the angular distribution map of the rock debris particles; In this embodiment of the invention, the step of encoding the angular features of the contour edges of the single-particle image to construct an angular distribution map of the rock debris particles includes: The single-particle image is traced to obtain the continuous edge contour lines of the rock debris particles; The continuous edge contour line is resampled with equal arc length contour points to obtain a discrete contour point sequence of the rock fragments. Local orientation analysis is performed on adjacent contour points in the discrete contour point sequence to obtain the contour curvature feature value of the contour points; Based on the contour curvature feature value, the contour points are calibrated for edge angle level to obtain the edge angle feature coding sequence of the rock fragments; Using the circumferential position of the contour points as the horizontal axis of the map and the corner feature encoding value as the vertical axis of the map, the corner feature encoding sequence is mapped to obtain the corner distribution map of the rock debris particles.

[0043] A pixel-by-pixel traversal tracking method is used to trace the contour edges of a single particle image. Starting from any edge pixel of the rock fragment particle contour, all edge pixels are traversed sequentially along the adjacent connection direction of the contour pixels. The coordinate information of each traversed edge pixel is recorded in real time. All the recorded coordinate information is drawn into continuous lines according to the traversal order to form a line structure that fits the edge of the rock fragment particle without interruption. This line structure is the continuous edge contour line of the rock fragment particle.

[0044] Starting with the first pixel of the continuous edge contour line as the sampling start position, contour pixels are selected along the natural extension direction of the continuous edge contour line at fixed arc length intervals. During the selection process, the arc length distance between any two adjacent sampling points is exactly the same. All contour pixels that meet the arc length interval requirements on the continuous edge contour line are extracted in sequence. These sampled contour pixels are then arranged in order according to their sequence on the continuous edge contour line to form an ordered set of pixels. This set is the discrete contour point sequence of rock debris particles.

[0045] For each contour point in the discrete contour point sequence, select its preceding and following adjacent contour points in the sequence, associate and integrate the coordinate information of these three contour points, analyze the degree of change of the direction of the polyline formed by these three points, determine the contour curvature at the central contour point based on the curvature of the polyline, and convert the contour curvature into the corresponding numerical representation result, which is the contour curvature characteristic value of the contour point.

[0046] The numerical values ​​of the contour curvature feature values ​​of all contour points are analyzed. Based on the magnitude of the contour curvature feature values, corresponding edge levels are divided. A unique fixed code value is assigned to each edge level. The contour curvature feature value of each contour point is matched to the corresponding edge level division interval to determine the edge level corresponding to each contour point. Then, the edge level is converted into the corresponding code value. The code values ​​of all contour points are combined in an orderly manner according to the original arrangement order of the discrete contour point sequence to form an ordered code set. This set is the edge feature code sequence of rock debris particles.

[0047] Using the circumferential position of each contour point in the discrete contour point sequence of rock fragments on the continuous edge contour line as the horizontal axis of the atlas, the position scale of the horizontal axis is marked sequentially according to the circumferential order of the contour points on the continuous edge contour line. Using the corner feature code value as the vertical axis of the atlas, the numerical scale of the vertical axis is marked sequentially according to the magnitude of the corner feature code value. Each code value in the corner feature code sequence is matched one-to-one with the circumferential position of its corresponding contour point, and the corresponding point of each code value is marked at the corresponding coordinate position of the atlas. All points are connected sequentially according to the circumferential order of the contour points to form a visualized atlas structure, which is the corner distribution atlas of rock fragments.

[0048] The beneficial effects are as follows: contour edge tracking can accurately obtain continuous edge contour lines that fit the edges of rock fragments, laying a precise contour foundation for the analysis of angular features; equal arc length resampling ensures the uniformity of the distribution of discrete contour point sequences, guaranteeing the consistency and fairness of subsequent angular feature analysis; local orientation analysis quantifies the contour curvature into contour curvature feature values, realizing the numerical representation of rock fragment angular features; angular level calibration and coding based on contour curvature feature values ​​form a standardized angular feature coding sequence, making the angular features of rock fragments easier to analyze and match in subsequent processes; and map mapping transforms the coding sequence into a visualized angular distribution map, which can intuitively and clearly present the angular feature distribution of rock fragment contours at various circumferential positions. The entire process realizes the systematic, quantitative, and visualized extraction and representation of rock fragment angular features, accurately capturing the detailed features of rock fragment angular morphology, providing accurate, intuitive, and standardized feature basis for subsequent rock fragment morphology matching, and effectively improving the detail, accuracy, and comparability of rock fragment morphology feature analysis.

