An automatic recognition system for drilling core images based on computer vision

Abstract

The invention discloses a computer vision-based automatic recognition system for drilling core images, relates to the field of image recognition, and comprises a data acquisition module, a preprocessing module, a database module, a data processing module, a verification module, and a data output module; the data processing module is used for extracting target feature data of an image; the target feature data is analyzed and an image recognition report is generated; the verification module is used for verifying the image recognition report and generating a verification feedback report; the invention firstly acquires a core image, analyzes the particle features and texture features of the surface of the core according to the core image information, and on the premise of ensuring the integrity of the core features, samples are respectively taken in the feature-dense area and the sparse area of ​​the core to judge the core type; finally, the core recognition result is verified by the core cross-section data, which can greatly reduce the probability of misjudgment of the core and greatly improve the automation level and recognition accuracy of the core analysis.

Description

Technical Field

[0001] The present invention relates to the field of image recognition, and in particular to an automatic recognition system for drilling core images based on computer vision. Background Art

[0002] Rock cores are key physical evidence for studying crustal structure, earth history, and mineral resource distribution. Through the identification and analysis of rock cores, we can gain a deeper understanding of the internal structure and evolution of the earth; it is not only the basis for geological exploration and mineral resource assessment, but also an important means to promote geological science research and technological development. At present, rock core identification mostly relies on the work experience of staff. Some identification equipment identifies the type of rock core by comparing the collected pictures of the rock core; however, the types of rock cores are diverse, and the appearance of some rock cores is relatively similar, resulting in a high misjudgment rate of the equipment and poor reliability. In order to improve the efficiency and reliability of rock core identification, an automatic recognition system for drilling core images based on computer vision is proposed. Summary of the invention

[0003] The technical problem to be solved by the present invention is: how to solve the problem that the existing safety protection system has a single protection type, resulting in poor protection effect and bringing certain impacts on the use of the safety protection system. A drilling core image automatic recognition system based on computer vision is provided.

[0004] The present invention solves the above technical problems through the following technical solutions, which include a data acquisition module, a preprocessing module, a database module, a data processing module, a verification module, and a data output module;

[0005] The data acquisition module is used to collect real-time image data of the core;

[0006] The preprocessing module is used to receive real-time image data, and perform denoising and enhancement on the real-time image data;

[0007] The database module is used to store standard characteristic data and historical data information of different cores;

[0008] The data processing module is used to receive real-time image data and extract the target feature data of the image; then import the standard feature data and historical data information, analyze the target feature data, and generate an image recognition report;

[0009] The verification module is used to receive the image recognition report and historical data information, process and analyze the historical data information, verify the image recognition report, and generate a verification feedback report;

[0010] The data output module is used to receive and output image recognition reports and verification feedback reports.

[0011] Preferably, the processing process of the preprocessing module is:

[0012] Acquire real-time image data;

[0013] Divide the image evenly into multiple block units;

[0014] Use the multithreaded programming library to create threads;

[0015] Allocate the divided block units evenly to each thread;

[0016] In each thread, the mean filtering algorithm and the histogram equalization algorithm are applied to the assigned block units for processing;

[0017] After all threads have completed processing, the processed block units are merged into a complete image and output.

[0018] Preferably, the historical data information includes historical characteristic data of each core, and the processing process of the data processing module is:

[0019] Get real-time image data;

[0020] Marking multiple unit areas at different positions of the real-time image, collecting image data in the unit areas and generating multiple target feature data;

[0021] Import standard characteristic data of different cores;

[0022] Count the number of target feature data A1;

[0023] Calculate the proportion of target characteristic data in the standard characteristic data of different cores, and use the core with the largest proportion as the pre-selected core;

[0024] Randomly import historical characteristic data of a set of pre-selected cores;

[0025] Counting the number of historical characteristic data of the pre-selected core A2, and then counting the number of occurrences of each characteristic data in the historical characteristic data A3;

[0026] Use the formula F=A3 / A2 to calculate the frequency F of each feature data;

[0027] When the frequency F ≥ the preset threshold Q1, it is defined as general characteristic data;

[0028] When the preset threshold Q2≤frequency F<preset threshold Q1, it is defined as accidental feature data;

[0029] When the frequency F is less than the preset threshold Q2, the characteristic data is cleared;

[0030] Counting the number of general feature data A4 and the number of accidental feature data A5 in the target feature data;

[0031] Calculate the comprehensive feature score E of the target feature data. The specific calculation process is:

[0032] ;

[0033] in, is the proportion of general feature data in the comprehensive feature score E, is the proportion of accidental feature data in the comprehensive feature score E, ;

[0034] When the comprehensive feature score E≥preset threshold Q3, the pre-selected core is changed to the output core, and an image recognition report is generated;

[0035] When the comprehensive characteristic score E is less than the preset threshold value Q3, the data of the pre-selected core is cleared.

