CN115100224B - Extraction method and system for coal mine roadway tunneling head-on cross fracture - Google Patents

Abstract

The application discloses a method and a system for extracting a head-on cross fracture in coal mine tunnel tunneling, wherein the method comprises the following steps: acquiring a coal roadway tunneling head-on image; preprocessing a coal roadway tunneling head-on image to obtain a first image; extracting the edge of the rock mass structural plane of the first image to obtain a second image; filtering the binary image pseudo edges of the second image to obtain a third image; extracting a linear edge of the third image; obtaining a fourth image; carrying out local connection processing on the fourth image according to the angle and distance criteria, and extracting a cross joint trace of a coal roadway tunneling head-on image; and overlapping the extracted cross joint trace on a coal roadway tunneling head-on image to obtain a mine roadway tunneling head-on cross fracture. The application provides an image segmentation method suitable for coal mine tunneling head-on cross joint trace detection by combining a global edge connection algorithm and a local edge connection algorithm on the basis of analyzing the characteristic of a coal mine tunneling head-on typical cross joint image.

Description

Extraction method and system for coal mine roadway tunneling head-on cross fracture

Technical Field

The application relates to the field of quantitative investigation, in particular to a method and a system for extracting a head-on cross fracture in coal mine tunnel tunneling.

Background

For the acquisition of the geometric dimension and distribution condition of the primary fracture, the traditional method adopts a field manual measurement mode, and is time-consuming, labor-consuming, low in accuracy and greatly influenced by human factors. And certain safety risks are also caused under complex mining conditions such as high ground stress, strong mining, fault areas and the like. In order to overcome the defects, in recent years, with the rapid development of computer vision technology, an image segmentation technology based on a digital image processing technology has been widely applied to crack trace detection and identification of a rock mass structural surface. Has the advantages of high efficiency, high speed, accuracy, safety and the like.

However, in the published literature, few people have carried out algorithm research on recognition conditions with a large number of structural faces and complex distribution, such as a coal mine roadway driving head-on cross joint fracture image. Therefore, in order to quantitatively evaluate the overhead self-stability performance of the tunneling head-on of the coal mine, the image segmentation algorithm suitable for the detection of the tunneling head-on cross joint trace of the coal mine is provided by combining the advantages of the global edge connection algorithm and the local edge connection algorithm on the basis of analyzing the characteristic of the typical cross joint image of the tunneling head-on of the coal mine, and experimental study is carried out by selecting the tunneling head-on cross joint image of the coal mine.

Disclosure of Invention

The application provides a method and a system for extracting a head-on cross fracture of coal mine tunnel tunneling, which combines the advantages of a global edge connection algorithm and a local edge connection algorithm on the basis of analyzing the characteristic of a typical cross joint image of the head-on cross joint of the coal mine tunnel tunneling, and provides an image segmentation algorithm suitable for detecting the trace of the head-on cross joint of the coal mine tunnel tunneling.

In order to achieve the purpose, the application provides a method for extracting a head-on cross fracture in coal mine tunnel tunneling, which comprises the following steps:

Acquiring a coal roadway tunneling head-on image;

preprocessing the coal roadway tunneling head-on image to obtain a first image;

extracting the edge of the rock mass structural plane of the first image to obtain a second image;

Filtering the binary image pseudo edges of the second image to obtain a third image;

extracting a linear edge of the third image; obtaining a fourth image;

Carrying out local connection processing on the fourth image according to an angle and distance criterion, and extracting a cross joint trace of the coal roadway tunneling head-on image; and overlapping the extracted cross joint trace on the coal roadway tunneling head-on image to obtain an ore roadway tunneling head-on cross fracture.

Optionally, the preprocessing method is to adaptively correct the illumination non-uniformity image by using a two-dimensional gamma function.

Optionally, the method for extracting the edge of the rock mass structural plane of the first image is to derive the gray value of the first image.

Optionally, the step of filtering the binary image pseudo edges of the second image includes:

An image description;

Removing branch points and corner points;

And filtering the noise of the non-structural surface.

Optionally, the method of extracting the linear edges of the third image includes using a hough transform.