[0049] S5. Perform long axis direction comparison on the single particle images, group single particle images with similar comparison results into orientation groups of the rock cutting particles, and quantify the dominant direction vector of the orientation group; In this embodiment of the invention, the step of performing long-axis direction comparison on the single-particle images, grouping single-particle images with similar comparison results into orientation groups of the rock debris particles, and quantifying the dominant direction vector of the orientation group includes: The principal axis orientation of the single-particle image is extracted to obtain the individual principal direction vector of the rock cutting particle; By mapping the principal direction vector of the individual rock fragments to angle values, the directional angle values ​​of the rock fragments are obtained. Density peak clustering is performed on the directional angle values ​​to obtain the orientation groups of the rock fragments; The dominant direction vector of the directional group is obtained by superimposing the individual principal direction vectors of the rock fragments in the directional group.

[0050] The step of performing density peak clustering on the directional angle values ​​to obtain the orientation groups of the rock debris particles includes: The distribution range of the direction angle value is analyzed to obtain the neighborhood cutoff distance of the direction angle value; The frequency of neighborhood points is statistically analyzed for the direction angle value to obtain the local density value of the direction angle value; wherein, the calculation formula for the local density value is as follows: ; in, Indicates the first Local density values ​​for each directional angle. Indicates the first Each direction angle value, Indicates the first Each direction angle value, Indicates the neighborhood cutoff distance, This represents the summation function. Indicates an indicator function, when hour, ,otherwise ; Based on the local density value, minimum distance addressing is performed on the direction angle value to obtain the relative distance value of the direction angle value; Based on the local density value and the relative distance value, cluster centers are identified for the direction angle value to obtain the cluster center point of the direction angle value; Using the cluster center point as the core, the directional angle values ​​are clustered and assigned to obtain the orientation groups of the rock fragments.

[0051] A full-dimensional geometric morphology analysis is performed on the contour of rock fragments corresponding to single-particle images. First, the geometric center of the rock fragment is determined. A Cartesian coordinate system is built with the geometric center as the origin. The coordinate distribution information of all pixels on the rock fragment contour relative to the geometric center is sorted out. The geometric axis that can penetrate the rock fragment contour and cover the longest spatial range of pixels is found, which is the principal axis. The planar extension direction and orientation of the principal axis are determined. The orientation feature of the principal axis is transformed into a vector form with directional attributes and amplitude features. This vector is the individual principal direction vector of the rock fragment.

[0052] Using the Cartesian coordinate system where the individual principal direction vector is located as the measurement reference, the positive direction of the horizontal axis of the coordinate system is set as the reference direction for angle measurement. The plane angle between the individual principal direction vector and the reference direction is accurately calculated, and the vector direction characteristics of the individual principal direction vector are completely converted into the corresponding angle value. This angle value is the direction angle value of the rock fragment.

[0053] Collect the directional angle values ​​corresponding to all rock fragments and organize them into a complete numerical set. Calculate the maximum and minimum values ​​of the directional angle values ​​in this numerical set, and calculate the numerical difference between the maximum and minimum values ​​to determine the overall distribution range of the directional angle values. Based on the numerical span of this distribution range, determine a fixed angle threshold, which is the neighborhood cutoff distance of the directional angle values.

[0054] For each directional angle value in the numerical set, with that directional angle value as the center, the neighborhood cutoff distance is used as the boundary of the neighborhood range. The number of other directional angle values ​​in the numerical set that are within the neighborhood range of the central angle value is counted one by one. The counted number is directly used as the local density value corresponding to the central directional angle value. The frequency statistics of neighborhood points of all directional angle values ​​are completed in sequence, and the corresponding local density values ​​are obtained.

