Uranium mining and metallurgy ore sorting method and system based on machine vision recognition

Abstract

The invention relates to the technical field of image processing, and discloses a uranium mining and metallurgy ore sorting method and system based on machine vision recognition, and the method comprises the steps: obtaining ore image data; preprocessing the ore image data, and calculating a gray level co-occurrence matrix of the image through a gray level formula; according to the gray-level co-occurrence matrix, a template storage library is obtained through a feature matching algorithm; performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; and obtaining a sorting result through a kernel density estimation clustering algorithm according to the target image and the template storage library. The method has the following effects that the ore area gray level co-occurrence matrix entropy is introduced to serve as the characteristic value, and the ore sorting accuracy and efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a uranium mining and metallurgy ore sorting method and system based on machine vision recognition. Background Art

[0002] In recent years, due to the country's strong investment in the development of the nuclear industry, more and more nuclear power plants have been built. As a fuel source for uranium power plants, the technology of uranium mining and metallurgy ore sorting has received widespread attention. Most of the uranium mines in southern my country are in granite vein type, and the ore and vein rock are not easy to separate, and a large amount of tailings will be produced during the mining process. There are a large number of exploitable ore resources remaining in these tailings. Therefore, the use of ore identification and sorting technology can realize the recycling of uranium ore resources in mining tailings, which can greatly improve the efficiency of ore mining and utilization, and conform to the concept of sustainable development. With the continuous development of computer technology, traditional ore identification, sorting and classification can also be automatically processed through visual recognition and deep learning. Compared with manual or infrared analysis, it not only achieves higher recognition accuracy, but also improves efficiency.

[0003] The existing technology for ore identification and sorting mainly uses manual analysis or machine vision recognition. Ore particles are screened and visually identified manually. This method mainly relies on the operating experience of professional technicians, which is time-consuming and labor-intensive, inefficient, and the accuracy is limited by the professional level of the technicians. With the development of machine vision technologies such as image recognition and deep learning, research and application based on machine vision technology has become a hot topic. However, there are still some shortcomings in ore sorting technology using machine vision recognition. It mainly uses deep learning for ore identification. Due to the complexity of ore types and the difficulty of identification, it is necessary to design a convolutional neural network with a deeper model. The identification of deep models will have problems with gradient explosion or gradient disappearance, which increases the difficulty of training, makes model convergence difficult, and has low recognition efficiency and accuracy. Summary of the invention

[0004] The present invention provides a uranium mining and metallurgy ore sorting method and system based on machine vision recognition, so as to realize accurate sorting of ore.

[0005] In the first aspect, in order to solve the above technical problems, the present invention provides a uranium mining and metallurgy ore sorting method based on machine vision recognition, comprising: Acquire ore image data; Preprocessing the ore image data, and calculating the grayscale co-occurrence matrix of the image through a grayscale formula; According to the gray level co-occurrence matrix, a template repository is obtained by a feature matching algorithm; Performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; A sorting result is obtained according to the target image and the template repository by using a kernel density estimation clustering algorithm.

[0006] In an optional implementation, the ore image data is preprocessed and the gray level co-occurrence matrix of the image is calculated by a gray level formula, including: Performing median filtering on the ore image data to obtain noise-reduced ore image data; According to the de-noised ore image data, a gray level co-occurrence matrix of the image is established, including: in, is the gray-level co-occurrence matrix, is the pixel in the image, Representation and Adjacent pixels, Represents all the pixels in the image. is the Kronecker function, when =0, and =0, the function value is 0; The entropy of the gray-level co-occurrence matrix is ​​calculated by the following formula: in, is the entropy of the gray-level co-occurrence matrix, It is the intermediate process quantity.

[0007] In an optional implementation, obtaining a template repository through a feature matching algorithm based on the gray level co-occurrence matrix includes: Extracting first feature points of a pre-stored typical ore region image at a preset first angle, wherein the feature points include grayscale variance, grayscale average gradient, grayscale average and entropy; Extracting second feature points of a pre-stored typical ore region image at a preset second angle, wherein the feature points include grayscale variance, grayscale average gradient, grayscale average and entropy; inputting the first feature point and the pixel coordinates corresponding to the feature point into a template repository; Calculating the Euclidean distance between the first feature point and the second feature point; According to the Euclidean distance and the entropy of the gray-level co-occurrence matrix as eigenvalues, the feature point coordinates and the Euclidean distance are saved in the template repository through an optical flow tracking algorithm.

[0008] In an optional implementation, the performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted includes: Performing threshold segmentation processing on the image data according to a preset image threshold; Obtain separate target images through watershed segmentation; The individual target images are subjected to burr filtering processing to obtain target images to be sorted.

