Image-based intelligent prediction system for strength of mine filling body

Abstract

The invention discloses a mine filling body strength intelligent prediction system based on images, and relates to the technical field of mine filling bodies, and the mine filling body strength intelligent prediction system comprises a data acquisition module, a data processing module, an image texture extraction module, a prediction module and an analysis module. The texture feature information of the image is extracted based on the image data, and then comprehensive prediction is performed by analyzing multiple texture feature data of the image data and combining the texture feature data with granularity information, so that the overall structural characteristics of the mine filling body are better reflected, and the prediction accuracy of the density of the filling body is improved; according to the system, multi-dimensional information such as image texture, granularity and density is combined to construct an accurate and efficient comprehensive prediction system.

Description

Technical Field

[0001] The invention relates to the technical field of mine filling bodies, and in particular to an image-based intelligent prediction system for the strength of mine filling bodies. Background Art

[0002] The application of mine filling bodies in the mining process is becoming more and more extensive. Filling bodies are not only used to reduce the void space in mines, but also can prevent the occurrence of geological disasters such as landslides and subsidence in mines. Therefore, the quality of filling bodies is crucial to the production safety of mines. Among them, the strength of filling bodies is an important indicator to quantify the stability of filling bodies, and its strength detection is the basis for evaluating the stability of filling bodies.

[0003] Existing mine filling strength prediction systems may mostly use drilling rigs to drill holes on site to cut filling samples and process them into standard sizes, and then transport them to the laboratory for strength testing to obtain the physical and mechanical parameters of the samples. However, this method may have certain limitations. General mines may not have equipment such as high-precision presses, and the integrity of the samples may be damaged during the drilling, coring, handling, sampling, and transportation processes. In addition, sample maintenance may be required in the laboratory stage, and laboratory maintenance may not be able to restore the actual complex temperature, humidity, and overburden pressure environment of the filling. As a result, the final strength prediction test results may have large deviations due to the influence of multiple factors. Summary of the invention

[0004] The purpose of the present invention is to provide an image-based intelligent prediction system for mine filling strength, which solves the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: an image-based intelligent prediction system for mine filling strength, comprising: Data acquisition module: The data acquisition module randomly selects multiple samples from the mine filling bodies in different areas for shooting to obtain high-definition images of the mine filling bodies; Data processing module: The high-definition image is denoised through the data processing module, and then the high-definition image is uniformly converted into a grayscale image, and then the grayscale value of the high-definition image is normalized; Image texture extraction module: The image texture extraction module extracts features from the high-definition image processed by the data processing module to output texture features, and then normalizes the extracted texture features to output the value of the i-th texture feature in the j-th filling body sample; Prediction module: the value of the i-th texture feature in the j-th filling sample is input into the prediction module, and the prediction module outputs the predicted density of the j-th filling sample, the preliminary strength prediction value of the j-th filling sample, and the integrated strength prediction value of the j-th filling sample; Analysis module: The predicted density of the j-th filling body sample, the preliminary strength prediction value of the j-th filling body sample and the integrated strength prediction value of the j-th filling body sample are input into the analysis module, and the analysis module visualizes the input data in the form of charts.

[0006] Optionally, the prediction module includes: a density analysis submodule, an intensity analysis submodule and a comprehensive prediction submodule.

[0007] Optionally, the calculation formula of the density analysis submodule is as follows: ; in: PGW j refers to the predicted density of the jth filling sample, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample, PD j Refers to the average granularity of the jth sample, MPT j Refers to the maximum texture eigenvalue of the jth sample, PQ j Refers to the standard deviation of the texture feature of the jth sample, AS i，j Refers to the weight of the i-th texture feature in the j-th sample, PC j Refers to the constant adjustment factor of the jth sample, PC j Set to 0.5, PPm refers to the minimum density value of the filling body, PPm is set to 1, PE j Refers to the adjustment constant of the jth sample, PE j Set to 0.5; Refers to the combined effects of texture features and granularity; Refers to the maximum texture feature value MPT of the jth sample j and the standard deviation PQ of the texture feature of the jth sample j Used to normalize texture features; Refers to the degree of influence of image texture features on the density of the filling; Refers to smoothing the extreme values ​​of density; The processing process of the density analysis submodule is as follows: the value PT of the i-th texture feature in the j-th filling body sample is i，jInput to the density analysis submodule, the density analysis submodule outputs the predicted density PGW of the jth filling sample j .

