Method for obtaining the size of agglomerated ore particles

Abstract

The application discloses a method for obtaining the size of agglomerated ore particles, and belongs to the technical field of machine vision, which comprises the following steps: step S01, positioning the area where a conveying belt is located in a full image, wherein the area where the conveying belt is located is a pure-color conveying belt background image containing ores and not containing surrounding environment information; step S02, performing threshold segmentation on the image obtained in step S01 by using the difference between the gray values of the ores and the conveying belt background, and extracting all the areas where the ores are located; step S03, separating individual ores from the overall image of the agglomerated ores obtained in step S02, and removing the agglomeration of the ores; and step S04, calculating the number of the ores separated in step S03, and obtaining the center and the radius of the ores. The method for obtaining the size of the agglomerated ore particles adopts a morphological method to realize the separation of the agglomerated ores, is more simple than other methods, has less operation time, and has higher detection efficiency in large-scale real-time application.

Description

Technical Field

[0001] This invention belongs to the field of machine vision technology, and in particular relates to a method for obtaining the size of adhered ore particles. Background Technology

[0002] Machine vision uses machines to replace human eyes for measurement and judgment. A machine vision system converts the captured target into an image signal, which is then transmitted to an image processing system to obtain the target's shape information. Based on pixel distribution, brightness, color, and other information, this information is converted into a digital signal. The image system performs various calculations on these signals to extract the target's features, and then controls the equipment's actions based on the judgment results. With the ever-increasing processing speed of computers, the application of machine vision in various industrial fields has been increasing year by year, and it has now become an indispensable part of many production lines.

[0003] Calculating ore particle size is a crucial indicator in ore particle analysis, significantly impacting ore processing, beneficiation, and ore property evaluation. During ore processing and beneficiation, particle size analysis helps determine parameters for ore selection and grinding, which is vital. Due to the unique nature of beneficiation, real-time manual acquisition is difficult, while machine vision methods can greatly improve efficiency. The most challenging aspect of ore particle size analysis is separating and calculating the particle size of aggregated ores. Aggregated ores are typically formed by multiple particles or clumps adhering together, exhibiting complex and irregular morphologies. This makes particle size analysis of aggregated ores difficult, as accurately defining the boundaries of each individual particle is challenging. Existing image segmentation methods based on corner detection and curvature for aggregated ore particles are commonly used. However, these methods require multiple computational steps and parameter adjustments, resulting in high algorithm complexity. For large-scale image data or real-time applications, timeliness issues may lead to missed detections and errors. Summary of the Invention

[0004] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides a method for obtaining the size of cohesive ore particles, which solves the technical problems of low algorithm efficiency and inaccurate results in large-scale real-time applications. To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0005] A method for obtaining the size of cohesive ore particles includes the following steps:

[0006] Step S01: Locate the area where the conveyor belt is located in the entire image. The area where the conveyor belt is located is a solid color conveyor belt background image that contains ore but does not contain information about the surrounding environment.

[0007] Step S02: Use the difference in grayscale values ​​between the ore and the conveyor belt background to perform threshold segmentation on the image obtained in step S01, extract all the ore areas, extract all the ore into the same area, fill the area, fill the holes in the ore, so that the ore area becomes a whole without holes.

[0008] Step S03: Segment individual ore pieces from the overall image of the adhered ore obtained in step S02 and remove the adhesion of the ore pieces.

[0009] Step S04: Calculate the number of ore pieces after segmentation in step S03, take the circumcircle of the contour fitting for each region, calculate the particle size of the ore, and obtain the center and radius of the circle.

[0010] Furthermore, in step S01, locating the area where the conveyor belt is located involves threshold segmentation of the entire image, separating unconnected areas into several independent regions, selecting the largest of these regions as the area where the conveyor belt is located, and cropping the area where the conveyor belt is located from the original image to obtain an image that contains the conveyor belt and the ore without interference.