Optionally, the method for performing local connection processing on the fourth image includes: and (3) designing an algorithm according to the distance criterion and the angle criterion, and connecting the segment lines meeting the allowable values of the distance criterion and the angle criterion.

The application also provides a coal mine roadway tunneling head-on cross fracture extraction system which comprises an image collection module, a preprocessing module, an edge detection module, a false edge filtering module, an edge fitting module and a local edge connecting module;

the image collection module is used for acquiring a coal roadway tunneling head-on image;

The preprocessing module is used for preprocessing the collected coal roadway tunneling head-on image to obtain a first image;

the edge detection module is used for extracting the edge of the rock mass structural plane of the first image to obtain a second image;

the false edge filtering module is used for filtering the false edges of the binary image of the second image to obtain a third image;

the edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image;

The local edge connection module is used for extracting a cross joint trace of the coal roadway tunneling head-on image; the local edge connection module is further used for superposing the extracted cross joint trace on the coal roadway tunneling head-on image to obtain an ore roadway tunneling head-on cross crack.

Optionally, the false edge filtering module comprises an image description unit, a clearing unit and a noise filtering unit; the image description unit is used for image description; the clearing unit is used for clearing branch points and corner points; the noise filtering unit is used for filtering the noise of the non-structural surface.

Drawings

In order to more clearly illustrate the technical solutions of the present application, the drawings that are needed in the embodiments are briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic flow diagram of an embodiment;

FIG. 2 is a schematic diagram of two-dimensional gamma correction results in a bit embodiment;

FIG. 3 is a schematic diagram of edge detection in an embodiment;

FIG. 4 is a schematic diagram of corner monitoring and removal in an embodiment;

FIG. 5 is a schematic diagram of the aspect ratio statistics of connected domains in the embodiment;

FIG. 6 is a schematic diagram of morphological feature processing results in the embodiment, wherein (a) is a schematic diagram before processing and (b) is a schematic diagram after processing;

Fig. 7 is a schematic diagram of a hough transform principle and a processing result, and a hough transform processing result in an embodiment, where (c) is a schematic diagram of a hough matrix, and (d) is a schematic diagram of a detected straight line

FIG. 8 is a schematic diagram of a segment line connection in an embodiment;

FIG. 9 is a schematic structural diagram of a coal mine roadway entry head-on cross fracture extraction system.

Detailed Description

Technical aspects of embodiments of the present application will be clearly and fully described in the following description of the embodiments of the present application with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present application without making any inventive effort, are intended to fall within the scope of the present application.

In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description.

Example 1

The general flow diagram of the extraction method of the coal mine tunnel tunneling head-on cross fracture in the embodiment of the application is shown in the figure I, and comprises the following steps:

S1, acquiring a coal roadway tunneling head-on image.

S2, adaptive correction of uneven illumination:

In the process of collecting a head-on image of a coal mine tunnel tunneling, noise is often generated due to the influence of factors such as non-uniform illumination, floating coal dust and the like, so that a satisfactory visual effect cannot be achieved. In order to reduce the influence of uneven illumination on an image and improve the quality of the image, it is necessary to enhance the original image.

The correction method of the illumination non-uniform image is mainly divided into two main types of correction methods with reference and correction methods without reference. In practical applications, reference-free illumination non-uniformity correction algorithms are of great interest. The existing non-reference illumination non-uniformity correction method mainly comprises an algorithm based on the Retinex theory, a histogram equalization method, a non-sharpening mask method, a morphological filtering method, a method based on a space illumination map and the like. The image is preprocessed by using an adaptive correction algorithm for the illumination non-uniformity image based on a two-dimensional gamma function.

For an image with uneven illumination, due to uneven distribution of illumination components in a scene, the brightness value of the image in a region with strong illumination in the image is enough or too strong, and the brightness value of the image in a region with weak illumination is insufficient.

