Multi-algorithm fused muck pile boulder rate detection method

Abstract

The invention provides a multi-algorithm fused muck pile boulder rate detection method. The method comprises the following steps: S1, acquiring a muck pile image to be detected; s2, filtering and noise reduction are carried out on the to-be-detected muck pile image; s3, performing segmentation processing on the image obtained in the S2 by using an adaptive threshold algorithm and an overall nested edge detection algorithm, and fusing output results of the segmentation processing by using a fusion algorithm; and S4, identifying the number of pixel points of each target rock area of the segmented image, if the number of pixel points is greater than a set pixel threshold value, determining that the rock is a large rock, and calculating the boulder rate according to the number of pixel points of the large rock area. The method has the advantages that preprocessing and smooth image processing are carried out on the detected muck pile image, gray iterative calculation is carried out on the muck pile image to determine a threshold value and reduce man-made interference, statistical filtering and denoising are carried out according to the mean value and the variance, then a statistical filtering method, a natural iteration method and an edge detection method are fused, image noise can be accurately removed, and the detection accuracy is improved. The accuracy of a detection result is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of coal mining, and in particular relates to a method for detecting large block rate of coal pile explosion by integrating multiple algorithms. Background Art

[0002] Image recognition technology is used to detect the large block rate of mine blasting piles to guide the parameters of subsequent blasting operations. Large pieces of ore will cause the crusher equipment to jam and affect the crusher's working efficiency. The ore image recognition and segmentation effect is restricted by factors such as lighting conditions, target density and low edge contrast, resulting in low large block rate detection accuracy. Summary of the invention

[0003] In view of this, the present invention aims to propose a multi-algorithm fusion method for detecting large block rate of burst piles, in order to solve at least one of the above-mentioned technical problems.

[0004] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0005] A method for detecting large block rate of burst piles by integrating multiple algorithms, comprising the following steps:

[0006] S1, collecting images of the explosion pile to be detected;

[0007] S2, filtering and denoising the image of the explosion pile to be detected;

[0008] S3, using the adaptive threshold algorithm and the overall nested edge detection algorithm to segment the image obtained in S2,

[0009] Use fusion algorithm to fuse the output results of segmentation processing;

[0010] S4, identifying the number of pixels in each target rock area of ​​the segmented image, if the number of pixels is greater than a set pixel threshold, it is determined to be a large rock, and the large rock rate is calculated according to the number of pixels in the large rock area.

[0011] Furthermore, in S1, a drone is used to collect images of the explosion pile to be detected by vertical shooting at a specified height.

[0012] Further, the S2 comprises the following steps:

[0013] S21, performing grayscale conversion on the image of the explosion pile to be detected;

[0014] S22, calculating the variance and mean of adjacent pixels;

[0015] S23, using the variance obtained in S22 as a smoothing denoising parameter to perform smoothing denoising on the image.

[0016] Furthermore, the adaptive threshold algorithm in S3 is as follows:

[0017] The first-order difference operator in the sliding window is used to calculate the gradient amplitude and gradient direction of each pixel;

[0018] Calculate the mean and variance of the pixels in the sliding window;

[0019] Calculate the difference between the grayscale value of the pixel and the mean value of the pixels in the sliding window, and take the absolute value of the difference as the first comparison value;

[0020] Calculate the product of the variance of the pixels in the sliding window and the statistical filtering threshold as the second comparison value;

[0021] If the first contrast value is not greater than the second contrast value, retain the pixel;

[0022] If the first contrast value is greater than the second contrast value, the mean value of the pixels in the sliding window replaces the original pixel point.

[0023] Furthermore, the overall nested edge detection algorithm in S3 is as follows:

[0024] The VGG network model is used, a lateral output layer is added to the convolutional layer of the VGG network model, and a deeper output layer is used to replace the pooling layer and the fully connected layer.

[0025] And add pixel-based classification balance weights.

[0026] Furthermore, the method of using the fusion algorithm to fuse the output results of the segmentation process is as follows:

[0027] Normalize the segmentation outputs of the adaptive threshold algorithm and the overall nested edge detection algorithm;

[0028] If both results are 1, the output fusion algorithm outputs 1, otherwise the output fusion algorithm outputs 0.