[0011] Furthermore, the specific steps for locating the area where the conveyor belt is located in step S01 are as follows:

[0012] Step 1.1: Perform grayscale processing on the image of the ore on the conveyor belt, select all pixels within the specified grayscale threshold range into the selected area, and then segment the selected area and the remaining area. The threshold segmentation formula is as follows:

[0013] S = {r, c} T ∈R|g min ≤f r，c ≤g max}

[0014] Where: S represents the set of pixels in the segmented region, r and c represent the pixel coordinates, T is the coordinate transpose symbol, R represents the set of all pixels in the image, f represents the gray value mapping of the point with coordinates (r, c), and g represents the range of gray values.

[0015] Step 1.2: Divide the region calculated in Step 1.1 into a series of independent small regions. These regions are composed of interconnected pixels. For connected regions, use run-length representation to describe the region. Use 4-neighborhood to describe the foreground and 8-neighborhood to describe the background. Use depth-first search algorithm to decompose unconnected regions into a series of independent small regions.

[0016] Step 1.3: Using feature extraction, select the region with the largest area from a series of regions. This region is where the conveyor belt is located. Remove small regions caused by local noise. All points in the region are represented by binary images and run-length encoding. The formula for calculating the area of ​​the region using the region run-length encoding algorithm is as follows:

[0017]

[0018] Where: a represents the area of ​​the region, R represents the number of points in the region, and ce i cs represents the starting coordinate point. i Indicates the end coordinate point;

[0019] Step 1.4: Crop the original image according to the obtained region to obtain an image containing only the effective area inside the conveyor belt. The formula for region cropping is as follows:

[0020]

[0021] Where: S represents the set of cropped points; c represents the grayscale value of each pixel; r and c represent the pixel coordinates.

[0022] Furthermore, the specific steps of step S02 are as follows:

[0023] Step 2.1: Threshold segmentation is performed using the difference in grayscale values ​​between the ore and the background to divide the area containing the ore. The formula for threshold segmentation is:

[0024] S = {r, c} T ∈R|g min ≤f r，c ≤g max}

[0025] Where r and c represent pixel coordinates, S represents the set of pixels in the segmented region, g represents the range of gray values, R represents the set of all pixels in the image, and f represents the gray value mapping of the point with coordinates (r, c).

[0026] Step 2.2: Divide the region into multiple independent small regions, select regions with an area greater than a certain value to eliminate interference from factors such as noise, separate and extract all ore regions from the background conveyor belt, and remove noise;

[0027] Step 2.3: Fill the holes inside the segmented ore area. Points within the area correspond to pixel 1 in the area image, and holes within the area correspond to pixel 0 in the area image. Fill the hole pixels with 1, remove the holes, and iterate using the following process to find and fill all hole areas:

[0028]

[0029] Where B is a symmetric structural element, if X k =X k+1 The algorithm terminates after k iterations, X k Includes all filled holes, and I represents the region of interest, which is the ore region that does not contain the background.

[0030] Furthermore, the specific steps of step S03 are as follows:

[0031] Step 3.1: Transform the area where the ore is located into a distance map. First, calculate the distance value. Then, add the mask covering the area to the known distance value and the mask at this location. Select the minimum value as the distance value of the current pixel. The masks from left to right and top to bottom are as follows:

[0032] d2 d1 d2

[0033] d1 0

[0034] Where d1 and d2 represent the values ​​of the mask, when d1 = 3 and d2 = 4, the distance graph is divided by 3 to obtain the approximate Euclidean distance;

[0035] Step 3.2: Calculate the maximum and minimum gray values ​​within the region. First, obtain the gray-level histogram of the region, and let r... k k = 0, 1, 2, ..., L-1 represents the gray levels of an L-level grayscale digital image f(x, y), and the formula for the non-normalized histogram of f is:

[0036] h(r k ) = n k k = 0, 1, 2, ..., L-1

[0037] In the formula n k The grayscale value of f is r k The number of pixels, similarly, the formula for the normalized histogram is:

[0038]

[0039] Where MN are the number of rows and columns of the image, respectively, and n k The grayscale value of f is r k The number of pixels; based on the histogram, obtain the maximum and minimum values ​​at both ends, and get the maximum and minimum gray levels;

[0040] Step 3.3: After scaling the maximum value obtained in Step 3.2 proportionally, perform erosion to break up the adhered parts of the ore. In morphological processing, the use of transposed structural elements in Minkowski subtraction is called erosion, and the formula is as follows:

[0041]

[0042] Where S represents the structuring element, R represents the image to be eroded, the structuring element is convolved with the image or set to be processed, and t represents the element remaining after convolution. When the structuring element matches an element in the image or set, the element is retained; otherwise, the element is eroded.

