Iron ore grade rapid identification method based on image processing

Abstract

This invention relates to a rapid iron ore grade identification method based on image processing for on-site application during mining. The method comprises the following steps: 1) acquiring digital images of iron ore samples of different grades from the mining area and establishing an iron ore grade identification model using a deep learning algorithm; 2) using this identification model to rapidly identify the iron ore grade within the boreholes of each blasting unit block, then estimating the iron ore grade of each unit block using the inverse power of distance method based on this grade, and finally calculating the iron ore grade of the entire blasting area using the inverse power of distance method for each unit block. The advantages of this invention are: it overcomes the shortcomings of complex procedures, large computational load, and long time consumption in determining iron ore grade in mining areas; the identification time for a single image is only 0.3 seconds. Compared with laboratory test values, the identification accuracy can reach 90%, achieving rapid on-site identification and estimation of iron ore grade, laying the foundation for building intelligent mines.

Description

Technical Field

[0001] This invention belongs to the field of deep learning application technology, and in particular relates to a rapid identification method for iron ore grade based on image processing for on-site application during mining. Background Technology

[0002] Traditional iron ore grade determination employs chemical titration. While accurate and widely used, this method suffers from drawbacks such as cumbersome sample collection and delivery, complex measurement procedures, long processing times, and poor timeliness. To address these shortcomings, domestic and international scholars have proposed various algorithms and estimation models. However, these algorithms and models still suffer from complex procedures, high computational load, and long processing times for ore grade determination. In recent years, methods such as spectroscopic analysis and X-ray transmission have also been applied to ore grade identification. However, these methods require processing of ore samples before laboratory testing to obtain results, failing to meet the requirements for rapid, efficient, immediate, and on-site determination of iron ore grade. Furthermore, manual estimation of ore grade by experienced technicians is subject to subjective factors, leading to uncertainties and potential for large or small errors in grade estimation.

[0003] Currently, the weighted average method is used in mines to calculate iron ore grade. This method is simple to operate, requires little calculation, and predicts relatively small fluctuations in grade at mining sites. However, because it cannot be verified in a timely manner, it is not conducive to accurately predicting the overall ore grade of the mining area. This invention utilizes the inverse power law method of distance, which can more accurately estimate ore grade. Compared to the weighted average method, the inverse power law method of distance is more accurate and requires only a few parameters to be set, thus providing a more scientific estimate of regional ore grade and further improving identification accuracy.

[0004] With the rapid development of artificial intelligence, image recognition technology based on deep learning neural networks has spawned a series of fast and accurate algorithms. Iron ore powder of different grades varies in color, making it feasible to estimate ore grade through image recognition. Applying these algorithms to the recognition of images of ores of different grades, and establishing and optimizing recognition models, can achieve rapid, efficient, real-time, and on-site determination of iron ore grade, laying the foundation for building intelligent mines. Summary of the Invention

[0005] The purpose of this invention is to provide a rapid iron ore grade identification method based on image processing for on-site application during mining. By establishing an iron ore grade identification model based on image processing technology, images of iron ore powder inside the blast holes of each unit block are captured within the entire blasting area of ​​the mining site. Then, the Keras neural network learning algorithm is used in the identification model for rapid grade identification. Subsequently, the iron ore grade value of each unit block is calculated using the inverse power law of distance method. Finally, the iron ore grade of the entire blasting area is estimated using the iron ore grade values ​​of each unit block. This invention can solve the problems of rapid, efficient, real-time, and on-site determination of iron ore grade in mining sites, thereby improving the intelligence of mining operations.

[0006] The objective of this invention is achieved through the following technical solution:

[0007] The present invention provides a method for rapid identification of iron ore grade based on image processing, characterized by comprising the following steps:

[0008] Step 1: Establish an iron ore grade identification model based on image processing technology

[0009] S1.1 Preparation of iron ore standard samples of different grades

[0010] First, based on the geological data of the iron deposit, the iron grade range of the iron ore is determined. Then, representative ore samples are collected in the mining area at grade intervals of 1% and numbered. In a laboratory environment, each sample is crushed to produce iron ore powder with a particle size of 0≤d≤1.5mm. Finally, equal amounts of 3~5g of each sample of iron ore powder are taken and spread in the center of an A4 sheet of paper to form a circular iron ore powder pile sample with the same area and a flat surface.

[0011] S1.2 Acquiring digital images of circular iron ore powder pile samples.