[0008] Optionally, the calculation formula of the strength analysis submodule is as follows: ; in: SLO j Refers to the initial strength prediction value of the jth filling sample, SA j Refers to the weight coefficient of the jth sample density, SB j Refers to the weight coefficient of the jth sample granularity, SP j Refers to the particle shape factor of the jth sample, SK j refers to the attenuation coefficient of the jth sample; Refers to the extent to which density and granularity contribute to the strength prediction; Refers to the combined effect of particles on strength prediction; Refers to the nonlinear relationship between particle size and density described by a logarithmic function; The processing process of the strength analysis submodule is as follows: the predicted density PGW of the jth filling body sample is j and the average particle size PD of the jth sample j Input to the intensity analysis submodule, and based on the weight coefficient SA of the jth sample density j and the weight coefficient SB of the jth sample granularity j Output the initial strength prediction value SLO of the jth filling sample j .

[0009] Optionally, the calculation formula of the comprehensive prediction submodule is as follows: ; in: FNL j Refers to the integrated strength prediction value of the jth filling sample, FA j Refers to SLO j The adjustment factor of Refers to the initial strength SLO used to adjust j The prediction results; Refers to the predicted density PGW of the jth filling sample j Smoothing out the effects of final intensity; Refers to the average particle size PD through the jth sample jThe square root of is used to reduce the effect of large particle size on the final strength; The processing process of the comprehensive prediction submodule is as follows: the predicted density PGW of the jth filling body sample is j , the initial strength prediction value SLO of the jth filling sample j , the attenuation coefficient SK of the jth sample j , the standard deviation PQ of the jth sample texture feature j , the average particle size PD of the jth sample j and the constant adjustment factor PC of the jth sample j Input to the comprehensive prediction submodule, which outputs the integrated strength prediction value FNL of the jth filling sample j .

[0010] Optionally, the calculation process of the standard deviation of the j-th sample texture feature in the density analysis submodule is as follows: ; Where: PQ j Refers to the standard deviation of the texture feature of the jth sample, PSS j refers to the mean value of the jth texture feature, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample.

[0011] Optionally, the method of performing feature extraction on the high-definition image in the image texture extraction module is: First, when using OpenCV and scikit-image to process mine filling images, the following texture feature extraction methods are used; The means of extracting texture features include: gray-level co-occurrence matrix extraction, local binary pattern extraction, edge density extraction and spectrum feature extraction.

[0012] Optionally, the gray level co-occurrence matrix extraction means is a statistical method for describing the spatial relationship between image gray levels, and is used to extract contrast texture features, homogeneity texture features, energy texture features and entropy texture features in the image.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention outputs the predicted density of the jth filling sample through the density analysis submodule, analyzes the texture features in the mine filling image to predict the density of the filling, and uses the image acquisition device to obtain the image data of the filling surface. This process is simple and fast, avoiding the tedious sampling and experimental steps in the traditional method. The density of the mine filling directly affects its strength and the image texture features are a non-invasive and intuitive quality assessment method. The texture features in the image represent the microstructure of the surface and interior of the filling. The quality, structure and density of the filling can be inferred by the texture changes. The system can realize real-time and automatic filling density prediction by extracting texture features through image analysis, and image analysis can realize real-time density prediction to provide timely feedback for mine operations. For example, by collecting image data at the mine site, an accurate estimate of the filling density can be quickly obtained, and the filling configuration can be adjusted in real time.

[0014] 2. The present invention outputs the preliminary strength prediction value of the j-th filling sample through the strength analysis submodule. This submodule combines the two key factors of particle size and density to estimate the strength of the filling. Particle size and density are the two basic physical properties that affect the strength of the filling. By combining their relationship, the strength of the filling can be more accurately predicted. This system greatly improves the prediction efficiency by using image data to extract particle size information and texture features to predict these properties. By considering particle size and combining density as the whole system, a more refined strength prediction is made. This submodule can dynamically give a more accurate strength estimate under different particle size and density conditions through comprehensive analysis. The particle size data is directly extracted through image analysis and combined with the density information extracted from the image texture to predict the strength, avoiding the cumbersome particle size measurement and strength experiments in traditional laboratories.

[0015] 3. The present invention outputs the integrated strength prediction value of the j-th filling body sample through the comprehensive prediction submodule. This submodule integrates the previous density and preliminary strength, and introduces factors such as particle shape factor and standard deviation for adjustment. It not only enhances the adjustment and prediction capabilities of density and particle size, but also smoothes the mutual influence between different features to ensure higher accuracy and stability of strength prediction. The model integrating multiple variables and adjustment factors is more accurate than a single formula when processing complex mine filling body data. Through this comprehensive adjustment, the comprehensive prediction submodule can provide the most accurate strength value and can be flexibly adjusted according to the actual mine environment, such as particle size, density and shape, to ensure that high-accuracy predictions can be provided under various circumstances.

