Core rock sampling automatic identification system and method based on image processing and machine learning

Abstract

The invention discloses a core rock sampling automatic identification system and method based on image processing and machine learning. The system comprises a drilling assembly, a sensor assembly, a measurement and control lamp assembly, an identification assembly, a relay assembly and a control and display assembly. The method comprises the steps that core rock drilling is conducted, and a core rock sample is obtained through extraction; core rock sample images are collected, and parameters in the drilling process are monitored; sorting physical and appearance characteristics of the obtained core rock sample and the core rock sample image, and storing the physical and appearance characteristics in a database; preprocessing the acquired core rock sample image, and extracting key features; training by using a machine learning algorithm to obtain a core rock category recognition model; and applying the core rock category identification model obtained by training to newly obtained core rock sample image data to obtain classification and identification results. According to the method, the sensor, image processing and the convolutional neural network are combined, automatic and real-time rock core recognition is realized, the recognition efficiency and accuracy are improved, and the method can be flexibly applied to a field environment.

Description

Technical Field

[0001] The present invention relates to the field of geological drilling technology, and in particular to a core rock sampling automatic identification system and method based on image processing and machine learning. Background Art

[0002] In the field of geological exploration, core identification is one of the key technologies. Traditional core identification mainly relies on manual operation, requiring geological researchers to observe rock slices through a microscope to determine the core type and its parameters. This method relies heavily on the experience of experts, which is not only time-consuming, but also costly, and has low identification efficiency and accuracy. In addition, traditional methods are difficult to achieve real-time monitoring and rapid feedback, especially in complex field environments, and it is even more difficult to ensure timely and accurate acquisition of core information.

[0003] To solve these problems, image processing technology has been gradually introduced into the field of core identification in recent years. By using rock thin section imaging under a microscope, combined with image analysis and feature extraction technology, researchers have made significant progress in core texture, rock composition, grain size and lithology. However, this type of technology is mainly limited to laboratory environments and requires large equipment for high-precision imaging, making it difficult to promote and apply in field environments.

[0004] With the development of artificial intelligence technology, the automatic identification method of rock thin section minerals based on deep learning has greatly improved the accuracy of identification by automatically extracting image features through convolutional neural networks (CNN). In particular, the application of lightweight convolutional neural networks has greatly reduced model parameters, improved recognition speed and model portability. However, existing identification methods mostly rely on microscope imaging or require expensive and bulky equipment for image acquisition, which is not suitable for real-time operation in the field.

[0005] Traditional core identification technology relies on the subjective judgment of experts, which makes it difficult to ensure the reliability and consistency of the identification results; traditional methods require complex experimental equipment and long sample processing, resulting in high costs and low efficiency, and it is difficult to meet the needs of large-scale geological exploration; existing technologies are difficult to achieve real-time and convenient core identification and information feedback in the field, and cannot timely guide the next step of geological exploration; some advanced image recognition technologies require expensive and large imaging equipment and cannot be widely used on site. Summary of the invention

[0006] The purpose of the present invention is to provide a core rock sampling automatic identification system and method based on image processing and machine learning, which is used to solve at least one of the above-mentioned technical problems. By combining sensors, image processing, convolutional neural networks and other technologies, it realizes dynamic and real-time core recognition, improves the efficiency and accuracy of recognition, and can be flexibly applied in field environments.

[0007] The embodiment of the present invention is achieved as follows:

[0008] An automatic identification system for core rock sampling based on image processing and machine learning, comprising:

[0009] The drilling assembly is used for core rock drilling and extracting core rock samples.

[0010] Sensor assembly for capturing images of core rock samples and monitoring parameters during drilling.

[0011] The measurement and control light assembly is used to provide lighting to provide clear vision for image acquisition of the sensor assembly.

[0012] The recognition component is used to process and recognize the collected core rock sample image to obtain a recognition result of the core rock category.

[0013] Relay components are used to control the switching status of various electrical components.

[0014] The control and display component is used to display and adjust various parameters in the drilling process, and control the operation of the measurement and control light component, the drilling component, the sensor component, the identification component and the relay component.

[0015] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling component includes a drilling tool module, a power output module, a battery module and a drilling propulsion module.

[0016] The drilling tool module includes a drill bit, a drill rod, a drill collar, a wing blade, a blade releaser and a sampling inner tube. The drill bit is installed at the front end of the drill rod, the drill bit and the drill rod are arranged in the drill collar, the rear end of the drill rod is connected to the drill collar, the wing blade is arranged at the front end of the drill bit, the rear end of the drill bit is connected to the blade releaser, and the sampling inner tube is arranged inside the front end of the drill bit.

[0017] The power output module comprises a motor, a reducer, a coupling and a bearing. The output end of the motor is connected to the reducer, and the output end of the reducer is connected to the drill collar and the drill rod through the coupling.

[0018] The battery module is connected to the electric motor and also connected to the control and display component.