[0012] The circular iron ore powder pile samples were illuminated with LED lights and photographed with a camera at color temperatures of 2700K~3500K and 4000K~6000K respectively. The number of digital images of each circular iron ore powder pile sample was 150~200.

[0013] S1.3 Digital Image Preprocessing and Sample Set Establishment

[0014] Denoising and data augmentation preprocessing were performed on digital images of circular iron ore powder pile samples to expand the digital image data and establish a sample set.

[0015] S1.4 Establish an iron ore grade identification model

[0016] A model for identifying iron ore grade was generated using the Keras neural network learning algorithm. The model was anchored to the middle part of the ore powder and trained using a sample set. Features such as color and particle size of iron ore of different grades were extracted from the preprocessed digital image data and input into the Keras neural network learning algorithm. The model was then trained using the sample set to establish the iron ore grade identification model.

[0017] Step 2: First, use the recognition model to identify the iron ore grade in the borehole. Then, use the inverse power of distance method to estimate the iron ore grade of each unit block. Finally, use the inverse power of distance method again to obtain the iron ore grade of the entire blasting area based on the iron ore grade of each unit block.

[0018] S2.1, Unit block of planned mining blasting area

[0019] First, the blasting area in the mining area is set as several target unit blocks, and the size of the unit blocks is calculated. Then, the radius R of the inscribed circle of the unit block is calculated with the center of the unit block to be identified as the center, and the area inside the inscribed circle is determined as the influence range. Finally, the distance between the center of each blast hole falling within the influence range and the center of the unit block is measured.

[0020] S2.2 Identify the grade value of iron ore powder in each borehole within the unit block using an iron ore grade identification model.

[0021] Digital images of iron ore powder taken from each borehole are captured by a camera and input into the iron ore grade identification model established in step 1 to identify and determine the grade of iron ore powder in each borehole.

[0022] S2.3. Based on the iron ore powder grade obtained from each borehole, the iron ore grade of the unit block is estimated using the following formula:

[0023]

[0024] In formulas (1) and (2), —Weighting coefficient, which is the inverse ratio of the k-th power of the distance from the borehole to the center of the unit block to be identified; gi—grade value of the i-th borehole; di—distance from the i-th borehole to the center of the unit block to be estimated; g—ore grade of the unit block to be identified, that is, the iron ore grade of each unit block is equal to the weighted average of the iron ore grades of each borehole within its influence range.

[0025] S2.4. The iron ore grade of the entire blasting zone is obtained by recalculating the iron ore grade of each unit block using the distance power inverse ratio method as follows:

[0026]

[0027] A—Area of ​​the irregular blasting zone; Xc—Abscissa of the centroid; Yc—ordinate of the centroid;

[0028] If the blasting zone is an irregular shape, the centroid is the center. Select a point on the shape as the origin. The x-axis coordinate is obtained in formula (3), and the y-axis coordinate is obtained in formula (4). A can be solved by formula (5) to be the area of ​​the irregular shape. After determining the center of the blasting zone, measure the distance from the center of each unit block to the center of the blasting zone.

[0029]

[0030] In formulas (6) and (7), —Weight coefficient, the distance from the center of the unit block to the center of the explosion zone to be identified, i.e., the inverse ratio of the distance to the power of k; g j — The grade value of the j-th unit block; Di — The distance of the i-th unit block from the center of the blasting area to be estimated; G — The ore grade of the blasting area to be identified, that is, the iron ore grade of each blasting area is equal to the weighted average of the iron ore grades of each unit block within its influence range.

[0031] Furthermore, in step 1.2, the camera is used to photograph the sample, with the camera's top-down angle forming a 90° angle with the plane where the mineral powder is spread; the camera's side-down angle forming a 40°~50° angle with the plane where the mineral powder is spread.

[0032] Furthermore, in S1.3, the digital image of the circular iron ore powder pile sample undergoes noise reduction and data augmentation preprocessing to expand the digital image data and establish a sample set. The specific process is as follows:

[0033] First, the digital images of iron ore samples of different grades were numbered in groups, with 1% intervals, sequentially numbered from low grade to high grade. Then, the digital images were augmented by geometric transformations such as sample translation, flipping, and rotation, while randomly adjusting the brightness and contrast of the LED lights. WidsMob Denoise software was used to preprocess the images by adjusting chromatic noise, luminance noise, and sharpness parameters. Finally, a sample set of no less than 400 digital images was established.