[0016] Fourth, the present invention affects the attenuation coefficient of the jth sample in the strength analysis submodule through the integrated strength prediction value of the jth filling sample to form a cyclic iterative system. In the strength analysis submodule, the attenuation coefficient mainly affects the adjustment of the strength of the relationship between particle size and density. By introducing the integrated strength prediction value to update the attenuation coefficient, the joint influence of particle size and density on the strength of the filling can be more accurately reflected. By introducing the adjustment of the comprehensive prediction submodule, the attenuation coefficient in the strength analysis submodule will be corrected according to the actual prediction of strength, so that the nonlinear relationship between particle size and density can be more accurately reflected in the strength estimation. Each iteration makes the prediction result closer to the true value. The comprehensive prediction submodule can obtain a more accurate comprehensive strength prediction after each round of iteration. Through this iterative method, the system can automatically correct its prediction deviation and continuously improve its accuracy. The attenuation coefficient is selected as a parameter affecting the iteration process because the attenuation coefficient controls the strength adjustment of the relationship between particle size and density. Its role in strength analysis is crucial. It determines the influence of particle size and density on the strength of the filling. Therefore, by adjusting the attenuation coefficient, we can more flexibly control the predicted strength, thereby improving the accuracy of the entire model. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of the method steps of the image-based intelligent prediction system for the strength of mine filling bodies; Figure 2 This is a schematic diagram of the overall structure of the image-based intelligent prediction system for mine filling strength; Figure 3 This is a schematic diagram of the structure of the prediction module in the image-based intelligent prediction system for mine filling strength. DETAILED DESCRIPTION

[0018] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] Regarding this image-based intelligent prediction system method for the strength of mine filling bodies, it is different from the existing mine filling body strength prediction system, which mostly uses drilling equipment to drill holes on site to cut filling body samples and process them into standard sizes, and then transport them to the laboratory for strength testing to obtain the physical and mechanical parameters of the samples. First of all, general mines may not have high-precision presses and other equipment, and the integrity of the samples may be damaged during the process of drilling, coring, handling, sampling, and transportation. In addition, the samples need to be maintained in the laboratory stage, and constant temperature and humidity maintenance may not be able to restore the actual complex temperature and humidity and the pressure of the overlying rock formation, resulting in the final strength prediction test results. The unfavorable situation of large deviations may occur under the influence of multiple factors; The module of this prediction system takes images of mine filling bodies and extracts image texture feature information based on the image data. It then analyzes multiple texture feature data of the image data and combines it with the particle size information for comprehensive prediction to better reflect the overall structural characteristics of the mine filling body and improve the prediction accuracy of the filling body density. This system combines multi-dimensional information such as image texture, particle size, density, etc. to construct an accurate and efficient comprehensive prediction system.

[0020] Example 1: Please refer to Figures 1 to 3 This implementation provides an image-based intelligent prediction system for mine filling strength, including: Data acquisition module: The data acquisition module randomly selects multiple samples from the mine filling bodies in different areas for shooting to obtain high-definition images of the mine filling bodies; Data processing module: The high-definition image is denoised through the data processing module, and then the high-definition image is uniformly converted into a grayscale image, and then the grayscale value of the high-definition image is normalized; Image texture extraction module: The image texture extraction module extracts features from the high-definition image processed by the data processing module to output texture features, and then normalizes the extracted texture features to output the value of the i-th texture feature in the j-th filling body sample; Prediction module: the value of the i-th texture feature in the j-th filling sample is input into the prediction module, and the prediction module outputs the predicted density of the j-th filling sample, the preliminary strength prediction value of the j-th filling sample, and the integrated strength prediction value of the j-th filling sample; Analysis module: The predicted density of the j-th filling body sample, the preliminary strength predicted value of the j-th filling body sample and the integrated strength predicted value of the j-th filling body sample are input into the analysis module. The analysis module visualizes the input data in the form of charts to help mine managers quickly understand the strength distribution of the filling body and provide optimization suggestions for the filling body based on the predicted results; The prediction module includes: a density analysis submodule, an intensity analysis submodule and a comprehensive prediction submodule.