Therefore, firstly, a multi-scale Gaussian function is adopted to extract illumination components in a scene, and the expression of the function is as follows:

Where G is the illumination component, c is the scale factor, and λ is the normalization constant. x is the abscissa of the input image pixel and y is the ordinate of the input image pixel. And convolving the original image with the Gaussian function to obtain an estimated value of the illumination component, wherein the estimated value is obtained as follows:

I(x,y)＝F(x,y)G(x,y)

wherein F (x, y) is an input image; i (x, y) is the estimated illumination component. And then, respectively extracting illumination components of the scene by utilizing a multi-scale Gaussian function, and weighting to finally obtain an estimated value of the illumination components, wherein the expression is as follows:

Wherein, I (x, y) is the illumination component value extracted and weighted by a plurality of Gaussian functions with different scales at the point (x, y); omega _i is the weighting coefficient of the illumination component extracted by the ith scale Gaussian function; i=1, 2, ··, N is the number of the used ruler degrees, here we take n=3, the scale factor c values selected herein are 15, 80 and 250, respectively. The illumination component of the image is extracted using a 3-scale gaussian function, the result of which is shown in fig. 2 (b).

After the illumination component is extracted, an illumination non-uniformity correction function is constructed according to the distribution characteristic of the illumination component, the illumination non-uniformity image is corrected, the brightness value of the area with over-illumination is reduced, and the brightness value of the area with over-illumination is improved. The parameters of the two-dimensional gamma function are adaptively adjusted by utilizing the distribution characteristics of illumination components of the image, so that the aim of improving the overall quality of the illumination non-uniform image is fulfilled. The two-dimensional gamma function expression is as follows:

Wherein O (x, y) is the brightness value of the corrected output image, and gamma is the index value for brightness enhancement; m is the luminance average of the illumination component.

Fig. 2 (c) and (d) are images enhanced by the illumination non-uniformity adaptive correction algorithm, and it can be seen that there is a significant enhancement for darker targets.

S3, edge detection:

the edge refers to the portion of the image where the local intensity variation is most significant, as shown in fig. 3 (a). Edges are the result of grey value discontinuities that can typically be detected using derivative methods. Common edge detection operators are Canny, prewitt and Sobel operators. The Canny detection operator is applied to coal roadway tunneling head-on image analysis, and the processing result is shown in fig. 3 (b).

S4 mathematical morphology processing:

For the edge detection of the coal roadway tunneling head-on image, the result not only comprises complex and numerous structural surfaces, but also comprises a plurality of non-structural surface noises. This is related to factors such as edge detection algorithm and parameter selection, special textures in the original image, darker rock mass, etc., which are difficult to avoid in practical situations. Therefore, for large images containing complex structural surfaces, non-structural surface filtering is required, comprising the steps of:

1. Image description

For the binary image detected by the edge detection operator, each edge curve is a pixel connected region. Each connected region can be marked in Matlab, so that information of all edge curves, such as main shaft length, main shaft width and the like, can be obtained. It is observed that the structural surface is generally of a "long linear" configuration. Therefore, the aspect ratio is taken as an important index for distinguishing noise such as a long linear structure and square or round noise, which is an important basis for filtering background noise and extracting a fault structural surface.

r＝D/R

Wherein D is the length of the main axis of the communicating domain, R is the width of the main axis of the communicating domain, and R is the aspect ratio.

2. Clearing branch points and corner points

After edge detection is performed on an image, as shown in fig. 4, there is often a branching point in the image edge detection result. This prevents the morphological filtering of non-structural surface noise such as bare rock mass in the binary image. To obtain more accurate trace parameters, it is necessary to clear the branch points in the binary image obtained by edge detection.

In addition, for broken rock face images, linear targets formed after edge detection are mutually connected to form a broken line, and then the broken line is required to be split into two or more chain codes from the corner point on the basis of corner point identification. The key to splitting is corner detection. The corner points are high curvature points on the contour curves, the curvature of the chain codes can be estimated according to the chain code differences, the corner points are further judged, and the detection result is shown in fig. 4.

3. Non-structural face noise filtering

According to the selected mathematical characteristic attribute, the length-width ratio and the area of all connected domains can be obtained as shown in fig. 5. According to statistical laws, the ratio of the length of the principal axis to the width of the principal axis of a "long-line" structure is generally greater than 3. For this reason, the connected domain is traversed, and if the ratio of the main axis length to the main axis width of the region is greater than 3 and the 8 connected domain area is greater than 10 pixels, the connected domain is determined to be a long line structure. Otherwise, the pixels in the connected domain are set as foreground colors and used as non-structural surface noise filtering. The processing results obtained according to the above steps are shown in fig. 6.