[0034] Further, in step 2.1, the blasting area of ​​the stope is set as several target unit blocks, and the size of the unit block and the radius of the inscribed circle of the unit block are calculated using the following formula:

[0035]

[0036] n x — The number of boreholes in the x-direction within the unit block; n y —Number of blast holes in the y-direction within a unit block; N—Total number of blast holes within the blast zone; T—Number of target unit blocks.

[0037] Substituting the data N and T into formula (8), we get That is, the number of blast holes in each unit block;

[0038] Let n x =n y , to obtain n x n y , will n x n y Substituting these values ​​into formulas (9) and (10) respectively, we obtain the unit block size and the radius R of the inscribed circle as follows:

[0039]

[0040] L x — Unit block length; L y —Unit block width; a—Hole spacing; b—Hole row spacing; R—Radius of the inscribed circle.

[0041] Compared with the prior art, the advantages of the present invention are:

[0042] 1) This invention uses image processing technology to quickly identify iron ore grade. By preparing samples of iron ore with different grades in the mining area, digital images of the samples are collected. Then, a deep learning algorithm is used to establish an iron ore grade identification model. Subsequently, the identification model is used to quickly identify the iron ore grade in the boreholes of the blasting area unit block. Then, based on the grade, the iron ore grade of each unit block is estimated using the distance power inverse ratio method. Finally, the iron ore grade of the entire blasting area is estimated from the iron ore grade of each unit block.

[0043] 2) This invention solves the problems of complex process, large amount of calculation and long time consumption in the determination of iron ore grade in the mining area, and realizes rapid on-site identification and estimation of iron ore grade, laying the foundation for building intelligent mines.

[0044] 3) This invention provides real-time on-site estimation and determination of iron ore grade. This process does not require manual testing, and the recognition time for a single image is only 0.3 seconds. Compared with laboratory test values, the absolute error of iron ore grade identification is 0.01, meaning that the model's accuracy in identifying ore grade can reach 90%. Attached Figure Description

[0045] Figure 1 This is a top view diagram of a square-hole blasting method;

[0046] Figure 2 This is a top-view diagram of a triangular hole-blasting method.

[0047] exist Figure 1 and Figure 2 In the diagram, 1—blast hole; 2—slope top line; 3—unit block; 4—distance from blast hole to center of unit block; 5—inscribed circle; 6—blasting zone. Detailed Implementation

[0048] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0049] Example 1

[0050] like Figure 1 As shown, in a certain iron ore mine, the blasting design is a square hole layout, using shallow hole blasting with a hole diameter of 40mm, a hole spacing of 1.5m, a hole row spacing of 1.5m, and a total of 81 blast holes.

[0051] The present invention provides a method for rapid identification of iron ore grade based on image processing, characterized by comprising the following steps:

[0052] Step 1: Establish an iron ore grade identification model based on image processing technology

[0053] S1.1 Preparation of iron ore standard samples of different grades

[0054] Based on the geological data of the iron ore deposit, the iron grade range of the iron ore was determined to be 25%≤α≤35%. Therefore, representative ore samples with grades of 25%≤α≤35% were first collected in the mining area at grade intervals of 1%, totaling ten groups, with three samples in each group. The sample number and grade value are shown in Table 1. Then, in a laboratory environment, each sample was crushed to produce iron ore powder with a particle size of 0≤d≤1.5mm. Finally, 5g of each sample of iron ore powder was taken and spread in the center of an A4 sheet of paper to form circular iron ore powder piles with the same area, flat surface, and a thickness of 1mm.

[0055] Table 1 Sample Numbers and Grade Values

[0056]

[0057] S1.2 Acquiring digital images of circular iron ore powder pile samples.

[0058] Circular iron ore powder pile samples were illuminated using LED lights with controllable light intensity. The samples were photographed using a camera at color temperatures of 2700K, 3500K, 4000K, 5000K, and 6000K. The camera was used to photograph the circular iron ore powder pile samples from a top-down angle of 90° to the plane containing the powder; and from a side-view angle of 40°, 45°, and 50° to the plane containing the powder. Each circular iron ore powder pile sample yielded 200 digital images.