[0036] Preferably, the real-time image data includes real-time size information, and the process of generating the target feature data further includes:

[0037] Randomly mark a rectangular area and count the number of particles in the rectangular area B1;

[0038] Randomly select n particles, and establish the circumscribed circle Y1 and the inscribed circle Y2 of each particle respectively, and obtain the radius size R1 of the circumscribed circle Y1 and the radius size R2 of the inscribed circle Y2;

[0039] Use the formula R3=(R1+R2) / 2 to find the particle radius;

[0040] Calculate the average particle radius of the n particles ;

[0041] Using the formula Calculate the average area S1 of the particles;

[0042] Get the size of the rectangular area and calculate the area S2 of the rectangular area;

[0043] Use the formula K=nS1 / S2 to calculate the ratio K of the total area of ​​the particle to the area of ​​the rectangular area;

[0044] Set the total area of ​​particles nS1 ≥ the preset threshold Q3, and obtain the minimum area of ​​the unit area KQ3;

[0045] Set the shape of the unit area to a rectangle and get the size of the unit area;

[0046] Randomly obtain multiple unit areas and collect image data in the unit areas;

[0047] Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

[0048] Preferably, the real-time image data also includes real-time texture data, and the process of generating the target feature data also includes:

[0049] Acquire real-time image data and draw a virtual straight line L parallel to the axis of the core in the height direction;

[0050] Acquire real-time texture data and mark the intersections between the virtual straight line and the texture. The marked points are C1, C2, C3, ..., CV, where V is the number of marked points.

[0051] Divide the virtual straight line into X sub-segments L` with a preset size;

[0052] Count the number of marked points on each sub-segment L'. The number of marked points on each sub-segment L' is , , ,..., ;

[0053] Calculate the difference P of the marked points on two adjacent sub-segments. The specific calculation process is:

[0054] ;

[0055] When the difference P ≤ the preset threshold Q4, the two sub-lines L` are merged;

[0056] When the difference P>preset threshold Q4, the two sub-lines L' are split to obtain multiple sub-cores;

[0057] Then, multiple unit areas are randomly obtained on each sub-core, and image data within the unit areas are collected;

[0058] Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

[0059] Preferably, when the difference P>preset threshold Q4, the process of generating the target feature data further includes:

[0060] Select any two adjacent marking points in the sub-core, and make a standard circle along the outer wall of the sub-core through the two marking points;

[0061] Get the local stripes between two standard circles;

[0062] Obtain the arc length Lr of the local stripes and the radius R4 of the core;

[0063] Calculate the radian number β of the central angle corresponding to the local stripes. The specific calculation process is:

[0064] ;

[0065] When the number of radians β> the preset threshold Q5, the number of marking points on the atomic core is increased by one.

[0066] Preferably, the real-time image data also includes real-time cross-section information, and the historical data information also includes historical cross-section dispersion data. The specific processing process of the verification module is:

[0067] Import image recognition reports to obtain real-time cross-section information of output cores;

[0068] Obtain the contour line of the core section, randomly select three non-overlapping first virtual points W on the contour line, and make a first virtual plane M through the three first virtual points W;

[0069] S reference points are randomly selected at positions within the core region on the first virtual plane M, and axial dimensions Lc1, Lc2, Lc3, ..., LcS of the S reference points from the core section are measured respectively;

[0070] Calculate the dispersion J1 of multiple axial dimensions. The specific calculation process is:

[0071] ;

[0072] in, is the average axial size, Lci is the axial size of the i-th reference point from the core section;

[0073] Then obtain multiple historical cross-section dispersion data of the output core and calculate their average value ;

[0074] Calculate the deviation Jc of the output core. The specific calculation process is:

[0075] ;

[0076] When the deviation Jc is less than or equal to the preset threshold Q6, an image recognition report is output;

[0077] When the deviation Jc> preset threshold Q6, a verification feedback report is output and a warning report is generated;