[0009] In an optional implementation, obtaining a sorting result by a kernel density estimation clustering algorithm according to the target image and the template repository includes: According to the target image, calculating the sorting parameters of the target image, the sorting parameters including grayscale variance, grayscale average gradient, grayscale average and grayscale entropy; According to the sorting parameters, sample points are obtained by weighted averaging; The probability density function value of the kernel density estimate is calculated by the following formula: in, is the probability density function value, is the bandwidth, is the number of sample points, is the kernel function, satisfying , for i The horizontal coordinate of the sample point number; According to the probability density function value, a kernel density histogram is obtained, including: in is the value of the density matrix in the kernel density histogram, and are the coordinates of the kernel density histogram, and They are i The horizontal and vertical coordinates of the sample points; Distinguishing high-density areas from low-density areas according to the kernel density histogram in combination with a preset density threshold; The local maximum value of the high-density area is taken as the cluster center: The cluster at the center of the cluster is taken as the sorting result.

[0010] In an optional implementation, the watershed segmentation includes: Divide the image into different regions based on the gradient information of the image; The areas with higher grayscale values ​​in the image are taken as peaks, and the areas with lower grayscale values ​​are taken as valleys, which are input into the water flow simulation model to obtain the boundary lines of different areas.

[0011] In an optional implementation, after obtaining the sorting result by using a kernel density estimation clustering algorithm according to the target image and the template repository, the method further includes: The sorting result is compared with the manual sorting result, and the density threshold is corrected.

[0012] In a second aspect, the present invention provides a uranium mining and metallurgy ore sorting system based on machine vision recognition, comprising: A data acquisition module, used for acquiring ore image data; A grayscale matrix module is used to pre-process the ore image data and calculate the grayscale co-occurrence matrix of the image through a grayscale formula; A template generation module, used to obtain a template repository according to the gray level co-occurrence matrix through a feature matching algorithm; A target extraction module is used to perform threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; The clustering sorting module is used to obtain a sorting result through a kernel density estimation clustering algorithm according to the target image and the template repository.

[0013] In a third aspect, the present invention further provides an electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the uranium mining and metallurgy ore sorting method based on machine vision recognition as described above is implemented.

[0014] In a fourth aspect, the present invention further provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute any one of the above-mentioned uranium mining and metallurgy ore sorting methods based on machine vision recognition.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention discloses a uranium mining and metallurgy ore sorting method and system based on machine vision recognition, the method comprising acquiring ore image data; preprocessing the ore image data, and calculating the grayscale co-occurrence matrix of the image through a grayscale formula; obtaining a template repository through a feature matching algorithm according to the grayscale co-occurrence matrix; performing threshold segmentation processing on the image data to obtain a number of target images to be sorted; and obtaining a sorting result through a kernel density estimation clustering algorithm according to the target image and the template repository. This method aims at the problem that the ore particles are of different sizes and mixed with a large amount of ashes in a complex mine environment, adopts a median filter denoising process to improve the image quality, and uses the LK matching algorithm combined with the entropy of the grayscale co-occurrence matrix of the ore region as the eigenvalue for optimization, thereby realizing the accurate matching and segmentation of typical ore region images at different angles. In the target extraction stage, by implementing threshold segmentation and watershed segmentation on the image, a single ore target can be accurately separated, and holes and burrs can be removed at the same time, thereby ensuring the accuracy of subsequent analysis. Furthermore, in the ore sorting process, the present invention adopts a method combining kernel density estimation with K-means clustering to automatically determine the optimal K value, overcoming the shortcomings of the traditional K-means method, such as strong subjectivity in K value selection and unstable sorting results, and ensuring the objectivity and consistency of ore classification. This method significantly enhances the accuracy of ore sorting and reduces the influence of human factors. At the same time, by comparing the sorting results with the actual mineral image features, the reliability of the sorting effect is verified, and the efficiency and accuracy of mineral identification are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic flow chart of a method for uranium mining and metallurgy ore separation based on machine vision recognition provided by the first embodiment of the present invention; Figure 2 It is a schematic diagram of the structure of a uranium mining and metallurgy ore separation system based on machine vision recognition provided by the second embodiment of the present invention. DETAILED DESCRIPTION

[0017] 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.

[0018] In recent years, due to the country's strong investment in the development of the nuclear industry, more and more nuclear power plants have been built. As a fuel source for uranium power plants, the technology of uranium mining and metallurgy ore sorting has received widespread attention. Most of the uranium mines in southern my country are in granite vein type, and the ore and vein rock are not easy to separate, and a large amount of tailings will be produced during the mining process. There are a large number of exploitable ore resources remaining in these tailings. Therefore, the use of ore identification and sorting technology can realize the recycling of uranium ore resources in mining tailings, which can greatly improve the efficiency of ore mining and utilization, and conform to the concept of sustainable development. With the continuous development of computer technology, traditional ore identification, sorting and classification can also be automatically processed through visual recognition and deep learning. Compared with manual or infrared analysis, it not only achieves higher recognition accuracy, but also improves efficiency.