[0059] S1.3 Digital Image Preprocessing and Sample Set Establishment

[0060] The digital images of the samples undergo noise reduction and data augmentation preprocessing to expand the digital image data and establish a sample set. The specific process is as follows:

[0061] First, the digital images are augmented by performing geometric transformations such as translation, flipping, and rotation of the samples, while randomly adjusting the brightness and contrast of the LED lights. Then, WidsMob Denoise software is used to perform image noise reduction preprocessing by adjusting the chromatic noise, luminance noise, and sharpness parameters. Finally, the digital images of each sample are expanded to 420 images to establish a digital image sample set.

[0062] S1.4 Establish an iron ore grade identification model

[0063] A model for identifying iron ore grade was generated using the Keras neural network learning algorithm and trained using a sample set. The specific process is as follows:

[0064] Features such as color and particle size of iron ore of different grades are extracted from preprocessed digital image data and input into Keras neural network learning algorithm. The algorithm is then trained using a sample set to establish an iron ore grade identification model.

[0065] Step 2: First, use the recognition model to identify the iron ore grade in borehole 1. Then, use the distance power inverse ratio method to estimate the iron ore grade of each unit block. Finally, use the iron ore grade of each unit block to estimate the iron ore grade of the entire blasting zone 6.

[0066] The following steps S2.1, S2.2, and S2.3 all use the test unit block 3 as an example to estimate the iron ore grade of the unit block. The measurement and calculation process for other unit blocks is the same.

[0067] S2.1, Unit block of planned mining blasting zone 6

[0068] like Figure 1 As shown, blasting zone 6 is located inside the top line 2 of an iron ore mine. The blasting design uses a square hole layout with shallow hole blasting. The hole diameter is 40 mm, the hole spacing a is 1.5 m, the hole row spacing b is 1.5 m, and the number of blast holes N is 81. The number of target unit blocks T in this blasting zone 6 is set to nine unit blocks.

[0069] The dimensions of the unit block 3 to be measured are calculated according to formulas (8) and (9); then, taking the center of the unit block 3 to be measured as the center, the radius R of the inscribed circle 5 of the unit block is calculated according to formula (10), and the area inside the inscribed circle 5 is determined as the influence range; finally, the distance 4 between the center of each borehole 1 falling within the influence range and the center of the unit block is measured, as follows:

[0070]

[0071] nx —The number of boreholes in the x-direction within the unit block; n y —The number of boreholes in the y-direction within the unit block is 1;

[0072] N—Total number of boreholes 1 within blast zone 6; T—Number of target unit blocks;

[0073] Substituting the data into formula (8), we get =9, meaning there are 9 blast holes in each unit block.

[0074] Because the blasting design uses a square hole arrangement, then n x =n y =3

[0075]

[0076] L x — Unit block length; L y —Unit block width; a—Hole spacing; b—Hole row spacing; R—Radius of the inscribed circle 5

[0077] n x =n y Substituting 3 into formulas (9) and (10), we get:

[0078] L x =L y =4.5m, meaning the size of the unit block 3 to be measured is 4.5 m. If the radius of the inscribed circle is 4.5 m, then the radius of the inscribed circle 5 is R = 2.25 m.

[0079] The blasting design uses a square hole layout, with n=7 blast holes falling within the influence range. These 7 blast holes are represented by N1, N2, N3, N4, N5, N6, and N7, respectively. The specific distances di between the centers of the 7 blast holes and the center of the unit block are shown in Table 2.

[0080] Table 2. Measurement results of the distance 4 between each borehole and the center of the unit block.

[0081]

[0082] S2.2. Use the recognition model to identify the grade value of iron ore powder in each borehole 1 within the unit block to be tested 3.

[0083] Digital images of iron ore powder inside seven boreholes (N1, N2, N3, N4, N5, N6, and N7) were captured using a camera and input into the iron ore grade identification model established in step 1 to determine the iron ore powder grade. The identification results are shown in Table 3.

[0084] Table 3. Iron ore powder grade values ​​in each borehole

[0085]

[0086] S2.3. Based on the iron ore powder grade of each borehole 1, the iron ore grade of the test unit block 3 is estimated using formulas (1) and (2), and the iron ore grade of the test unit block 3 is obtained. The specific process is as follows:

[0087] Substitute the distance di between each borehole 1 and the center of the unit block into formula (1), where k=1;

[0088]

[0089] get ,Right now =0.05, =0.05, =0.06, =0.68, =0.05, =0.06, =0.05;

[0090] Then Compared with Table 3 Substitute into formula (2):

[0091]

[0092] get =0.2951;

[0093] That is, the iron ore grade of the unit block 3 to be tested is 29.51%.