[0078] Preferably, the processing of the verification module further includes:

[0079] Obtain the contour line of the core section and the geometric center point Z1 of the contour line;

[0080] Draw a virtual line through the center point Z1, and the intersection points of the virtual line and the contour line are Z2 and Z3 respectively;

[0081] A point Z4 is randomly selected on the core, and a second virtual plane perpendicular to the axis of the core is made through Z4;

[0082] Obtain the diameter size Ld of the core, and then respectively obtain the distance sizes of the second virtual plane from the intersection point Z2 and the intersection point Z3, the two distance sizes are Lh1 and Lh2 respectively;

[0083] Calculate the inclination angle θ of the virtual line. The specific calculation process is:

[0084] ;

[0085] Draw virtual lines in different directions through the geometric center point Z1, calculate the inclination angles of the virtual lines, and obtain the average value θ1 of the multiple virtual angles;

[0086] Get three virtual points W and calculate the inclination angle μ between any two virtual points W. The specific calculation process is:

[0087] ;

[0088] in is the vertical distance between two virtual points W, is the horizontal distance between two virtual points W;

[0089] When μ>θ1, the corresponding virtual point W is deleted and reselected.

[0090] Preferably, when μ≤θ1, the processing of the verification module further includes:

[0091] Randomly select three second virtual points on the core end surface, and make the third virtual plane by making the three second virtual points;

[0092] Select the maximum and minimum points of the distance from the third virtual plane on the second virtual plane and use the preset rule to calculate the inclination η1 of the normal line of the third virtual plane, and then use the same maximum and minimum points to calculate the inclination η2 of the normal line of the first virtual plane M using the preset rule;

[0093] The default rules are:

[0094] Obtain the maximum value point and the minimum value point of the size from the target plane on the second virtual plane and calculate the difference Lm between the two;

[0095] Read the diameter size Ld of the core;

[0096] Calculate the inclination η` of the target plane normal. The specific calculation process is:

[0097] Tanη`=Ld / Lm;

[0098] When the target plane is the normal of the third virtual plane, η1 is calculated; when the target plane is the first virtual plane M, η2 is calculated;

[0099] Calculate the anomaly G of the two normal inclinations. The specific calculation process is:

[0100] G = |η1-η2| / η1;

[0101] When the abnormality G>the preset threshold Q7, the data of the first virtual plane is cleared and the virtual point W is reselected.

[0102] Preferably, the system further comprises a warning module, wherein the warning module is used to receive a verification report and to take a warning action when the verification report is a warning report.

[0103] Compared with the prior art, the present invention has the following advantages: the system first collects core images, analyzes the particle characteristics and texture characteristics of the surface of the core according to the core image information, and on the premise of ensuring the integrity of the core characteristics, samples are respectively taken in the feature-dense area and sparse area of ​​the core to determine the core type, thereby improving the reference value of the feature data; finally, the core identification result is verified by the core cross-section data, which can greatly reduce the probability of misjudgment of the core, greatly improve the automation level and identification accuracy of the core analysis, promote the progress and innovation of geological exploration technology, and make the system more reliable. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] Figure 1 It is the overall module diagram of the present invention. DETAILED DESCRIPTION

[0105] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0106] like Figure 1 As shown, this embodiment provides a technical solution: a drilling core image automatic recognition system based on computer vision, including a data acquisition module, a preprocessing module, a database module, a data processing module, a verification module, and a data output module;

[0107] The data acquisition module is used to collect real-time image data of the core;

[0108] The preprocessing module is used to receive real-time image data, and perform denoising and enhancement on the real-time image data;

[0109] The database module is used to store standard characteristic data and historical data information of different cores;

[0110] It should be noted that the standard characteristic data are characteristic specimen data prepared in advance according to the core characteristics.

[0111] The data processing module is used to receive real-time image data and extract the target feature data of the image; then import the standard feature data and historical data information, analyze the target feature data, and generate an image recognition report;

[0112] The verification module is used to receive the image recognition report and historical data information, process and analyze the historical data information, verify the image recognition report, and generate a verification feedback report;

[0113] The data output module is used to receive and output image recognition reports and verification feedback reports.