[0029] Furthermore, in said S4, the pixel threshold is calculated and set according to the crusher diameter and the proportional coefficient;

[0030] The ratio of the pixels in the large rock area to the total number of pixels in the image is the large rock rate.

[0031] A second aspect of the present invention provides an electronic device, comprising a processor and a memory communicatively connected to the processor and used to store instructions executable by the processor, wherein the processor is used to execute the method described in the first aspect above.

[0032] The third aspect of the present invention provides a server, comprising at least one processor, and a memory communicatively connected to the processor, the memory storing instructions executable by the at least one processor, the instructions being executed by the processor so that the at least one processor executes the method described in the first aspect.

[0033] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method described in the first aspect.

[0034] Compared with the prior art, the multi-algorithm fusion method for detecting the rate of large blocks in a pile explosion described in the present invention has the following beneficial effects:

[0035] The multi-algorithm fusion method for detecting large block rate of explosive piles described in the present invention preprocesses and smoothes the detected explosive pile images, enhances the images at the same time, highlights the key points of the images to facilitate edge detection, performs grayscale iterative calculation on the explosive pile images to determine the threshold to reduce human interference, performs statistical filtering and denoising based on the mean and variance, and then integrates the statistical filtering method and the natural iteration method edge detection method, which can accurately remove image noise, improve the accuracy of coal rock image segmentation, and improve the accuracy of detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] The accompanying drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the accompanying drawings:

[0037] Figure 1 This is a schematic diagram of the detection method according to an embodiment of the present invention;

[0038] Figure 2 Schematic diagram of image segmentation using different thresholds according to an embodiment of the present invention. DETAILED DESCRIPTION

[0039] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0040] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and the like are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, features defined as "first", "second", and the like may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "multiple" means two or more.

[0041] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood by specific circumstances.

[0042] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0043] like Figure 1 As shown, a multi-algorithm fusion method for detecting large block rate of burst piles includes the following steps:

[0044] S1, collecting images of the explosion pile to be detected;

[0045] In S1, a drone is used to collect images of the explosion pile to be detected by vertical shooting at a specified height.

[0046] On the basis of ensuring the effectiveness and accuracy of the test, three different blast piles were selected for testing, and DJI Air2S drone was used to conduct aerial surveys of the blast piles respectively. The drone's built-in camera lens was adjusted vertically and horizontally to shoot the blast pile mineral rocks.

[0047] The parameters of the DJI Ai r2S drone camera image sensor are 1 inch, effective pixels are 20 million, and the lens parameters are: FOV (field of view) 88°, equivalent focal length 22mm, aperture f / 2.8, focus range 0.6m to infinity. Maximum photo resolution: 5472×3078, 16:9 or 5472×3648, 3:2. The camera's frame control points are set within the test area, and corresponding encryption processing points are set at the corners of the area to track and check the accuracy within the test area in a timely manner. To ensure the safety of the drone during data collection, combined with the actual operating conditions of the shooting area, the flight altitude will eventually be designated as 60-100m, and 80m is preferred in this application.

[0048] S2, filtering and denoising the image of the explosion pile to be detected;

[0049] S2 includes the following steps:

[0050] In view of the situation where false edges or edge loss occur in the edge detection algorithm of explosive rock images, a Gaussian filter algorithm is used to smooth the graphics and retain the real edges. The traditional algorithm is improved through smoothing denoising and threshold determination methods, and the mean and variance are used for filtering and denoising. The iterative method is used to determine the threshold of statistical filtering between 0.7-1.5 instead of manually determining the threshold to reduce human factors.

[0051] Gaussian function is used as the basis function for image edge processing. Common RGB digital images are used for processing. Each pixel information in the image is composed of three color shades of red, green and blue channels, and is represented by three discrete functions f(k, i, j), (i, j) represents the position in the Cartesian system), and the formula is as follows:

[0052]

[0053] In the image RGB model,

[0054] Let R(curco l or.R)=G(curco l or.G)=R(curco l or.B);

[0055] S21, grayscale conversion is performed on the image of the explosion pile to be detected; the grayscale conversion formula is as follows:

[0056] Gray(i,j)=0.299R+0.589G+0.114B;

[0057] Gray(i, j) is the gray value of the point (i, j) in the grayscale image after the image is converted.