S5, edge fitting:

If the edges are sharp and the noise is extremely low, the edge image can be binarized and refined into a single pixel wide closed connected boundary map. However, edge discontinuities due to noise, uneven illumination, and other effects due to the introduction of false brightness discontinuities, the resulting group of pixels rarely describe an edge in its entirety. Therefore, it is typical to use a connection process to name the edge pixel combinations as intentional edges immediately after using the edge detection algorithm. Edge connections can be generally divided into two types: global edge connections and local edge connections. Typically, we are at a specific location where the object of interest is not known. In this case, the global characteristics of the image can be utilized to directly detect the target contour.

Hough transforms were proposed by Hough in 1962. As shown in fig. 7 (a), in a two-dimensional plane, a straight line passing through a point (x, y) can be expressed as:

y_i＝ax_i+b

where a is the slope and b is the intercept. As shown in fig. 7 (b), in the polar coordinate system, ρ represents the normal distance of the straight line from the far point, θ represents the angle between the normal line and the x axis, and this equation can be transformed into:

ρ＝xcosθ+ysinθ

The transformation is the Hough transformation of the point (x, y) in the polar coordinate system. By hough transform, a straight line in a rectangular coordinate system can be corresponded to a point in a polar coordinate system. Thus, the problem of detecting the straight line in the image space can be converted into the problem of detecting the point in the parameter space. The result of using the hough transform is shown in fig. 7 (d).

S6, local edge connection:

by observing the distribution characteristics of the fault trace, the sectional lines which are arranged in a row or are similar are found to have a high probability of being the same structural surface, namely the sectional lines have the characteristic of similar distance and trend on a coordinate system, so that an algorithm can be designed according to a distance criterion and an angle criterion, and the sectional lines which simultaneously meet the allowable values of the two sectional lines are connected.

As shown in fig. 8, assuming that the two line segments L _i and L _j, the criterion for determining that the segment lines belong to the same structural plane includes two of:

① Criterion 1: distance criteria. The criteria may be described as

D_min＜D

Wherein D _min is the distance between the nearest two ends P ₂,Q₁ of the two line segments L _i and L _j, and D is the distance threshold. And if the distance criterion is met, performing the next judgment.

② Criterion 2: an angle criterion. The criteria may be described as

|k_i-k_j|＜tan α

Where k _i,k_j is the slope of line segments L _i and L _j, respectively, and α is the angle threshold. If the angle criterion is satisfied, two points P ₂,Q₁ are connected.

And carrying out local edge connection processing according to the algorithm.

Example two

Fig. 9 is a schematic structural diagram of the extraction system for the cross fissure of the heading head-on of the coal mine tunnel, which mainly comprises a preprocessing module, an edge detection module, a false edge filtering module, an edge fitting module and a local edge connecting module.

The preprocessing module is used for preprocessing the collected coal roadway tunneling head-on image to obtain a first image; the edge detection module is used for extracting the edge of the rock mass structural surface of the first image to obtain a second image; the false edge filtering module is used for filtering the false edges of the binary image of the second image to obtain a third image; the edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image; the local edge connection module is used for extracting the cross joint trace of the coal roadway tunneling head-on image. The local edge connection module is also used for superposing the extracted cross-section management trace on the coal roadway tunneling head-on image to obtain the mine roadway tunneling head-on cross crack.

Next, the structural composition and functional implementation of each system portion will be specifically described in connection with the present embodiment.

In this embodiment, the preprocessing module is configured to preprocess the acquired coal roadway driving head-on image, so as to obtain a first image.

In the process of collecting a head-on image of a coal mine tunnel tunneling, noise is often generated due to the influence of factors such as non-uniform illumination, floating coal dust and the like, so that a satisfactory visual effect cannot be achieved. In order to reduce the influence of uneven illumination on an image and improve the quality of the image, it is necessary to enhance the original image.