[0114] The system first collects core images, analyzes the surface particle characteristics and texture characteristics of the core based on the core image information, and on the premise of ensuring the integrity of the core characteristics, samples are taken in the feature-dense and sparse areas of the core to determine the core type, thereby improving the reference value of the feature data. Finally, the core identification results are verified through the core cross-section data, which can greatly reduce the probability of misjudgment of the core, greatly improve the automation level and identification accuracy of core analysis, promote the advancement and innovation of geological exploration technology, and make the system more reliable.

[0115] Among them, the processing process of the preprocessing module is:

[0116] Acquire real-time image data;

[0117] Divide the image evenly into multiple block units;

[0118] Use the multithreaded programming library to create threads;

[0119] It should be noted that the multi-threaded programming library can be C++'s std::thread module, Python's threading module, etc., which are existing technologies and will not be described in detail.

[0120] Allocate the divided block units evenly to each thread;

[0121] In each thread, the mean filtering algorithm and the histogram equalization algorithm are applied to the assigned block units for processing;

[0122] After all threads have completed processing, the processed block units are merged into a complete image and output.

[0123] Among them, the historical data information includes the historical characteristic data of each core, and the processing process of the data processing module is:

[0124] Get real-time image data;

[0125] Marking multiple unit areas at different positions of the real-time image, collecting image data in the unit areas and generating multiple target feature data;

[0126] Import standard characteristic data of different cores;

[0127] Count the number of target feature data A1;

[0128] Calculate the proportion of target characteristic data in the standard characteristic data of different cores, and use the core with the largest proportion as the pre-selected core;

[0129] Randomly import historical characteristic data of a set of pre-selected cores;

[0130] Counting the number of historical characteristic data of the pre-selected core A2, and then counting the number of occurrences of each characteristic data in the historical characteristic data A3;

[0131] Use the formula F=A3 / A2 to calculate the frequency F of each feature data;

[0132] When the frequency F ≥ the preset threshold Q1, it is defined as general characteristic data;

[0133] When the preset threshold Q2≤frequency F<preset threshold Q1, it is defined as accidental feature data;

[0134] When the frequency F is less than the preset threshold Q2, the characteristic data is cleared;

[0135] Counting the number of general feature data A4 and the number of accidental feature data A5 in the target feature data;

[0136] Calculate the comprehensive feature score E of the target feature data. The specific calculation process is:

[0137] ;

[0138] in, is the proportion of general feature data in the comprehensive feature score E, is the proportion of accidental feature data in the comprehensive feature score E, ;

[0139] When the comprehensive feature score E≥preset threshold Q3, the pre-selected core is changed to the output core, and an image recognition report is generated;

[0140] When the comprehensive characteristic score E is less than the preset threshold value Q3, the data of the pre-selected core is cleared.

[0141] It should be noted that the target characteristic data of the core include the characteristics of color, particles, texture, pores and cracks.

[0142] In this embodiment, the type of the core is preliminarily determined based on the image features of the core, and then the historical feature data of the core of this category is obtained. According to the frequency of occurrence of different features in the historical feature data, the features are divided into general features that appear frequently and accidental features that appear in small quantities. The general features are feature data that appear frequently in the acquisition of this type of core, and serve as the main reference for identifying the type of core. The accidental features are feature data that do not appear frequently in the acquisition of this type of core, and serve as an important reference for identifying the type of core. There are also some feature data that appear in the core with a very small probability and have great instability. Such data has no reference value and is removed. Then, the core type identified by the system is verified based on these features. Through the above technical means, the type of the core can be comprehensively identified and judged, so the accuracy and reliability of the system's identification can be improved.

[0143] In the above embodiment, the unit area is randomly marked for sampling to obtain target feature data. Due to the diverse core structures, the position distribution of different features is uneven and unstable; random sampling will reduce the reference value of the data. In order to improve the reference value of the data and the reliability of system recognition, the following further solutions are proposed:

[0144] Furthermore, the real-time image data includes real-time size information, and the process of generating the target feature data also includes:

[0145] Randomly mark a rectangular area and count the number of particles in the rectangular area B1;

[0146] Randomly select n particles, and establish the circumscribed circle Y1 and the inscribed circle Y2 of each particle respectively, and obtain the radius size R1 of the circumscribed circle Y1 and the radius size R2 of the inscribed circle Y2;

[0147] Use the formula R3=(R1+R2) / 2 to find the particle radius;

[0148] Calculate the average particle radius of the n particles ;

[0149] Using the formula Calculate the average area S1 of the particles;

