A method for estimating the volume of blasted blocks in the field based on image processing

Abstract

This invention relates to a method for estimating the volume of blasted ore blocks based on image processing, belonging to the field of blasted block size recognition technology. The invention measures the volume of blasted ore samples to form an ore sample volume dataset; it acquires images of the ore samples and uses a deep learning UNet model for image segmentation, extracting the projected area and longest particle size of the blocks, and calculating two-dimensional morphological feature values. f and three-dimensional morphological feature values k Quantitative two-dimensional morphological feature values j With three-dimensional morphological feature values k The correlation; based on the theory of the equal-area sphere assumption and the projected area of ​​the ore sample block, the volume of the ore sample block is estimated to obtain the volume of the equal-area sphere. V b ; Construct a volume estimation model for blasted blocks; Build a volume estimation system for blasted blocks based on the PyCharm platform to estimate the volume of blasted blocks on site, and obtain the estimated volume of blasted blocks. V E ; Draw the longest particle size d Projected area A and estimated volume of blasted blocks V E The cumulative gradation curve is used to achieve automated volume estimation.

Description

Technical Field

[0001] This invention relates to a method for estimating the volume of blasted blocks in the field based on image processing, belonging to the field of blasted block size recognition technology. Background Technology

[0002] The size distribution of blasted ore piles is of significant importance for optimizing the production process. Larger ore sizes can lead to overload of downstream crushing equipment, increasing energy consumption and potentially causing equipment damage. Furthermore, different size distributions can affect the efficiency of ore transportation and loading. Therefore, accurately measuring the size distribution of blasted piles is crucial for timely adjustment of blasting parameters and achieving optimal economic benefits.

[0003] In existing technologies, improving the accuracy of explosive block size recognition mainly relies on optimizing image segmentation algorithms to obtain more accurate two-dimensional shape parameters, such as the longest particle diameter and projected area of ​​the explosive block. However, research on methods for measuring the volume of explosive blocks is still limited, and existing methods are mostly based on assumptions of regular shapes, such as the assumption of an equal-area sphere, an equal-area triangular pyramid, and an equal-area cube. Since explosive blocks are often irregular in shape, this leads to significant measurement errors. Although 3D point cloud technology can achieve 3D modeling of the blocks and measure their volume, the sheer number of explosive blocks in the field necessitates a significant investment of time and effort to model all of them. Summary of the Invention

[0004] To address the problems of irregular shapes in blasted blocks leading to significant errors in volume estimation, and the inefficiency of 3D modeling for measuring blasted blocks, this invention proposes an image processing-based method for estimating the volume of blasted blocks in the field. The method employs a displacement method to measure the volume of blasted ore samples, creating an ore sample volume dataset. Images of the ore samples are acquired, and a deep learning UNet model is used for image segmentation to extract the projected area and longest particle size of the blocks, calculating two-dimensional morphological feature values. And three-dimensional morphological feature value k; quantized two-dimensional morphological feature value The correlation with the three-dimensional morphological characteristic value k; based on the theory of the equal-area sphere assumption and the projected area of ​​the ore sample block, the volume of the ore sample block is estimated, and the volume V of the equal-area sphere is obtained. b Construct a volume estimation model for blasted blocks; build a volume estimation system for blasted blocks based on the PyCharm platform to estimate the volume of blasted blocks on site, and obtain the estimated volume V of the blasted blocks. E ; Draw the longest particle diameter d, projected area A, and estimated volume V of the blasted block. E The cumulative gradation curve is obtained; the entire process from image acquisition to volume estimation and volume gradation curve plotting is automated to improve the efficiency of volume estimation.

[0005] A method for estimating the volume of blasted blocks in a field based on image processing, the specific steps of which are as follows:

[0006] S1. After blasting in the mine, ore samples are taken to obtain several ore samples. These samples are then sieved according to the particle size range of small, medium, and large to obtain ore sample sets with different particle size ranges.

