A Method and Apparatus for Selecting Foam Image Features in Flotation Process Based on Sensitive Mutual Information

Abstract

This invention provides a method for selecting froth image features in a flotation process based on sensitive mutual information (mRMR). The method includes: preprocessing the obtained froth flotation image feature data by removing outliers and normalizing the data; calculating the sensitivity coefficient between image features and concentrate grade using the Pearson correlation coefficient; constructing an optimization objective function for selecting sensitive features of the froth image based on the mRMR criterion; and optimizing feature selection using the proposed multi-group cooperative search particle swarm optimization algorithm. This method can construct a selection objective function for sensitive froth image features based on existing data and optimize feature selection using a multi-group cooperative search particle swarm optimization algorithm. This enables the selection of numerous extracted image features, reducing data redundancy and avoiding complex calculations caused by data redundancy, thus improving the accuracy and efficiency of concentrate grade prediction in the froth flotation process.

Description

Technical Field

[0001] This application relates to the field of foam image feature selection technology, and specifically discloses a method and apparatus for selecting foam image features in a flotation process based on sensitive mutual information. Background Technology

[0002] Foam images can serve as an important indicator of flotation performance, as the flotation production conditions can be directly reflected on the foam surface.

[0003] As the production conditions during the flotation process change, the relevant characteristics of the foam surface will also change. A large number of image features can be extracted during foam flotation, such as the mean red, mean green, mean blue, and mean gray values. These diverse features can be used to identify the production status, but not all initially obtained features contain important information, nor can all features be used to accurately identify the production status.

[0004] Redundant and irrelevant features significantly increase the difficulty of model training. Therefore, feature selection is required for the foam image features to obtain a selected feature subset, so as to maximize the mutual information between the extracted features and the concentrate grade, and reduce the redundancy between the features.

[0005] Feature selection can improve the accuracy of feature subset identification of industrial production status and simplify model training. Therefore, an accurate feature selection method for flotation process foam images is crucial for evaluating the working conditions of the flotation process.

[0006] To address this problem, this invention proposes a novel method and apparatus for selecting foam image features in a flotation process based on sensitive mutual information. Summary of the Invention

[0007] To address the shortcomings of the prior art, this invention proposes a method and apparatus for selecting foam image features in a flotation process based on sensitive mutual information.

[0008] The technical solution proposed in this invention is:

[0009] A method for selecting features from flotation images in a flotation process based on sensitive mutual information includes the following steps:

[0010] S1, Extract foam image features based on the collected foam image video;

[0011] S2, preprocessing the features of the bubble image;

[0012] S3, calculate the Pearson coefficient between each image feature and the corresponding concentrate grade;

[0013] S4. Construct an optimization objective function for sensitive bubble image feature selection based on the mRMR criterion;

[0014] S5 employs a multi-group collaborative search particle swarm optimization algorithm for feature selection.

[0015] In one possible design, step S1 specifically involves extracting image features using image processing methods after preliminary consideration and screening.

[0016] In one possible design, the extracted image features in step S1 include at least the RGB mean, RGB red channel value, RGB green channel value, RGB blue channel value, grayscale mean, grayscale variance, hue, saturation, brightness, mean foam movement speed, variance of foam movement speed, foam shape, mean foam size, variance of foam size, kurtosis of foam size, skewness of foam size, load-bearing capacity, mean burst rate, variance of burst rate, energy, entropy, correlation, degree of foam stacking, inverse moment, moment of inertia, contrast, uniformity, second moment of angle, grayscale co-occurrence matrix, and roughness.

[0017] In one possible design, step S2 includes: outlier removal, removal of data items that are clearly inconsistent with the actual situation, and dimensionless processing, with the specific steps as follows:

[0018]

[0019] Among them, X i It is the i-th image feature data after dimensionless processing. It is the i-th original image feature data. and These are the maximum and minimum values ​​of the i-th image feature data, respectively.