[0150] Get the size of the rectangular area and calculate the area S2 of the rectangular area;

[0151] Use the formula K=nS1 / S2 to calculate the ratio K of the total area of ​​the particle to the area of ​​the rectangular area;

[0152] Set the total area of ​​particles nS1 ≥ the preset threshold Q3, and obtain the minimum area of ​​the unit area KQ3;

[0153] Set the shape of the unit area to a rectangle and get the size of the unit area;

[0154] Randomly obtain multiple unit areas and collect image data in the unit areas;

[0155] Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

[0156] In this scheme, the size of the particles and the proportion of the space occupied by the particles in the core are calculated first, and then the size of the unit area is determined based on the preset threshold of the area required for particle distribution in core identification. Through the above method, the integrity of the core surface data can be greatly preserved; at the same time, the distribution state of the particles in the core is preserved; avoid misjudgment caused by excessive limitation of samples in feature identification, and improve the accuracy of system identification.

[0157] Furthermore, the real-time image data also includes real-time texture data, and the generation process of the target feature data also includes:

[0158] Acquire real-time image data and draw a virtual straight line L parallel to the axis of the core in the height direction;

[0159] Acquire real-time texture data and mark the intersections between the virtual straight line and the texture. The marked points are C1, C2, C3, ..., CV, where V is the number of marked points.

[0160] Divide the virtual straight line into X sub-segments L` with a preset size;

[0161] Count the number of marked points on each sub-segment L'. The number of marked points on each sub-segment L' is , , ,..., ;

[0162] Calculate the difference P of the marked points on two adjacent sub-segments. The specific calculation process is:

[0163] ;

[0164] When the difference P ≤ the preset threshold Q4, the two sub-lines L` are merged;

[0165] When the difference P>preset threshold Q4, the two sub-lines L' are split to obtain multiple sub-cores;

[0166] Then, multiple unit areas are randomly obtained on each sub-core, and image data within the unit areas are collected;

[0167] Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

[0168] In this scheme, the distribution data of the core texture is obtained, and the core is divided according to the sparseness and density of the texture distribution. The unit area is marked with a more targeted scheme to obtain more valuable feature data, thereby improving the system recognition accuracy and reducing recognition errors.

[0169] Furthermore, when the difference P>preset threshold Q4, the process of generating target feature data further includes:

[0170] Select any two adjacent marking points in the sub-core, and make a standard circle along the outer wall of the sub-core through the two marking points;

[0171] Get the local stripes between two standard circles;

[0172] Obtain the arc length Lr of the local stripes and the radius R4 of the core;

[0173] Calculate the radian number β of the central angle corresponding to the local stripes. The specific calculation process is:

[0174] ;

[0175] When the number of radians β> the preset threshold Q5, the number of marking points on the atomic core is increased by one.

[0176] Some textures in the core are incomplete, resulting in unclear imprints of some textures and inability to stably identify them, which will affect the division of sub-cores. In this scheme, a standard circle is established on two adjacent marking points, and the continuity of the feature is judged based on the arc length of the local texture in the standard circle. When the arc length of the texture is long, it means that the texture is partially blurred but has not disappeared. This texture needs to be taken into consideration, which can improve the accuracy of sub-core division and thus improve the accuracy of feature data collection.

[0177] The core section has a relatively important reference value in the process of identifying the core. Due to the different results inside the core, its physical parameters such as hardness, toughness, brittleness, etc. are also quite different, and the shape of its section will be different. For example, hard and brittle rocks are more likely to form regular sections during drilling, while soft and plastic rocks may form irregular sections. Regular sections are relatively smooth. This characteristic is used to further verify the core identification results to reduce the system misjudgment rate. The following solutions are proposed:

[0178] Among them, the real-time image data also includes real-time cross-section information, and the historical data information also includes historical cross-section dispersion data. The specific processing process of the verification module is:

[0179] Import image recognition reports to obtain real-time cross-section information of output cores;

[0180] Obtain the contour line of the core section, randomly select three non-overlapping first virtual points W on the contour line, and make a first virtual plane M through the three first virtual points W;

[0181] S reference points are randomly selected at positions within the core region on the first virtual plane M, and axial dimensions Lc1, Lc2, Lc3, ..., LcS of the S reference points from the core section are measured respectively;

[0182] Calculate the dispersion J1 of multiple axial dimensions. The specific calculation process is:

[0183] ;