[0055] No. The local density values ​​of the directional angle values ​​are derived from the results of neighborhood frequency statistics for all directional angle values. The directional angle values ​​are derived from the directional angle values ​​obtained by mapping the principal direction vector of individual rock fragment particles, and are angular representations of the directional characteristics of individual rock fragment particles. The directional angle values ​​are derived from the set of directional angle values ​​obtained by mapping the principal direction vectors of individual rock fragments. These are used to count the frequency of other directional angle values ​​in the neighborhood. The neighborhood cutoff distance is derived from the distribution range analysis of all directional angle values ​​and is the boundary of the neighborhood range determined based on the overall distribution range of the directional angle values. The summation function is an operation that iterates through and counts all other directional angle values, used to accumulate the number of directional angle values ​​in the neighborhood that meet the conditions. The indicator function is a judgment function used to determine whether a directional angle value is within the neighborhood range. When the neighborhood condition is met, a valid count is output; otherwise, an invalid count is output.

[0056] The significance of this calculation method is based on the first... Centered on the first directional angle value, count the number of other directional angle values ​​within the neighborhood cutoff distance range. This count is the number of directional angle values. The local density value of a direction angle is used to characterize the density of that direction angle value in the overall distribution. The higher the density value, the more direction angle values ​​are gathered near that direction angle value, which means that the rock debris particles are more concentrated in that direction.

[0057] When the neighborhood cutoff distance is fixed, the first The local density value of the i-th directional angle value increases with the increase of the number of other directional angle values ​​in the neighborhood, and decreases when the number of other directional angle values ​​in the neighborhood decreases. When the i-th... When there are no other directional angle values ​​in the neighborhood of a given directional angle value, the local density value is zero, indicating that the directional angle value is isolated in the overall distribution. As the neighborhood cutoff distance increases, the neighborhood range of each directional angle value expands, and the corresponding local density value increases accordingly. Conversely, as the neighborhood cutoff distance decreases, the neighborhood range shrinks, and the corresponding local density value decreases accordingly.

[0058] For each directional angle value, firstly, select all directional angle values ​​in the numerical set whose local density value is greater than the local density value of the current analysis angle value. Calculate the angle difference between the current directional angle value and these selected angle values. Select the smallest difference from all calculated angle differences. This difference is the relative distance value of the current directional angle value. If there is no directional angle value with a larger local density value in the numerical set, then the maximum angle difference between the current directional angle value and all other angle values ​​in the numerical set is taken as its relative distance value.

[0059] A comprehensive analysis is conducted on the local density and relative distance values ​​of all directional angle values ​​in the numerical set. The characteristic performance of the two values ​​is considered simultaneously. Directional angle values ​​with outstanding local density and relative distance are selected. These angle values ​​are located in the core distribution position in the numerical set, and the directional angle values ​​in these core positions are determined as the cluster centers of directional angle values.

[0060] Using each cluster center point as the core attribution benchmark, the neighborhood cutoff distance is used as the boundary of the attribution determination range. All directional angle values ​​in the numerical set that are within the neighborhood of the cluster center point are uniformly merged into the group corresponding to the cluster center point. The attribution determination and assignment of all cluster centers are completed in sequence to ensure that each directional angle value in the numerical set is assigned to the corresponding unique group. The set of single-particle images of rock debris corresponding to each group is the orientation group of rock debris particles.

[0061] For each rock fragment orientation group, the principal direction vectors of all individual rock fragments within the group are collected. All individual principal direction vectors are placed in the same Cartesian coordinate system. With the origin of the coordinate system as the vector superposition center, a comprehensive vector field superposition calculation is performed on the directional attributes and amplitude characteristics of all individual principal direction vectors. The resultant vector obtained after superposition is normalized to form a vector that can represent the overall directional characteristics of the orientation group. This vector is the dominant direction vector of the orientation group.

[0062] The beneficial effects are as follows: by extracting the principal axis orientation, the long axis direction of rock fragments is transformed into a quantified individual principal direction vector, realizing a precise numerical representation of the individual directional characteristics of rock fragments. Angle value mapping further transforms the vector characteristics into easily analyzable angle values. Density peak clustering achieves precise merging of rock fragments with similar directional characteristics through multi-dimensional numerical analysis, making the division of directional groups conform to the actual directional distribution law of rock fragments. The dominant direction vector obtained by vector field superposition accurately represents the overall directional trend of each directional group. The entire process completes the systematic and quantitative analysis of rock fragments from individual direction to group directional distribution, accurately capturing the directional distribution characteristics and laws of rock fragments, providing a standardized and comparable directional feature basis for subsequent rock fragment morphology matching, effectively enriching the feature dimensions of rock fragment morphology analysis, and further improving the comprehensiveness and accuracy of rock fragment morphology analysis.