[0019] Its technical effects are: the structural design of the drilling tool module, the coordination of the drill bit, drill rod, side blades, sampling inner tube and other components, enable the system to accurately obtain core rock samples from the formation without human intervention, realize automated core rock sampling, and greatly improve the sampling efficiency; the measurement and control light assembly provides stable and uniform lighting conditions, ensuring visual clarity during image acquisition and improving image quality.

[0020] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling tool module also includes a rubber vibration damping ring, a drill sleeve and a shear pin.

[0021] The rubber vibration damping ring is arranged at the connection between the drill rod and the drill collar.

[0022] The drill sleeve is arranged outside the drill collar, and the drill collar and the drill rod move along the drill sleeve.

[0023] The shear pin is arranged at the rear end of the drill rod, and when a predetermined load is reached, the shear pin breaks.

[0024] Its technical effects are: the rubber vibration damping ring is set at the connection between the drill pipe and the drill collar, which can effectively absorb and relieve the energy generated by mechanical vibration and impact during drilling, reduce the adverse effects on the integrity of the core rock sample, and ensure the stability of the drilling process; the drill sleeve is set outside the drill collar, so that the drill collar and the drill pipe can move smoothly along the drill sleeve, providing a guiding function, ensuring that the drill tool maintains a stable path during the drilling process, avoiding deflection, protecting the core components of the drill tool, and reducing the direct wear of the drill collar and the drill rod by the external environment; the shear pin is designed to break when the predetermined load is reached, thereby preventing equipment damage caused by overload. As a load protection device, when the drill tool encounters abnormal resistance or load during drilling, the shear pin will break first, cutting off the connection between the drill pipe and the power transmission part, avoiding more serious mechanical failures.

[0025] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling component also includes a cooling and lubrication module.

[0026] The cooling and lubrication module includes a cooling pump, a water inlet pipe and a recovery pipe, wherein the water inlet pipe is connected to the output port of the cooling pump, the recovery pipe is connected to the input port of the cooling pump, the liquid outlet of the water inlet pipe extends into the drill collar, and the liquid inlet of the recovery pipe extends into the drill collar.

[0027] Its technical effects are: through the cooling and lubrication module, the temperature around the drill bit and drill rod is effectively reduced, ensuring the normal operation of the drill tool under high load conditions, and providing lubrication for the drill bit and drill rod to reduce the friction between the drill bit and the formation, thereby reducing wear and extending the service life of the drill tool; the rock cuttings accumulated around the drill bit are taken away by the coolant and discharged out of the drill collar through the recovery pipe, thereby keeping the drill bit clean and preventing the drill bit from getting stuck or decreasing efficiency due to rock cuttings accumulation. Under various complex geological conditions, such as high temperature or high friction geological environments, the cooling and lubrication module can effectively adjust and control the temperature and lubrication state during the drilling process, allowing the system to adapt to different working environments and enhancing the system's diversified application capabilities.

[0028] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the measurement and control lamp assembly includes a protective cover, a measurement and control lamp bracket, a light source and a heat sink, the protective cover and the measurement and control lamp bracket are sealed and connected to form a sealed space inside, and the light source and the heat sink are both installed on the measurement and control lamp bracket and are located in the sealed space.

[0029] The sensor component includes an image sensor, a temperature sensor, a pressure sensor, a vibration sensor and a data processing unit. The image sensor captures an image of the front end of the drill bit, the temperature sensor monitors the temperature of the power output module, the pressure sensor monitors the pressure applied to the drill bit, and the vibration sensor monitors the vibration of the drill rod. The data processing unit connects the image sensor, the temperature sensor, the pressure sensor and the vibration sensor through wires to transmit the collected data to the identification component and the control and display component.

[0030] The relay assembly comprises a relay, a wiring terminal and a connecting wire, one end of the wiring terminal is connected to the relay via the connecting wire, and the other end is connected to a component to be controlled.

[0031] Its technical effect is: combined with the protection design of the measurement and control light components and the real-time monitoring of the sensor components, the system can maintain an efficient and stable working state under a variety of complex and harsh environmental conditions. Whether it is strong light, dust environment, or high temperature, high pressure, and severe vibration drilling scenes, the system can adaptively adjust to ensure the normal operation of the equipment.

[0032] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the identification component includes:

[0033] The feature acquisition module is used to receive the core rock sample image acquired by the sensor component, organize the physical and appearance features of the core rock sample extracted by the drilling component, and store them in a database.

[0034] The preprocessing module is used to preprocess the collected core rock sample image and extract key features.

[0035] The model training module is used to train a core rock category recognition model using a machine learning algorithm.

[0036] The real-time recognition module is used to apply the core rock category recognition model obtained through training to the newly acquired core rock sample image data, obtain the recognition result of the core rock category, and feed it back to the control and display component.