[0020] In one possible design, step S3 specifically involves calculating the sensitivity of the foam image features to the concentrate grade using the Pearson coefficient, where the Pearson coefficient r between the i-th image feature and the concentrate grade is... i The calculation formula is as follows:

[0021] in

[0022] in, For the j-th sample of the i-th image feature, m is the number of samples for the i-th image feature, and yj is the concentrate grade value corresponding to the j-th image sample. for The mean, For y j The mean.

[0023] In one possible design, in step S4, the mRMR criterion is used to construct an optimization objective function for bubble image feature selection, incorporating the sensitivity coefficient. The detailed steps are as follows:

[0024] The first step is to calculate the mutual information between each foam image feature and the concentrate grade, using the following formula:

[0025]

[0026] Where, Y = {y 1 ,y 2 ,...,y j} represents concentrate grade data. The image feature value is represented as Meanwhile, the concentrate grade is y. j The probability, The image feature value is represented as The probability, p(y) j ) indicates that the concentrate grade is y j The probability of;

[0027] The second step involves further processing the selected features and calculating the mutual information between the selected features and the selected feature set. The calculation formula is as follows:

[0028]

[0029] Among them, S k For the k selected features, The image feature value is represented as Meanwhile, the selected image feature value is The probability, The image feature value is represented as The probability, The selected image feature value is The probability of;

[0030] The third step involves combining the sensitivity coefficients of image features and concentrate grade to select image features with high mutual information to concentrate grade, while avoiding redundancy among selected features. An optimization objective function for sensitive foam image feature selection is constructed, and the calculation formula is as follows:

[0031]

[0032] Where S is the number of selected image features.

[0033] In one possible design, step S5 specifically involves:

[0034] The first step involves using three populations with different inertia weight update methods, defined as global search, normal search, and fine search, respectively. Each population has 6 particles. The inertia weight ω1 update method for the global search population is shown below.

[0035]

[0036] The method for updating the inertia weight ω2 of a normal search population is as follows:

[0037] ω2=ω min +(ω max -ω min (k-1) / Maxgen;

[0038] The method for updating the inertia weight ω3 of the fine-search population is shown below, and the specific calculation is as follows:

[0039]

[0040] Where, ω max ω is the maximum value of the set inertia weight. min Here, k is the minimum set inertia weight, k is the current iteration number, Maxgen is the total number of iterations, cos(·) is the cosine function, and sig(·) is the sigmoid function.

[0041] The particle velocity and position update formulas for the three populations are as follows:

[0042] v i (k+1)=ωv i (k)+c1r1(pbest i -x i (k))+c2r2(gbest-x i (k));

[0043] x i (k+1)=x i (k)+v i (k+1);

[0044] Among them, v i (k+1) represents the search rate of the i-th particle in the (k+1)-th iteration, v i (k) The search rate of the i-th particle in the k-th iteration, where c1 and c2 are learning factors, c1 = c2 = 1.49445, and r1 and r2 are random numbers between (0,1). pbest i Let be the individual optimal solution for the i-th particle in the k-th iteration, and gbest be the swarm's overall optimal solution in the k-th iteration. ω represents the inertia weight, and x... i (k+1) represents the position of the i-th particle in the (k+1)-th iteration, x i (k) represents the position of the i-th particle in the k-th iteration;

[0045] The second step is to normalize the particle's position value to [-5, 5], and then encode the particle's position into binary using the following formula:

[0046]

[0047]

[0048] in, z represents the particle position normalized to the range [-5, 5]. i =1 indicates that X is selected. i Image features, z i =0 means X is not selected. i Image features;

[0049] The third step is to calculate the fitness function f(z) for each particle. i )

[0050]

[0051] Based on the above formula, find the particle in each of the three populations that maximizes the fitness function value, denoted as x. max1 (k), x max2 (k) and x max3 (k).

[0052] The fourth step involves mutating these three particles, generating 50 particles centered on each of them using a Gaussian distribution function, as shown below:

[0053]

[0054] Where σ is the standard deviation of the population particles, and μ is the mean of the population particles;

[0055] Fifth step, based on the fitness function f(z) from step three. i Calculate the fitness function of the 50 newly generated particles, select the particles with the largest fitness function values ​​from the three populations, and randomly select 5 particles from each of the three populations. Replace the particles in the original three populations with the selected 6 particles.

