An intelligent identification and analysis system for coal gangue based on machine vision and image processing

Abstract

The present invention belongs to the field of coal gangue identification, and specifically is a coal gangue intelligent identification and analysis system based on machine vision and image processing. The system analyzes the compatibility score of each target object relative to the coal gangue by combining the first, second and third order matching degrees of each target object relative to the coal gangue, so as to identify whether each target object is coal gangue. The first order matching degree comprehensively considers the color features, texture features and morphological features in the appearance image of each target object, the second order matching degree comprehensively considers the thermal distribution effect, thermal boundary effect and thermal conductivity effect fed back in the thermal imaging image of each target object, and the third order matching degree comprehensively considers the uniformity within the X-ray imaging partition of each target object and the uniformity between partitions. Matching degrees of different orders evaluate the target object from different angles, so that the coal gangue identification has strong complementarity and accuracy, and helps to understand the degree of compatibility between the target object and the coal gangue more deeply and comprehensively.

Description

Technical Field

[0001] The invention belongs to the field of coal gangue identification, and in particular relates to a coal gangue intelligent identification and analysis system based on machine vision and image processing. Background Art

[0002] With the continuous growth of global energy demand, coal, as one of the main fossil energy sources, has become particularly important in terms of mining and utilization efficiency. However, a large amount of coal gangue, a hard rock with low carbon content and low calorific value, is produced during coal mining, accounting for up to 15% of coal production. The escape of sulfides inside it will also pollute the atmosphere, farmland and water sources, causing serious impacts on the environment. Therefore, how to effectively separate coal and gangue in the raw coal in the early stage through technical means has become an urgent problem to be solved in the coal industry.

[0003] There are also some solutions related to gangue identification and sorting in the prior art. For example, China Patent Publication No. CN116452506A is an intelligent visual identification and sorting method for underground gangue based on machine learning. It obtains ore images on the underground conveyor belt through image acquisition equipment, performs segmentation, smoothing filtering and sharpening enhancement to filter out Gaussian noise interference, uses histogram equalization and morphological opening and closing operations to perform grayscale transformation, and divides the acquired image into the gradient or edge direction histogram of each pixel point in the cell unit to form the image HOG feature. SVM is used to classify the ore HOG features to help identify the category to which the ore belongs, thereby improving the system image processing speed and recognition efficiency.

[0004] Another Chinese patent publication number CN118314128B is a method for identifying coal gangue based on comprehensive utilization of coal gangue. It uses high-sensitivity machine vision equipment to acquire images of the coal involved in the screening, and performs image color comparison based on the acquired images of the corresponding coal. Based on the results of the proportion of black areas in the image, it effectively locks the internal coal gangue or abnormal coal, thereby improving the recognition efficiency of the coal and shortening the recognition time.

[0005] Although the above scheme proposes some solutions related to gangue identification and sorting, the existing technology still has the following limitations, specifically: 1. The existing technology for gangue identification is limited to a single feature dimension. For example, the above scheme uses texture features and color features as single entry points to identify gangue, ignoring the full feature analysis and comparison at the gangue image level, resulting in the lack of accuracy, scientificity and reliability of gangue identification results.

[0006] 2. The existing technology usually adopts a partitioning device acquisition mode for multi-dimensional image acquisition of conveyor belt gangue. However, the partitioning device image acquisition not only causes significant differences in shooting light, affecting the accuracy of image acquisition, but also may cause images of different partitions to be misaligned or overlapped when spliced, making it difficult to locate the coal mine sorting objects, thereby affecting the coal gangue identification and sorting effect. Summary of the invention

[0007] In order to overcome the shortcomings of the background technology, an embodiment of the present invention provides a coal gangue intelligent recognition and analysis system based on machine vision and image processing, which can effectively solve the problems involved in the above-mentioned background technology.

[0008] The purpose of the present invention can be achieved through the following technical solutions: a coal gangue intelligent identification and analysis system based on machine vision and image processing, including: a centralized image acquisition module, a first-order matching degree evaluation module, a second-order matching degree evaluation module, a third-order matching degree evaluation module, a coal gangue transmission module and a cloud database.

[0009] The centralized image acquisition module is connected to the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module respectively; the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module are all connected to the gangue conveying module; and the cloud database is connected to the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module respectively.

[0010] The centralized image acquisition module is used to locate each mineral object at the conveyor belt laying location in the target coal mine sorting area and carry out centralized image acquisition, including appearance image acquisition, thermal imaging acquisition and X-ray imaging acquisition.