[0043] Step 3.4: Using the connectivity of the eroded areas as a criterion, divide the region into different independent areas. Perform separate operations on each separated area, and then perform an expansion operation to restore the eroded areas, making the ore region return to its normal size. In morphological processing, the use of transposed structural elements in the Minkowski addition method is called expansion, and the formula is as follows:

[0044]

[0045] Here, R is the region to be processed, S is called the structuring element, and t represents the element left after convolution. For each structuring element S that overlaps with the image R, if at least one corresponding element satisfies the condition, the output pixel at that position is retained. The same dilation operation with the same structure is performed on each independent region to restore the eroded region. The independent region is the region where each piece of ore is located.

[0046] Furthermore, the specific steps of step S04 are as follows:

[0047] First, obtain the outline of each ore region. The formula for obtaining the outline is as follows:

[0048] β(A) = A - (AB)

[0049] Where: A represents the region where the contour needs to be calculated, B represents the structuring element used for erosion, β(A) represents the obtained contour, A is eroded using structuring element B, and the difference between A and B is taken as the result. Then, the minimum circumcircle in any direction is calculated. The method for fitting the circumcircle is based on the convex hull of the region. In a specific region, the convex hull of a set of points is the minimum convex set that contains all points in the region. The circumcircle diameter is fitted using the region boundary distance with the largest distance, and the center and radius are obtained. After calculating the particle size of all ore, the average value is calculated as the average particle size of the current ore. The formula for the average particle size is as follows:

[0050]

[0051] N represents the quantity of ore, D i This indicates the particle size of each ore, with Dv50 representing the average particle size.

[0052] The beneficial effects of this invention are:

[0053] The method for obtaining the size of cohesive ore particles in this invention uses a morphological approach to segment cohesive ore, which is simpler than other methods, requires less computation time, and has higher detection efficiency in large-scale real-time applications. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of ore adhesion;

[0055] Figure 2 This is a schematic diagram of the ore region after threshold segmentation;

[0056] Figure 3 This is a schematic diagram of the area where the ore was located after filling;

[0057] Figure 4 This is a distance map after regional distance transformation;

[0058] Figure 5 This is a schematic diagram of the area where the ore is separated from the aggregate.

[0059] Figure 6 It is a schematic diagram of the circumcircle of each ore. Detailed Implementation

[0060] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0061] This invention provides a method for obtaining the size of cohesive ore particles, comprising the following steps:

[0062] Step S01: Locate the area where the conveyor belt is located in the entire image. The area where the conveyor belt is located is a solid color conveyor belt background image that contains ore but does not contain information about the surrounding environment.

[0063] Specifically, locating the area where the conveyor belt is located involves threshold segmentation of the entire image, separating unconnected areas into several independent regions, selecting the largest of these regions as the area where the conveyor belt is located, cropping the area where the conveyor belt is located from the original image, and obtaining an image that contains the conveyor belt and the ore without interference.

[0064] Furthermore, the specific steps for locating the area where the conveyor belt is located in step 1 are as follows:

[0065] Step 1.1: Perform grayscale processing on the image of the ore on the conveyor belt, select all pixels within the specified grayscale threshold range into the selected area, and then segment the selected area and the remaining area. The threshold segmentation formula is as follows:

[0066] S = {r, c} T ∈R|g min ≤f r，c ≤gmax}

[0067] Where: S represents the set of pixels in the segmented region, r and c represent the pixel coordinates, T is the coordinate transpose symbol, R represents the set of all pixels in the image, f represents the gray value mapping of the point with coordinates (r, c), and g represents the range of gray values.