[0037] Its technical effects are as follows: the recognition component realizes high-precision and efficient recognition of core rock sample categories through the mutual cooperation of feature collection, preprocessing, model training and real-time recognition modules. The system can automatically adapt to changes in different environments and core rock types, thereby improving the stability and accuracy of recognition. The system significantly reduces the workload and errors of manual identification by automatically identifying the categories of core rock samples, thereby improving work efficiency and safety. At the same time, the automated recognition process can realize continuous and uninterrupted core rock category analysis, providing strong support for intelligent decision-making in the drilling process.

[0038] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the control and display component includes:

[0039] The data receiving module is used to receive the recognition result of the core rock category obtained by the recognition component.

[0040] The data analysis module is used to extract sample categories, feature values ​​and confidence information according to the recognition results.

[0041] The regulating and controlling module is used to adjust the working parameters and working modes of the drilling assembly, the sensor assembly and the measurement and control light assembly.

[0042] The status monitoring module is used to monitor the working status of the measurement and control light assembly, the drilling assembly, the sensor assembly, the identification assembly and the relay assembly, and to issue an alarm if any abnormality occurs.

[0043] The display module is used to display the recognition result, the working parameters, the working mode and the monitoring information on the user interface.

[0044] Its technical effects are: the control and display components realize intelligent control, real-time data processing and all-round monitoring of the system through the close cooperation of various modules; it can automatically adjust the working parameters according to real-time data, optimize the operating status of each component, and significantly improve the efficiency of drilling operations; the display module provides operators with an intuitive and easy-to-understand interface, which is convenient for real-time understanding of the system status and operation progress, and improves the user's operating experience and the operability of the system.

[0045] A core rock sampling automatic identification method based on image processing and machine learning comprises: performing core rock drilling and extracting core rock samples.

[0046] Capture images of core rock samples and monitor parameters during drilling.

[0047] For the obtained core rock samples and core rock sample images, physical and appearance features are sorted out and stored in a database.

[0048] The collected core rock sample images are preprocessed and key features are extracted.

[0049] The core rock category recognition model was trained using a machine learning algorithm.

[0050] The core rock category recognition model obtained through training is applied to newly acquired core rock sample image data to obtain classification and recognition results.

[0051] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification method based on image processing and machine learning, the structure of the core rock category recognition model includes a convolutional layer, a pooling layer, a fully connected layer and a loss function.

[0052] The input of the convolutional layer is X∈R W×H×C , is a multidimensional matrix representing the color, texture and morphological characteristics of the core rock sample image, where W is the width of the image, H is the height of the image, and C is the number of channels of the image.

[0053] The convolution layer performs a convolution operation on the input feature map X and the convolution kernel W, and outputs a generated feature map Z, which represents the local features of the core rock sample image.

[0054] The input of the pooling layer is the feature map Z∈R output by the convolutional layer W′×H′×K , where W′ is the width after pooling, H′ is the height after pooling, and K is the number of convolution kernels.

[0055] The pooling layer downsamples the feature map Z and outputs a generated reduced-size feature map P, which represents the significant local features of the core rock sample image.

[0056] The input of the fully connected layer is a one-dimensional vector f=Flatten(P) obtained by flattening the reduced-size feature map P, which is used to describe the core rock category as a whole, and has a dimension size of N.

[0057] The fully connected layer maps the one-dimensional vector to the category space and outputs the generated category score y representing the possibility of each category.

[0058] The input of the loss function is the category score y and the corresponding core rock actual category label y true .

[0059] The loss function calculates the difference between the predicted value and the actual value, and outputs a loss value for updating the model parameters.

[0060] Its technical effect is: using the organic combination of convolutional layers, pooling layers, fully connected layers and loss functions to perform feature extraction, dimensionality reduction, category mapping and error optimization on core rock sample images, and ultimately achieve high-precision core rock category recognition. Through the local receptive field and weight sharing mechanism of the convolutional layer, the color, texture and morphological features of the core rock sample image are effectively extracted, and local features in the image, such as the subtle texture and edge features of the rock, are captured; the pooling layer retains the significant local features of the image by downsampling the feature map output by the convolutional layer, pays more attention to the key features in the image, and significantly reduces the data dimension, which not only reduces the computational complexity and improves the training efficiency of the model, but also effectively reduces the risk of model overfitting; the fully connected layer flattens the pooled feature map into a one-dimensional vector and maps it to the category space to generate a probability score for each category, so that the model can accurately judge the category of the entire core rock sample based on the local features extracted by the pooling layer, enhances the overall expression ability of the model, and improves the accuracy of core rock category recognition; the loss function calculates the difference between the model prediction value and the actual label, and uses the error feedback to update the model parameters. By continuously optimizing the loss function, the convolution kernel, pooling parameters and weights of the fully connected layer are gradually adjusted to make the prediction results more accurate.

[0061] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification method based on image processing and machine learning, the core rock category identification model obtained by training is applied to the newly acquired core rock sample image data to obtain the classification and identification results including:

[0062] The outputs of the multiple trained core rock category recognition models are y 1 ,y 2 ,...,y n .