[0007] S2. The volume V of each ore sample in the ore sample set with different particle size ranges is measured by the displacement method to form an ore sample volume dataset;

[0008] S3. Collect images of each ore sample from a ore sample set with different particle size ranges, and annotate the images of each ore sample to form an ore sample image dataset;

[0009] S4. Train the UNet deep learning model on the labeled images in the ore sample image dataset with different particle size ranges to complete image segmentation, extract the projected area A and the longest particle size d of each ore sample block, and calculate the two-dimensional morphological feature value. and three-dimensional morphological feature value k;

[0010] S5. Two-dimensional morphological feature values and μ k Nonlinear fitting is performed to obtain the fitting function relationship, thereby quantifying the two-dimensional morphological feature values. The correlation with the three-dimensional morphological feature value k; the μ k The relationship between μ and the three-dimensional morphological feature value k is: k = lgk;

[0011] S6. Based on the assumption of an equal-area sphere and the projected area of ​​the ore sample block, estimate the volume of the ore sample block to obtain the volume V of the equal-area sphere. b ;

[0012] S7. Volume V of a sphere with equal area b Linear fitting was performed with the measured volume V of each ore sample in S2. The fitting coefficients m and n are obtained respectively;

[0013] S8. Use Pearson correlation analysis to evaluate the fit coefficients m and n and the morphological eigenvalues. The correlation between k;

[0014] S9. Perform nonlinear fitting on the three-dimensional morphological feature value k and the fitting coefficient m to obtain the functional relationship, and then apply the two-dimensional morphological feature value from step S5. and μ k Logarithmic transformation is performed on the fitted function relationship to construct a volume estimation model for blasted blocks;

[0015] S10. Based on the UNet image segmentation algorithm and the blasted block volume estimation model, a blasted block volume estimation system was built on the PyCharm platform to estimate the volume of blasted blocks on site, obtaining the estimated volume V of the blasted blocks. E And plot the longest particle size d, projected area A, and estimated volume V of the blasted block. E The cumulative gradation curve.

[0016] In step S1, the range of small particle size D is D < 50 mm, the range of medium particle size is 50 ≤ D ≤ 100 mm, and the range of large particle size is 100 < D ≤ 150 mm.

[0017] The two-dimensional morphological feature value in step S4 The calculation formula is:

[0018] (1);

[0019] The formula for calculating the three-dimensional morphological feature value k is:

[0020] (2)

[0021] In the formula, A is the projected area of ​​the ore sample block, in mm. 2 ; d is the longest particle size of the ore sample block, mm; V is the measured volume of the ore sample block, mm. 3 .

[0022] The fitting function relationship in step S5 is as follows:

[0023] (3)

[0024] In the formula, μ k = lgk, where k is the three-dimensional morphological feature value. These are two-dimensional morphological feature values.

[0025] The formula for estimating the volume of the ore sample block in step S6 is as follows:

[0026] (4)

[0027] In the formula, V b Let A be the volume of a sphere with equal surface area, and let A be the projected area of ​​the ore sample block in mm. 2 .

[0028] The functional relationship in step S9 is as follows:

[0029] (5)

[0030] In the formula: m is the fitting coefficient, and k is the three-dimensional morphological feature value.

[0031] The volume estimation formula for the blasted block volume estimation model in step S9 is as follows:

[0032] (6)

[0033] In the formula: V E Estimate the volume of the blasted block in mm 3 A represents the projected area of ​​the ore sample block, in mm². 2 ; These are two-dimensional morphological parameters.

[0034] The beneficial effects of this invention are:

[0035] (1) The present invention constructs the conversion relationship between the two-dimensional projection shape feature value and the three-dimensional shape feature value of the rock block. Only by extracting the block projection area (A) and the longest diameter (d) in the blasting image through image processing technology can the volume of the blasted block be estimated.

[0036] (2) The method of the present invention is simple to operate and easy to program. Compared with the volume calculation of three-dimensional point clouds, it is more suitable for the statistical analysis of the volume of large-scale blasted blocks. Attached Figure Description

[0037] Figure 1 This is a flowchart of the present invention;

[0038] Figure 2 These are some ore samples collected on-site;

[0039] Figure 3 This is a device for acquiring images of mineral rocks;

[0040] Figure 4 Extraction of two-dimensional morphological feature parameters from ore samples;

[0041] Figure 5 UNet model network architecture;

[0042] Figure 6 for and μ k Perform nonlinear fitting plot;

[0043] Figure 7 A comparison diagram of the volume of a sphere with equal area and its actual volume;

[0044] Figure 8 The following are line graphs of the fitting coefficients m and n, where (a) is the fitting coefficient m and (b) is the fitting coefficient n.