[0011] The first-order matching evaluation module is used to record the mineral objects located at the conveyor belt laying location in the target coal mine sorting area as target objects, and evaluate the first-order matching degree of each target object relative to the coal gangue based on the appearance image collected from each target object and the comprehensive color features, texture features and morphological features.

[0012] The second-order matching evaluation module is used to evaluate the second-order matching degree of each target object relative to the coal gangue based on the thermal imaging images collected from each target object, comprehensively considering the thermal distribution effect, thermal boundary effect and thermal conductivity effect.

[0013] The third-order matching evaluation module is used to evaluate the third-order matching degree of each target object relative to the coal gangue according to the X-ray imaging images collected from each target object.

[0014] The gangue transmission module is used to analyze the compatibility score of each target object relative to the gangue based on the first, second and third order matching degree of each target object relative to the gangue, and use this as the basis for judging whether to assign the gangue identity label to each target object, and transmit each target object with the identity label of gangue to the gangue storage area.

[0015] The cloud database is used to store the color feature indicators, texture feature indicators and morphological feature indicators of the gangue appearance image corresponding to the preset reference, store the thermal distribution effect indicators, thermal boundary effect indicators and thermal conductivity effect indicators of the gangue thermal imaging image corresponding to the preset reference, store the grayscale variance of the internal pixel of the unit partition and the grayscale variance of the pixel between the partitions of the gangue X-ray imaging image corresponding to the preset reference, and store the display brightness value corresponding to each temperature color scale of the thermal imaging image.

[0016] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention eliminates the defects of inconsistent light in the acquisition mode of the existing partitioning equipment and the high difficulty of positioning and tracking by performing centralized image acquisition work on each mineral object at the conveyor belt laying point in the sorting area, thereby ensuring the continuity and stability of the sorting process.

[0017] (2) The present invention evaluates the first-order matching degree of each target object relative to the coal gangue by comprehensively considering the color features, texture features and morphological features in the appearance image. The multi-angle features at the appearance level provide more valuable information for the matching evaluation, which more truly reflects the similarity between the target object and the coal gangue, thereby making the evaluation result more scientific and convincing.

[0018] (3) The present invention evaluates the second-order matching degree of each target object relative to coal gangue by integrating the thermal distribution effect, thermal boundary effect and thermal conductivity effect fed back in the thermal imaging image, and more accurately understands the similarity between the target object and coal gangue in thermal properties, thereby improving the accuracy of recognition.

[0019] (4) The present invention evaluates the third-order matching degree of each target object relative to the coal gangue based on the internal uniformity and inter-partition uniformity of the X-ray imaging partitions collected from each target object, comprehensively considers the internal structural heterogeneity of the coal gangue due to the presence of other minerals, and enhances the comprehensiveness of the identification.

[0020] (5) The present invention combines the first, second and third order matching degrees of each target object with respect to coal gangue, analyzes the compatibility score of each target object with respect to coal gangue, and thereby identifies whether each target object is coal gangue. The matching degrees of different orders evaluate the target object from different angles, making coal gangue identification highly complementary, helping to understand the similarities between the target object and coal gangue more deeply and comprehensively, and avoiding the one-sidedness and errors caused by relying solely on a single type of feature for judgment. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0022] Figure 1 It is a schematic diagram of module connection of the present invention.

[0023] Figure 2 It is a schematic diagram of the planning of each circular area in the thermal imaging image of the target object of the present invention. DETAILED DESCRIPTION

[0024] 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.

[0025] Reference Figure 1 As shown, the present invention provides a coal gangue intelligent recognition and analysis system based on machine vision and image processing, including: a centralized image acquisition module, a first-order matching degree evaluation module, a second-order matching degree evaluation module, a third-order matching degree evaluation module, a coal gangue transmission module and a cloud database.

[0026] The centralized image acquisition module is connected to the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module respectively; the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module are all connected to the gangue conveying module; and the cloud database is connected to the first-order matching degree evaluation module, the second-order matching degree evaluation module, and the third-order matching degree evaluation module respectively.

[0027] The centralized image acquisition module is used to locate each mineral object at the conveyor belt laying location in the target coal mine sorting area and carry out centralized image acquisition work, including appearance image acquisition, thermal imaging acquisition and X-ray imaging acquisition.

[0028] The embodiment of the present invention performs centralized image acquisition work on each mineral object at the conveyor belt laying location in the sorting area, thereby eliminating the defects of inconsistent light in the acquisition mode of the existing partitioning equipment and the high difficulty of positioning and tracking, thereby ensuring the continuity and stability of the sorting process.