[0058] S22, calculate the variance and mean of adjacent pixel points; based on the premise that the pixel point x(i, j) is known, solve the variance and mean of the surrounding edge pixel points, which are expressed as μ (mean) and σ2 (variance) respectively. The calculation formula is as follows:

[0059]

[0060] Among them, i, j are the gray levels of pixels, and N is the number of pixels in the neighborhood.

[0061] S23, using the variance obtained in S22 as a smoothing denoising parameter to perform smoothing denoising on the image. Eliminate the interference of external factors on smoothing denoising. The image of the explosive pile rock block will inevitably be disturbed by the surrounding environment during the acquisition process. Therefore, a filtering denoising algorithm should be selected for edge detection, which can achieve denoising and make the image edge clearly visible.

[0062] S3, using the adaptive threshold algorithm and the overall nested edge detection algorithm to segment the image obtained in S2,

[0063] Use fusion algorithm to fuse the output results of segmentation processing;

[0064] The adaptive threshold algorithm in S3 is as follows:

[0065] Taking the sliding window centered at (x, y) as an example, under the premise of changing the sliding window (3×3, 5×5, 7×7), by comparing the image segmentation effects of different windows, the influence of the moving window on the final effect of the algorithm segmentation is analyzed. The window length with the best segmentation effect is set to 3×3, which is also the window for the subsequent coal rock image binarization processing.

[0066] The first-order difference operator in the sliding window is used to calculate the gradient amplitude and gradient direction of each pixel; the calculation formula is as follows:

[0067] G1(x,y)=f(x+1,y+1)-f(x-1,y-1)+f(x,y+1)

[0068] -f(x-1,y)+f(x+1,y)-f(x,y-1);

[0069] G2(x,y)=f(x-1,y+1)-f(x+1,y-1)+f(x-1,y)

[0070] -f(x,y-1)+f(x,y+1)-f(x+1,y);

[0071]

[0072] Among them, G1(x, y) is the partial derivative in the 45° direction, G2(x, y) is the partial derivative in the 135° direction, M is the gradient amplitude, and θ is the gradient direction.

[0073] Calculate the mean and variance of the pixels in the sliding window;

[0074]

[0075] All pixels in the image are filtered and denoised, and the 3×3 neighborhood matrix of pixel x(i, j) is expressed as:

[0076]

[0077] Calculate the difference between the gray value of the pixel and the mean value of the pixels in the sliding window, and take the absolute value of the difference as the first contrast value A. The formula is as follows:

[0078] A=|x(i,j)-μ|;

[0079] The product of the variance of the pixels in the sliding window and the statistical filtering threshold is calculated as the second comparison value. The formula is as follows:

[0080] B = σM;

[0081] If the first contrast value is not greater than the second contrast value, retain the pixel;

[0082] If the first contrast value is greater than the second contrast value, the mean value of the pixels in the sliding window replaces the original pixel point.

[0083] The threshold M of statistical filtering is determined based on the grayscale iteration method. The average maximum and minimum grayscale levels in the grayscale range [0, i-1] of the image are taken and set as the initial threshold T. The image is divided into the maximum grayscale level and the minimum grayscale level to calculate the grayscale mean of the two parts as the new threshold Tn value. The iteration continues until |Tn-T| < 0.5. The threshold Tn is the iteration threshold. The average grayscale μ1, μ2 can be expressed as:

[0084]

[0085] Among them, i is the gray level of the pixel, and ni is the number of pixels with gray level i.

[0086] The overall nested edge detection algorithm in S3 is as follows:

[0087] The VGG network model is used, a lateral output layer is added to the convolutional layer of the VGG network model, and a deeper output layer is used to replace the pooling layer and the fully connected layer;

[0088] And add pixel-based classification balance weights. The classification balance weight method plays a more significant role in solving the problem of positive and negative sample loss balance. The loss function used is as follows:

[0089]

[0090] Among them, W is the standard layer network parameter set, w(m) is the side output layer parameter, m represents a certain layer, β = |Y_| / Y+|, 1-β = |Y+| / Y_|, |Y_| and |Y+| are the edge and non-edge pixel data sets of the data set.