[0045] Figure 9 The fitting coefficients m and n are the sum of the morphological feature values. Correlation analysis of ĸ;

[0046] Figure 10 A nonlinear fitting plot of the three-dimensional morphological feature value k and the fitting coefficient m;

[0047] Figure 11 These are some ore samples collected on-site in Example 1;

[0048] Figure 12 Extraction of two-dimensional morphological feature parameters from ore samples in Example 1;

[0049] Figure 13 The cumulative gradation curve for the longest particle size d in Example 1;

[0050] Figure 14 The cumulative gradation curve for projected area A in Example 1;

[0051] Figure 15 Estimated volume V of the blasted block in Example 1 E The cumulative gradation curve;

[0052] Figure 16 Estimated volume V of the blasted block in the indoor test of Example 1 E Verification was performed by comparing the measured ore sample volume V using the displacement method;

[0053] Figure 17 Images of blasted blocks at the blasting site in Example 2;

[0054] Figure 18 Estimating the volume V of the blasted block at the blasting site in Example 2. E The cumulative gradation curve. Detailed Implementation

[0055] The present invention will be further described in detail below with reference to specific embodiments, but the scope of protection of the present invention is not limited to the content described. Invention Overview

[0057] A method for estimating the volume of blasted blocks in the field based on image processing (see...) Figure 1 The specific steps are as follows:

[0058] S1. After blasting in the mine, ore samples were taken, obtaining 500 ore samples. These samples were then sieved according to particle size ranges of small, medium, and large to obtain ore sample sets for different particle size ranges (see...). Figure 2 The range of small particle size D is D < 50 mm, the range of medium particle size is 50 ≤ D ≤ 100 mm, and the range of large particle size is 100 < D ≤ 150 mm.

[0059] S2. The volume V of each ore sample in the ore sample set with different particle size ranges is measured by the displacement method to form an ore sample volume dataset;

[0060] S3. Using a cylinder with a diameter of 50 mm as a reference, images of each ore sample from different particle size ranges were collected (see...). Figure 3 ), and used labelme software to annotate the images of each ore sample (see Figure 4 (a) forms a dataset of ore sample images;

[0061] S4. Divide the labeled dataset into training and testing sets in an 8:2 ratio, and use the UNet deep learning model (see...). Figure 5 The image segmentation is trained using labeled images from a dataset of ore sample images with different particle size ranges (training set). (See...) Figure 4 In (b), the OpenCV findContours function is used to detect contours in the image, extract the projected area A and the longest particle size d of each ore sample block, and calculate the two-dimensional morphological feature value. and the three-dimensional morphological feature value k; the two-dimensional morphological feature value The calculation formula is:

[0062] (1)

[0063] The formula for calculating the three-dimensional morphological feature value k is:

[0064] (2)

[0065] In the formula, A is the projected area of ​​the ore sample block, in mm. 2 ; d is the longest particle size of the ore sample block, mm; V is the measured volume of the ore sample block, mm. 3 ;

[0066] S5. For different two-dimensional morphological feature values Analyze two-dimensional morphological feature values With μ k The distribution pattern, such as Figure 6 As shown; for two-dimensional morphological feature values and μ k Nonlinear fitting is performed to obtain the fitting function relationship, thereby quantifying the two-dimensional morphological feature values. The correlation with the three-dimensional morphological feature value k; the μ k The relationship between μ and the three-dimensional morphological feature value k is: k = lgk;

[0067] The fitting function relationship is as follows:

[0068] , (3)

[0069] In the formula, μk = lgk, where k is the three-dimensional morphological feature value. These are two-dimensional morphological feature values;

[0070] S6. Based on the assumption of an equal-area sphere and the projected area of ​​the ore sample block, estimate the volume of the ore sample block to obtain the volume V of the equal-area sphere. b The formula for estimating the volume of the ore sample block is:

[0071] (4);

[0072] In the formula, V b Let A be the volume of a sphere with equal surface area, and let A be the projected area of ​​the ore sample block in mm. 2 ;