[0063] S6. The dominant direction vector and the angular distribution map are morphologically matched with the standard rock debris map of the rock debris particles to obtain the rock debris morphology analysis results of the rock debris sample.

[0064] In this embodiment of the invention, the step of performing morphological matching between the dominant direction vector and the angular distribution map and the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphology analysis results of the rock fragment sample includes: The dominant direction vector and the edge angle distribution map are fused to obtain the feature vector to be matched for the rock fragments; Based on the feature vector to be matched, a similarity search is performed in the standard rock debris map library of the rock debris particles to obtain the candidate matching map of the rock debris particles; The candidate matching map and the feature vector to be matched are compared with local similarity to obtain the local similarity sequence of the rock fragments; Based on the local fit sequence, the candidate matching maps are evaluated for comprehensive confidence, and the rock fragment morphology analysis results of the rock fragment samples are obtained.

[0065] The directional representation information of the dominant directional vector is extracted, and then the edge feature encoding information corresponding to each circumferential position in the edge angle distribution map is extracted. The directional representation information is used as the starting part of the feature vector, and then all edge feature encoding information is sequentially spliced ​​in the circumferential order of the edge angle distribution map to form an ordered set of feature information. This set is the feature vector to be matched for the rock debris particles.

[0066] The feature vector to be matched is compared one by one with the corresponding feature vector of each standard rock debris map in the standard rock debris map library. During the comparison, the matching of the corresponding feature information is checked sequentially at each position of the feature vector. The number of matching feature points between the feature vector of each standard rock debris map and the feature vector to be matched is counted. All standard rock debris maps are sorted according to the number of matching feature points, and the top-ranked standard rock debris maps are selected as candidate matching maps for rock debris particles.

[0067] The corresponding feature vectors of the feature vector to be matched and the candidate matching map are both divided into several local feature segments. Each local feature segment corresponds to a circumferential local region of the rock fragment contour. For each candidate matching map, each local feature segment is compared with the corresponding local feature segment of the feature vector to be matched one by one. The number of matching feature points in each local feature segment is counted. The matching status of each local feature segment is converted into the corresponding consistency representation. The consistency representations of all local feature segments are arranged in an orderly manner in circumferential order to form a local consistency sequence of rock fragments.

[0068] For each candidate matching map corresponding to a local consistency sequence, all local consistency representations in the sequence are accumulated and integrated to obtain the comprehensive confidence representation of the candidate matching map. The comprehensive confidence representations of all candidate matching maps are sorted, and the candidate matching map with the highest comprehensive confidence representation is selected. The rock fragment morphology information corresponding to the candidate matching map is used as the rock fragment morphology analysis result of the rock fragment sample.

[0069] The beneficial effects are as follows: by fusing the dominant direction vector with the angular distribution map, the dimensions of rock fragment morphology features are enriched, enabling the feature vector to be matched to simultaneously reflect the directional distribution features and angular details of rock fragments. Similarity retrieval in the standard rock fragment map library can quickly narrow down the matching range and improve matching efficiency. Local fit comparison can accurately capture the morphological matching of various local areas of rock fragments, avoiding overall matching deviation. The comprehensive confidence assessment comprehensively considers the local fit, ensuring the accuracy and reliability of the final rock fragment morphology analysis results. The entire process achieves accurate and efficient matching of rock fragment morphology, providing standardized and quantifiable results for the morphological analysis of rock fragment samples, and effectively improving the accuracy and application value of rock fragment morphology analysis.

[0070] like Figure 2 The diagram shown is a functional block diagram of an online rock debris morphology analysis system based on image recognition, provided in an embodiment of the present invention.

[0071] The image recognition-based online rock debris morphology analysis system 100 of this invention can be installed in an electronic device. Depending on the functions implemented, the image recognition-based online rock debris morphology analysis system 100 may include an image acquisition and edge response extraction module 101, a contour connectivity repair module 102, a single-particle image extraction module 103, an edge feature encoding module 104, an orientation group identification and dominant direction quantification module 105, and a morphology matching analysis module 106. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, stored in the memory of the electronic device.