The correction method of the illumination non-uniform image is mainly divided into two main types of correction methods with reference and correction methods without reference. In practical applications, reference-free illumination non-uniformity correction algorithms are of great interest. The existing non-reference illumination non-uniformity correction method mainly comprises an algorithm based on the Retinex theory, a histogram equalization method, a non-sharpening mask method, a morphological filtering method, a method based on a space illumination map and the like. The image is preprocessed by using an adaptive correction algorithm for the illumination non-uniformity image based on a two-dimensional gamma function.

For an image with uneven illumination, due to uneven distribution of illumination components in a scene, the brightness value of the image in a region with strong illumination in the image is enough or too strong, and the brightness value of the image in a region with weak illumination is insufficient.

Therefore, firstly, a multi-scale Gaussian function is adopted to extract illumination components in a scene, and the expression of the function is as follows:

Where G is the illumination component, c is the scale factor, and λ is the normalization constant. x is the abscissa of the input image pixel and y is the ordinate of the input image pixel. And convolving the original image with the Gaussian function to obtain an estimated value of the illumination component, wherein the estimated value is obtained as follows:

I(x,y)＝F(x,y)G(x,y)

wherein F (x, y) is an input image; i (x, y) is the estimated illumination component. And then, respectively extracting illumination components of the scene by utilizing a multi-scale Gaussian function, and weighting to finally obtain an estimated value of the illumination components, wherein the expression is as follows:

Wherein, I (x, y) is the illumination component value extracted and weighted by a plurality of Gaussian functions with different scales at the point (x, y); omega _i is the weighting coefficient of the illumination component extracted by the ith scale Gaussian function; i=1, 2, ··, N is the number of the used ruler degrees, here we take n=3, the scale factor c values selected herein are 15, 80 and 250, respectively. The illumination component of the image is extracted using a 3-scale gaussian function, the result of which is shown in fig. 2 (b).

After the illumination component is extracted, an illumination non-uniformity correction function is constructed according to the distribution characteristic of the illumination component, the illumination non-uniformity image is corrected, the brightness value of the area with over-illumination is reduced, and the brightness value of the area with over-illumination is improved. The parameters of the two-dimensional gamma function are adaptively adjusted by utilizing the distribution characteristics of illumination components of the image, so that the aim of improving the overall quality of the illumination non-uniform image is fulfilled. The two-dimensional gamma function expression is as follows:

Wherein O (x, y) is the brightness value of the corrected output image, and gamma is the index value for brightness enhancement; m is the luminance average of the illumination component.

Fig. 2 (c) and (d) are images enhanced by the illumination non-uniformity adaptive correction algorithm, and it can be seen that there is a significant enhancement for darker targets.

The edge detection module is used for extracting the edge of the rock mass structural plane of the first image to obtain a second image. Edges refer to the portions of the image where local intensity variations are most pronounced, as shown in fig. 3 (a). Edges are the result of grey value discontinuities that can typically be detected using derivative methods. Common edge detection operators are Canny, prewitt and Sobel operators. The Canny detection operator is applied to coal roadway tunneling head-on image analysis, and the processing result is shown in fig. 4 (b).

The false edge filtering module is used for filtering the false edge of the binary image of the second image to obtain a third image.

For the edge detection of the coal roadway tunneling head-on image, the result not only comprises complex and numerous structural surfaces, but also comprises a plurality of non-structural surface noises. This is related to factors such as edge detection algorithm and parameter selection, special textures in the original image, darker rock mass, etc., which are difficult to avoid in practical situations. Therefore, for large images containing complex structural surfaces, non-structural surface filtering is required, comprising the steps of:

1. Image description

For the binary image detected by the edge detection operator, each edge curve is a pixel connected region. Each connected region can be marked in Matlab, so that information of all edge curves, such as main shaft length, main shaft width and the like, can be obtained. It is observed that the structural surface is generally of a "long linear" configuration. Therefore, the aspect ratio is taken as an important index for distinguishing noise such as a long linear structure and square or round noise, which is an important basis for filtering background noise and extracting a fault structural surface.

r＝D/R

Wherein D is the length of the main axis of the communicating domain, R is the width of the main axis of the communicating domain, and R is the aspect ratio.