[0056] The sixth step is to compare the fitness function values ​​of the three population particles. If the search results of the global search or normal search population are better than those of the fine search population, then the positions of the global search or normal search population particles and the fine search population particles are swapped; otherwise, they are not swapped, and the process proceeds to the next iteration.

[0057] Step 7: Repeat steps 1 through 6 to obtain the final best search result, and thus obtain the selected image feature set.

[0058] In one possible design, step S5, the seventh step is to repeat steps one through six until the final optimal search result is obtained after 5000 iterations, thus obtaining the selected image feature set.

[0059] The present invention also provides a flotation process foam image feature selection device based on sensitive mutual information, including a memory, a control processor, and a computer program stored in the memory and executable on the control processor. The control processor executes the program to implement the aforementioned flotation process foam image feature selection method based on sensitive mutual information.

[0060] The present invention also provides a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the aforementioned method for selecting foam image features in a flotation process based on sensitive mutual information.

[0061] Compared with the prior art, the advantages of the present invention are:

[0062] This invention proposes a precise and efficient method for optimizing the selection of flotation bubble image features based on sensitive mutual information. This method comprehensively extracts bubble image attributes to obtain an appropriate number of image features. Addressing the large number and complex correlations of bubble image features during flotation, the method calculates the sensitivity coefficient between features and label data. Using the mRMR criterion, an optimization objective function for sensitive bubble image feature selection is constructed, significantly reducing redundancy among selected features, thereby lowering computational difficulty and model complexity, and greatly facilitating the final feature selection. A multi-population collaborative search particle swarm optimization algorithm is employed to optimize feature selection, upgrading the traditional single-population search to a three-population search. Mutation processing is performed on the results of each iteration, increasing the probability of the optimal value and improving iteration accuracy. Finally, several bubble image features with the highest mutual information with concentrate grade and the lowest redundancy are obtained.

[0063] The present invention utilizes a feature selection function combined with a multi-group collaborative search particle swarm optimization algorithm for feature selection. It comprehensively considers the sensitivity coefficients between various image features and label data, thus solving the problem of redundancy among image features in traditional feature selection. It also improves the search efficiency and accuracy in the optimization process, optimizes the traditional flotation process, and is suitable for further promotion in engineering practice. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0065] Figure 1 This is a flowchart of the present invention;

[0066] Figure 2 This is a flowchart of the multi-swarm collaborative search particle swarm algorithm of the present invention. Detailed Implementation

[0067] In at least one embodiment, Figure 1 As shown, a method for selecting features from foam images in a flotation process based on sensitive mutual information is proposed. This method includes the following steps:

[0068] S1: Extract foam image features based on the acquired foam image video.

[0069] In the flotation process, several image features are typically extracted as evaluation criteria to characterize the performance of the flotation process. In this invention, after preliminary consideration and screening, image processing methods were used to extract 30 image features: RGB mean, RGB red channel value, RGB green channel value, RGB blue channel value, grayscale mean, grayscale variance, hue, saturation, brightness, mean foam movement speed, variance of foam movement speed, foam shape, mean foam size, variance of foam size, kurtosis of foam size, skewness of foam size, load-bearing capacity, mean burst rate, variance of burst rate, energy, entropy, correlation, degree of foam stacking, inverse moment, moment of inertia, contrast, uniformity, second moment of angle, grayscale co-occurrence matrix, and roughness.

[0070] S2: Preprocess the features of the bubble image.

[0071] The large amount of foam image feature data obtained may contain data that does not reflect reality and may be noisy due to equipment failure, noise interference, or image recording errors. Therefore, these data need to be preprocessed. The detailed steps are as follows:

[0072] The first step is to remove outlier data; remove data items that are clearly inconsistent with the actual situation.

[0073] The second step involves processing the foam to be dimensionless, as different foam characteristics imply different physical meanings and have different dimensions and numerical ranges. The specific steps are as follows:

[0074]

[0075] Among them, X i It is the i-th image feature data after dimensionless processing. It is the i-th original image feature data. and These are the maximum and minimum values ​​of the i-th image feature data, respectively.