[0063] The newly acquired core rock sample image data is combined with the output of several models to obtain the final classification and recognition result y final =∑j α j y j , where α j is the weight of the j-th model.

[0064] The technical effect is: by introducing multiple trained core rock category recognition models, each model is good at identifying different types of core rock samples, and the outputs of these models are weighted and fused, the overall performance of the recognition system is effectively improved, which helps to improve the generalization ability of the recognition system for unseen samples. By combining the prediction results of multiple models, the system can show better stability and accuracy when dealing with different geological conditions, different image qualities or different sample types.

[0065] The beneficial effects of the embodiments of the present invention are:

[0066] The present invention integrates drilling, image acquisition, parameter monitoring and identification processing into one system. Through the collaborative work of various components, the automatic sampling and identification of core rock samples is realized, forming a complete set of sampling image acquisition and sample classification and identification processes. The efficient collaboration of each module enables the system to respond quickly to changes in practical applications, ensuring adaptability in different environments and conditions. By automating the traditional manual sampling and identification process, the work efficiency and identification accuracy are significantly improved. The core rock sampling automatic identification system based on image processing and machine learning of the present invention can perform parallel sampling at multiple drilling positions at the same time, greatly improving the overall drilling efficiency. The parallel sampling method not only improves the utilization rate of resources, but also effectively shortens the geological exploration cycle.

[0067] In terms of recognition methods, the present invention adopts an advanced deep learning network structure, and uses advanced machine learning algorithms such as convolutional neural networks (CNN) to train core rock sample images, including convolutional layers, pooling layers, and fully connected layers, etc., to improve the recognition ability of the model. The self-learning ability of image features is fully utilized, so that the system can gradually improve the recognition rate in the process of continuously accumulating data. Through the fusion recognition of multiple models, the present invention effectively overcomes the limitations of a single model in feature learning, thereby further improving the recognition accuracy of core rock categories. Compared with traditional feature extraction and classification algorithms, it has stronger adaptability and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0069] Figure 1 It is a structural schematic diagram of a drilling component in the core rock sampling automatic identification system based on image processing and machine learning of the present invention;

[0070] Figure 2 The figure is a flow chart of the automatic identification method for core rock sampling based on image processing and machine learning of the present invention.

[0071] In the figure: 1-drill bit; 2-drill rod; 3-drill collar; 4-flank blade; 5-blade releaser; 6-sampling inner tube; 7-rubber vibration damping ring; 8-drill sleeve; 9-shear pin; 10-water inlet pipe; 11-recovery pipe. DETAILED DESCRIPTION

[0072] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0073] Please refer to Figure 1 The first embodiment of the present invention provides an automatic identification system for core rock sampling based on image processing and machine learning, which includes: a drilling component, used to perform core rock drilling and extract core rock samples; a sensor component, used to collect core rock sample images and monitor parameters during the drilling process; a measurement and control light component, used to provide lighting and provide a clear vision for the image collection of the sensor component; an identification component, used to process and identify the collected core rock sample images to obtain an identification result of the core rock category; a relay component, used to control the switching state of each electrical component; a control and display component, used to display and adjust various parameters during the drilling process, and control the operation of the measurement and control light component, the drilling component, the sensor component, the identification component and the relay component.

[0074] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling component includes a drilling tool module, a power output module, a battery module and a drilling propulsion module; the drilling tool module includes a drill bit 1, a drill rod 2, a drill collar 3, a flank blade 4, a blade releaser 5 and a sampling inner tube 6, the drill bit 1 is installed at the front end of the drill rod 2, the drill bit 1 and the drill rod 2 are arranged in the drill collar 3, the rear end of the drill rod 2 is connected to the drill collar 3, the flank blade 4 is arranged at the front end of the drill bit 1, the rear end of the drill bit 1 is connected to the blade releaser 5, and the sampling inner tube 6 is arranged inside the front end of the drill bit 1; the power output module includes a motor, a reducer, a coupling and a bearing, the output end of the motor is connected to the reducer, and the output end of the reducer is connected to the drill collar 3 and the drill rod 2 through a coupling; the battery module is connected to the motor and also connected to the control and display component.

[0075] Its technical effects are: the structural design of the drilling tool module, the coordination of the drill bit, drill rod, side blades, sampling inner tube and other components, enable the system to accurately obtain core rock samples from the formation without human intervention, realize automated core rock sampling, and greatly improve the sampling efficiency; the measurement and control light assembly provides stable and uniform lighting conditions, ensuring visual clarity during image acquisition and improving image quality.

[0076] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling tool module also includes a rubber vibration damping ring 7, a drill sleeve 8 and a shear pin 9; the rubber vibration damping ring 7 is arranged at the connection between the drill rod 2 and the drill collar 3; the drill sleeve 8 is arranged outside the drill collar 3, and the drill collar 3 and the drill rod 2 move along the drill sleeve 8; the shear pin 9 is arranged at the rear end of the drill rod 2, and when the predetermined load is reached, the shear pin 9 breaks.