2. Clearing branch points and corner points

After edge detection is performed on an image, as shown in fig. 4, there is often a branching point in the image edge detection result. This prevents the morphological filtering of non-structural surface noise such as bare rock mass in the binary image. To obtain more accurate trace parameters, it is necessary to clear the branch points in the binary image obtained by edge detection.

In addition, for broken rock face images, linear targets formed after edge detection are mutually connected to form a broken line, and then the broken line is required to be split into two or more chain codes from the corner point on the basis of corner point identification. The key to splitting is corner detection. The corner points are high curvature points on the contour curves, the curvature of the chain codes can be estimated according to the chain code differences, the corner points are further judged, and the detection result is shown in fig. 4.

3. Non-structural face noise filtering

According to the selected mathematical characteristic attribute, the length-width ratio and the area of all connected domains can be obtained as shown in fig. 5. According to statistical laws, the ratio of the length of the principal axis to the width of the principal axis of a "long-line" structure is generally greater than 3. For this reason, the connected domain is traversed, and if the ratio of the main axis length to the main axis width of the region is greater than 3 and the 8 connected domain area is greater than 10 pixels, the connected domain is determined to be a long line structure. Otherwise, the pixels in the connected domain are set as foreground colors and used as non-structural surface noise filtering. The processing results obtained according to the above steps are shown in fig. 6.

And the edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image. If the edges are sharp and the noise is extremely low, the edge image can be binarized and thinned into a single pixel wide closed connected boundary map. However, edge discontinuities due to noise, uneven illumination, and other effects due to the introduction of false brightness discontinuities, the resulting group of pixels rarely describe an edge in its entirety. Therefore, it is typical to use a connection process to call edge pixel combinations meaningful edges immediately after using an edge detection algorithm. Edge connections can be generally divided into two types: global edge connections and local edge connections. Typically, we are at a specific location where the object of interest is not known. In this case, the global characteristics of the image can be utilized to directly detect the target contour.

Hough transforms were proposed by Hough in 1962. As shown in fig. 7 (a), in a two-dimensional plane, a straight line passing through a point (x, y) can be expressed as:

y_i＝ax_i+b

where a is the slope and b is the intercept. As shown in fig. 7 (b), in the polar coordinate system, ρ represents the normal distance of the straight line from the far point, θ represents the angle between the normal line and the x axis, and this equation can be transformed into:

ρ＝xcosθ+ysinθ

The transformation is the Hough transformation of the point (x, y) in the polar coordinate system. By hough transform, a straight line in a rectangular coordinate system can be corresponded to a point in a polar coordinate system. Thus, the problem of detecting the straight line in the image space can be converted into the problem of detecting the point in the parameter space. The result of using the hough transform is shown in fig. 7 (d).

The local edge connection module is used for extracting the cross joint trace of the coal roadway tunneling head-on image.

By observing the distribution characteristics of the fault trace, the sectional lines which are arranged in a row or are similar are found to have a high probability of being the same structural surface, namely the sectional lines have the characteristic of similar distance and trend on a coordinate system, so that an algorithm can be designed according to a distance criterion and an angle criterion, and the sectional lines which simultaneously meet the allowable values of the two sectional lines are connected.

As shown in fig. 8, assuming that the two line segments L _i and L _j, the criterion for determining that the segment lines belong to the same structural plane includes two of:

① Criterion 1: distance criteria. The criteria may be described as

D_min＜D

Wherein D _min is the distance between the nearest two ends P ₂,Q₁ of the two line segments L _i and L _j, and D is the distance threshold. And if the distance criterion is met, performing the next judgment.

② Criterion 2: an angle criterion. The criteria may be described as

|k_i-k_j|＜tan α

Where k _i,k_j is the slope of line segments L _i and L _j, respectively, and α is the angle threshold. If the angle criterion is satisfied, two points P ₂,Q₁ are connected.

The local edge connection module is also used for superposing the extracted cross joint trace on the coal roadway tunneling head-on image to obtain the mine roadway tunneling head-on cross fracture.