[0076] S3: Calculate the Pearson coefficient between each image feature and the corresponding concentrate grade.

[0077] In the process of using foam image features for modeling, condition identification, soft sensing, and optimization control of the foam flotation process, the large variety of image features can easily lead to problems such as computational complexity, low efficiency, and implementation difficulties. Therefore, the Pearson coefficient is used to calculate the sensitivity of foam image features to concentrate grade. The Pearson coefficient r between the i-th image feature and the concentrate grade is... i The calculation formula is as follows:

[0078] in in, For the j-th sample of the i-th image feature, m is the number of samples for the i-th image feature, y j Let j be the concentrate grade value corresponding to the j-th image sample. for The mean, For y j The mean.

[0079] The Pearson correlation coefficient measures the linear relationship between two quantities, with values ​​ranging from -1 to 1. The magnitude of the Pearson correlation coefficient indicates a linear correlation between image feature data and concentrate grade. If r... i A value ≥0.6 indicates a strong linear correlation between the feature and the label. This means that as the feature changes in a certain direction, the label data will change in that direction or the opposite direction, indicating that the concentrate grade is sensitive to this feature. If 0.4 ≤ r i If r < 0.6, it indicates a generally linear relationship between the feature and the label, and that the concentrate grade is generally sensitive to this feature. i If the value is less than 0.4, it indicates a weak linear relationship between this characteristic and concentrate grade. Therefore, it can be considered that this characteristic is insensitive.

[0080] S4: Constructing an optimized objective function for sensitive foam image feature selection based on the mRMR criterion.

[0081] Mutual information is a useful information metric in information theory, measuring the amount of information contained in one random variable about another. In the flotation process, the correlation between each flotation feature and concentrate grade varies. Some features have high mutual information with concentrate grade, while others have low mutual information. In such cases, the former features should be retained while the latter should be appropriately discarded to select the most useful features. However, in the feature set strongly correlated with concentrate grade, there are often cases where image features have high mutual information. To avoid redundancy of feature information, such features should also be discarded. In this invention, the sensitivity coefficient is combined with the mRMR (Minimal-Redundancy-Maximal-Relevance) criterion to construct an optimization objective function for flotation image feature selection. Detailed steps are as follows:

[0082] The first step is to calculate the mutual information between each foam image feature and the concentrate grade, using the following formula:

[0083]

[0084] Where, Y = {y 1 ,y 2 ,...,y j} represents concentrate grade data. The image feature value is represented as Meanwhile, the concentrate grade is y j The probability, The image feature value is represented as The probability, p(y) j ) indicates that the concentrate grade is y j The probability of.

[0085] The second step, to avoid data redundancy between selected features, involves further processing the selected features and calculating the mutual information between the selected features and the selected feature set. The calculation formula is as follows:

[0086]

[0087] Where Sk represents the k selected features, The image feature value is represented as Meanwhile, the selected image feature value is The probability, The image feature value is represented as The probability, The selected image feature value is The probability of.

[0088] The third step involves combining the sensitivity coefficients of image features and concentrate grade to select image features with high mutual information to concentrate grade, while avoiding redundancy among selected features. An optimization objective function for sensitive foam image feature selection is constructed, and the calculation formula is as follows:

[0089]

[0090] Where S is the number of selected image features.

[0091] S5: Feature optimization selection is performed using a multi-group cooperative search particle swarm optimization algorithm.

[0092] like Figure 2 As shown in the figure, in order to solve the optimization objective function of feature selection, as shown in formula (5), a particle swarm optimization algorithm based on multi-group cooperative search is proposed. The algorithm flowchart is shown in the figure. Figure 2 As shown.

[0093] The first step involves using three populations with different inertia weight update methods, defined as global search, normal search, and fine search, with 6 particles in each population. The inertia weight ω1 update method for the global search population is shown in formula (6), the inertia weight ω2 update method for the normal search population is shown in formula (7), and the inertia weight ω3 update method for the fine search population is shown in formula (8). The specific calculations are as follows:

[0094]

[0095] ω2=ω min +(ω max -ω min (k-1) / Maxgen (7)

[0096]

[0097] Where, ω max ω is the maximum value of the set inertia weight. min Here, k is the minimum set inertia weight, k is the current iteration number, Maxgen is the total number of iterations, cos(·) is the cosine function, and sig(·) is the sigmoid function.