[0091] In some embodiments, Pr(y j =1X; W, w(m)) is calculated by the Sigmoid kernel function and the Laplace kernel function.

[0092] The formula for the Sigmoid kernel function is:

[0093] K(x,x i ) = tanh(k(x×x i )+θ);

[0094] It is used to enhance the nonlinear expression ability of the model;

[0095] The formula of the Laplace kernel function is:

[0096]

[0097] It is used to capture local distance information;

[0098] Combining the two kernel functions, we get the corresponding composite kernel function:

[0099] K(x,x i ) = tanh(αK laplace (x,x i )+θ);

[0100] Where K laplace (x,x i ) is the Laplace kernel function, so that the model can use the local distance information of the Laplace kernel function and the nonlinear change ability of the Sigmoid kernel function at the same time.

[0101] In other embodiments, Pr(y j =1X; W, w(m)) is calculated by the following formula:

[0102] The formula for the Sigmoid kernel function is:

[0103] K(x,x i ) = tanh(k(x×xi )+θ);

[0104] In other embodiments, Pr(y j =1X; W, w(m)) is calculated by the following formula:

[0105] The formula of the Laplace kernel function is:

[0106]

[0107] The adaptive threshold algorithm has a better segmentation effect for local thresholds, but for images with poor edge contrast, this traditional adaptive threshold algorithm is difficult to achieve ideal segmentation results. In comparison, although the overall nested edge detection algorithm can achieve better segmentation effects for such low-contrast images, it is difficult to effectively identify small and medium-sized targets in the image.

[0108] The method of using the fusion algorithm to fuse the output results of the segmentation process is as follows:

[0109] Normalize the segmentation outputs of the adaptive threshold algorithm and the overall nested edge detection algorithm;

[0110] If both results are 1, the output fusion algorithm outputs 1, otherwise the output fusion algorithm outputs 0.

[0111] The formula is as follows:

[0112]

[0113] f merge is the pixel value at position (i, j) of the image segmentation and fusion algorithm, f adapt (i, j) is the output result of the adaptive threshold algorithm, f HED (i, j)——The output result of HED algorithm.

[0114] S4, identifying the number of pixels in each target rock area of ​​the segmented image, if the number of pixels is greater than a set pixel threshold, it is determined to be a large rock, and the large rock rate is calculated according to the number of pixels in the large rock area.

[0115] The image segmentation algorithm separates the target area from the background, thereby achieving accurate detection of pixels in the coal and rock block images.

[0116] In S4, the pixel threshold is calculated and set according to the crusher diameter and the proportional coefficient;

[0117] The ratio of the pixels in the large rock area to the total number of pixels in the image is the large rock rate.

[0118] Gray threshold noise filtering is less, although the smoothing effect of noise is better, but there are many false edges when the edges of the rock surface are connected ( Figure 2 A threshold that is too low will result in poor visual effects, pseudo-edges, and inability to correctly distinguish edges; a threshold that is too high may result in a lack of image edge information, which will further prevent researchers from accurately determining the edge features of the image. A threshold between 0.7 and 1.2 is used to replace the manually determined threshold to reduce human factors. The image with the threshold determined by the iterative method ( Figure 2 c) The original image is restored well, the edge area is protected, the local noise of the rock image is smoothed, the edge is clear after image processing, the edge judgment is more accurate, the target area of ​​the rock image is protected, the edge details and texture features of the open-pit mine image are retained, and a good noise reduction effect is achieved. The continuity characteristics of the line segment are also reflected, which is similar to people's daily visual habits and is similar to the original image. The edge details and texture features of the open-pit mine image are retained well, and a good noise reduction effect is achieved.

[0119] use Figure 2 The number of pixels of coal rock obtained by converting the target contour of coal rock is used to detect target rocks with more pixels than large blocks. Among them, there are rocks with more than 245,390 pixels (calibrated as large blocks), with 314,517 pixels. The results show that the number of pixels in the measurement area reaches 2,315,410 pixels. Further, the coal rock large block rate can be obtained to be 13.58%.