[0184] in, is the average axial size, Lci is the axial size of the i-th reference point from the core section;

[0185] Then obtain multiple historical cross-section dispersion data of the output core and calculate their average value ;

[0186] Calculate the deviation Jc of the output core. The specific calculation process is:

[0187] ;

[0188] When the deviation Jc is less than or equal to the preset threshold Q6, an image recognition report is output;

[0189] When the deviation Jc> preset threshold Q6, a verification feedback report is output and a warning report is generated;

[0190] This scheme first randomly selects three points on the cross section of the core to make a reference surface, selects reference points on the reference surface, calculates the distances and dispersions of multiple reference points from the core cross section, and when the distance deviations of multiple reference points from the core cross section are large, it indicates that the cross section is irregular; then the historical cross section dispersion data of the core cross section is retrieved, and the deviation between the two is calculated. When the deviation is large, it indicates that there is a large difference in the regularity of the two cross sections, and there is a large risk of misjudgment, and a warning is issued.

[0191] Furthermore, the processing of the verification module also includes:

[0192] Obtain the contour line of the core section and the geometric center point Z1 of the contour line;

[0193] Draw a virtual line through the center point Z1, and the intersection points of the virtual line and the contour line are Z2 and Z3 respectively;

[0194] A point Z4 is randomly selected on the core, and a second virtual plane perpendicular to the axis of the core is made through Z4;

[0195] Obtain the diameter size Ld of the core, and then respectively obtain the distance sizes of the second virtual plane from the intersection point Z2 and the intersection point Z3, the two distance sizes are Lh1 and Lh2 respectively;

[0196] Calculate the inclination angle θ of the virtual line. The specific calculation process is:

[0197] ;

[0198] Draw virtual lines in different directions through the geometric center point Z1, calculate the inclination angles of the virtual lines, and obtain the average value θ1 of the multiple virtual angles;

[0199] Get three virtual points W and calculate the inclination angle μ between any two virtual points W. The specific calculation process is:

[0200] ;

[0201] in is the vertical distance between two virtual points W, is the horizontal distance between two virtual points W;

[0202] When μ>θ1, the corresponding virtual point W is deleted and reselected.

[0203] Since the core section may be partially incomplete or deformed, there is a certain risk in randomly selecting virtual points to make the reference surface. Here, the selection of virtual points is verified and adjusted to avoid large deviations between the reference plane and the core section, which may affect the results of the secondary verification.

[0204] In the above verification process, there are situations where the slope of the virtual point meets the requirements, but the inclination direction of the established reference surface and the core section deviates. To solve this problem, a solution is proposed to verify through the reference surface and the normal line of the core section:

[0205] Furthermore, when μ≤θ1, the processing of the verification module also includes:

[0206] Randomly select three second virtual points on the core end surface, and make the third virtual plane by making the three second virtual points;

[0207] Select the maximum and minimum points of the distance from the third virtual plane on the second virtual plane and use the preset rule to calculate the inclination η1 of the normal line of the third virtual plane, and then use the same maximum and minimum points to calculate the inclination η2 of the normal line of the first virtual plane M using the preset rule;

[0208] The default rules are:

[0209] Obtain the maximum value point and the minimum value point of the size from the target plane on the second virtual plane and calculate the difference Lm between the two;

[0210] Read the diameter size Ld of the core;

[0211] Calculate the inclination η` of the target plane normal. The specific calculation process is:

[0212] Tanη`=Ld / Lm

[0213] When the target plane is the normal of the third virtual plane, η1 is calculated; when the target plane is the first virtual plane M, η2 is calculated;

[0214] Calculate the anomaly G of the two normal inclinations. The specific calculation process is:

[0215] G=|η1-η2| / η1

[0216] When the abnormality G>the preset threshold Q7, the data of the first virtual plane is cleared and the virtual point W is reselected.

[0217] When the virtual point selection is reasonable, the anomaly between the normal inclination angle of the core section and the inclination angle of the normal of the first virtual plane M is calculated. When the anomaly of the two is small, it means that the reference surface is reasonably selected and the verification accuracy is high. The double verification provides a guarantee for the reliability of system identification.

[0218] The system also includes a warning module, which is used to receive a verification report and take a warning action when the verification report is a warning report.

[0219] It should be noted that the average value in the present invention is obtained by dividing the sum of the quantities by the number of quantities, which is the most common average calculation scheme and is a prior art and will not be described in detail.