The foregoing embodiments are merely illustrative of the preferred embodiments of the present application, and the scope of the present application is not limited thereto, but various modifications and improvements made by those skilled in the art to which the present application pertains without departing from the spirit of the present application, should be made within the scope of protection defined in the claims of the present application.

Claims (2)

1. The extraction method of the coal mine tunnel tunneling head-on cross fracture comprises the following steps:

Acquiring a coal roadway tunneling head-on image;

preprocessing the coal roadway tunneling head-on image to obtain a first image;

extracting the edge of the rock mass structural plane of the first image to obtain a second image;

Filtering the binary image pseudo edges of the second image to obtain a third image;

extracting a linear edge of the third image; obtaining a fourth image;

Carrying out local connection processing on the fourth image according to an angle and distance criterion, and extracting a cross joint trace of the coal roadway tunneling head-on image; overlapping the extracted cross joint trace on the coal roadway tunneling head-on image to obtain an ore roadway tunneling head-on cross crack; the method for carrying out local connection processing on the fourth image comprises the following steps: designing an algorithm according to a distance criterion and an angle criterion, and connecting segment lines meeting allowable values of the distance criterion and the angle criterion at the same time; if the edge is obvious and the noise is extremely low, binarizing the edge image and thinning the edge image into a closed connected boundary diagram with single pixel width; directly detecting the target contour by utilizing the global characteristic of the image;

in a two-dimensional plane, a straight line passing through a point (x, y) can be expressed as:

yi＝axi+b

Wherein a is a slope and b is an intercept; in the polar coordinate system, ρ represents the normal distance of the straight line from the far point, θ represents the included angle between the normal line and the x axis, and this equation can be transformed into:

ρ＝xcosθ+ysinθ

The transformation is Hough transformation of the point (x, y) in the polar coordinate system; through Hough transform, a straight line in a rectangular coordinate system can be corresponding to a point in a polar coordinate system; converting the problem of detecting the straight line in the image space into the problem of detecting the point in the parameter space;

The preprocessing method is to adaptively correct the illumination non-uniform image by utilizing a two-dimensional gamma function;

The method for extracting the edge of the rock mass structural plane of the first image is to conduct derivation on the gray value of the first image;

The step of filtering the false edges of the binary image of the second image comprises the following steps:

An image description;

Removing branch points and corner points;

Filtering the noise of the non-structural surface;

The method for extracting the linear edge of the third image comprises the steps of adopting Hough transformation;

in the process of collecting a coal mine roadway heading head-on image, noise is generated due to the influence of non-uniform illumination and floating coal dust factors, and in order to reduce the influence of non-uniform illumination on the image, the quality of the image is improved, and the original image is enhanced;

preprocessing an image by using a self-adaptive correction algorithm of the illumination non-uniform image based on a two-dimensional gamma function;

for an image with uneven illumination, due to uneven distribution of illumination components in a scene, the brightness value of the image in a region with strong illumination in the image is enough or too strong, and the brightness value of the image in the region with weak illumination is insufficient;

Extracting illumination components in a scene by adopting a multi-scale Gaussian function, wherein the function expression is as follows:

wherein G is an illumination component, c is a scale factor, lambda is a normalization constant, x is an abscissa of an input image pixel, y is an ordinate of the input image pixel, and an estimated value of the illumination component can be obtained by convolving an original image with a Gaussian function, and the result is as follows:

I(x,y)＝F(x,y)G(x,y)

Wherein F (x, y) is an input image; i (x, y) is the estimated illumination component; and then, respectively extracting illumination components of the scene by utilizing a multi-scale Gaussian function, and weighting to finally obtain an estimated value of the illumination components, wherein the expression is as follows:

Wherein, I (x, y) is the illumination component value extracted and weighted by a plurality of Gaussian functions with different scales at the point (x, y); omega _i is the weighting coefficient of the illumination component extracted by the ith scale Gaussian function; i=1, 2,..n is the number of scales used, n=3, where the scale factor c has values of 15, 80 and 250, respectively; extracting illumination components of the image by using a 3-scale Gaussian function;