[0073] S7. Volume V of a sphere with equal area b Linear fitting was performed with the measured volume V of each ore sample in S2. The fitting coefficients m and n are obtained respectively; the volume V of the equal-area sphere is obtained. b A comparison with the measured volume V of each ore sample in S2 is shown below. Figure 7 The line graphs of the fitting coefficients m and n are shown below. Figure 8 ;

[0074] S8. Use Pearson correlation analysis to evaluate the fit coefficients m and n and the morphological eigenvalues. The correlation between k and (see) Figure 9 );

[0075] S9. Perform nonlinear fitting on the three-dimensional morphological feature value k and the fitting coefficient m (see...) Figure 10 The functional relationship is obtained, and the two-dimensional morphological feature values ​​from step S5 are used to... and μ k Logarithmic transformation is performed on the fitted function relationship to construct a volume estimation model for blasted blocks;

[0076] The functional relationship is as follows:

[0077] (5);

[0078] In the formula: m is the fitting coefficient, and k is the three-dimensional morphological feature value;

[0079] The volume estimation formula for the blasted block volume estimation model is as follows:

[0080] (6);

[0081] In the formula: V E Estimate the volume of the blasted block in mm 3A represents the projected area of ​​the ore sample block, in mm². 2 ; Two-dimensional morphological parameters;

[0082] S10. Based on the UNet image segmentation algorithm and the blasted block volume estimation model, a blasted block volume estimation system was built on the PyCharm platform to estimate the volume of blasted blocks on site, obtaining the estimated volume V of the blasted blocks. E And plot the longest particle size d, projected area A, and estimated volume V of the blasted block. E The cumulative gradation curve.

[0083] Example 1: A method for estimating the volume of blasted blocks in a field based on image processing, the specific steps of which are as follows:

[0084] S1. Ore samples were taken from a mine in Yunnan Province, and 500 ore samples were obtained. These samples were then sieved according to particle size ranges of small, medium, and large to obtain ore sample sets with different particle size ranges (see...). Figure 11 The range of small particle size D is D < 50 mm, the range of medium particle size is 50 ≤ D ≤ 100 mm, and the range of large particle size is 100 < D ≤ 150 mm.

[0085] S2. The volume V of each ore sample in the ore sample set with different particle size ranges is measured by the displacement method to form an ore sample volume dataset;

[0086] S3. Using a 50 mm diameter cylinder as a reference, images of each ore sample from different particle size ranges were collected, and the image of each ore sample was annotated using LabelMe software (see...). Figure 12 (a) forms a dataset of ore sample images;

[0087] S4. Divide the labeled dataset into training and testing sets in an 8:2 ratio, and use the UNet deep learning model (see...). Figure 5 The image segmentation is trained using labeled images from a dataset of ore sample images with different particle size ranges (training set). (See...) Figure 12 In (b), the OpenCV findContours function is used to detect contours in the image, extract the projected area A and the longest particle size d of each ore sample block, and calculate the two-dimensional morphological feature value. and three-dimensional morphological feature value k;

[0088] S5. Based on the UNet image segmentation algorithm and the blasted block volume estimation model, and using a blasted block volume estimation system built on the PyCharm platform, the volume of the blasted block at the scene is estimated, and the estimated volume V of the blasted block is obtained. EAnd plot the longest particle size d, projected area A, and estimated volume V of the blasted block. E The cumulative gradation curve is shown below. Figures 13-15 ;

[0089] The volume estimation formula for the blasted block volume estimation model is as follows:

[0090] (6);

[0091] In the formula: V E Estimate the volume of the blasted block in mm 3 A represents the projected area of ​​the ore sample block, in mm². 2 ; Two-dimensional morphological parameters;

[0092] S6. Estimated volume V of the blasted block in the indoor test E This is compared and verified with step S2, which uses the water displacement method to measure the actual volume V of the ore sample (see [link]). Figure 16 Estimated volume V of the blasted block E The absolute average error between the measured volume V and the actual volume is 8.98%, and the gradation curves of the two are very close, which proves the accuracy of the volume estimation of the blasted block and meets the measurement requirements of the block volume at the blasting site.