[0019] The existing technology for ore identification and sorting mainly uses manual analysis or machine vision recognition. Ore particles are screened and visually identified manually. This method mainly relies on the operating experience of professional technicians, which is time-consuming and labor-intensive, inefficient, and the accuracy is limited by the professional level of the technicians. With the development of machine vision technologies such as image recognition and deep learning, research and application based on machine vision technology has become a hot topic. However, there are still some shortcomings in ore sorting technology using machine vision recognition. It mainly uses deep learning for ore identification. Due to the complexity of ore types and the difficulty of identification, it is necessary to design a convolutional neural network with a deeper model. The identification of deep models will have problems with gradient explosion or gradient disappearance, which increases the difficulty of training, makes model convergence difficult, and has low recognition efficiency and accuracy.

[0020] In order to solve the above technical problems, refer to Figure 1 The first embodiment of the present invention provides a method for uranium mining and metallurgy ore separation based on machine vision recognition, comprising the following steps: S11, acquiring ore image data; S12, preprocessing the ore image data, and calculating the grayscale co-occurrence matrix of the image by using a grayscale formula; S13, obtaining a template repository through a feature matching algorithm according to the gray level co-occurrence matrix; S14, performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; S15, obtaining a sorting result by a kernel density estimation clustering algorithm according to the target image and the template repository.

[0021] In step S11, ore image data is acquired.

[0022] In one embodiment, a high-resolution industrial CCD or CMOS camera is used to capture images of ore samples at an actual ore mining or processing site; for stationary ore samples, a high-resolution scanner can be used to obtain detailed image information. In order to ensure image quality, good lighting conditions are required, and in special cases, specific light sources are required to highlight the characteristics of the ore. At the same time, ensure that the parameters in the image acquisition process are consistent, such as lighting intensity, angle, and camera settings, to reduce the impact of external factors on image quality. Furthermore, the collected original images often contain noise and other interference information, and need to be denoised, contrast enhanced, and the image size adjusted to a uniform size.

[0023] In step S12, the ore image data is preprocessed, and the grayscale co-occurrence matrix of the image is calculated by a grayscale formula.

[0024] In one embodiment, the ore image data is preprocessed and the gray level co-occurrence matrix of the image is calculated by a gray level formula, including: Performing median filtering on the ore image data to obtain noise-reduced ore image data; According to the de-noised ore image data, a gray level co-occurrence matrix of the image is established, including: in, is the gray-level co-occurrence matrix, is the pixel in the image, Representation and Adjacent pixels, Represents all the pixels in the image. is the Kronecker function, when =0, and =0, the function value is 0; The entropy of the gray-level co-occurrence matrix is ​​calculated by the following formula: in, is the entropy of the gray-level co-occurrence matrix, It is the intermediate process quantity.

[0025] It is worth mentioning that median filtering is a nonlinear digital image processing technology that is widely used to remove noise in images, especially for salt and pepper noise, that is, random bright or dark spots. The basic principle of median filtering is to replace the gray value of each pixel with the median of the gray values ​​of all pixels in its neighborhood. Therefore, it has better edge protection ability and will not easily cause edge blurring like mean filtering.

[0026] It is worth mentioning that the Gray-Level Co-occurrence Matrix (GLCM). GLCM is a statistical method used to describe the spatial relationship between pixels in an image. It captures the texture information of an image by recording the co-occurrence frequency between different gray levels. Specifically, each element P(i, j) in GLCM represents the probability that one pixel has gray level i and the other has gray level j, which is d away from each other in a specific direction. If the original image is in color, it needs to be converted into a grayscale image. The formula for grayscale conversion is: . Then determine the calculated distance and direction, and traverse the entire image, count the number of occurrences of each pair of adjacent pixel grayscale combinations, and fill these counts into a two-dimensional array, which is the grayscale co-occurrence matrix. In order to make the grayscale co-occurrence matrix a probability distribution, it needs to be normalized, that is, each element is divided by the sum of all elements. The resulting matrix reflects the relative frequency of occurrence of different grayscale pairs at a given direction and distance. The entropy of the grayscale co-occurrence matrix is ​​an indicator of image complexity and randomness. The higher the entropy, the higher the complexity and randomness of the image.

[0027] In step S13, a template repository is obtained according to the gray level co-occurrence matrix through a feature matching algorithm.