[0029] The first-order matching evaluation module is used to record the mineral objects located at the conveyor belt laying location in the target coal mine sorting area as target objects, and evaluate the first-order matching degree of each target object relative to the coal gangue based on the appearance image collected from each target object and the comprehensive color features, texture features and morphological features.

[0030] Specifically, the evaluation of the first-order matching degree of each target object relative to the gangue includes: obtaining the color channel intensity value of each pixel after the appearance image of each target object is converted into the RGB color space , is the number of each target object, , The number of each pixel after the target object appearance image is converted to RGB color space, , calculate the pixel centroid intensity value after the appearance image of each target object is converted into the RGB color space, compare it with the preset color channel intensity value corresponding to the black pixel in the RGB color space, and obtain the relative offset of the color centroid of the appearance image of each target object.

[0031] It should be noted that the pixel centroid intensity value of the above target object appearance images after conversion to RGB color space is .

[0032] It should also be noted that the relative offset of the color center of gravity of the above-mentioned appearance images of each target object is obtained by substituting the pixel center of gravity intensity value after converting the appearance images of each target object into the RGB color space and the preset color channel intensity value corresponding to the black pixel in the RGB color space into the Euclidean distance formula in three-dimensional form.

[0033] It is particularly noted that the preset color channel intensity value corresponding to the black pixel in the above RGB color space is The data is derived from the RGB color space mapping principle, that is, the RGB color space is based on the combination of the three basic colors of red, green and blue to create various colors. The intensity value range of the three color channels is , where 0 represents colorless, that is, this color channel has no light contribution. When the intensity values ​​of the three color channels are all 0, it means that no light is emitted, and it appears black. For details, please refer to the professional RGB color comparison table.

[0034] Set the color channel intensity reference threshold of the noise pixel in RGB color space , retrieve the various noise pixels after the appearance image of each target object is converted into RGB color space, count the number of noise pixels and perform ratio analysis with the total number of pixels in the appearance image, and obtain the relative proportion of noise in the appearance image of each target object.

[0035] It should be noted that the color channel intensity value of each color pixel after the appearance image of each target object is converted into RGB color space satisfy Conditions, among which The number of each noise pixel after the target object appearance image is converted to RGB color space. .

[0036] The cumulative value of the relative deviation of the color center of gravity and the relative proportion of the noise is used as the color feature index to obtain the color feature index of the appearance image of each target object. .

[0037] The appearance image of each target object is converted into a grayscale image, and each pixel in the image is traversed at a preset pixel distance and preset angle. The grayscale co-occurrence matrix of the appearance image of each target object is constructed and its various feature indicators, including contrast, correlation and entropy, are obtained. The contrast and entropy are accumulated, and further ratio analysis is performed with the correlation. The ratio analysis result is used as a texture feature indicator to obtain the texture feature indicator of the appearance image of each target object. .

[0038] It should be noted that the above-mentioned preset pixel distance and preset angle are the basic building elements of the grayscale co-occurrence matrix. The preset pixel distance can be set with reference to the resolution of the target object appearance image. If the resolution of the target object appearance image is greater than or equal to the conventional high-definition resolution standard value, it means that the target object appearance image is a high-resolution image, and the preset pixel distance can be 3. If the resolution of the target object appearance image is less than the conventional high-definition resolution standard value, it means that the target object appearance image is a low-resolution image, and the preset pixel distance can be 1. It should be noted that the conventional high-definition resolution standard value is derived from the image or video industry experience specifications, and the specific value can be 1920×1080.

[0039] Typical options for preset angles include The four directions can fully describe the spatial relationship between pixels in the image, so as to better capture the texture features of the image. The selection of the above preset angles does not come from a single data source or literature, but is the result of a long-term research and development, and gradually formed a consensus in multiple academic papers and practices. For example, the paper "Textural Features for Image Classification" systematically proposed the method of using GLCM for texture analysis for the first time, and discussed the importance of different directions. The four directions are widely used and recognized in the field of imaging, and the present invention specifically selects the value of the preset angle , that is, the horizontal direction.