[0021] In this embodiment: the density analysis submodule predicts the density by analyzing the texture features in the mine filling body image, and uses the image acquisition device to obtain the image data of the filling body surface. This process is not only simple and fast, but also avoids the cumbersome sampling and experimental steps in the traditional method. Compared with the traditional laboratory density determination method, image analysis can realize real-time density prediction and provide timely feedback for mine operations. For example, by collecting image data at the mine site, an accurate estimate of the filling body density can be quickly obtained, and the filling body configuration can be adjusted in real time; The strength analysis submodule combines two key factors, particle size and density, to estimate the strength of the filling. Particle size and density are two basic physical properties that affect the strength of the filling. By combining their relationship, the strength of the filling can be predicted more accurately. This system directly uses image data to extract particle size information and texture features to predict these properties, greatly improving the prediction efficiency. This method provides a more refined strength prediction by considering particle size and combining it with density, a conventional strength indicator. Fillings with larger particle sizes usually have higher porosity, which affects their strength, while fillings with higher density usually have greater strength. Through this comprehensive analysis, the strength analysis submodule can dynamically give more accurate strength estimates under different particle size and density conditions. The particle size data is directly extracted through image analysis, and combined with the density information extracted from the image texture to predict the strength, avoiding the cumbersome particle size determination and strength experiments in traditional laboratories; The comprehensive prediction submodule combines the results of the first two and introduces factors such as particle shape factor and standard deviation for adjustment. This not only enhances the adjustment and prediction capabilities of density and particle size, but also smoothes the mutual influence between different features, ensuring higher accuracy and stability of strength prediction. This model that integrates multiple variables and adjustment factors is more accurate than a single formula when processing complex mine filling data. Through this comprehensive adjustment, the comprehensive prediction submodule can provide the most accurate strength value and can be flexibly adjusted according to the actual mining environment, such as particle size, density, shape, etc., to ensure high-accuracy predictions in all situations.

[0022] See also Figures 1 to 3 , the processing process of the density analysis submodule is as follows: ; in: PGW j refers to the predicted density of the jth filling sample, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample; PD j Refers to the average particle size of the jth sample, the average size of the sample particles, which directly affects the density and strength of the ore; MPT j Refers to the maximum texture feature value of the jth sample, indicating the maximum value of all texture features in the sample, and is used to normalize other texture features; PQ j Refers to the standard deviation of the texture feature of the jth sample; ; Where: PQ j Refers to the standard deviation of the texture feature of the jth sample, PSS j refers to the mean value of the jth texture feature, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample; AS i，j Refers to the weight of the i-th texture feature in the j-th sample; PC j Refers to the constant adjustment factor of the jth sample, PC j Set to 0.5, used to adjust the constant factor of density prediction, ranging from 0.1 to 1.0; PPm refers to the minimum density value of the filling body. PPm is set to 1 to avoid zero division errors in density division; PE j Refers to the adjustment constant of the jth sample, PE j Set to 0.5 to adjust the denominator. If PPm is too small, the constant term may dominate the result, resulting in unstable calculations. If PPm is too large, the effect of the summation term may be ignored, causing the result to be biased towards the constant term. j Balance this denominator term, the value range is [0.5, 10]; Refers to the combined effect of texture features and granularity, the average granularity PD of the jth sample j The square root operation reflects the nonlinear effect of particle size on density. This means that the larger the particle size, the greater the contribution of texture features to density, but the growth rate is nonlinear. The square root operation can avoid the extreme effect on the result when the particle size is too large. Refers to the maximum texture feature value MPT of the jth sample j and the standard deviation PQ of the texture feature of the jth sample j Used to normalize texture features. The standard deviation reflects the degree of texture change, while the maximum value normalization helps to reduce the impact of extreme values. This allows the calculation range of each texture feature to remain consistent, thereby ensuring the smoothness of the model; Refers to the degree of influence of image texture features on the density of the filling; Refers to smoothing the extreme values ​​of the density. This term is used to adjust the predicted density by subtracting the constant adjustment factor PC of the jth sample. j With adjustment constant PE j The ratio of ppm to avoid the density prediction value being too large or too small is used to smoothly adjust the extreme density values ​​to ensure the stability of the model. Specifically, PPm is the minimum density value of the filling body. Using this form can avoid numerical instability caused by dividing by zero or a very small value. The value PT of the i-th texture feature in the j-th filling sample i，j Input to the density analysis submodule, the density analysis submodule outputs the predicted density PGW of the jth filling sample j .

[0023] In this embodiment: This submodule predicts the density of mine fillings by image texture features. This part is crucial in the intelligent prediction system because the density of mine fillings directly affects their strength, and image texture features are a non-invasive and intuitive quality assessment method. The texture features in the image represent the microstructure of the surface and interior of the filling. The quality, structure and density of the filling can be indirectly inferred through changes in texture. By extracting texture features through image analysis, the system can achieve real-time and automatic prediction of filling density without relying on expensive experiments or long-term monitoring. The density and strength of the filling are significantly related. The density of the filling is predicted by the weighted average of texture features. The density analysis submodule can accurately predict the density of the filling without traditional physical measurements, which provides a basis for subsequent strength prediction. Density prediction through image texture features avoids complex physical sampling processes and has a higher level of automation. Combining image processing with the prediction of mine filling density is an innovative combination, especially in practical applications, which reduces dependence on actual sampling and greatly improves the applicability and efficiency of the system.