[0068] Step 1.2: Divide the region calculated in Step 1.1 into a series of independent small regions. These regions are composed of interconnected pixels. To calculate the connected regions, we first define pixel connectivity. A pixel is considered connected if it shares a common edge with another pixel in the four directions (top, bottom, left, and right), and this is called a 4-neighborhood. If four elements on the diagonal are also considered connected, it is called an 8-neighborhood. We use run-length representation to describe the regions, using 4-neighborhood for the foreground and 8-neighborhood for the background. We then use a depth-first search algorithm to decompose the disconnected regions into a series of independent small regions.

[0069] Step 1.3: Using feature extraction, select the region with the largest area among a series of regions. This region is where the conveyor belt is located. Remove small regions caused by local noise. All points in the region are represented by binary images using run-length encoding. The formula for calculating the area is as follows:

[0070]

[0071] Where: a represents the area of ​​the region, R represents the number of points in the region, and ce i cs represents the starting coordinate point. i Indicates the end coordinate point;

[0072] The formula for calculating the common encoding is as follows:

[0073]

[0074] It is easy to see from the two formulas above that ordinary encoding requires reading each point, while travel calculation involves adding and subtracting coordinates. Therefore, the cumulative formula using travel encoding requires much less computation than the cumulative formula using ordinary encoding. Thus, the use of the region travel algorithm in this invention to represent the region can significantly improve the speed of calculating the area.

[0075] Step 1.4: Crop the original image according to the previously obtained region to obtain an image containing only the effective area inside the conveyor belt, which facilitates the next step of processing. The formula for region cropping is as follows:

[0076]

[0077] Where: S represents the set of cropped points; c represents the grayscale value of each pixel; r and c represent the pixel coordinates.

[0078] Points within the specified region in the image are retained, while points outside the region are set to a grayscale value of 0. For example... Figure 1 As shown, the region-cropped image is obtained.

[0079] Step S02: Use the difference in grayscale values ​​between the ore and the conveyor belt background to perform threshold segmentation on the image obtained in step S01, extract all the ore areas, extract all the ore into the same area, fill the area, fill the holes in the ore, so that the ore area becomes a whole without holes.

[0080] Step 2.1: Threshold segmentation is performed using the difference in grayscale values ​​between the ore and the background to divide the area containing the ore. The formula for threshold segmentation is:

[0081] S = {r, c} T ∈R|g min ≤f r，c ≤g max}

[0082] Where r and c represent pixel coordinates, S represents the set of pixels in the segmented region, and g represents the range of gray values. R represents the set of all pixels in the image, and f represents the gray value mapping obtained from the point at coordinates (r, c). Figure 2 The image shown.

[0083] Step 2.2: Similar to Step 1.2, the area is divided into multiple independent small areas. The largest area is selected to eliminate interference from factors such as noise. All areas where the ore is located are separated from the background conveyor belt and extracted to remove noise.

[0084] Step 2.3: Since a fixed threshold segmentation is used, grayscale value changes and camera noise in some areas inside the ore may cause grayscale values ​​to fall below the threshold, resulting in holes within the segmented ore area. This can introduce errors into subsequent processing and calculations, so these holes need to be filled. A hole is a background pixel surrounded by a bounding box connected by foreground pixels. Points within the area correspond to region image pixel 1, and holes within the area correspond to region image pixel 0. The hole pixels are filled with 1, thus removing the holes. The following iterative process is used to find and fill all hole areas:

[0085]

[0086] Where B is a symmetric structural element, if X k =X k+1 The algorithm terminates after k iterations, X kIncludes all filled holes, where I represents the region of interest, i.e., the ore region excluding the background. X is defined in each iteration. k with I c The intersection ensures that the area of ​​interest is filled, i.e., the pores inside the ore, while the background is not filled. By iteratively filling all the pores within the area, a solid region is formed, resulting in the entire area containing the ore without pores, as shown in the example. Figure 3 The image shown.