[0098] The second step involves updating the particle velocity and position formulas for the three populations as follows:

[0099] v i (k+1)=ωv i (k)+c1r1(pbest i -x i (k))+c2r2(gbest-x i (k)) (9)

[0100] x i (k+1)=x i (k)+v i (k+1) (10)

[0101] Among them, v i (k+1) represents the search rate of the i-th particle in the (k+1)-th iteration, v i (k) The search rate of the i-th particle in the k-th iteration, where c1 and c2 are learning factors, c1 = c2 = 1.49445, and r1 and r2 are random numbers between (0,1). pbest i Let x be the optimal solution for the i-th particle in the k-th iteration, gbest be the optimal solution for the swarm in the k-th iteration, and ω be the inertia weight. The inertia weight update formulas for the three swarms are shown in (6)-(8). i (k+1) represents the position of the i-th particle in the (k+1)-th iteration, x i (k) represents the position of the i-th particle in the k-th iteration.

[0102] The third step is to normalize the particle's position value to [-5,5] and encode the particle's position into binary using the following formula.

[0103]

[0104]

[0105] in, z represents the particle position normalized to the range [-5, 5]. i =1 indicates that X is selected. i Image features, z i =0 means X is not selected. i Image features.

[0106] The fourth step is to calculate the fitness function f(z) for each particle. i )

[0107]

[0108] According to formula (12), find the particle that maximizes the fitness function value in each of the three populations, denoted as x. max1 (k), x max2 (k) and x max3 (k).

[0109] The fifth step is to mutate these three particles. Using each of the three particles as a center, generate 50 particles using a Gaussian distribution function. The Gaussian distribution function is as follows:

[0110]

[0111] Where σ is the standard deviation of the population particles, and μ is the mean of the population particles.

[0112] Step 6: Calculate the fitness function of the newly generated 50 particles according to formula (12), select the particle with the largest fitness function value from the three populations, and randomly select 5 particles from each population to replace the particles in the original three populations with 6 selected particles.

[0113] Step 7: Compare the fitness function values ​​of the three population particles. If the search results of the global search or normal search population are better than those of the fine search population, then swap the positions of the global search or normal search population particles with those of the fine search population particles; otherwise, do not swap and proceed to the next iteration.

[0114] Step 8: Repeat step (1) until the final best search result is obtained after 5000 iterations, and then the selected image feature set is obtained.

[0115] The present invention also provides a flotation process foam image feature selection device based on sensitive mutual information, including a memory, a control processor, and a computer program stored in the memory and executable on the control processor. The control processor executes the program to implement the aforementioned flotation process foam image feature selection method based on sensitive mutual information.

[0116] The present invention also provides a control system, including the aforementioned flotation process foam image feature selection device based on sensitive mutual information.

[0117] The modulation method according to embodiments of the present disclosure can be programmed into a computer program and stored on a computer-readable storage medium. When the computer program is executed by a processor, the fault-tolerant control method described above can be implemented.

[0118] Examples of computer-readable storage media include: read-only memory (ROM), random access programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-RLTH, etc. BD-RE, Blu-ray or optical disc storage, hard disk drive (HDD), solid-state drive (SSD), card storage (such as multimedia cards, secure digital (SD) cards, or ultra-fast digital (XD) cards), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state drive, and any other device configured to store a computer program and any associated data, data files, and data structures in a non-transitory manner and to provide the computer program and any associated data, data files, and data structures to a processor or computer so that the processor or computer can execute the computer program. In one example, the computer program and any associated data, data files, and data structures are distributed across a networked computer system, such that the computer program and any associated data, data files, and data structures are stored, accessed, and executed in a distributed manner through one or more processors or computers.