[0028] In one embodiment, the template repository is obtained through a feature matching algorithm based on the grayscale co-occurrence matrix, including: extracting a first feature point of a typical ore area image pre-stored at a preset first angle, the feature point including grayscale variance, grayscale average gradient, grayscale average and entropy; extracting a second feature point of a typical ore area image pre-stored at a preset second angle, the feature point including grayscale variance, grayscale average gradient, grayscale average and entropy; inputting the first feature point and the pixel coordinates corresponding to the feature point into the template repository; calculating the Euclidean distance between the first feature point and the second feature point; and using the Euclidean distance and the entropy of the grayscale co-occurrence matrix as eigenvalues, saving the feature point coordinates and the Euclidean distance to the template repository through an optical flow tracking algorithm.

[0029] It is worth noting that since the ore may show different visual characteristics at different angles, selecting multiple angles helps to establish a comprehensive feature library. For several dimensions of feature points, the grayscale variance reflects the degree of change of grayscale values ​​in the local area of ​​the image. A higher variance usually means that the area contains more details or edge information; the grayscale average gradient indicates the speed of change of the grayscale value of pixels in the image, which can be used to detect the edge and texture information in the image; the grayscale average is the arithmetic average of the grayscale values ​​of all pixels in the local area of ​​the image, providing information about the brightness level; entropy can measure the complexity of information in the local area of ​​the image. The higher the entropy, the richer or more random the information in the area. The entropy of the grayscale co-occurrence matrix reflects the texture complexity in the local area of ​​the image. When calculating the Euclidean distance, the entropy of the grayscale co-occurrence matrix is ​​used as a weight factor to adjust the distance between different feature points. The purpose of this is to make the area with richer texture information have higher importance in the matching process. According to the weighted Euclidean distance calculated above, the movement of the corresponding feature points between the two images is determined, and the change of pixels in the image sequence over time can be estimated through the optical flow tracking algorithm, so as to find the matching feature point pairs. The template repository includes feature information such as the location coordinates, grayscale variance, grayscale average gradient, grayscale average, and entropy of feature points. It also stores the Euclidean distance between feature points calculated by the optical flow tracking algorithm.

[0030] It is worth noting that the optical flow tracking algorithm is the Lucas-Kanade (LK) optical flow method, which is used to estimate the movement of pixels in an image sequence. The LK optical flow method is based on a series of assumptions and uses these assumptions to calculate the displacement of pixels between adjacent frames. It is assumed that the brightness value of the object surface will not change in a short period of time, which means that when a pixel moves from one frame to the next, its grayscale value remains unchanged; the movement of the pixel is smooth, that is, the displacement of the pixel is small in a very short time interval; adjacent pixels have similar movements, that is, all pixels in a small window have the same optical flow vector. Based on the above assumptions, the LK optical flow method estimates the optical flow by comparing the grayscale changes between two adjacent frames, and estimates the optical flow vector through the gradient information in the local window. This method uses the optical flow tracking algorithm to obtain the Euclidean distance between feature points.

[0031] In step S14, threshold segmentation processing is performed on the image data to obtain a plurality of target images to be sorted.

[0032] In one embodiment, the image data is subjected to threshold segmentation processing to obtain a plurality of target images to be sorted, including: the image data is subjected to threshold segmentation processing according to a preset image threshold; a separate target image is obtained by watershed segmentation; the separate target image is subjected to burr filtering processing to obtain a target image to be sorted.

[0033] It is worth mentioning that watershed segmentation is a morphological image segmentation method that simulates topographic features in geography, such as mountains, valleys, and catchment basins. In image processing, this algorithm treats the image as a topological surface, where the height of each pixel represents its grayscale value. The watershed algorithm defines different regions by identifying local minima in the image, and these regions expand as the "flood" rises until they meet at the boundary. This process eventually forms closed contours that are used to distinguish different objects. This method selects some initial markers as foreground or background. Using the gradient image as input, as well as the above-mentioned markers, the watershed algorithm is applied. The algorithm simulates the process of flooding on the gradient image, and when two regions with different labels are about to merge, a boundary is established to prevent them from merging. The result of the watershed segmentation algorithm is a labeling matrix, in which each region has a unique integer label. The individual segmented regions can be extracted based on these labels, and the results can be processed for subsequent target sorting.

[0034] It is worth noting that for images with uneven illumination or complex backgrounds, local adaptive threshold methods such as local binary pattern (LBP) or adaptive threshold methods based on neighborhood average can be used. For images with uniform illumination, the Otsu algorithm can automatically find the optimal threshold to maximize the inter-class variance and segment the entire image using a fixed threshold. After threshold segmentation, in order to eliminate small noise points or fill holes, morphological operations such as opening or closing operations can be used, and connected region marking can be used to identify and separate different objects.