[0220] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0221] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0222] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A computer vision-based automatic recognition system for drilling core images, characterized in that: It includes data acquisition module, preprocessing module, database module, data processing module, verification module and data output module; The data acquisition module is used to collect real-time image data of the core; The preprocessing module is used to receive real-time image data, and perform denoising and enhancement on the real-time image data; The database module is used to store standard characteristic data and historical data information of different cores; The data processing module is used to receive real-time image data and extract the target feature data of the image; then import the standard feature data and historical data information, analyze the target feature data, and generate an image recognition report; The verification module is used to receive the image recognition report and historical data information, process and analyze the historical data information, verify the image recognition report, and generate a verification feedback report; The data output module is used to receive and output image recognition reports and verification feedback reports; The historical data information includes the historical characteristic data of each core, and the processing process of the data processing module is: Get real-time image data; Marking multiple unit areas at different positions of the real-time image, collecting image data in the unit areas and generating multiple target feature data; Import standard characteristic data of different cores; Count the number of target feature data A1; Calculate the proportion of target characteristic data in the standard characteristic data of different cores, and use the core with the largest proportion as the pre-selected core; Randomly import historical characteristic data of a set of pre-selected cores; Counting the number of historical characteristic data of the pre-selected core A2, and then counting the number of occurrences of each characteristic data in the historical characteristic data A3; Use the formula F=A3 / A2 to calculate the frequency F of each feature data; When the frequency F ≥ the preset threshold Q1, it is defined as general characteristic data; When the preset threshold Q2≤frequency F<preset threshold Q1, it is defined as accidental feature data; When the frequency F is less than the preset threshold Q2, the characteristic data is cleared; Counting the number of general feature data A4 and the number of accidental feature data A5 in the target feature data; Calculate the comprehensive feature score E of the target feature data. The specific calculation process is: ； in, is the proportion of general feature data in the comprehensive feature score E, is the proportion of accidental feature data in the comprehensive feature score E, ; When the comprehensive feature score E≥preset threshold Q3, the pre-selected core is changed to the output core, and an image recognition report is generated; When the comprehensive characteristic score E is less than the preset threshold Q3, the data of the pre-selected core is cleared; The real-time image data includes real-time size information, and the generation process of the target feature data also includes: Randomly mark a rectangular area and count the number of particles in the rectangular area B1; Randomly select n particles, and establish the circumscribed circle Y1 and the inscribed circle Y2 of each particle respectively, and obtain the radius size R1 of the circumscribed circle Y1 and the radius size R2 of the inscribed circle Y2; Use the formula R3=(R1+R2) / 2 to find the particle radius; Calculate the average particle radius of the n particles ; Using the formula Calculate the average area S1 of the particles; Get the size of the rectangular area and calculate the area S2 of the rectangular area; Use the formula K=nS1 / S2 to calculate the ratio K of the total area of ​​the particle to the area of ​​the rectangular area; Set the total area of ​​particles nS1 ≥ the preset threshold Q3, and obtain the minimum area of ​​the unit area KQ3; Set the shape of the unit area to a rectangle and get the size of the unit area; Randomly obtain multiple unit areas and collect image data in the unit areas; Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

2. The computer vision-based automatic recognition system for drilling core images according to claim 1, characterized in that: The processing process of the preprocessing module is: Acquire real-time image data; Divide the image evenly into multiple block units; Use the multithreaded programming library to create threads; Allocate the divided block units evenly to each thread; In each thread, the mean filtering algorithm and the histogram equalization algorithm are applied to the assigned block units for processing; After all threads have completed processing, the processed block units are merged into a complete image and output.

3. The computer vision-based automatic drilling core image recognition system according to claim 1, characterized in that: The real-time image data also includes real-time texture data, and the generation process of the target feature data also includes: Acquire real-time image data and draw a virtual straight line L parallel to the axis of the core in the height direction; Acquire real-time texture data and mark the intersections between the virtual straight line and the texture. The marked points are C1, C2, C3, ..., CV, where V is the number of marked points. Divide the virtual straight line into X sub-segments L` with a preset size; Count the number of marked points on each sub-segment L'. The number of marked points on each sub-segment L' is , , ,..., ; Calculate the difference P of the marked points on two adjacent sub-segments. The specific calculation process is: ； When the difference P ≤ the preset threshold Q4, the two sub-lines L` are merged; When the difference P>preset threshold Q4, the two sub-lines L' are split to obtain multiple sub-cores; Then, multiple unit areas are randomly obtained on each sub-core, and image data within the unit areas are collected; Acquire feature data from multiple image data, and merge the same feature data to generate multiple target feature data.