[0072] In this embodiment, the functions of each module / unit are as follows: The image acquisition and edge response extraction module 101 is used to perform linear array scanning on the acquired rock debris sample and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. The contour connectivity repair module 102 is used to repair the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment. The single-particle image extraction module 103 is used to use the complete closed contour as a mask to remove the background from the original rock debris strip image to obtain a single-particle image of the rock debris particle. The corner feature encoding module 104 is used to encode the corner features of the contour edge of the single particle image to construct the corner distribution map of the rock debris particles. The orientation group identification and dominant direction quantification module 105 is used to compare the single particle images along their long axis, group single particle images with similar comparison results into orientation groups of the rock debris particles, and quantify the dominant direction vector of the orientation group. The morphological matching analysis module 106 is used to perform morphological matching between the dominant direction vector and the angular distribution map and the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphological analysis results of the rock fragment sample.

[0073] In the several embodiments provided by this invention, it should be understood that the disclosed methods and systems can be implemented in other ways. For example, the system embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0074] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0075] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0076] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0077] This application embodiment can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0078] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. An online method for analyzing rock debris morphology based on image recognition, characterized in that, The method includes: S1. Perform linear array scanning on the collected rock debris samples and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. S2. Perform boundary connectivity repair on adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment particles; S3. Using the complete closed contour as a mask, the background of the original rock debris strip image is removed to obtain a single-particle image of the rock debris particles; S4. Encode the angular features of the contour edges of the single-particle image to construct the angular distribution map of the rock debris particles; S5. Perform long axis direction comparison on the single particle images, group single particle images with similar comparison results into orientation groups of the rock cutting particles, and quantify the dominant direction vector of the orientation group; S6. The dominant direction vector and the angular distribution map are morphologically matched with the standard rock debris map of the rock debris particles to obtain the rock debris morphology analysis results of the rock debris sample.

2. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The process of performing linear array scanning on the collected rock debris samples and grayscale gradient detection on the original rock debris strip images obtained from the scanning to obtain the edge response regions of the rock debris particles in the rock debris samples includes: The collected rock debris samples were dispersed and spread out, and the resulting sample distribution layer was subjected to linear array push-broom imaging to obtain the original rock debris strip image of the rock debris sample. Gradient magnitude reconstruction is performed on the pixels in the original rock debris strip image to obtain the edge intensity characterization of the pixels; Based on the edge intensity characterization, local extremum screening is performed on the pixels to obtain candidate edge points of rock debris particles in the rock debris sample; Eight-neighbor connectivity analysis is performed on the candidate edge points to merge interconnected candidate edge points into connected regions. The connected regions are then delineated using the minimum bounding rectangle to obtain the edge response region of the rock fragments.

3. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The step of repairing the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment includes: Edge discontinuity detection is performed on the edge response region to obtain the contour fracture endpoints of the rock fragments; The contour fracture endpoint with the smallest included angle between the direction vectors is taken as the endpoint pair to be connected for the rock cutting particles; Edge direction interpolation fitting is performed on the gap region between the endpoints to be connected to obtain the transition edge segment of the endpoints to be connected. The transition edge segment is merged with the edge response area to obtain the complete closed contour of the rock fragment.

4. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The step of using the complete closed contour as a mask to remove the background from the original rock debris strip image to obtain a single-particle image of the rock debris grain includes: The complete closed contour is filled with a binarized region to obtain a binarized mask layer of the rock fragments; A morphological closure operation is performed on the small cavities in the binarized mask layer to obtain a repaired mask layer for the rock debris particles; The restorative mask layer is registered pixel-by-pixel with the original rock debris strip image to obtain the mask extraction image of the rock debris particles; Isolated noise points are removed from the mask-extracted image to obtain a purified image of the rock fragments. The cleaned rock debris image is subjected to connected component separation and cropping to obtain single-particle images of the rock debris particles.

5. The online analysis method for rock debris morphology based on image recognition as described in claim 4, characterized in that, The step of removing isolated noise points from the mask-extracted image to obtain a purified image of the rock fragments includes: Topological labeling is performed on the connected components in the mask-extracted image to obtain a connected component labeling image of the rock debris particles; The number of pixels in the connected component is measured to obtain the scale characterization value of the connected component. Based on the scale characterization value, scale-level stripping is performed on the connected components to obtain the isolated noise connected domains of the rock fragments. The isolated noise connected components are removed from the connected component labeled image to obtain the purified rock debris image of the rock debris particles.

6. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The step of encoding the angular features of the contour edges of the single-particle image to construct an angular distribution map of the rock debris particles includes: The single-particle image is traced to obtain the continuous edge contour lines of the rock debris particles; The continuous edge contour line is resampled with equal arc length contour points to obtain a discrete contour point sequence of the rock fragments. Local orientation analysis is performed on adjacent contour points in the discrete contour point sequence to obtain the contour curvature feature value of the contour points. Based on the contour curvature feature value, the contour points are calibrated for edge angle level to obtain the edge angle feature coding sequence of the rock debris particles; Using the circumferential position of the contour points as the horizontal axis of the map and the corner feature encoding value as the vertical axis of the map, the corner feature encoding sequence is mapped to obtain the corner distribution map of the rock debris particles.

7. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The step of performing long-axis alignment on the single-particle images, grouping single-particle images with similar alignment results into orientation groups for the rock debris particles, and quantifying the dominant direction vector of the orientation group includes: The principal axis orientation of the single-particle image is extracted to obtain the individual principal direction vector of the rock cutting particle; By mapping the principal direction vector of the individual rock fragments to angle values, the directional angle values ​​of the rock fragments are obtained. Density peak clustering is performed on the directional angle values ​​to obtain the orientation groups of the rock fragments; The dominant direction vector of the directional group is obtained by superimposing the individual principal direction vectors of the rock fragments in the directional group.

8. The online analysis method for rock debris morphology based on image recognition as described in claim 7, characterized in that, The step of performing density peak clustering on the directional angle values ​​to obtain the orientation groups of the rock debris particles includes: The distribution range of the direction angle value is analyzed to obtain the neighborhood cutoff distance of the direction angle value; The frequency of neighborhood points is statistically analyzed for the direction angle value to obtain the local density value of the direction angle value; wherein, the calculation formula for the local density value is as follows: ； in, Indicates the first Local density values ​​for each directional angle. Indicates the first Each direction angle value, Indicates the first Each direction angle value, Indicates the neighborhood cutoff distance, This represents the summation function. Indicates an indicator function, when hour, ,otherwise ; Based on the local density value, minimum distance addressing is performed on the direction angle value to obtain the relative distance value of the direction angle value; Based on the local density value and the relative distance value, cluster centers are identified for the direction angle value to obtain the cluster center point of the direction angle value; Using the cluster center point as the core, the directional angle values ​​are clustered and assigned to obtain the orientation groups of the rock fragments.

9. The online analysis method for rock debris morphology based on image recognition as described in claim 1, characterized in that, The step of matching the dominant direction vector and the angular distribution map with the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphology analysis results of the rock fragment sample includes: The dominant direction vector and the edge angle distribution map are fused to obtain the feature vector to be matched for the rock fragments; Based on the feature vector to be matched, a similarity search is performed in the standard rock debris map library of the rock debris particles to obtain the candidate matching map of the rock debris particles; The candidate matching map and the feature vector to be matched are compared with local similarity to obtain the local similarity sequence of the rock fragments; Based on the local fit sequence, the candidate matching maps are evaluated for comprehensive confidence, and the rock fragment morphology analysis results of the rock fragment samples are obtained.

10. An online rock debris morphology analysis system based on image recognition, characterized in that, The system for implementing the online rock cuttings morphology analysis method based on image recognition as described in claim 1 includes: The image acquisition and edge response extraction module is used to perform linear array scanning on the acquired rock debris samples and perform grayscale gradient detection on the original rock debris strip image obtained by scanning to obtain the edge response region of the rock debris particles in the rock debris sample. The contour connectivity repair module is used to repair the boundary connectivity of adjacent response regions in the edge response region to obtain the complete closed contour of the rock fragment particles. The single-particle image extraction module is used to remove the background from the original rock debris strip image by using the complete closed contour as a mask, so as to obtain the single-particle image of the rock debris particle. An edge feature encoding module is used to encode the edge features of the contour edges of the single particle image in order to construct an edge angle distribution map of the rock debris particles; The orientation group identification and dominant direction quantification module is used to compare the single particle images along their long axis, group single particle images with similar comparison results into orientation groups of the rock cutting particles, and quantify the dominant direction vector of the orientation group. The morphological matching analysis module is used to perform morphological matching between the dominant direction vector and the angular distribution map and the standard rock fragment map of the rock fragment particles to obtain the rock fragment morphological analysis results of the rock fragment sample.

Images