[0077] The technical effect is as follows: the rubber vibration damping ring 7 is arranged at the connection between the drill rod 2 and the drill collar 3, which can effectively absorb and alleviate the energy generated by mechanical vibration and impact during the drilling process, reduce the adverse effects on the integrity of the core rock sample, and ensure the stability of the drilling process; the drill sleeve 8 is arranged outside the drill collar 3, so that the drill collar 3 and the drill rod 2 can move smoothly along the drill sleeve 8, providing a guiding function, ensuring that the drill tool maintains a stable path during the drilling process, avoiding deflection, protecting the core components of the drill tool, and reducing the direct wear of the drill collar and the drill rod by the external environment; the shear pin is designed to break when the predetermined load is reached, thereby preventing equipment damage caused by overload. As a load protection device, the shear pin 9 will break first when the drill tool encounters abnormal resistance or load during the drilling process, cutting off the connection between the drill rod and the power transmission part, avoiding more serious mechanical failures.

[0078] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the drilling component also includes a cooling and lubrication module; the cooling and lubrication module includes a cooling pump, a water inlet pipe 10 and a recovery pipe 11, the water inlet pipe 10 is connected to the output port of the cooling pump, the recovery pipe 11 is connected to the input port of the cooling pump, the liquid outlet of the water inlet pipe 10 extends into the drill collar 3, and the liquid inlet of the recovery pipe 11 extends into the drill collar 3.

[0079] The technical effect is as follows: the temperature around the drill bit and drill rod is effectively reduced through the cooling and lubrication module, ensuring the normal operation of the drill tool under high load conditions, and providing lubrication for the drill bit and drill rod to reduce the friction between the drill bit and the formation, thereby reducing wear and extending the service life of the drill tool; the rock cuttings accumulated around the drill bit are taken away by the coolant, and discharged out of the drill collar 3 through the recovery pipe 11, thereby keeping the drill bit clean and preventing the drill bit from getting stuck or decreasing efficiency due to the accumulation of rock cuttings. Under various complex geological conditions, such as high temperature or high friction geological environments, the cooling and lubrication module can effectively adjust and control the temperature and lubrication state during the drilling process, so that the system can adapt to different working environments and enhance the system's diversified application capabilities.

[0080] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the measurement and control lamp assembly includes a protective cover, a measurement and control lamp bracket, a light source and a radiator. The protective cover and the measurement and control lamp bracket are sealed and connected to form a sealed space inside. The light source and the radiator are both installed on the measurement and control lamp bracket and are located in the sealed space. Specifically, the protective cover is installed on the measurement and control lamp bracket through a sealing ring and a threaded structure, and the edges are tightly matched to ensure waterproof and dustproof performance. The transparent part of the protective cover faces the light source to ensure unimpeded passage of light. The radiator and the light source are designed as an integrated whole and are directly installed on the back of the light source. They are in close contact with the light source through thermal paste or a heat pipe to ensure effective conduction and dissipation of heat.

[0081] The sensor assembly includes an image sensor, a temperature sensor, a pressure sensor, a vibration sensor and a data processing unit. The image sensor collects images of the front end of the drill bit 1, the temperature sensor monitors the temperature of the power output module, the pressure sensor monitors the pressure of the drill bit 1, and the vibration sensor monitors the vibration of the drill rod 2. The data processing unit connects the image sensor, the temperature sensor, the pressure sensor and the vibration sensor through wires to transmit the collected data to the identification component and the control and display component. Specifically, the installation position of the image sensor needs to be consistent with the direction of the light source in the measurement and control light assembly to ensure that a clear image can be obtained. The brightness and angle adjustment of the light source must be coordinated with the viewing angle of the image sensor. The temperature sensor is installed near the heat-prone components such as the motor and the reducer, fixed by thermal conductive glue or a clamp, and monitors temperature changes in real time. The pressure sensor is installed near the drill rod or drill bit, fixed by threaded connection or flange, and monitors the pressure of the drill tool in real time. The vibration sensor is installed on the drill rod or the supporting structure, fixed by bolts or clamps, and is used to detect vibration during drilling to ensure the stability of the drilling process.

[0082] The relay assembly includes a relay, a terminal block and a connecting wire, one end of the terminal block is connected to the relay through the connecting wire, and the other end is connected to the element to be controlled. The terminal block fixes the connecting wire through a screw or a spring connector, and is connected to the motor, the measurement and control light assembly, the sensor and other equipment. The functions realized by the relay assembly include: controlling the switch and brightness adjustment of the light through the relay to ensure the best lighting under different drilling conditions; controlling the start and stop of the motor, adjusting the drilling speed and direction; controlling the power supply status of the sensor to ensure data collection when needed and saving energy consumption in the non-working state; controlling the power supply and operating status of the image processing equipment to ensure recognition after the image collection is completed; and connecting with the control and display components to realize centralized control of each module. Through reasonable installation, connection and layout design, the relay assembly can effectively manage the various electrical components in the system and ensure the efficient operation of the entire core rock sampling automatic identification system based on image processing and machine learning.