[0035] In one embodiment, in the field of image processing, during the target extraction process, due to noise, illumination changes, blurred object boundaries, etc., burrs (i.e., irregular protruding points or small isolated pixels) may appear on the edge of the extracted target. These burrs will interfere with subsequent analysis, such as feature extraction, shape recognition, etc., so filtering is required to remove them. Gaussian filtering is a linear smoothing filter that can effectively remove high-frequency noise and smooth the image by using a Gaussian function as a weight template to calculate the weighted average of pixels in the neighborhood, but it may blur image details. Wavelet transform can be used for multi-resolution analysis, and by decomposing the image into different frequency bands, noise and burrs can be processed at different scales. This method helps to protect important structural information of the image. For specific types of burrs, templates can be defined for matching and correction. For example, in image processing after document scanning, predefined templates can be used to detect and remove small burrs on the edge of the text.

[0036] It is worth noting that the opening operation is a process of corrosion followed by expansion. Erosion is a process of eliminating boundary points and shrinking the boundary inward. It is achieved by comparing the structural element with every pixel in the image and removing those areas that cannot completely contain the structural element. This process helps to remove small noise points in the image and disconnect narrower connected parts. Dilation is to expand the highlight area in the image so that its boundary extends outward. The dilation operation is achieved by comparing the structural element with every pixel in the image and filling the parts covered by the structural element. This process can help fill small holes and broken parts in the image. Since corrosion first removes these small points and subsequent dilation cannot restore them, the opening operation can effectively clean up these isolated noise points; for very narrow connections between two objects, corrosion may disconnect them, thereby separating the two objects; although corrosion will reduce the size of the object, through subsequent dilation, larger connected domains are usually restored, thereby ensuring the integrity of the main information. The closing operation is a process of dilation followed by corrosion, which has the opposite effect.

[0037] In step S15, a sorting result is obtained by using a kernel density estimation clustering algorithm according to the target image and the template repository.

[0038] In one implementation, based on the target image, the sorting parameters of the target image are calculated, and the sorting parameters include grayscale variance, grayscale average gradient, grayscale average and grayscale entropy; According to the sorting parameters, sample points are obtained by weighted averaging; The probability density function value of the kernel density estimate is calculated by the following formula: in, is the probability density function value, is the bandwidth, is the number of sample points, is the kernel function, satisfying , for i The horizontal coordinate of the sample point number; According to the probability density function value, a kernel density histogram is obtained, including: in is the value of the density matrix in the kernel density histogram, and are the coordinates of the kernel density histogram, and They are i The horizontal and vertical coordinates of the sample points; Distinguishing high-density areas from low-density areas according to the kernel density histogram in combination with a preset density threshold; The local maximum value of the high-density area is taken as the cluster center: The cluster at the center of the cluster is taken as the sorting result.

[0039] It is worth mentioning that Kernel Density Estimation (KDE) is a non-parametric probability density function estimation method that can estimate the probability density function of data without knowing the distribution form of the data. Features are extracted from the target image, and the features of all images are merged into a large feature set to form a high-dimensional feature space. Kernel density estimation is used to estimate the probability density in the feature space. The appropriate bandwidth controls the smoothness of the estimation. According to the results of kernel density estimation, different cluster centers can be defined by finding local maximum points. Alternatively, each feature point can be directly assigned to the nearest high-density area to achieve clustering. If you know which categories the template belongs to, you can also classify it by comparing the density distribution of the target image features with the template features. Finally, based on the clustering or classification results, the images are sorted. For example, if the goal is to find the target image that is most similar to a specific template, the similarity between the target image features and the density distribution of each template category can be calculated and sorted accordingly.

[0040] It is worth noting that the kernel function is a symmetric smooth function with an area of ​​1, and a Gaussian kernel function can be used. For each observed data point, the kernel function is applied as the probability density value, and all these kernel functions are superimposed to form an overall probability density estimate. A bandwidth that is too small will cause the estimate to be too volatile and may capture a lot of noise; a bandwidth that is too large will make the estimate too smooth and lose important details. The bandwidth can be automatically determined by statistical methods such as cross-validation, or set according to empirical rules. The kernel density histogram is actually a graph drawn using the probability density function obtained by kernel density estimation. Compared with the traditional histogram, the kernel density histogram is not affected by the bin selection, so it can more accurately reflect the true distribution characteristics of the data.