[0093] Example 2: Application of blasting block volume measurement at a blasting site in Yunnan: An image processing-based method for estimating the volume of blasting blocks at the site is adopted. The specific steps are as follows:

[0094] Images of the blasted debris at the blast site are shown below. Figure 17 Based on the UNet image segmentation algorithm and the blasted block volume estimation model, and using a blasted block volume estimation system built on the PyCharm platform, the volume of the blasted block at the scene was estimated, yielding the estimated volume V of the blasted block. E And draw the estimated volume V of the blasted blocks at the blasting site. E The cumulative gradation curve (see) Figure 18 );

[0095] The blasting block volume estimation system can estimate the volume of blocks at the blasting site and quickly plot the volume distribution curve of the blasting blocks, improving the efficiency of block size identification at the blasting site.

[0096] The specific embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A method for estimating the volume of a block of rock based on image processing of a blast site, characterized in that, The specific steps are as follows: S1. After the mine blasting, the ore sampling is carried out to obtain a plurality of ore samples, and the ore samples are screened according to the small particle size, the medium particle size and the large particle size to obtain ore sample sets of different particle size ranges; S2. The volume V of each ore sample in the ore sample set of different particle size ranges is measured by the drainage method to form an ore sample volume data set; S3. The image of each ore sample in the ore sample set of different particle size ranges is collected, and the image of each ore sample is labeled to form an ore sample image data set; S4. The UNet model using deep learning is used to train the labeled images in the ore sample image dataset of different particle size ranges, complete image segmentation, extract the projection area A and the longest particle size d of each ore sample block, and calculate the two-dimensional morphological characteristic value and the three-dimensional morphological characteristic value k; The two-dimensional morphological characteristic value The calculation formula is: ； The calculation formula of the three-dimensional morphological characteristic value k is: ； In the formula, A is the projected area of the ore sample block, mm 2 ; d is the longest particle diameter of the ore sample block, mm; and V is the measured volume of the ore sample block, mm 3 ; S5. performing nonlinear fitting on the two-dimensional morphological characteristic value and μ k to obtain a fitting function relationship, so as to quantify the correlation between the two-dimensional morphological characteristic value and the three-dimensional morphological characteristic value k; the relationship between the three-dimensional morphological characteristic value k and μ k is: μ k = lgk; The fitting function relationship is: ； wherein k is a three-dimensional shape feature value, is a two-dimensional shape feature value; S6. Estimate the volume of the ore sample block based on the theory of equal-area sphere and the projected area of the ore sample block, and obtain the volume of the equal-area sphere V b ; S7. Volume of an equi-area sphere V b Linear fitting is performed on the measured volume V of each ore sample in S2. , respectively, to obtain the fitting coefficients m and n. S8. Evaluate the correlation of the fitting coefficients m, n and the morphological feature values k with Pearson correlation analysis. , k; S9. Nonlinear fitting is performed on the three-dimensional morphological characteristic value k and the fitting coefficient m to obtain a functional relationship formula, and the two-dimensional morphological characteristic value k of step S5. is logarithmically converted into a fitting function relationship formula of μ and m k , to construct a blasting block volume estimation model; the functional relationship formula is: ； In the formula, m is a fitting coefficient, and k is a three-dimensional morphological characteristic value. S10. Based on the UNet image segmentation algorithm and the blasting block volume estimation model, a blasting block volume estimation system is built on the Pycharm platform to estimate the volume of the blasting block on site, and the estimated volume of the blasting block V is obtained E ; The longest particle diameter d, the projected area A and the blasting block volume estimation V are plotted E The cumulative grading curve of the blasting block volume estimation model is plotted; the volume estimation formula of the blasting block volume estimation model is: ； where: V E Volume of the blasted block, mm 3 ; A is the projected area of the ore sample block, mm 2 ; is the two-dimensional morphological parameter.

2. The method for estimating the volume of a block of rock based on image processing of a blast site according to claim 1, characterized in that: Step S1. The range of the small particle size D is D<50mm, the range of the medium particle size is 50≤D≤100mm, and the range of the large particle size is 100<D≤150mm.

3. The method of claim 1, wherein: Step S6. The estimation relationship of the volume of the ore sample block is: ； wherein V b is the volume of an equal-area sphere, A is the projected area of the ore sample block, mm 2 .

Images