[0083] Its technical effect is: combined with the protection design of the measurement and control light components and the real-time monitoring of the sensor components, the system can maintain an efficient and stable working state under a variety of complex and harsh environmental conditions. Whether it is strong light, dust environment, or high temperature, high pressure, and severe vibration drilling scenes, the system can adaptively adjust to ensure the normal operation of the equipment.

[0084] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the identification component includes: a feature acquisition module, used to receive the core rock sample image acquired by the sensor component, organize the physical and appearance characteristics of the core rock sample extracted by the drilling component, and store them in a database; a preprocessing module, used to preprocess the acquired core rock sample image and extract key features; a model training module, used to use a machine learning algorithm to train a core rock category identification model; a real-time identification module, used to apply the trained core rock category identification model to newly acquired core rock sample image data, obtain a core rock category identification result, and feed it back to the control and display component.

[0085] Its technical effects are as follows: the recognition component realizes high-precision and efficient recognition of core rock sample categories through the mutual cooperation of feature collection, preprocessing, model training and real-time recognition modules. The system can automatically adapt to changes in different environments and core rock types, thereby improving the stability and accuracy of recognition. The system significantly reduces the workload and errors of manual identification by automatically identifying the categories of core rock samples, thereby improving work efficiency and safety. At the same time, the automated recognition process can realize continuous and uninterrupted core rock category analysis, providing strong support for intelligent decision-making in the drilling process.

[0086] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification system based on image processing and machine learning, the control and display component includes: a data receiving module, used to receive the identification result of the core rock category obtained by the identification component; a data parsing module, used to extract the sample category, characteristic value and confidence information according to the identification result; an adjustment and control module, used to adjust the working parameters and working mode of the drilling component, the sensor component and the measurement and control light component; a status monitoring module, used to monitor the working status of the measurement and control light component, the drilling component, the sensor component, the identification component and the relay component, and if an abnormality occurs, an alarm is issued; a display module, used to display the identification result, the working parameters, the working mode and monitoring information to the user interface.

[0087] Its technical effects are: the control and display components realize intelligent control, real-time data processing and all-round monitoring of the system through the close cooperation of various modules; it can automatically adjust the working parameters according to real-time data, optimize the operating status of each component, and significantly improve the efficiency of drilling operations; the display module provides operators with an intuitive and easy-to-understand interface, which is convenient for real-time understanding of the system status and operation progress, and improves the user's operating experience and the operability of the system.

[0088] Please refer to Figure 2The second embodiment of the present invention provides an automatic identification method for core rock sampling based on image processing and machine learning, which includes: performing core rock drilling to extract core rock samples; collecting core rock sample images and monitoring parameters during the drilling process; sorting out the physical and appearance characteristics of the acquired core rock samples and the core rock sample images, and storing them in a database; preprocessing the acquired core rock sample images and extracting key features; using a machine learning algorithm to train a core rock category recognition model; applying the trained core rock category recognition model to newly acquired core rock sample image data to obtain classification and recognition results.

[0089] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification method based on image processing and machine learning, the structure of the core rock category identification model includes a convolution layer, a pooling layer, a fully connected layer and a loss function; the input of the convolution layer is X∈R W×H×C , is a multidimensional matrix representing the color, texture and morphological features of the core rock sample image, where W is the width of the image, H is the height of the image, and C is the number of channels of the image; the convolution layer performs a convolution operation on the input feature map X and the convolution kernel W, and outputs the generated feature map Z, which represents the local features of the core rock sample image, Z i,j,k =Σ m,n,l W m,n,l,k ·X i+m,j+n,l +b k , where Z i,j,k is the value of the kth feature map at position (i, j), W m,n,l,k is the weight of the convolution kernel, b k is the bias, X i+m,j+n,l is the value of the input image at position (i+m, j+n) and the lth channel; the input of the pooling layer is the feature map Z∈R output by the convolutional layer W′×H′×K , where W′ is the width after pooling, H′ is the height after pooling, and K is the number of convolution kernels; the pooling layer downsamples the feature map Z and outputs the generated size-reduced feature map P, which represents the significant local features of the core rock sample image, P i,j,k =max(Z i+m,j+n,k ), is the value of the reduced-size feature map P at position (i, j) and the kth feature map; the input of the fully connected layer is a one-dimensional vector f=Flatten(P) obtained by flattening the reduced-size feature map P for an overall description of the core rock category, with a dimension size of N; the fully connected layer maps the one-dimensional vector to the category space, and outputs the generated category score y representing the possibility of each category, y=W f ·f+b f , where W f is the weight matrix of the fully connected layer, b fis the bias vector; the input of the loss function is the category score y and the corresponding core rock actual category label y true ; The loss function calculates the difference between the predicted value and the actual value, and outputs the loss value used to update the model parameters, L = -Σ i y true,i log(y i ), y i is the prediction score of the i-th category, y true,i is the actual label of the i-th category.