[0041] It is worth noting that for a given cluster, the cluster center is the arithmetic mean of all sample points in the cluster in each feature dimension. If the data point is described by multiple features, the cluster center is also a vector with the same number of features. Before the K-means algorithm starts, k initial cluster centers need to be randomly selected. In each iteration, the data point is assigned to the cluster corresponding to the nearest cluster center according to the distance between each data point and all current cluster centers. The commonly used distance metric here is Euclidean distance, but other distance metrics can also be used. Once all data points are assigned to the corresponding clusters, the new cluster center of each cluster needs to be recalculated. The new cluster center is the mean of all data points in the cluster. This process will be repeated until the cluster center no longer changes significantly or the preset maximum number of iterations is reached. Further, in order to improve the selection of the initial cluster center, the K-means++ algorithm was proposed. Its main idea is to select cluster centers more dispersedly in the initialization stage, thereby reducing the possibility of the algorithm falling into a local optimal solution. Specifically, the first center point is randomly selected, and subsequent center points tend to be selected away from the existing center points.

[0042] In one embodiment, after obtaining the sorting result by the kernel density estimation clustering algorithm according to the target image and the template repository, it also includes: comparing the sorting result with the manual sorting result and correcting the density threshold. After sorting the objects in the target image using the kernel density estimation clustering algorithm, a domain expert or an experienced operator manually classifies the same set of data according to professional knowledge, and inputs the classification results into the system. The system compares the results of automatic sorting with the results of manual sorting to evaluate the accuracy of automatic sorting. Consider the proportion of objects correctly identified as a certain class by automatic sorting to all objects automatically identified as that class as accuracy. Display the misclassification between different categories as a confusion matrix. Based on the results of the comparative analysis, adjust the density threshold in the kernel density estimation. If it is found that some low-density areas are misjudged as new clusters, the occurrence of this situation can be reduced by increasing the density threshold; on the contrary, if the objects in the high-density area are not correctly classified into the same cluster, the density threshold needs to be appropriately lowered. Until the results of automatic sorting and the results of manual sorting reach a satisfactory degree of matching.

[0043] In summary, the present invention proposes a uranium mining and smelting ore sorting method based on machine vision recognition, aiming to solve multiple key problems existing in current ore sorting technology, including insufficient image segmentation accuracy in complex mine environments, difficulty in deep learning model training, and lack of objectivity in K value selection in K-means clustering algorithm. By integrating advanced image processing technology and pattern recognition algorithm, the method can significantly improve the recovery efficiency and resource utilization of ore in the uranium mining and smelting process. In terms of image processing, the present invention adopts grayscale co-occurrence matrix entropy as the eigenvalue, and optimizes it in combination with the LK matching algorithm, thereby improving the accuracy and efficiency of image matching. The LK matching algorithm uses the concept of optical flow to track the displacement of feature points in the image, and calculates these displacements to achieve accurate matching of typical ore area images at different angles. In addition, the present invention also introduces grayscale co-occurrence matrix entropy as the eigenvalue, which enhances the robustness and adaptability in the image matching process. By preprocessing the collected standard ore images, including median filtering denoising, feature extraction, and establishing an LK matching template repository based on grayscale co-occurrence matrix entropy, efficient conversion from original images to effective target images is achieved. In view of the difficulties in ore sorting, the present invention adopts a method combining kernel density estimation with K-means clustering algorithm to automatically determine the K value, thereby avoiding the uncertainty caused by the need to manually set the K value in the traditional method. This method first constructs a probability distribution map based on the kernel density estimation of the sample points, and then determines the optimal number of clusters according to the number of peaks in the distribution map. This adaptive way of determining the K value not only improves the accuracy of cluster analysis, but also reduces the influence of human factors. At the same time, the present invention also uses a variety of sorting parameter factors, such as grayscale variance, grayscale average gradient, grayscale average and grayscale entropy, to comprehensively evaluate each target to ensure that the sorting results are more accurate and reliable. Furthermore, the present invention realizes the effective separation of a single target through threshold segmentation and watershed segmentation technology, and takes into account the removal of holes and burrs and other details in the processing process, further ensuring the quality of the target image. The whole scheme not only improves the accuracy of ore sorting, but also shows extremely high stability and reliability in practical applications. The uranium mining and metallurgy ore sorting method based on machine vision recognition proposed by the present invention not only improves the image matching efficiency, but also greatly improves the accuracy of ore sorting.

[0044] Reference Figure 2 The second embodiment of the present invention provides a uranium mining and metallurgy ore sorting system based on machine vision recognition, comprising: A data acquisition module, used for acquiring ore image data; A grayscale matrix module is used to pre-process the ore image data and calculate the grayscale co-occurrence matrix of the image through a grayscale formula; A template generation module, used to obtain a template repository according to the gray level co-occurrence matrix through a feature matching algorithm; A target extraction module is used to perform threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; The clustering sorting module is used to obtain a sorting result through a kernel density estimation clustering algorithm according to the target image and the template repository.