[0119] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments. Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the various illustrative components, blocks, modules, circuits, and steps are described above in a generalized form in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. Those skilled in the art may implement the described functionality in different ways for each particular application, but such implementation decisions should not be construed as departing from the scope of the invention. The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein may be implemented or performed using a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. The general-purpose processor may be a microprocessor, but in alternatives, it may be any conventional processor, battery compartment control board, micro battery compartment control board, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, multiple microprocessors, one or more microprocessors cooperating with a DSP core, or any other such configuration. The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of both. The software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information to / from the storage medium. In an alternative, the storage medium may be integrated into the processor. The processor and storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and storage medium may reside as discrete components in the user terminal. In one or more exemplary embodiments, the described functionality may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functionality may be stored or transmitted therefrom as one or more instructions or code on a computer-readable medium.Computer-readable media includes both computer storage media and communication media, encompassing any medium that facilitates the transfer of a computer program from one location to another. Storage media can be any available medium accessible to a computer. By way of example and not limitation, such computer-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disc storage, disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and is accessible to a computer. Any connection is also legitimately referred to as computer-readable media. For example, if software is transmitted from a website, a central control computer, or other remote source using coaxial cable, fiber optic cable, twisted pair, digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwave, then that coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of media. As used in this article, disk and disc include compact discs (CDs), laser discs, optical discs, digital multi-purpose discs (DVDs), floppy disks, and Blu-ray discs. Disks typically reproduce data magnetically, while discs reproduce data optically using lasers. Combinations of these should also be included within the scope of computer-readable media.

[0120] Although the methods described above are illustrated and depicted as a series of actions for the sake of simplicity, it should be understood and appreciated that these methods are not limited by the order of the actions, as some actions may occur in a different order and / or concurrently with other actions from the illustrations and descriptions herein or not illustrated and described herein but which may be understood by those skilled in the art, according to one or more embodiments.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

[0122] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for selecting features from flotation images in a flotation process based on sensitive mutual information, characterized in that, Includes the following steps: S1, extract foam image features based on the acquired foam image video; Step S1 specifically involves extracting image features using image processing methods after preliminary consideration and screening; Step In S1, the extracted image features include at least the RGB mean, RGB red channel value, RGB green channel value, RGB blue channel value, gray level mean, gray level variance, hue, saturation, brightness, foam movement speed mean, foam movement speed variance, foam shape, foam size mean, foam size variance, foam size kurtosis, foam size skewness, load-bearing capacity, burst rate mean, burst rate variance, energy, entropy, correlation, foam stacking degree, inverse moment, moment of inertia, contrast, uniformity, angular second moment, gray-level co-occurrence matrix, and roughness. S2, preprocessing the bubble image features; step S2 includes: outlier removal, removing data items that are obviously inconsistent with the actual situation, and dimensionless processing, the specific steps are as follows: Among them, X i It is the i-th image feature data after dimensionless processing. It is the i-th original image feature data. and These are the maximum and minimum values ​​of the i-th image feature data, respectively; S3, calculate the Pearson coefficient between each image feature and the corresponding concentrate grade; S4, Constructing an optimization objective function for sensitive foam image feature selection based on the mRMR criterion; In step S4, the optimization objective function for foam image feature selection is constructed using the mRMR criterion in conjunction with the sensitivity coefficient. The detailed steps are as follows: The first step is to calculate the mutual information between each foam image feature and the concentrate grade, using the following formula: Where, Y = {y 1 ,y 2 ,...,y j } represents concentrate grade data. The image feature value is represented as Meanwhile, the concentrate grade is y j The probability, The image feature value is represented as The probability, p(y) j ) indicates that the concentrate grade is y j The probability of; The second step involves further processing the selected features and calculating the mutual information between the selected features and the selected feature set. The calculation formula is as follows: Among them, S k For the k selected features, The image feature value is represented as Meanwhile, the selected image feature value is The probability, The image feature value is represented as The probability, The selected image feature value is The probability of; The third step involves combining the sensitivity coefficients of image features and concentrate grade to select image features with high mutual information to concentrate grade, while avoiding redundancy among selected features. An optimization objective function for sensitive foam image feature selection is constructed, and the calculation formula is as follows: Where S is the number of selected image features; S5: Feature optimization selection is performed using a multi-group collaborative search particle swarm algorithm.