[0024] See also Figures 1 to 3 , the processing process of the strength analysis submodule is as follows: ; in: SLO j Refers to the initial strength prediction value of the jth filling sample, SA j Refers to the weight coefficient of the jth sample density, SB j Refers to the weight coefficient of the jth sample granularity; SP jRefers to the particle shape factor of the jth sample, which reflects the effect of particle shape on ore strength; SP j =4AL / (π×LL 2 ); Where: AL refers to the area of ​​the sample particles, LL refers to the maximum length of the sample particles; SK j Refers to the attenuation coefficient of the jth sample, which is used to adjust the average particle size PD of the jth sample. j and the predicted density PGW of the jth filling sample j The effect of interaction between them on strength; Refers to the contribution of density and granularity to strength prediction. The influence of density is determined by the weight coefficient SA of the jth sample density. j Adjustment: The effect of particle size is enhanced by taking the square root, because the increase of particle size usually significantly affects the strength of the filling body. The square operation can make the effect of particle size on strength more prominent. The square term of particle size amplifies the effect of particle size on strength in the model, reflecting the nonlinear effect of particle size on strength. This processing method can significantly improve the accuracy of strength prediction; Refers to the comprehensive effect of particles on strength prediction, the particle shape factor SP of the jth sample j Affects the arrangement and stacking of particles and thus affects the strength. The influence of particle shape factor will be smoothed by square root operation to avoid the influence of particle shape on strength being too extreme; Refers to the use of a logarithmic function to describe the nonlinear relationship between particle size and density. This form of logarithmic operation will reduce the impact of the proportional changes between particle size and density on the intensity prediction, making the interaction between the two smoother and more stable; The predicted density PGW of the jth filling sample j and the average particle size PD of the jth sample j Input to the intensity analysis submodule, and based on the weight coefficient SA of the jth sample density j and the weight coefficient SB of the jth sample granularity j Output the initial strength prediction value SLO of the jth filling sample j .

[0025] In this embodiment: This submodule further predicts the strength of mine filling by considering the relationship between particle size and density. Particle size and density are two key factors affecting the strength of filling. This submodule quantifies the nonlinear relationship between particle size and density to more accurately estimate the strength of filling. This submodule provides a more comprehensive perspective for strength prediction by combining the effects of particle size and density on strength. Particle size and density respectively reflect the microstructural characteristics and macroscopic properties of the filling. Through the interaction between particle size and density, the strength of the filling can be predicted more accurately, especially the quantitative analysis of its nonlinear relationship, which makes the system's strength prediction more accurate under different conditions. By considering the square of particle size and density on strength, this submodule significantly improves the traditional prediction method and avoids the problem of ignoring the complex relationship between the two. By nonlinearly modeling the relationship between particle size and density, this submodule can better reflect the true situation of filling strength than traditional linear models.

[0026] See also Figures 1 to 3 , the processing process of the comprehensive prediction submodule is as follows: ; in: FNL j Refers to the predicted integrated strength value of the jth filling sample; FA j Refers to SLO j The adjustment factor is used to adjust the impact of the initial strength; Refers to the initial strength prediction value SLO used to adjust the jth filling sample j The exponential decay operation is to reduce the influence of the standard deviation of texture features on the intensity prediction. The standard deviation PQ of the texture features of the jth sample j The larger the exponential term, the smaller its influence, so that the influence of the variability of texture features on intensity prediction gradually weakens; Refers to the predicted density PGW of the jth filling sample j Smooth the effect on the final strength by adding 1 to prevent problems when the density is zero. This function adjusts the range of strength prediction according to the density, and the contribution to the final strength gradually increases when the density is higher; Refers to the average particle size PD through the jth sample j The square root of is used to reduce the impact of large particle size on the final strength. The larger the particle size, the smaller the negative impact on the final strength. This operation can effectively avoid unreasonable predictions caused by excessively large particles. The predicted density PGW of the jth filling sample j, the initial strength prediction value SLO of the jth filling sample j , the attenuation coefficient SK of the jth sample j , the standard deviation PQ of the jth sample texture feature j , the average particle size PD of the jth sample j and the constant adjustment factor PC of the jth sample j Input to the comprehensive prediction submodule, which outputs the integrated strength prediction value FNL of the jth filling sample j .

[0027] In this embodiment: This submodule combines the results of the first two submodules and introduces more adjustment factors to comprehensively predict the final strength of the filling body. Formula three can comprehensively consider various factors, correct the limitations of a single prediction model, and output a more accurate and reliable filling body strength value. By combining the density analysis submodule with the strength analysis submodule, this submodule can effectively integrate various influencing factors and improve the comprehensiveness and accuracy of strength prediction. The innovation of this submodule is that it integrates multiple factors that affect the strength of the filling body. Through a comprehensive analysis of factors such as texture, granularity, and density, a more accurate and stable strength prediction can be achieved.