[0087] Step S03: Segment individual ore pieces from the overall image of the adhered ore obtained in step S02 and remove the adhesion of the ore pieces.

[0088] First, the image size is obtained, and the region is transformed into a distance map using distance transformation. Then, the grayscale range of the distance map is obtained, and the region is eroded based on this range. Because the distance between bonded ore sections is smaller than that of normal ore, appropriate erosion structuring elements are selected to erode the joints. The region is then further divided into independent segments for each ore section. Finally, dilation is applied to restore the original size and dimensions of the ore.

[0089] Step 3.1: Transform the region containing the ore into a distance map. The pixels in the distance map represent distances, specifically the minimum distance from each point within R to all points outside the region. Therefore, the distance to all points on the boundary is 1. When using a 4-connected region, if (r1, c1) T and (r2, c2) T If there are two points, then the 4-connected distance can be calculated using the following formula:

[0090] d4 = |r2 - r1| + |c2 - c1|

[0091] Where d4 represents the distance between four connected regions, and r1, r2, c1, and c2 represent the coordinates of the pixels.

[0092] If 8-connectivity is used, then the corresponding distance is called the 8-connected distance, which is calculated using the following formula:

[0093] d8 = max(|r2-r1|, |c2-c1|)

[0094] Where d8 represents the distance of a 4-connected region, and r1, r2, c1, and c2 represent the coordinates of the pixels. Both 4-connected and 8-connected regions approximate the Euclidean distance, which is calculated using the formula:

[0095]

[0096] Where, d e R1 represents the Euclidean distance, and R2, C1, and C2 represent the coordinates of the pixels.

[0097] The distance transformation first sets points outside the region to 0, and then sets all distance values ​​within the region to a reasonably chosen maximum distance value, i.e., 2. b -1, where b is the number of pixels from the image, such as 2. 16 -1 is used as the initial distance value for the entire region. Then, a line-by-line scan is performed across the entire image. A small mask is placed at the current pixel. The existing distance values ​​in the mask-covered area are added to the element at that position in the mask. The minimum value from all calculated results is then selected as the distance value for the current pixel. The masks from left to right and top to bottom are as follows:

[0098] d2 d1 d2

[0099] d1 0

[0100] Where d1 and d2 represent the mask values, as can be seen from the previous formula, when d1 = 1 and d2 = ∞, the calculated result is a 4-connected region; when d1 = 1 and d2 = 1, the calculated result is an 8-connected region; when d1 = 3 and d2 = 4, dividing the distance graph by 3 gives a good approximation of the Euclidean distance. For example, as shown... Figure 4 The image shown.

[0101] Step 3.2: Calculate the maximum and minimum gray values ​​within the region. First, obtain the gray-level histogram of the region, and let r... k k = 0, 1, 2, ..., L-1 represents the gray levels of an L-level grayscale digital image f(x, y), and the formula for the non-normalized histogram of f is:

[0102] h(r k ) = n k k = 0, 1, 2, ..., L-1

[0103] In the formula n k The grayscale value of f is r k The number of pixels, similarly, the formula for the normalized histogram is:

[0104]

[0105] Where MN are the number of rows and columns of the image, respectively, and n k The grayscale value of f is r k The number of pixels. Based on the histogram, obtain the maximum and minimum values ​​at both ends, and then obtain the maximum and minimum gray levels.

[0106] Step 3.3: Scale the maximum value obtained in Step 3.2 proportionally and then perform erosion to break up the adhered parts of the ore. In morphological processing, the use of transposed structural elements in the Minkowski subtraction is called erosion, and the formula is as follows:

[0107]

[0108] Where S represents the structuring element, R represents the image to be eroded, and the structuring element is convolved with the image or set to be processed. t represents the element remaining after convolution. When the structuring element matches an element in the image or set, the element is retained; otherwise, the element is eroded.