2. The method for selecting foam image features in a flotation process based on sensitive mutual information as described in claim 1, characterized in that, In step S3, specifically, the sensitivity of foam image features to concentrate grade is calculated using the Pearson coefficient. The Pearson coefficient r between the i-th image feature and the concentrate grade is... i The calculation formula is as follows: in in, For the j-th sample of the i-th image feature, m is the number of samples for the i-th image feature, y j Let j be the concentrate grade value corresponding to the j-th image sample. for The mean, For y j The mean.

3. The method for selecting foam image features in a flotation process based on sensitive mutual information as described in claim 1, characterized in that, Step S5 specifically includes: The first step involves using three populations with different inertia weight update methods, defined as global search, normal search, and fine search, respectively. Each population has 6 particles. The inertia weight ω1 update method for the global search population is shown below. The method for updating the inertia weight ω2 of a normal search population is as follows: ω2=ω min +(ω max -oh min )(k-1) / Maxgen; The method for updating the inertia weight ω3 of the fine-search population is shown below, and the specific calculation is as follows: Where, ω max ω is the maximum value of the set inertia weight. min Here, k is the minimum set inertia weight, k is the current iteration number, Maxgen is the total number of iterations, cos(·) is the cosine function, and sig(·) is the sigmoid function. The particle velocity and position update formulas for the three populations are as follows: v i (k+1)=ωv i (k)+c1r1(pbest i -x i (k))+c2r2(gbest-x i (k)); x i (k+1)=x i (k)+v i (k+1); Among them, v i (k+1) represents the search rate of the i-th particle in the (k+1)-th iteration, v i (k) represents the search rate of the i-th particle in the k-th iteration, c1 and c2 are learning factors, c1 = c2 = 1.49445, and r1 and r2 are random numbers between (0,1). i Let be the individual optimal solution for the i-th particle in the k-th iteration, and gbest be the swarm's overall optimal solution in the k-th iteration. ω represents the inertia weight, and x... i (k+1) represents the position of the i-th particle in the (k+1)-th iteration, x i (k) represents the position of the i-th particle in the k-th iteration; The second step is to normalize the particle's position value to [-5, 5], and then encode the particle's position into binary using the following formula: in, z represents the particle position normalized to the range [-5, 5]. i =1 indicates that X is selected. i Image features, z i =0 means X is not selected. i Image features; The third step is to calculate the fitness function f(z) for each particle. i ) Based on the above formula, find the particle in each of the three populations that maximizes the fitness function value, denoted as x. max1 (k), x max2 (k) and x max3 (k); The fourth step involves mutating these three particles, generating 50 particles centered on each of them using a Gaussian distribution function, as shown below: Where σ is the standard deviation of the population particles, and μ is the mean of the population particles; Fifth step, based on the fitness function f(z) from step three. i Calculate the fitness function of the 50 newly generated particles, select the particle with the largest fitness function value from the three populations, and randomly select 5 particles from each of the three populations to replace the original particles in the three populations with 6 selected particles; in the sixth step, compare the fitness function values ​​of the particles in the three populations. If the search result of the global search or normal search population is better than the search result of the fine search population, then swap the positions of the particles in the global search or normal search population with the positions of the particles in the fine search population; otherwise, do not swap, and proceed to the next iteration. Step 7: Repeat steps 1 through 6 to obtain the final best search result, and thus obtain the selected image feature set.

4. The method for selecting foam image features in a flotation process based on sensitive mutual information as described in claim 3, characterized in that, In step S5, the seventh step is to repeat steps one through six until the final best search result is obtained after 5000 iterations, thus obtaining the selected image feature set.

5. A flotation process foam image feature selection device based on sensitive mutual information, characterized in that, The method includes a memory, a control processor, and a computer program stored in the memory and executable on the control processor, wherein the control processor executes the program to implement the flotation process foam image feature selection method based on sensitive mutual information as described in any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions for causing a computer to perform the flotation process foam image feature selection method based on sensitive mutual information as described in any one of claims 1-4.

Images