[0028] It is worth noting that the integrated strength prediction value FNL of the jth filling sample j Further calculations are performed to affect the attenuation coefficient SK of the jth sample in the intensity analysis submodule j , with the initial strength prediction value SLO of the jth filling sample j and the integrated strength prediction value FNL of the jth filling sample j Continuously optimize, the specific processing process is as follows: First up: SK j，new =SK j，old +NA×(FNL j - SLO j ); Second: Set the iteration termination condition: Termination condition 1: The number of iterations is 100; Termination condition 2: |FNL j，new -FNL j，old |<0.001; in: NA refers to the learning rate, which indicates the step size adjusted at each iteration; SK j，new Refers to the attenuation coefficient of the j-th sample after update; SK j，old Refers to the attenuation coefficient of the jth sample before the update; FNL j，newRefers to the updated integrated strength prediction value of the jth filling sample; FNL j，old Refers to the predicted integrated strength value of the jth filling sample before updating; In this embodiment: In the strength analysis submodule SK j The main influence of the relationship between particle size and density on the strength adjustment is introduced by introducing the integrated strength prediction value FNL of the jth filling sample j SK j Updated to more accurately reflect the combined effects of particle size and density on filling strength, FNL j It can be used as an adjustment factor to ensure that the impact of particle size and density can be more accurately reflected in each iteration. By introducing the adjustment of the comprehensive prediction submodule, the SK j It will be corrected according to the actual prediction of strength, so that the nonlinear relationship between particle size and density can be more accurately reflected in the strength estimation. Each iteration makes the prediction result closer to the true value. The comprehensive prediction submodule can obtain a more accurate comprehensive strength prediction after each iteration, especially when the relationship between particle size and density is more finely adjusted in the strength analysis submodule. The comprehensive prediction submodule can give a more optimized strength prediction based on the updated information. Through this iterative method, the system can automatically correct its prediction deviation and continuously improve its accuracy. j As a parameter affecting the iteration process, it is because SK j The strength adjustment controls the relationship between particle size and density. It plays a crucial role in the strength analysis submodule. It determines the influence of particle size and density on the strength of the filling. Therefore, by adjusting SK j We can control the prediction strength more flexibly to improve the accuracy of the entire model. The comprehensive prediction submodule combines multiple factors when calculating the strength, and SK j As a parameter that controls the relationship between granularity and density, it directly affects the prediction results in the comprehensive prediction submodule. j This enables the prediction results of the comprehensive prediction submodule to more accurately reflect the actual strength situation, thereby better adjusting and optimizing the final prediction results.