[0094] Table 4 shows the estimated iron ore grades for the nine unit blocks within the blasting zone of the stope.

[0095] Table 4. Estimated iron ore grade of nine blocks in the blasting zone.

[0096]

[0097] S2.4 The iron ore grade of the entire blast zone 6 is obtained by secondary calculation using the distance power inverse ratio method on the iron ore grade of the nine unit blocks. The specific process is as follows:

[0098]

[0099] A—Area of ​​the irregular blasting zone; Xc—Abscissa of the centroid; Yc—ordinate of the centroid;

[0100] First, determine the lower left corner as the origin of the coordinate system. Use formula (3) to calculate the centroid x-coordinate of the irregular blasting area as 6.8. Use formula (4) to calculate the centroid y-coordinate of the irregular blasting area as 4.7. The area of ​​the irregular shape of the blasting area is 104 according to formula (5).

[0101] The distance from each unit block to the center of the blast zone was measured. There are nine unit blocks in the blast zone, which are labeled B1, B2, B3, B4, B5, B6, B7, B8, and B9. The specific distances are shown in Table 5.

[0102] Table 5. Measurement results of the distance between each unit block and the center of the blast zone.

[0103]

[0104] The iron ore grade of the test blasting area is estimated using formulas (6) and (7), and the specific process is as follows:

[0105] Substitute the distance Di between the center of each unit block and the center of the explosion zone into formula (6), where k=1;

[0106]

[0107] get ,Right now =0.068, =0.102, =0.185, =0.101, =0.070, =0.106, =0.239, =0.090, =0.044;

[0108] Then Compared with g in Table 4 j Substitute into formula (7):

[0109]

[0110] get =0.2949;

[0111] That is, the iron ore grade of the entire blasting area is 29.49%.

[0112] Example 2

[0113] Because these are different blasting zones 6 within the same iron ore deposit mining area, the iron ore grade identification model based on image processing technology established in step 1 of Example 1 is still applicable to identifying the iron ore grade of this blasting zone. Therefore, the rapid iron ore grade identification method based on image processing of the present invention is characterized in that Example 2 is executed starting from step 2 of the present invention, and the specific process is as follows:

[0114] Step 2: First, use the recognition model to identify the iron ore grade in borehole 1. Then, use the inverse power distance method to estimate the iron ore grade of each unit block. Finally, use the inverse power distance method to estimate the iron ore grade of the entire blasting zone 6 a second time based on the iron ore grade of each unit block.

[0115] The following steps S2.1, S2.2, and S2.3 all use the test unit block 3 as an example to estimate the iron ore grade of the unit block. The measurement and calculation process for other unit blocks is the same.

[0116] S2.1, Unit block of planned mining blasting zone 6

[0117] like Figure 2 As shown, blasting zone 6 is located inside the top line 2 of a certain iron mine. The blasting design is a triangular hole layout, using deep hole blasting with a hole diameter of 178 mm, a hole spacing a of 8 m, a hole row spacing b of 4.8 m, and a number of blast holes N of 98. The number of target unit blocks T in this blasting zone 6 is set to seven unit blocks.

[0118] The dimensions of unit block 3 are calculated according to formulas (8) and (9); then, taking the center of the unit block 3 to be measured as the center, the radius R of the inscribed circle 5 of the unit block is calculated according to formula (10), and the area inside the inscribed circle 5 is determined as the influence range; finally, the distance 4 between the center of each borehole falling within the influence range and the center of the unit block is measured, as follows;

[0119]

[0120] n x —The number of boreholes in the x-direction within the unit block; n y —The number of boreholes in the y-direction within the unit block is 1;

[0121] N—Total number of blast holes in blast zone 6; T—Number of target unit blocks.

[0122] Substituting the data into formula (8), we get =14, meaning there are 14 blast holes in each unit block;

[0123] Let n x = = 3.74; then n y =14 / nx =3.74;

[0124]

[0125] L x — Unit block length; L y —Unit block width; a—Hole spacing; b—Hole row spacing; R—Radius of the inscribed circle 5.

[0126] n x =n y Substituting 3.74 into formulas (9) and (10), we get:

[0127] L x = =3.74 8 = 29.92 30;

[0128] L y = =3.74 4.5 = 16.83 17;

[0129] That is, the size of the unit block is 30 m If the radius of the inscribed circle is 17 m, then the radius R of the inscribed circle 5 is 8.5 m.