[0120] The detected explosion pile image is preprocessed and smoothed, and the image is enhanced at the same time to highlight the key points of the image for easy edge detection. The grayscale of the explosion pile image is iteratively calculated to determine the threshold to reduce human interference. Statistical filtering and denoising are performed according to the mean and variance. Then, the statistical filtering method and the natural iteration method edge detection method are integrated to accurately remove image noise, improve the accuracy of coal rock image segmentation, and improve the accuracy of detection results.

[0121] Embodiment 2:

[0122] An electronic device comprises a processor and a memory which is communicatively connected to the processor and is used to store instructions executable by the processor, wherein the processor is used to execute the method described in the first embodiment above.

[0123] Embodiment three:

[0124] A server includes at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the method described in Example 1.

[0125] Embodiment 4:

[0126] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in embodiment 1 is implemented.

[0127] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

[0128] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A multi-algorithm fusion method for detecting large block rate of pile explosion, characterized in that: The following steps are involved: S1, collecting images of the explosion pile to be detected; S2, filtering and denoising the image of the explosion pile to be detected; S3, using the adaptive threshold algorithm and the overall nested edge detection algorithm to segment the image obtained in S2, Use fusion algorithm to fuse the output results of segmentation processing; S4, identifying the number of pixels in each target rock area of ​​the segmented image, if the number of pixels is greater than a set pixel threshold, it is determined to be a large rock, and the large rock rate is calculated according to the number of pixels in the large rock area.

2. According to the multi-algorithm fusion method for detecting the rate of large blocks of pile explosions according to claim 1, it is characterized in that: In S1, a drone is used to collect images of the explosion pile to be detected by vertical shooting at a specified height.

3. According to the multi-algorithm fusion method for detecting large block rate of pile explosion according to claim 1, it is characterized in that: The S2 comprises the following steps: S21, performing grayscale conversion on the image of the explosion pile to be detected; S22, calculating the variance and mean of adjacent pixels; S23, using the variance obtained in S22 as a smoothing denoising parameter to perform smoothing denoising on the image.

4. The method for detecting large block rate of pile explosion by integrating multiple algorithms according to claim 1 is characterized in that: The adaptive threshold algorithm in S3 is as follows: The first-order difference operator in the sliding window is used to calculate the gradient amplitude and gradient direction of each pixel; Calculate the mean and variance of the pixels in the sliding window; Calculate the difference between the grayscale value of the pixel and the mean value of the pixels in the sliding window, and take the absolute value of the difference as the first comparison value; Calculate the product of the variance of the pixels in the sliding window and the statistical filtering threshold as the second comparison value; If the first contrast value is not greater than the second contrast value, retain the pixel; If the first contrast value is greater than the second contrast value, the mean value of the pixels in the sliding window replaces the original pixel point.

5. According to the multi-algorithm fusion method for detecting large block rate of pile explosion according to claim 4, it is characterized in that: The overall nested edge detection algorithm in S3 is as follows: The VGG network model is used, a lateral output layer is added to the convolutional layer of the VGG network model, and a deeper output layer is used to replace the pooling layer and the fully connected layer. And add pixel-based classification balancing weights.

6. The method for detecting the rate of large blocks of pile explosion by integrating multiple algorithms according to claim 5 is characterized in that: The method of using the fusion algorithm to fuse the output results of the segmentation process is as follows: Normalize the segmentation outputs of the adaptive threshold algorithm and the overall nested edge detection algorithm; If both results are 1, the output fusion algorithm outputs 1, otherwise the output fusion algorithm outputs 0.

7. The method for detecting large block rate of pile explosion by integrating multiple algorithms according to claim 1 is characterized by: In said S4, the pixel threshold is calculated and set according to the crusher diameter and the proportional coefficient; The ratio of the pixels in the large rock area to the total number of pixels in the image is the large rock rate.

8. An electronic device, comprising a processor and a memory connected to the processor for storing instructions executable by the processor, characterized in that: The processor is used to execute the method described in any one of claims 1 to 7.

9. A server, characterized in that: It includes at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the processor so that the at least one processor executes the method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images