[0109] Step 3.4, similar to step 1.2, uses the connectivity of the etched areas as a criterion to divide the region into different independent areas. The subsequent operations will treat these separated areas individually; in effect, these separate areas represent the separation of the ore's bonded areas. After isolating the areas, an expansion operation is performed to restore the etched areas, returning the ore region to its normal size. In morphological processing, the use of transposed structural elements in the Minkowski addition is called expansion, and the formula is as follows:

[0110]

[0111] Here, R is the region to be processed, S is called the structuring element, and t represents the element remaining after convolution. For each position where the structuring element S overlaps with the image R, if at least one corresponding element satisfies the condition, the output pixel at that position is retained. Different effects can be obtained by using structuring elements with different structures for erosion or dilation. By performing the same dilation operation on each independent region, the eroded region is restored; the independent regions are the areas where each piece of ore is located. For example,... Figure 5 The image shown.

[0112] Step S04: Calculate the number of ore pieces after segmentation in step S03, take the circumcircle of the contour fitting for each region, calculate the particle size of the ore, and obtain the center and radius of the circle.

[0113] Furthermore, the number of ores in the image and the minimum circumcircle of each ore are calculated to obtain the radius of the circumcircle of the ores. The average radius of the ores in the current image is then calculated. Based on the ratio of the size of the pixels within the camera's field of view to the actual physical size, the average physical particle size of the ores can be obtained.

[0114] Specifically, the number of ores in the image is calculated, the region where each segmented ore is located is read, and the minimum circumcircle of each ore region is fitted.

[0115] First, obtain the outline of each ore region. The formula for obtaining the outline is:

[0116] β(A) = A - (AB)

[0117] A represents the region where the contour needs to be calculated, B represents the structuring element used for erosion, and β(A) represents the obtained contour. A is eroded using structuring element B, and the difference between A and B is taken as the result. Then, the minimum circumcircle in any direction is calculated. The method for fitting the circumcircle is based on the convex hull of the region. Within a specific region, the convex hull of a set of points is the minimum convex set containing all points in the region. Using the region boundary with the largest distance as the circumcircle diameter, the circumcircle is fitted, and the center and radius are calculated. The fitted circumcircle is as follows: Figure 6 As shown. The average particle size of the current ore is calculated after calculating the particle size of all ore samples. The formula for the average particle size is as follows:

[0118]

[0119] N represents the quantity of ore, D i This indicates the particle size of each ore. Dv50 represents the average particle size.

[0120] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Any modifications, alterations, substitutions, and variations made by those skilled in the art to the above embodiments are within the scope of the present invention.

Claims

1. A method for obtaining the size of cohesive ore particles, characterized in that, Includes the following steps: Step S01: Locate the area where the conveyor belt is located in the entire image. The area where the conveyor belt is located is a solid color conveyor belt background image that contains ore but does not contain information about the surrounding environment. Step S02: Use the difference in grayscale values ​​between the ore and the conveyor belt background to perform threshold segmentation on the image obtained in step S01, extract all the ore areas, extract all the ore into the same area, fill the area, fill the holes in the ore, so that the ore area becomes a whole without holes. Step S03: Segment individual ore pieces from the overall image of the adhered ore obtained in step S02 and remove the adhesion of the ore pieces. The specific steps of step S03 are as follows: Step 3.1: Transform the area where the ore is located into a distance map using distance transformation; Step 3.2: Obtain the grayscale range of the distance map, calculate the maximum and minimum grayscale values. First, obtain the grayscale histogram of the region, let... Represents an L-level grayscale digital image grayscale, The formula for a nonnormalized histogram is: ； In the formula yes medium grayscale The number of pixels, similarly, the formula for the normalized histogram is: ; Where MN are the number of rows and columns of the image, respectively. yes medium grayscale The number of pixels; Based on the histogram, obtain the maximum and minimum values ​​at both ends, and then obtain the maximum and minimum gray levels; Step 3.3: Scale the maximum value obtained in Step 3.2 proportionally and then erode the area to break up the ore that is stuck together. Step 3.4: Using whether the eroded areas are connected as the standard, divide the area into different independent areas, perform separate operations on the separated areas, and perform an expansion operation on the independent areas to restore the eroded areas and restore the ore area to its normal size. Perform the same expansion operation on each independent area to restore the eroded areas. The independent areas are the areas where each piece of ore is located. Step S04: Calculate the number of ore pieces after segmentation in step S03, take the circumcircle of the contour fitting for each region, calculate the particle size of the ore, and obtain the center and radius of the circle.