[0040] It should also be noted that the correlation of the above grayscale co-occurrence matrix is ​​used to measure the linear relationship of the grayscale values ​​in the image in space. It is a specific numerical indicator. The texture feature index of the target object appearance image is actually a relative analysis of the texture roughness of the appearance image, so as to facilitate the subsequent analysis of the first-order matching degree of the coal gangue with a texture relatively rough in the coal mine. The correlation is inversely proportional to the texture roughness of the appearance image. Therefore, it is used as the denominator to participate in the texture feature index analysis. The specific calculation process for the correlation of the grayscale co-occurrence matrix is: according to the grayscale of the appearance image Constructing gray-level co-occurrence matrix for The matrix of Indicates that at the preset pixel distance and preset angle, the gray value is With gray value The probability of occurrence of pixel pairs is given by the formula , Calculate the gray value respectively The probability of occurrence of pixels in the horizontal direction and the gray value are The probability of the pixel appearing in the vertical direction is given by the formula , , , Calculate the row mean, column mean, row standard deviation and column standard deviation of the gray level co-occurrence matrix, according to the formula

[0041] Analyze the correlation of the gray-level co-occurrence matrix.

[0042] It should be noted that there are existing mature calculation formulas for the characteristic indicators of the gray-level co-occurrence matrix, which will not be elaborated here. In addition, in the process of further analyzing the ratio of the above-mentioned contrast and entropy accumulation values ​​to the correlation, a preset constant is added to the denominator to avoid the denominator being 0.

[0043] Specifically, the evaluation of the first-order matching degree of each target object relative to the gangue also includes: identifying the appearance image and outlining the outer contour of each target object, obtaining the position coordinates of each pixel point on the outer contour to form a discrete point set of the outer contour of each target object, and performing discrete Fourier transform on the horizontal coordinate sequence and the vertical coordinate sequence in the discrete point set of the outer contour to form the discrete Fourier coefficients corresponding to the horizontal and vertical coordinate sequences of the discrete point set of the outer contour of each target object, which are respectively recorded as , is the number of each pixel point in the outer contour discrete point set, .

[0044] It should be noted that the above mentioned horizontal coordinate sequence of the discrete point set of the outer contour of each target object can be recorded as , is the number of each pixel point in the outer contour discrete point set, , so we can use the formula

[0045] Obtain the discrete Fourier coefficients corresponding to the horizontal coordinate sequence of the discrete point set of the outer contour of each target object, where is the total number of pixels in the discrete point set of the outer contour, are all natural constants, is the imaginary unit. Similarly, the discrete Fourier coefficients corresponding to the ordinate sequence of the discrete point set of the outer contour of each target object are obtained. They are all numerical values ​​of pixel numbers in the discrete point set of the outer contour.

[0046] Set the low frequency discrete point value , record the pixel points in the discrete point set of the outer contour of each target object and located before the low-frequency discrete point value as low-frequency pixel points, and obtain the Fourier descriptor of each low-frequency pixel point, including the model and argument , is the number of each low-frequency pixel, , sequentially generate the low-frequency Fourier descriptor feature vectors of the outer contour of each target object, and use them as the morphological feature indicators of the appearance image of each target object .

[0047] It should be noted that the dimension of the feature vector of the low-frequency Fourier descriptor of the target object outline is The basis for the present invention to apply discrete Fourier transform to the discrete point set of the outer contour of each target object to obtain morphological feature indicators is that the Fourier transform can decompose the shape information of the outer contour of the target object into different frequency components. The low-frequency part corresponds to the overall characteristics of the shape, such as the general shape of the object, the main concave-convex structure, etc., and the high-frequency part corresponds to the detailed changes of the contour. For the recognition of coal gangue with irregular shapes and more tortuous changes, we can directly focus on the low-frequency part of the Fourier descriptor of the target object to summarize the matching status of the target object relative to the coal gangue at the morphological feature level.

[0048] The color feature index, texture feature index, and morphological feature index corresponding to the preset reference of the gangue appearance image stored in the cloud database are extracted and recorded as , according to the formula

[0049] Obtain the first-order matching degree of each target object relative to coal gangue, where It is a reasonable threshold for presetting the similarity of morphological features.

[0050] It should be noted that the value of the above-mentioned preset morphological feature similarity reasonable threshold is comprehensively affected by multiple factors such as application scenarios and coal gangue characteristics. Specifically, the system developers can use the known coal gangue appearance image data set of the target coal mine sorting area to conduct experimental tests on morphological similarity recognition, and use the reasonable test values ​​of each morphological feature similarity as input to observe the accuracy and recall rate of the experimental sorting results identified with morphological feature similarity as the entry point, and accumulate the accuracy and recall rate of the experimental sorting results with the product of their corresponding preset weights, respectively, and use the accumulated value as the output adaptation score of the reasonable test value of morphological feature similarity, thereby obtaining the output adaptation score of each morphological feature similarity reasonable test value, and selecting the morphological feature similarity reasonable test value corresponding to the maximum output adaptation score as the preset morphological feature similarity reasonable threshold of the present invention.