[0045] Preferably, the data acquisition module is used to: Acquire ore image data.

[0046] Preferably, the grayscale matrix module is used for: The ore image data is preprocessed, and the grayscale co-occurrence matrix of the image is calculated by the grayscale formula, including: Performing median filtering on the ore image data to obtain noise-reduced ore image data; According to the de-noised ore image data, a gray level co-occurrence matrix of the image is established, including: in, is the gray-level co-occurrence matrix, is the pixel in the image, Representation and Adjacent pixels, Represents all the pixels in the image. is the Kronecker function, when =0, and =0, the function value is 0; The entropy of the gray-level co-occurrence matrix is ​​calculated by the following formula: in, is the entropy of the gray-level co-occurrence matrix, It is the intermediate process quantity.

[0047] Preferably, the template generation module is used to: According to the grayscale co-occurrence matrix, a template repository is obtained through a feature matching algorithm, including: extracting a first feature point of a typical ore area image pre-stored at a preset first angle, the feature point including grayscale variance, grayscale average gradient, grayscale average and entropy; extracting a second feature point of a typical ore area image pre-stored at a preset second angle, the feature point including grayscale variance, grayscale average gradient, grayscale average and entropy; inputting the first feature point and the pixel coordinates corresponding to the feature point into the template repository; calculating the Euclidean distance between the first feature point and the second feature point; based on the Euclidean distance and the entropy of the grayscale co-occurrence matrix as feature values, saving the feature point coordinates and the Euclidean distance to the template repository through an optical flow tracking algorithm.

[0048] Preferably, the target extraction module is used to: The image data is subjected to threshold segmentation processing to obtain a plurality of target images to be sorted, including: performing threshold segmentation processing on the image data according to a preset image threshold; obtaining a separate target image by watershed segmentation; and performing burr filtering processing on the separate target image to obtain a target image to be sorted.

[0049] Preferably, the watershed segmentation comprises: The image is divided into different regions based on its gradient information. The regions with higher grayscale values ​​are taken as peaks, and the regions with lower grayscale values ​​are taken as valleys. These regions are input into the water flow simulation model to obtain the dividing lines of different regions.

[0050] Preferably, the cluster sorting module is used to: According to the target image and the template repository, a sorting result is obtained by a kernel density estimation clustering algorithm, including: According to the target image, calculating the sorting parameters of the target image, the sorting parameters including grayscale variance, grayscale average gradient, grayscale average and grayscale entropy; According to the sorting parameters, sample points are obtained by weighted averaging; The probability density function value of the kernel density estimate is calculated by the following formula: in, is the probability density function value, is the bandwidth, is the number of sample points, is the kernel function, satisfying , for i The horizontal coordinate of the sample point number; According to the probability density function value, a kernel density histogram is obtained, including: in is the value of the density matrix in the kernel density histogram, and are the coordinates of the kernel density histogram, and They are i The horizontal and vertical coordinates of the sample points; According to the kernel density histogram combined with a preset density threshold, high-density areas and low-density areas are distinguished; the local maximum value of the high-density area is used as the cluster center; and the cluster at the cluster center is used as the sorting result.

[0051] Preferably, after obtaining the sorting result by a kernel density estimation clustering algorithm based on the target image and the template repository, the method further includes: comparing the sorting result with a manual sorting result and correcting the density threshold.

[0052] It should be noted that the uranium mining and metallurgy ore sorting system based on machine vision recognition provided in an embodiment of the present invention is used to execute all the process steps of the uranium mining and metallurgy ore sorting method based on machine vision recognition in the above-mentioned embodiment. The working principles and beneficial effects of the two correspond one to one, and therefore will not be repeated here.

[0053] The embodiment of the present invention also provides an electronic device. The electronic device includes: a processor, a memory, and a computer program stored in the memory and executable on the processor, such as a data acquisition program. When the processor executes the computer program, the steps in the above-mentioned embodiment of the method for uranium mining and metallurgy ore separation based on machine vision recognition are implemented, such as Figure 1 Alternatively, when the processor executes the computer program, the functions of the modules / units in the above-mentioned device embodiments are realized, such as the data acquisition module.

[0054] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to implement the present invention. The one or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, which are used to describe the execution process of the computer program in the electronic device.

[0055] The electronic device may be a computing device such as a desktop computer, a notebook, a PDA, and a smart tablet. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will appreciate that the above components are merely examples of electronic devices and do not constitute a limitation on the electronic device. The electronic device may include more or fewer components than the above components, or may combine certain components, or different components. For example, the electronic device may also include input and output devices, network access devices, buses, etc.

[0056] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device.