2. The method for obtaining the size of cohesive ore particles according to claim 1, characterized in that: In step S01, locating the area where the conveyor belt is located involves threshold segmentation of the entire image, separating unconnected areas into several independent regions, selecting the largest region as the area where the conveyor belt is located, and cropping the area where the conveyor belt is located from the original image to obtain an image that contains the conveyor belt and ore without interference.

3. The method for obtaining the size of cohesive ore particles according to claim 2, characterized in that: In step S01, the specific steps for locating the area where the conveyor belt is located are as follows: Step 1.1: Perform grayscale processing on the image of the ore on the conveyor belt, select all pixels within the specified grayscale threshold range into the selected area, and then segment the selected area and the remaining area. The threshold segmentation formula is as follows: ； Where: S represents the set of pixels in the segmented region, Represents the pixel coordinates, T is the coordinate transpose symbol, R represents the set of all pixels in the image, and f represents the coordinates of a pixel. Gray-scale value mapping of a point, where g represents the range of gray-scale values; Step 1.2: Divide the region calculated in Step 1.1 into a series of independent small regions. These regions are composed of interconnected pixels. For connected regions, use run-length representation to describe the region. Use 4-neighborhood to describe the foreground and 8-neighborhood to describe the background. Use depth-first search algorithm to decompose unconnected regions into a series of independent small regions. Step 1.3: Using feature extraction, select the region with the largest area from a series of regions. This region is where the conveyor belt is located. Remove small regions caused by local noise. All points in the region are represented by binary images and run-length encoding. The formula for calculating the area of ​​the region using the region run-length encoding algorithm is as follows: ； Where: ɑ represents the area of ​​the region, and Г represents the number of points within the region. Indicates the starting coordinate point. Indicates the end coordinate point; Step 1.4: Crop the original image according to the obtained area to obtain an image that only contains the effective area inside the conveyor belt.

4. The method for obtaining the size of cohesive ore particles according to claim 1, characterized in that, The specific steps of step S02 are as follows: Step 2.1: Threshold segmentation is performed using the difference in grayscale values ​​between the ore and the background to divide the area containing the ore. The formula for threshold segmentation is: ； in Let S represent the set of pixels in the segmented region, g represent the range of grayscale values, R represent the set of all pixels in the image, and f represent the coordinates of the pixel. Point grayscale value mapping; Step 2.2: Divide the region into multiple independent small regions, select regions with an area greater than a certain value to eliminate noise interference, separate and extract all ore areas from the background conveyor belt, and remove noise. Step 2.3: Fill the holes inside the segmented ore area. Points within the area correspond to pixel 1 in the area image, and holes within the area correspond to pixel 0 in the area image. Fill the hole pixels with 1, remove the holes, and iterate using the following process to find and fill all hole areas: ； Where B is a symmetric structural element, if The algorithm terminates after k iterations. Includes all filled holes, and I represents the region of interest, which is the ore region that does not contain the background.

5. The method for obtaining the size of cohesive ore particles according to claim 1, characterized in that, The specific steps of step S04 are as follows: First, obtain the outline of each ore region. The formula for obtaining the outline is as follows: ； Where: A represents the region where the contour needs to be calculated, and B represents the structural element used for erosion. The obtained contour is represented by the erosion of A using the structuring element B. The difference between A and B is then taken as the result. The minimum circumcircle in any direction is then calculated. The method for fitting the circumcircle is based on the convex hull of the region. Within a specific region, the convex hull of a set of points is the minimum convex set that contains all points in the region. The distance with the maximum distance from the region boundary is used as the diameter of the circumcircle. The circumcircle is fitted, and the center and radius are obtained. After calculating the particle size of all ore, the average value is calculated as the average particle size of the current ore.

Images