[0130] The blasting design uses a triangular hole layout, with n=8 blast holes falling within the influence range. These 8 blast holes are represented by N1, N2, N3, N4, N5, N6, N7, and N8, respectively. The specific distances 4 between the centers of the 8 blast holes and the center of the unit block are shown in Table 6.

[0131] Table 6. Results of distance measurement between each borehole and the center of the unit block.

[0132]

[0133] S2.2. Use the recognition model to identify the grade value of iron ore powder in each borehole 1 within the unit block to be tested 3.

[0134] Table 7 Iron ore powder grade values ​​in each borehole

[0135]

[0136] S2.3. Based on the iron ore powder grade of each borehole, the iron ore grade of the test unit block 3 is estimated using formulas (1) and (2). The specific process is as follows:

[0137] Substitute the distance di between each borehole 1 and the center of the unit into formula (1), where k=1;

[0138]

[0139] get ,Right now =0.10, =0.15, =0.09, =0.08, =0.28, =0.09, =0.08, =0.15;

[0140] Then Compared with Table 7 Substitute into formula (2)

[0141]

[0142] get =0.3076;

[0143] That is, the iron ore grade of unit block 3 to be estimated and tested is 30.76%.

[0144] Table 8 shows the estimated iron ore grades for the seven unit blocks within blast zone 6 of the stope.

[0145] Table 8. Estimated iron ore grade of seven blocks in the blasting zone.

[0146]

[0147] S2.4. The iron ore grade of the entire blasting zone is obtained by secondary calculation of the iron ore grade of each unit block using the distance power inverse ratio method. The specific process is as follows:

[0148]

[0149] A—Area of ​​the irregular blasting zone; Xc—Abscissa of the centroid; Yc—ordinate of the centroid;

[0150] Similar to step S2.4 in Example 1, use formula (3) to determine the lower left corner as the origin of the coordinate system, calculate the x-coordinate of the centroid of the irregular blasting area as 41.2, use formula (4) to calculate the y-coordinate as 19.6, and use formula (5) to obtain the area of ​​the irregular shape of the blasting area as 3450.

[0151] Measure the distance from each unit block to the center of the blast zone. There are seven unit blocks in the blast zone, which are labeled B1, B2, B3, B4, B5, B6, and B7. The specific distances are shown in Table 9.

[0152] Table 9 shows the measurement results of the distance between each unit block and the center of the blast zone.

[0153]

[0154] Similar to step S2.4 in Example 1, the iron ore grade of the test blasting area is estimated to obtain the iron ore grade of the test blasting area. The specific process is as follows:

[0155] Substitute the distance Di between each unit block 3 and the center of the explosion zone into formula (6), where k=1

[0156]

[0157] get ,Right now =0.079, =0.151, =0.172, =0.072, =0.076, =0.206, =0.244;

[0158] Then Compared with g in Table 8 j Substitute into formula (7)

[0159]

[0160] The calculated value is G = 0.2922;

[0161] The iron ore grade in the test zone is 29.22%.

Claims

1. A method for rapid identification of iron ore grade based on image processing, characterized in that, Includes the following steps: S1. Establish an iron ore grade identification model based on image processing technology. S1.1 Preparation of iron ore standard samples of different grades First, based on the geological data of the iron deposit, the iron grade range of the iron ore is determined. Then, representative ore samples are collected in the mining area at grade intervals of 1% and numbered. In a laboratory environment, each sample is crushed to produce iron ore powder with a particle size of 0≤d≤1.5mm. Finally, equal amounts of 3~5g of each sample of iron ore powder are taken and spread in the center of an A4 sheet of paper to form a circular iron ore powder pile sample with the same area and a flat surface. S1.2 Acquiring digital images of circular iron ore powder pile samples. The circular iron ore powder pile samples were illuminated with LED lights and photographed with a camera at color temperatures of 2700K~3500K and 4000K~6000K respectively. The number of digital images of each circular iron ore powder pile sample was 150~200. S1.3 Digital Image Preprocessing and Sample Set Establishment Denoising and data augmentation preprocessing were performed on digital images of circular iron ore powder pile samples to expand the digital image data and establish a sample set. S1.4 Establish an iron ore grade identification model A model for identifying iron ore grade was generated using the Keras neural network learning algorithm and trained using a sample set. Features such as color and particle size of iron ore of different grades are extracted from preprocessed digital image data and input into Keras neural network learning algorithm. The algorithm is trained using a sample set to establish an iron ore grade identification model. S2. First, the iron ore grade in the borehole is identified using the identification model. Then, the iron ore grade of each unit block is estimated using the inverse power of distance method. Finally, the iron ore grade of the entire blasting zone is estimated using the iron ore grade of each unit block. S2.1, Unit block of planned mining blasting area First, the blasting area in the mining area is set as several target unit blocks, and the size of the unit blocks is calculated. Then, the radius R of the inscribed circle of the unit block is calculated with the center of the unit block to be identified as the center, and the area inside the inscribed circle is determined as the influence range. Finally, the distance between the center of each blast hole falling within the influence range and the center of the unit block is measured. S2.2 Identify the grade value of iron ore powder in each borehole within the unit block using an iron ore grade identification model. Digital images of iron ore powder taken from each borehole are captured by a camera and input into the iron ore grade identification model established in step 1 to identify and determine the grade of iron ore powder in each borehole. S2.