4. The computer vision-based automatic drilling core image recognition system according to claim 3, characterized in that: When the difference P>preset threshold Q4, the target feature data generation process further includes: Select any two adjacent marking points in the sub-core, and make a standard circle along the outer wall of the sub-core through the two marking points; Get the local stripes between two standard circles; Obtain the arc length Lr of the local stripes and the radius R4 of the core; Calculate the radian number β of the central angle corresponding to the local stripes. The specific calculation process is: ； When the number of radians β> the preset threshold Q5, the number of marking points on the atomic core is increased by one.

5. The computer vision-based automatic drilling core image recognition system according to claim 1, characterized in that: The real-time image data also includes real-time cross-section information, and the historical data information also includes historical cross-section dispersion data. The specific processing process of the verification module is as follows: Import image recognition reports to obtain real-time cross-section information of output cores; Obtain the contour line of the core section, randomly select three non-overlapping first virtual points W on the contour line, and make a first virtual plane M through the three first virtual points W; S reference points are randomly selected at positions within the core region on the first virtual plane M, and axial dimensions Lc1, Lc2, Lc3, ..., LcS of the S reference points from the core section are measured respectively; Calculate the dispersion J1 of multiple axial dimensions. The specific calculation process is: ； in, is the average axial size, Lci is the axial size of the i-th reference point from the core section; Then obtain multiple historical cross-section dispersion data of the output core and calculate their average value ; Calculate the deviation Jc of the output core. The specific calculation process is: ； When the deviation Jc is less than or equal to the preset threshold Q6, an image recognition report is output; When the deviation Jc>preset threshold Q6, a verification feedback report is output and a warning report is generated.

6. The computer vision-based automatic drilling core image recognition system according to claim 5, characterized in that: The processing of the verification module also includes: Obtain the contour line of the core section and the geometric center point Z1 of the contour line; Draw a virtual line through the center point Z1, and the intersection points of the virtual line and the contour line are Z2 and Z3 respectively; A point Z4 is randomly selected on the core, and a second virtual plane perpendicular to the axis of the core is made through Z4; Obtain the diameter size Ld of the core, and then respectively obtain the distance sizes of the second virtual plane from the intersection point Z2 and the intersection point Z3, the two distance sizes are Lh1 and Lh2 respectively; Calculate the inclination angle θ of the virtual line. The specific calculation process is: ； Draw virtual lines in different directions through the geometric center point Z1, calculate the inclination angles of the virtual lines, and obtain the average value θ1 of the multiple virtual angles; Get three virtual points W and calculate the inclination angle μ between any two virtual points W. The specific calculation process is: ； in is the vertical distance between two virtual points W, is the horizontal distance between two virtual points W; When μ>θ1, the corresponding virtual point W is deleted and reselected.

7. The computer vision-based automatic recognition system for drilling core images according to claim 6, characterized in that: When μ≤θ1, the processing of the verification module also includes: Randomly select three second virtual points on the core end surface, and make the third virtual plane by making the three second virtual points; Select the maximum and minimum points of the distance from the third virtual plane on the second virtual plane and use the preset rule to calculate the inclination η1 of the normal line of the third virtual plane, and then use the same maximum and minimum points to calculate the inclination η2 of the normal line of the first virtual plane M using the preset rule; The default rules are: Obtain the maximum value point and the minimum value point of the size from the target plane on the second virtual plane and calculate the difference Lm between the two; Read the diameter size Ld of the core; Calculate the inclination η` of the target plane normal. The specific calculation process is: Tanη`=Ld / Lm; When the target plane is the normal of the third virtual plane, η1 is calculated; when the target plane is the first virtual plane M, η2 is calculated; Calculate the anomaly G of the two normal inclinations. The specific calculation process is: G = |η1-η2| / η1; When the abnormality G>the preset threshold Q7, the data of the first virtual plane is cleared and the virtual point W is reselected.

8. The computer vision-based automatic drilling core image recognition system according to claim 5, characterized in that: It also includes a warning module, which is used to receive the verification report and take a warning action when the verification report is a warning report.

Images