[0090] Its technical effect is: using the organic combination of convolutional layers, pooling layers, fully connected layers and loss functions to perform feature extraction, dimensionality reduction, category mapping and error optimization on core rock sample images, and ultimately achieve high-precision core rock category recognition. Through the local receptive field and weight sharing mechanism of the convolutional layer, the color, texture and morphological features of the core rock sample image are effectively extracted, and local features in the image, such as the subtle texture and edge features of the rock, are captured; the pooling layer retains the significant local features of the image by downsampling the feature map output by the convolutional layer, pays more attention to the key features in the image, and significantly reduces the data dimension, which not only reduces the computational complexity and improves the training efficiency of the model, but also effectively reduces the risk of model overfitting; the fully connected layer flattens the pooled feature map into a one-dimensional vector and maps it to the category space to generate a probability score for each category, so that the model can accurately judge the category of the entire core rock sample based on the local features extracted by the pooling layer, enhances the overall expression ability of the model, and improves the accuracy of core rock category recognition; the loss function calculates the difference between the model prediction value and the actual label, and uses the error feedback to update the model parameters. By continuously optimizing the loss function, the convolution kernel, pooling parameters and weights of the fully connected layer are gradually adjusted to make the prediction results more accurate.

[0091] In a preferred embodiment of the present invention, in the above-mentioned core rock sampling automatic identification method based on image processing and machine learning, the application of the trained core rock category identification model to the newly acquired core rock sample image data to obtain classification and identification results includes: the outputs of the multiple trained core rock category identification models are y 1 ,y 2 ,…,y n ; Combine the newly acquired core rock sample image data with the output of several models to obtain the final classification and recognition result y final =∑ j α j y j , where α j is the weight of the j-th model.

[0092] The technical effect is: by introducing multiple trained core rock category recognition models, each model is good at identifying different types of core rock samples, and the outputs of these models are weighted and fused, the overall performance of the recognition system is effectively improved, which helps to improve the generalization ability of the recognition system for unseen samples. By combining the prediction results of multiple models, the system can show better stability and accuracy when dealing with different geological conditions, different image qualities or different sample types.

[0093] It should be understood that the above specific embodiments of the present invention are only used to illustrate or explain the principles of the present invention, and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.

Claims

1. An automatic identification system for core rock sampling based on image processing and machine learning, characterized in that: include: A drilling assembly, used for core rock drilling and extracting core rock samples; Sensor components for capturing images of core rock samples and monitoring parameters during drilling; A measurement and control light assembly, used to provide illumination and provide clear vision for image acquisition of the sensor assembly; An identification component is used to process and identify the collected core rock sample image to obtain an identification result of the core rock category; Relay assembly, used to control the switch status of each electrical component; The control and display component is used to display and adjust various parameters in the drilling process, and control the operation of the measurement and control light component, the drilling component, the sensor component, the identification component and the relay component.

2. The core rock sampling automatic identification system based on image processing and machine learning according to claim 1 is characterized in that: The drilling assembly includes a drilling tool module, a power output module, a battery module and a drilling propulsion module; The drilling tool module comprises a drill bit (1), a drill rod (2), a drill collar (3), a wing blade (4), a blade releaser (5) and a sampling inner tube (6); the drill bit (1) is installed at the front end of the drill rod (2); the drill bit (1) and the drill rod (2) are arranged in the drill collar (3); the rear end of the drill rod (2) is connected to the drill collar (3); the wing blade (4) is arranged at the front end of the drill bit (1); the rear end of the drill bit (1) is connected to the blade releaser (5); and the sampling inner tube (6) is arranged inside the front end of the drill bit (1); The power output module comprises a motor, a reducer, a coupling and a bearing, the output end of the motor is connected to the reducer, and the output end of the reducer is connected to the drill collar (3) and the drill rod (2) via the coupling; The battery module is connected to the electric motor and also connected to the control and display component.

3. The core rock sampling automatic identification system based on image processing and machine learning according to claim 2 is characterized in that: The drilling tool module also includes a rubber vibration damping ring (7), a drill sleeve (8) and a shear pin (9); The rubber vibration damping ring (7) is arranged at the connection between the drill rod (2) and the drill collar (3); The drill sleeve (8) is sleeved outside the drill collar (3), and the drill collar (3) and the drill rod (2) move along the drill sleeve (8); The shear pin (9) is arranged at the rear end of the drill rod (2), and when a predetermined load is reached, the shear pin (9) breaks.

4. The core rock sampling automatic identification system based on image processing and machine learning according to claim 2 is characterized in that: The drilling assembly also includes a cooling and lubrication module; The cooling and lubrication module comprises a cooling pump, a water inlet pipe (10) and a recovery pipe (11); the water inlet pipe (10) is connected to the output port of the cooling pump, the recovery pipe (11) is connected to the input port of the cooling pump, the liquid outlet of the water inlet pipe (10) extends into the drill collar (3), and the liquid inlet of the recovery pipe (11) extends into the drill collar (3).