[0029] In the specific implementation process, multiple submodules in this method are used to form an image-based intelligent prediction system for mine filling strength. The value PT of the i-th texture feature in the j-th filling sample is i，j Input to the density analysis submodule, the density analysis submodule outputs the predicted density PGW of the jth filling sample jBy analyzing the texture features in the mine filling image to predict the density, the image data of the filling surface is obtained by image acquisition equipment. This process is not only simple and fast, but also avoids the tedious sampling and experimental steps in the traditional method. Compared with the traditional laboratory density determination method, image analysis can realize real-time density prediction and provide timely feedback for mine operations. For example, by collecting image data at the mine site, an accurate estimate of the filling density can be quickly obtained, and the filling configuration can be adjusted in real time. By taking the predicted density PGW of the jth filling sample j and the average particle size PD of the jth sample j Input to the intensity analysis submodule, and based on the weight coefficient SA of the jth sample density j and the weight coefficient SB of the jth sample granularity j Output the initial strength prediction value SLO of the jth filling sample j This submodule combines two key factors, particle size and density, to estimate the strength of the filling. Particle size and density are two basic physical properties that affect the strength of the filling. By combining their relationship, the strength of the filling can be predicted more accurately. This system greatly improves the prediction efficiency by using image data to extract particle size information and texture features to predict these properties. It provides a more refined strength prediction by considering particle size and combining it with density, a conventional strength indicator. Fillings with larger particle sizes usually have higher porosity, which affects their strength, while fillings with higher density usually have greater strength. Through this comprehensive analysis, this submodule can dynamically give more accurate strength estimates under different particle size and density conditions. It directly extracts particle size data through image analysis and combines it with density information extracted from image texture to predict strength, avoiding the cumbersome particle size measurement and strength experiments in traditional laboratories. By taking the predicted density PGW of the jth filling sample j , the initial strength prediction value SLO of the jth filling sample j , the attenuation coefficient SK of the jth sample j , the standard deviation PQ of the jth sample texture feature j , the average particle size PD of the jth sample j and the constant adjustment factor PC of the jth sample j Input to the comprehensive prediction submodule, which outputs the integrated strength prediction value FNL of the jth filling sample j, combining the results of the first two, and introducing factors such as particle shape factor and standard deviation for adjustment, not only enhances the adjustment and prediction capabilities of density and particle size, but also smoothes the mutual influence between different features to ensure higher accuracy and stability of strength prediction. This model that integrates multiple variables and adjustment factors is more accurate than a single formula when processing complex mine filling data. Through this comprehensive adjustment, the comprehensive prediction submodule can provide the most accurate strength value and can be flexibly adjusted according to the actual mine environment, such as particle size, density and shape, to ensure that high-accuracy predictions can be provided in various situations; The integrated strength prediction value FNL of the jth filling sample j Further calculations are performed to affect the attenuation coefficient SK of the jth sample in the intensity analysis submodule j , in the strength analysis submodule SK j The main influence of the relationship between particle size and density on the strength adjustment is introduced by introducing the integrated strength prediction value FNL of the jth filling sample j SK j The update can more accurately reflect the combined effect of particle size and density on the strength of the filling body, by introducing the SK in the adjustment strength analysis submodule of the comprehensive prediction submodule j It will be corrected according to the actual prediction of strength, so that the nonlinear relationship between particle size and density can be more accurately reflected in the strength estimation. Each iteration makes the prediction result closer to the true value. The comprehensive prediction submodule can obtain a more accurate comprehensive strength prediction after each round of iteration. Through this iterative method, the system can automatically correct its prediction deviation and continuously improve its accuracy. Select SK j As a parameter affecting the iteration process, SK j Strength adjustment controls the relationship between particle size and density. It plays a crucial role in the strength analysis submodule. It determines the influence of particle size and density on the strength of the filling. Therefore, by adjusting SK j We can more flexibly control the strength of the prediction, thereby improving the accuracy of the entire model; This allows the various sub-modules to cooperate with each other in calculations, and to perform overall cycles and iterations, so that the overall system has the effect of automatic optimization and updating, and thus better adaptability.

[0030] Example 2: Please refer to Figure 1 , Figure 2 and Figure 3 , the method of extracting features from high-definition images in the image texture extraction module is: First, when using OpenCV and scikit-image to process mine filling images, the following texture feature extraction methods are used; The extraction methods of texture features include: gray level co-occurrence matrix extraction method, local binary pattern extraction method, edge density extraction method and spectrum feature extraction method; The gray level co-occurrence matrix extraction method is a statistical method for describing the spatial relationship between image gray levels, and is used to extract contrast texture features, homogeneity texture features, energy texture features and entropy texture features in an image.

[0031] In this embodiment: wherein contrast represents the drastic degree of change of pixel grayscale value in the image, homogeneity represents the uniformity of pixel grayscale value, energy reflects the texture regularity of the image, and entropy represents the complexity of grayscale distribution in the image; Local binary pattern extraction is a method used to describe the local texture features of an image. It generates a binary number by comparing the neighborhood around each pixel with the central pixel value to extract texture information. The local texture features of an image are extracted by this extraction method. The edge density extraction method is to measure the density of the edge in the image by calculating the ratio of the edge pixels in the image to the pixels of the entire image. Edge extraction can be completed by common edge detection operators, such as Canny edge detection. The extracted edge density texture feature is the ratio of edge pixels in the image. Spectral feature extraction is performed by performing frequency domain analysis on the image. These features describe the frequency distribution of the image, such as low-frequency information, the large structure of the image, the smooth part and high-frequency information, the details and noise of the image. The frequency distribution of the extracted texture features is used to reflect the low-frequency and high-frequency components of the image, and the power spectral density is used to represent the energy distribution of different frequencies.

[0032] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Image-based intelligent prediction system for mine filling strength, characterized by: include: Data acquisition module: The data acquisition module randomly selects multiple samples from the mine filling bodies in different areas for shooting to obtain high-definition images of the mine filling bodies; Data processing module: The high-definition image is denoised through the data processing module, and then the high-definition image is uniformly converted into a grayscale image, and then the grayscale value of the high-definition image is normalized; Image texture extraction module: The image texture extraction module extracts features from the high-definition image processed by the data processing module to output texture features, and then normalizes the extracted texture features to output the value of the i-th texture feature in the j-th filling body sample; Prediction module: the value of the i-th texture feature in the j-th filling sample is input into the prediction module, and the prediction module outputs the predicted density of the j-th filling sample, the preliminary strength prediction value of the j-th filling sample, and the integrated strength prediction value of the j-th filling sample; Analysis module: The predicted density of the j-th filling body sample, the preliminary strength prediction value of the j-th filling body sample and the integrated strength prediction value of the j-th filling body sample are input into the analysis module. The analysis module visualizes the input data in the form of charts to help mine managers quickly understand the strength distribution of the filling body and provide optimization suggestions for the filling body based on the prediction results.