[0051] It should be noted that the analysis basis of the first-order matching degree of each target object relative to coal gangue in the present invention is that in the raw material mining scene, coal mines usually show black or nearly black color, its texture is relatively smooth, and its shape tends to be rounded. In contrast, due to the inclusion of other mineral components inside coal gangue, its appearance presents some non-black areas, the texture appears relatively rough, and the shape is not round enough. In view of this, when calculating the first-order matching degree relative to coal gangue for the appearance image of each target object, the greater the deviation of the color of each target object from black or the greater the proportion of non-black color, the rougher the texture, and the closer the shape to coal gangue, the greater its first-order matching degree relative to coal gangue.

[0052] Specifically, the calculation formulas of the Fourier descriptors of the low-frequency pixels are:

[0053] , ,in For the The first discrete point set of the target object's outer contour The horizontal and vertical coordinates of each low-frequency pixel point correspond to the discrete Fourier coefficients.

[0054] The embodiment of the present invention evaluates the first-order matching degree of each target object relative to the coal gangue by comprehensively considering the color features, texture features and morphological features in the appearance image. The multi-angle features at the appearance level provide more valuable information for the matching evaluation, and more realistically reflect the similarity between the target object and the coal gangue, thereby making the evaluation result more scientific and convincing.

[0055] The second-order matching evaluation module is used to evaluate the second-order matching degree of each target object relative to the coal gangue based on the thermal imaging images collected from each target object, comprehensively considering the thermal distribution effect, thermal boundary effect and thermal conductivity effect.

[0056] Reference Figure 2As shown, specifically, the evaluation of the second-order matching degree of each target object relative to the gangue includes: extracting the thermal imaging image of each target object, taking the center of the thermal imaging image as the center of the circle and superimposing them one by one with a preset size length as the radius to generate each circular area, obtaining the average heat source temperature value of each circular area in the thermal imaging image of each target object according to the temperature color scale presented by each circular area, calculating the temperature gradient value between each circular area and its adjacent circular area in the thermal imaging image of each target object, and performing mean calculation on the temperature gradient value between each circular area and its adjacent circular area to obtain the thermal distribution effect index of the thermal imaging image of each target object. .

[0057] It should be noted that the specific value of the above-mentioned preset size length can be obtained by retrieving the distance between the center point of the thermal imaging image of the target object and each point on the boundary contour of the target object, screening the maximum distance, and taking the ratio of the maximum distance to the preset number as the preset size length. The preset number can be directly set by the system developer. In order to facilitate the temperature comparison between each circular area, the preset number is usually set to 5.

[0058] Retrieve the temperature color code of each unit sampling point on the outer contour of each target object's thermal imaging image to obtain its corresponding heat source temperature value, calculate the standard deviation of the heat source temperature of each unit sampling point on the outer contour of each target object's thermal imaging image, and use it as the thermal boundary effect index of each target object's thermal imaging image .

[0059] According to the display brightness values ​​corresponding to the temperature color scales of the thermal imaging images stored in the cloud database, the display brightness values ​​corresponding to the temperature color scales in the thermal imaging images of each target object are obtained, and the thermal conductivity effect index of the thermal imaging images of each target object is obtained by calculating the average value of the display brightness values ​​corresponding to the temperature color scales in the thermal imaging images of each target object. .

[0060] Specifically, the second-order matching degree of each target object relative to the gangue is evaluated, and the second-order matching degree of each target object relative to the gangue is evaluated. The second-order matching degree of each target object relative to the gangue is evaluated. The second-order matching degree of each target object relative to the gangue is evaluated. The second-order matching degree of each target object relative to the gangue is evaluated. The second-order matching degree of each target object relative to the gangue is evaluated. , according to the formula Analyze the second-order matching degree of each target object relative to coal gangue.

[0061] It should be noted that when analyzing the second-order matching of each target object relative to coal gangue, the thermal conductivity effect index presents an inversely proportional state. The reasons are as follows: compared with the raw coal mine, under the same temperature conditions, the thermal conductivity characteristics of coal gangue will make its color code present a lower brightness. Since coal gangue contains many other minerals, the difference in its thermal conductivity affects the transfer and distribution of heat inside and on the surface of coal gangue, making coal gangue show a lower brightness in thermal imaging, which in turn makes the thermal conductivity effect index inversely proportional to the target matching degree.