[0057] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the electronic device by running or executing the computer program and / or module stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0058] Wherein, if the module / unit integrated in the electronic device is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0059] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0060] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for uranium mining and metallurgy ore separation based on machine vision recognition, characterized in that: include: Acquire ore image data; Preprocessing the ore image data, and calculating the grayscale co-occurrence matrix of the image through a grayscale formula; According to the gray level co-occurrence matrix, a template repository is obtained by a feature matching algorithm; Performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; A sorting result is obtained according to the target image and the template repository by using a kernel density estimation clustering algorithm.

2. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 1 is characterized in that: The preprocessing of the ore image data and the calculation of the grayscale co-occurrence matrix of the image by the grayscale formula include: Performing median filtering on the ore image data to obtain noise-reduced ore image data; According to the de-noised ore image data, a gray level co-occurrence matrix of the image is established, including: in, is the gray-level co-occurrence matrix, is the pixel in the image, Representation and Adjacent pixels, Represents all the pixels in the image. is the Kronecker function, when =0, and =0, the function value is 0; The entropy of the gray-level co-occurrence matrix is ​​calculated by the following formula: in, is the entropy of the gray-level co-occurrence matrix, It is the intermediate process quantity.

3. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 1 is characterized in that: The step of obtaining a template repository according to the gray level co-occurrence matrix through a feature matching algorithm includes: Extracting first feature points of a pre-stored typical ore region image at a preset first angle, wherein the feature points include grayscale variance, grayscale average gradient, grayscale average and entropy; Extracting second feature points of a pre-stored typical ore region image at a preset second angle, wherein the feature points include grayscale variance, grayscale average gradient, grayscale average and entropy; inputting the first feature point and the pixel coordinates corresponding to the feature point into a template repository; Calculating the Euclidean distance between the first feature point and the second feature point; According to the Euclidean distance and the entropy of the gray-level co-occurrence matrix as eigenvalues, the feature point coordinates and the Euclidean distance are saved in the template repository through an optical flow tracking algorithm.

4. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 1 is characterized in that: The step of performing threshold segmentation processing on the image data to obtain a plurality of target images to be sorted includes: Performing threshold segmentation processing on the image data according to a preset image threshold; Obtain separate target images through watershed segmentation; The individual target images are subjected to burr filtering processing to obtain target images to be sorted.

5. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 1 is characterized in that: The step of obtaining a sorting result by using a kernel density estimation clustering algorithm according to the target image and the template repository includes: According to the target image, calculating the sorting parameters of the target image, the sorting parameters including grayscale variance, grayscale average gradient, grayscale average and grayscale entropy; According to the sorting parameters, sample points are obtained by weighted averaging; The probability density function value of the kernel density estimate is calculated by the following formula: in, is the probability density function value, is the bandwidth, is the number of sample points, is the kernel function, satisfying , for i The horizontal coordinate of the sample point number; According to the probability density function value, a kernel density histogram is obtained, including: in is the value of the density matrix in the kernel density histogram, and are the coordinates of the kernel density histogram, and They are i The horizontal and vertical coordinates of the sample points; Distinguishing high-density areas from low-density areas according to the kernel density histogram in combination with a preset density threshold; The local maximum value of the high-density area is taken as the cluster center: The cluster at the center of the cluster is taken as the sorting result.

6. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 4 is characterized in that: The watershed segmentation comprises: Divide the image into different regions based on the gradient information of the image; The areas with higher grayscale values ​​in the image are taken as peaks, and the areas with lower grayscale values ​​are taken as valleys, which are input into the water flow simulation model to obtain the boundary lines of different areas.

7. The uranium mining and metallurgy ore separation method based on machine vision recognition according to claim 1 is characterized in that: After obtaining the sorting result by using the kernel density estimation clustering algorithm according to the target image and the template repository, the method further includes: The sorting result is compared with the manual sorting result, and the density threshold is corrected.

8. A uranium mining and metallurgy ore sorting system based on machine vision recognition, characterized in that: include: A data acquisition module, used for acquiring ore image data; A grayscale matrix module is used to pre-process the ore image data and calculate the grayscale co-occurrence matrix of the image through a grayscale formula; A template generation module, used to obtain a template repository according to the gray level co-occurrence matrix through a feature matching algorithm; A target extraction module is used to perform threshold segmentation processing on the image data to obtain a plurality of target images to be sorted; The clustering sorting module is used to obtain a sorting result through a kernel density estimation clustering algorithm according to the target image and the template repository.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, the method for uranium mining and metallurgy ore separation based on machine vision recognition as described in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute the uranium mining and metallurgy ore separation method based on machine vision recognition as described in any one of claims 1 to 7.

Images