2. The image-based intelligent prediction system for mine filling strength according to claim 1 is characterized in that: The prediction module includes: a density analysis submodule, an intensity analysis submodule and a comprehensive prediction submodule.

3. The image-based intelligent prediction system for mine filling strength according to claim 2 is characterized in that: The calculation formula of the density analysis submodule is as follows: ； in: PGW j refers to the predicted density of the jth filling sample, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample, PD j Refers to the average granularity of the jth sample, MPT j Refers to the maximum texture eigenvalue of the jth sample, PQ j Refers to the standard deviation of the texture feature of the jth sample, AS i，j Refers to the weight of the i-th texture feature in the j-th sample, PC j Refers to the constant adjustment factor of the jth sample, PC j Set to 0.5, PPm refers to the minimum density value of the filling body, PPm is set to 1, PE j Refers to the adjustment constant of the jth sample, PE j Set to 0.5; Refers to the combined effects of texture features and granularity; Refers to the maximum texture feature value MPT of the jth sample j and the standard deviation PQ of the texture feature of the jth sample j The square root of is used to normalize the texture features; Refers to the degree of influence of image texture features on the density of the filling; Refers to smoothing the extreme values ​​of density; The processing process of the density analysis submodule is as follows: the value PT of the i-th texture feature in the j-th filling body sample is i，j Input to the density analysis submodule, the density analysis submodule outputs the predicted density PGW of the jth filling sample j .

4. The image-based intelligent prediction system for mine filling strength according to claim 3 is characterized in that: The calculation formula of the strength analysis submodule is as follows: ； in: SLO j Refers to the initial strength prediction value of the jth filling sample, SA j Refers to the weight coefficient of the jth sample density, SB j Refers to the weight coefficient of the jth sample granularity, SP j Refers to the particle shape factor of the jth sample, SK j refers to the attenuation coefficient of the jth sample; Refers to the extent to which density and granularity contribute to the strength prediction; Refers to the combined effect of particles on strength prediction; Refers to the nonlinear relationship between particle size and density described by a logarithmic function; The processing process of the strength analysis submodule is as follows: the predicted density PGW of the jth filling body sample is j and the average particle size PD of the jth sample j Input to the intensity analysis submodule, and based on the weight coefficient SA of the jth sample density j and the weight coefficient SB of the jth sample granularity j Output the initial strength prediction value SLO of the jth filling sample j .

5. The image-based intelligent prediction system for mine filling strength according to claim 4 is characterized in that: The calculation formula of the comprehensive prediction submodule is as follows: ； in: FNL j Refers to the integrated strength prediction value of the jth filling sample, FA j Refers to SLO j The adjustment factor of Refers to the initial strength SLO used to adjust j The prediction results; Refers to the predicted density PGW of the jth filling sample j Smoothing out the effects of final intensity; Refers to the average particle size PD through the jth sample j The square root of is used to reduce the effect of large particle size on the final strength; The processing process of the comprehensive prediction submodule is as follows: the predicted density PGW of the jth filling body sample is j , the initial strength prediction value SLO of the jth filling sample j , the attenuation coefficient SK of the jth sample j , the standard deviation PQ of the jth sample texture feature j , the average particle size PD of the jth sample j and the constant adjustment factor PC of the jth sample j Input to the comprehensive prediction submodule, which outputs the integrated strength prediction value FNL of the jth filling sample j .

6. The image-based intelligent prediction system for mine filling strength according to claim 3 is characterized in that: The calculation process of the standard deviation of the j-th sample texture feature in the density analysis submodule is as follows: ； Where: PQ j Refers to the standard deviation of the texture feature of the jth sample, PSS j refers to the mean value of the jth texture feature, n refers to the number of image texture features, PT i，j Refers to the value of the i-th texture feature in the j-th filling sample.

7. The image-based intelligent prediction system for mine filling strength according to claim 1 is characterized in that: The method of extracting features from high-definition images in the image texture extraction module is as follows: First, when using OpenCV and scikit-image to process mine filling images, the following texture feature extraction methods are used; The means of extracting texture features include: gray-level co-occurrence matrix extraction, local binary pattern extraction, edge density extraction and spectrum feature extraction.

8. The image-based intelligent prediction system for mine filling strength according to claim 7 is characterized in that: The gray level co-occurrence matrix extraction method is a statistical method for describing the spatial relationship between image gray levels, and is used to extract contrast texture features, homogeneity texture features, energy texture features and entropy texture features in an image.

Images