[0062] Specifically, the temperature gradient value between each circular area and its adjacent circular area in the thermal imaging image of each target object is the ratio of the absolute difference of the average heat source temperature between each circular area and its adjacent circular area in the thermal imaging image of each target object to the preset size length.

[0063] The embodiment of the present invention evaluates the second-order matching degree of each target object relative to coal gangue by integrating the heat distribution effect, thermal boundary effect and thermal conductivity effect fed back in the thermal imaging image, and more accurately understands the similarity between the target object and coal gangue in thermal properties, thereby improving the accuracy of recognition.

[0064] The third-order matching evaluation module is used to evaluate the third-order matching degree of each target object relative to the coal gangue based on the X-ray imaging images collected from each target object.

[0065] Specifically, the third-order matching degree of each target object relative to the gangue includes: extracting the X-ray imaging image of each target object, dividing the X-ray imaging image according to a preset grid size to obtain each partition, and calculating the internal pixel grayscale variance of each partition of the X-ray imaging image of each target object according to the grayscale value of each pixel in each partition of the X-ray imaging image of each target object. and the pixel grayscale variance between X-ray image partitions , is the number of each partition of the X-ray image, , according to the formula Analyze the third-order matching degree of each target object relative to coal gangue, among which They are the grayscale variance of pixels within the unit partition and the grayscale variance of pixels between partitions corresponding to the preset reference of the coal gangue X-ray imaging image stored in the cloud database. The number of partitions of the X-ray image.

[0066] It should be noted that gangue is solid waste produced during coal mining, and its shape and size are usually irregular. However, since gangue does not require detailed internal structure analysis like precision mechanical parts or electronic components, for the application scenario of pixel grayscale difference analysis in gangue X-ray imaging, the preset unit grid size of 1 cm × 1 cm is relatively suitable, and can effectively distinguish the characteristics of different areas of gangue without over-emphasizing details.

[0067] The embodiment of the present invention evaluates the third-order matching degree of each target object relative to the gangue based on the internal uniformity of the X-ray imaging partitions and the uniformity between partitions collected from each target object, comprehensively considers the internal structural unevenness of the gangue due to the presence of other minerals, and enhances the comprehensiveness of identification.

[0068] The gangue transmission module is used to analyze the compatibility score of each target object relative to the gangue in combination with the first, second and third order matching degree of each target object relative to the gangue, and use this as the basis for judging whether to assign a gangue identity tag to each target object, and transmit each target object with an identity tag of gangue to the gangue storage area.

[0069] Specifically, the calculation formula for the compatibility score of each target object relative to coal gangue is:

[0070] ,in are the preset weights corresponding to the first, second and third order matching degrees of the target object relative to the gangue, for .

[0071] For example, the above The specific value can be The value is based on the following: the first, second and third order matching degrees of the target object relative to the coal gangue are obtained from the analysis of the corresponding appearance image, thermal imaging image and X-ray imaging image respectively, which play an indispensable role in fully understanding the fit score of the target object relative to the coal gangue. Therefore, considering the comprehensiveness and complementarity of information, the first, second and third order matching degrees can be set with the same weight.

[0072] Specifically, the judgment condition for assigning the target object with a gangue identity label is that the compatibility score of the target object relative to the gangue is greater than or equal to a preset compatibility score reaching threshold.

[0073] The embodiment of the present invention combines the first, second, and third order matching degrees of each target object with respect to coal gangue, analyzes the compatibility score of each target object with respect to coal gangue, and thereby identifies whether each target object is coal gangue. Matching degrees of different orders evaluate the target object from different angles, making coal gangue identification highly complementary, helping to more deeply and comprehensively understand the similarities between the target object and coal gangue, and avoiding the one-sidedness and errors caused by relying solely on a single type of feature for judgment.

[0074] The cloud database is used to store color feature indicators, texture feature indicators and morphological feature indicators corresponding to preset references of coal gangue appearance images, store thermal distribution effect indicators, thermal boundary effect indicators and thermal conductivity effect indicators corresponding to preset references of coal gangue thermal imaging images, store grayscale variances of pixels inside unit partitions and grayscale variances of pixels between partitions corresponding to preset references of coal gangue X-ray imaging images, and store display brightness values ​​corresponding to temperature color scales of thermal imaging images.

[0075] It should be noted that the above series of data stored in the cloud data are all implanted by the system development.

[0076] The above contents are merely examples and explanations of the structure of the present invention. The technicians in this technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the invention or exceed the scope defined by the present invention, they should all fall within the protection scope of the present invention.