3. Based on the iron ore powder grade obtained from each borehole, the iron ore grade of the unit block is estimated using the following formula: ； In formulas (1) and (2), —Weighting coefficient, which is the inverse ratio of the k-th power of the distance from the borehole to the center of the unit block to be identified; gi—grade value of the i-th borehole; di—distance from the i-th borehole to the center of the unit block to be estimated; g—ore grade of the unit block to be identified, that is, the iron ore grade of each unit block is equal to the weighted average of the iron ore grades of each borehole within its influence range. S2.4 The iron ore grade of the entire blast zone was calculated using the inverse power of distance method for each unit block, as follows: ； A—Area of ​​the irregular blast zone; Xc—Abscissa of the centroid; Yc—ordinate of the centroid; If the blasting zone is an irregular shape, the centroid is the center. Select a point of the shape as the origin. The x-axis coordinate is obtained in formula (3), and the y-axis coordinate is obtained in formula (4). A can be solved by formula (5) to be the area of ​​the irregular shape. After determining the center of the blasting zone, measure the distance from the center of each unit block to the center of the blasting zone. ； In formulas (6) and (7), —Weight coefficient, the distance from the center of the unit block to the center of the explosion zone to be identified, i.e., the inverse ratio of the distance to the power of k; g j — The grade value of the j-th unit block; Di — The distance of the i-th unit block from the center of the blasting area to be estimated; G — The ore grade of the blasting area to be identified, that is, the iron ore grade of each blasting area is equal to the weighted average of the iron ore grades of each unit block within its influence range.

2. The method for rapid identification of iron ore grade based on image processing according to claim 1, characterized in that, In S1.2, the camera is used to photograph the sample. The camera's top-down angle is 90° to the iron ore powder surface, and the camera's side-view angle is 40°~50° to the iron ore powder surface.

3. The method for rapid identification of iron ore grade based on image processing according to claim 1, characterized in that, In S1.3, the digital image of the circular iron ore powder pile sample undergoes noise reduction and data augmentation preprocessing to expand the digital image data and establish a sample set. The specific process is as follows: First, the digital images of iron ore samples of different grades were numbered in groups, with 1% intervals, sequentially numbered from low grade to high grade. Then, the digital images were enhanced by geometric transformations such as sample translation, flipping, and rotation, while randomly adjusting the brightness and contrast of the LED lights. WidsMob Denoise software was used to preprocess the images by adjusting chromatic noise, luminance noise, and sharpness parameters. Finally, a sample set of 350-400 digital images was established.

4. The method for rapid identification of iron ore grade based on image processing according to claim 1, characterized in that, In step 2.1, the blasting area in the stope is defined as several target unit blocks. The size of the unit block and the radius of its inscribed circle are calculated using the following formulas: ； n x — The number of boreholes in the x-direction within the unit block; n y —Number of blast holes in the y-direction within a unit block; N—Total number of blast holes within the blast zone; T—Number of target unit blocks; Substituting the data N and T into formula (8), we get That is, the number of blast holes in each unit block; Let n x =n y , to obtain n x n y , will n x n y Substituting these values ​​into formulas (9) and (10) respectively, we obtain the unit block size and the radius R of the inscribed circle as follows: ； L x — Unit block length; L y —Unit block width; a—Hole spacing; b—Hole row spacing; R—Radius of the inscribed circle.

Images