5. The core rock sampling automatic identification system based on image processing and machine learning according to claim 2 is characterized in that: The measuring and controlling lamp assembly comprises a protective cover, a measuring and controlling lamp bracket, a light source and a radiator. The protective cover and the measuring and controlling lamp bracket are sealed and connected to form a sealed space inside. The light source and the radiator are both mounted on the measuring and controlling lamp bracket and are located in the sealed space. The sensor assembly comprises an image sensor, a temperature sensor, a pressure sensor, a vibration sensor and a data processing unit, wherein the image sensor collects images of the front end of the drill bit (1), the temperature sensor monitors the temperature of the power output module, the pressure sensor monitors the pressure applied to the drill bit (1), and the vibration sensor monitors the vibration of the drill rod (2); the data processing unit connects the image sensor, the temperature sensor, the pressure sensor and the vibration sensor via wires and transmits the collected data to the identification component and the control and display component; The relay assembly comprises a relay, a wiring terminal and a connecting wire, one end of the wiring terminal is connected to the relay via the connecting wire, and the other end is connected to a component to be controlled.

6. The core rock sampling automatic identification system based on image processing and machine learning according to claim 1 is characterized in that: The identification component comprises: A feature acquisition module, used to receive the core rock sample image acquired by the sensor assembly, sort out the physical and appearance features of the core rock sample extracted by the drilling assembly, and store them in a database; A preprocessing module, used for preprocessing the collected core rock sample image and extracting key features; A model training module is used to train a core rock category recognition model using a machine learning algorithm; The real-time recognition module is used to apply the core rock category recognition model obtained through training to the newly acquired core rock sample image data, obtain the recognition result of the core rock category, and feed it back to the control and display component.

7. The core rock sampling automatic identification system based on image processing and machine learning according to claim 1 is characterized in that: The control and display components include: A data receiving module, used for receiving the identification result of the core rock category obtained by the identification component; A data analysis module, used to extract sample categories, feature values ​​and confidence information according to the recognition results; A regulating and controlling module, used for adjusting the working parameters and working modes of the drilling assembly, the sensor assembly and the measurement and control light assembly; A status monitoring module, used to monitor the working status of the measurement and control light assembly, the drilling assembly, the sensor assembly, the identification assembly and the relay assembly, and to issue an alarm if an abnormality occurs; The display module is used to display the recognition result, the working parameters, the working mode and the monitoring information on the user interface.

8. A core rock sampling automatic identification method based on image processing and machine learning, characterized in that: include: Conduct core rock drilling and extract core rock samples; Capture core rock sample images and monitor parameters during drilling; For the obtained core rock samples and core rock sample images, sort out the physical and appearance characteristics and store them in a database; Preprocessing the collected core rock sample image and extracting key features; A core rock category recognition model is obtained by training using a machine learning algorithm; The core rock category recognition model obtained through training is applied to newly acquired core rock sample image data to obtain classification and recognition results.

9. The method for automatic identification of core rock sampling based on image processing and machine learning according to claim 8, characterized in that: The structure of the core rock category recognition model includes a convolutional layer, a pooling layer, a fully connected layer and a loss function; The input of the convolutional layer is X∈R W×H×C , is a multidimensional matrix representing the color, texture and morphological characteristics of the core rock sample image, wherein W is the width of the image, H is the height of the image, and C is the number of channels of the image; The convolution layer performs a convolution operation on the input feature map X and the convolution kernel W, and outputs a generated feature map Z, which represents the local features of the core rock sample image; The input of the pooling layer is the feature map Z∈R output by the convolutional layer W′×H′×K , where W′ is the width after pooling, H′ is the height after pooling, and K is the number of convolution kernels; The pooling layer downsamples the feature map Z and outputs a generated reduced-size feature map P, which represents significant local features of the core rock sample image; The input of the fully connected layer is a one-dimensional vector f=Flatten(P) obtained by flattening the reduced-size feature map P for overall description of the core rock category, with a dimension size of N; The fully connected layer maps the one-dimensional vector to the category space, and outputs the generated category score y representing the possibility of each category; The input of the loss function is the category score y and the corresponding core rock actual category label y true ; The loss function calculates the difference between the predicted value and the actual value, and outputs a loss value for updating the model parameters.

10. The automatic identification method for core rock sampling based on image processing and machine learning according to claim 9, characterized in that: The core rock category recognition model obtained through training is applied to the newly acquired core rock sample image data to obtain classification and recognition results, including: The outputs of the multiple trained core rock category recognition models are y1, y2, ..., y n ; The newly acquired core rock sample image data is combined with the output of several models to obtain the final classification and recognition result y final =∑ j α j y j , where α j is the weight of the j-th model.

Images