Claims

1. A coal gangue intelligent identification and analysis system based on machine vision and image processing, characterized in that: The system includes: Centralized image acquisition module, used to locate each mineral object at the conveyor belt laying location in the target coal mine sorting area and carry out centralized image acquisition, including appearance image acquisition, thermal imaging acquisition and X-ray imaging acquisition; The first-order matching evaluation module is used to record the mineral objects located at the conveyor belt laying location in the target coal mine sorting area as target objects, and evaluate the first-order matching degree of each target object relative to the coal gangue based on the appearance image collected from each target object and the color, texture and morphological features; The second-order matching evaluation module is used to evaluate the second-order matching degree of each target object relative to the gangue based on the thermal imaging images collected from each target object, comprehensively considering the thermal distribution effect, thermal boundary effect and thermal conductivity effect; A third-order matching evaluation module is used to evaluate the third-order matching degree of each target object relative to the coal gangue according to the X-ray imaging images collected from each target object; The gangue transmission module is used to analyze the compatibility score of each target object relative to the gangue by combining the first, second and third order matching degree of each target object relative to the gangue, and use this as the basis for judging whether to give the gangue identity tag to each target object, and transmit each target object with the identity tag of gangue to the gangue storage area; A cloud database is used to store color feature indicators, texture feature indicators and morphological feature indicators corresponding to preset references of coal gangue appearance images, store thermal distribution effect indicators, thermal boundary effect indicators and thermal conductivity effect indicators corresponding to preset references of coal gangue thermal imaging images, store grayscale variance of pixels inside unit partitions and grayscale variance of pixels between partitions corresponding to preset references of coal gangue X-ray imaging images, and store display brightness values ​​corresponding to temperature color scales of thermal imaging images; The evaluation of the first-order matching degree of each target object relative to the gangue includes: obtaining the color channel intensity value of each pixel after the appearance image of each target object is converted into the RGB color space , The number of each pixel after the target object appearance image is converted to RGB color space, , calculate the pixel centroid intensity value after the appearance image of each target object is converted into the RGB color space, compare it with the preset color channel intensity value corresponding to the black pixel in the RGB color space, and obtain the relative offset of the color centroid of the appearance image of each target object; Set the color channel intensity reference threshold of the noise pixel in RGB color space , retrieve each noise pixel after the appearance image of each target object is converted into RGB color space, count the number of noise pixels and perform ratio analysis on it with the total number of pixels in the appearance image, and obtain the relative proportion of noise in the appearance image of each target object; The color channel intensity value of each noise pixel after the appearance image of each target object is converted to RGB color space satisfy Conditions, among which The number of each noise pixel after the target object appearance image is converted to RGB color space. ; The cumulative value of the relative deviation of the color center of gravity and the relative proportion of the noise is used as the color feature index to obtain the color feature index of the appearance image of each target object. , is the number of each target object, ; The appearance image of each target object is converted into a grayscale image, and each pixel in the image is traversed at a preset pixel distance and preset angle. The grayscale co-occurrence matrix of the appearance image of each target object is constructed and its various feature indicators, including contrast, correlation and entropy, are obtained. The contrast and entropy are accumulated, and further ratio analysis is performed with the correlation. The ratio analysis result is used as a texture feature indicator to obtain the texture feature indicator of the appearance image of each target object. ; After the appearance image of each target object is converted to RGB color space, the pixel centroid intensity value is ; The relative deviation of the color center of gravity of each target object appearance image is obtained by substituting the pixel center intensity value after the appearance image of each target object is converted into the RGB color space and the preset color channel intensity value corresponding to the black pixel in the RGB color space into the Euclidean distance formula in three-dimensional form; The specific calculation process of the correlation of the gray level co-occurrence matrix is ​​as follows: according to the gray level of the appearance image Constructing gray-level co-occurrence matrix for The matrix of Indicates that at the preset pixel distance and preset angle, the grayscale value is With gray value The probability of occurrence of pixel pairs is given by the formula , Calculate the gray value respectively The probability of occurrence of pixels in the horizontal direction and the gray value are The probability of the pixel appearing in the vertical direction is given by the formula , , , Calculate the row mean, column mean, row standard deviation and column standard deviation of the gray level co-occurrence matrix, according to the formula Analyze the correlation of gray-level co-occurrence matrix; The evaluation of the first-order matching degree of each target object relative to the gangue also includes: identifying the appearance image and outlining the outer contour of each target object, obtaining the position coordinates of each pixel point on the outer contour to form a discrete point set of the outer contour of each target object, and performing discrete Fourier transform on the horizontal coordinate sequence and the vertical coordinate sequence in the discrete point set of the outer contour to form discrete Fourier coefficients corresponding to the horizontal and vertical coordinate sequences of the discrete point set of the outer contour of each target object, respectively recorded as , is the number of each pixel point in the outer contour discrete point set, ; Set the low frequency discrete point value , record the pixel points in the discrete point set of the outer contour of each target object and located before the low-frequency discrete point value as low-frequency pixel points, and obtain the Fourier descriptor of each low-frequency pixel point, including the model and argument , is the number of each low-frequency pixel, , sequentially generate the feature vectors of the low-frequency Fourier descriptors of the outer contours of each target object, and use them as the morphological feature indicators of the appearance images of each target object ; The color feature index, texture feature index, and morphological feature index corresponding to the preset reference of the gangue appearance image stored in the cloud database are extracted and recorded as , according to the formula Obtain the first-order matching degree of each target object relative to coal gangue, where It is a reasonable threshold for presetting the similarity of morphological features.

2. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 1 is characterized in that: The calculation formulas of the Fourier descriptors of the low-frequency pixels are: , ,in For the The first discrete point set of the target object's outer contour The horizontal and vertical coordinates of each low-frequency pixel point correspond to the discrete Fourier coefficients.

3. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 1 is characterized in that: The evaluation of the second-order matching degree of each target object relative to the gangue includes: extracting the thermal imaging image of each target object, taking the center of the thermal imaging image as the center of the circle and superimposing them one by one with a preset length as the radius to generate each circular area, obtaining the average heat source temperature value of each circular area of ​​the thermal imaging image of each target object according to the temperature color scale presented by each circular area, calculating the temperature gradient value between each circular area and its adjacent circular area in the thermal imaging image of each target object, and performing mean calculation on the temperature gradient value between each circular area and its adjacent circular area to obtain the thermal distribution effect index of the thermal imaging image of each target object. ; Retrieve the temperature color code of each unit sampling point on the outer contour of each target object's thermal imaging image to obtain its corresponding heat source temperature value, calculate the standard deviation of the heat source temperature of each unit sampling point on the outer contour of each target object's thermal imaging image, and use it as the thermal boundary effect index of each target object's thermal imaging image ; According to the display brightness values ​​corresponding to the temperature color scales of the thermal imaging images stored in the cloud database, the display brightness values ​​corresponding to the temperature color scales in the thermal imaging images of each target object are obtained, and the thermal conductivity effect index of the thermal imaging images of each target object is obtained by calculating the average value of the display brightness values ​​corresponding to the temperature color scales in the thermal imaging images of each target object. .

4. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 3 is characterized by: The second-order matching degree of each target object relative to the gangue is evaluated, and the method further includes: extracting the thermal distribution effect index, thermal boundary effect index and thermal conductivity effect index corresponding to the preset reference of the thermal imaging image of the gangue stored in the cloud database, which are respectively recorded as , according to the formula Analyze the second-order matching degree of each target object relative to coal gangue.

5. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 3 is characterized by: The temperature gradient value between each circular area and its adjacent circular area in the thermal imaging image of each target object is the ratio of the absolute difference of the average heat source temperature between each circular area and its adjacent circular area in the thermal imaging image of each target object to the preset size length.

6. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 4 is characterized by: The third-order matching degree of each target object relative to the gangue includes: extracting the X-ray imaging image of each target object, dividing the X-ray imaging image according to a preset grid size to obtain each partition, and calculating the internal pixel grayscale variance of each partition of the X-ray imaging image of each target object according to the grayscale value of each pixel in each partition of the X-ray imaging image of each target object. and the pixel grayscale variance between X-ray image partitions , is the number of each partition of the X-ray image, , according to the formula Analyze the third-order matching degree of each target object relative to coal gangue, among which They are the grayscale variance of pixels within the unit partition and the grayscale variance of pixels between partitions corresponding to the preset reference of the coal gangue X-ray imaging image stored in the cloud database. The number of partitions of the X-ray image.

7. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 6 is characterized by: The calculation formula for the compatibility score of each target object relative to coal gangue is: ,in are the preset weights corresponding to the first, second and third order matching degrees of the target object relative to the gangue, for .

8. The intelligent identification and analysis system for coal gangue based on machine vision and image processing according to claim 1 is characterized by: The judgment condition for assigning the target object an identity label of coal gangue is that the compatibility score of the target object relative to the coal gangue is greater than or equal to a preset compatibility score reaching threshold.

Images