After the illumination component is extracted, an illumination non-uniformity correction function is constructed according to the distribution characteristic of the illumination component, correction processing is carried out on the illumination non-uniformity image, the brightness value of the area with over-illumination is reduced, and the brightness value of the area with over-illumination is improved; the parameters of the two-dimensional gamma function are adaptively adjusted by utilizing the distribution characteristics of illumination components of the image, so that the aim of improving the overall quality of the illumination non-uniform image is fulfilled; the two-dimensional gamma function expression is as follows:

Wherein 0 (x, Y) is a luminance value of the corrected output image, and Y is an index value for luminance enhancement; m is the luminance average of the illumination component;

Filtering the binary image pseudo edges of the second image to obtain a third image;

For the edge detection of the coal roadway tunneling head-on image, the result not only comprises complex and numerous structural surfaces, but also comprises non-structural surface noise; this is related to edge detection algorithms and parameter selection, special textures in the original image, and darker rock mass factors; for large images containing complex structural surfaces, non-structural surface filtering is required, and the method comprises the following steps:

Image description

For a binary image obtained by the detection of an edge detection operator, each edge curve is a pixel communication area; each communication area can be marked in Matlab, so that information of all edge curves is obtained, and the information of the edge curves is the length of a main shaft and the width of the main shaft; the structure is in a 'long line type' structure; the length-width ratio is used as an important index for distinguishing a long line structure from square or circular noise, which is an important basis for filtering background noise and extracting a fault structural surface;

r＝D/R

wherein D is the length of the main shaft of the communicating domain, R is the width of the main shaft of the communicating domain, and R is the length-width ratio;

clearing branch points and corner points

After the image is subjected to edge detection, branch points often exist in an image edge detection result; this prevents the noise of the unstructured surface of the bare rock in the binary image from being filtered by adopting a morphological method; in order to obtain more accurate trace parameters, branch points in the binary image obtained by edge detection need to be cleared;

For broken rock face images, linear targets formed after edge detection are mutually connected to form folding lines, and at the moment, the folding lines are split into two or more chain codes from corner points on the basis of corner point identification; the key point of the splitting is the detection of angular points; the corner points are high curvature points on the contour curves, and can estimate the curvature of the chain codes according to the chain code differences and judge the corner points;

Non-structural face noise filtering

Obtaining the length-width ratio and the area of all connected domains according to the selected mathematical characteristic attribute; according to the statistical rule, the ratio of the length of the main shaft of the 'long line type' structure to the width of the main shaft is more than 3; traversing the connected domain, and judging that the connected domain is of a long line structure if the ratio of the length of the main shaft to the width of the main shaft of the domain is more than 3 and the area of the connected domain is more than 10 pixels; otherwise, the pixels in the communication domain are set as foreground colors and used as non-structural surface noise for filtering;

the edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image; if the edges are obvious and the noise is extremely low, the edge image can be binarized and thinned into a single-pixel wide closed connected boundary map; edge discontinuities due to noise, uneven illumination, and other effects due to introducing false brightness discontinuities, the resulting group of pixels rarely describe an edge in its entirety; combining edge pixels to a meaningful edge using a connection process immediately after using the edge detection algorithm; edge connections can be generally divided into two types: global edge connections and local edge connections.

2. The extraction system is characterized by comprising an image collection module, a preprocessing module, an edge detection module, a false edge filtering module, an edge fitting module and a local edge connecting module;

the image collection module is used for acquiring a coal roadway tunneling head-on image;

The preprocessing module is used for preprocessing the collected coal roadway tunneling head-on image to obtain a first image;

The edge detection module is used for extracting the edge of the rock mass structural plane of the first image to obtain a second image;

the false edge filtering module is used for filtering the false edges of the binary image of the second image to obtain a third image;

The edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image;

Carrying out local connection processing on the fourth image according to an angle and distance criterion, and extracting a cross joint trace of the coal roadway tunneling head-on image; overlapping the extracted cross joint trace on the coal roadway tunneling head-on image to obtain an ore roadway tunneling head-on cross crack; the method for carrying out local connection processing on the fourth image comprises the following steps: designing an algorithm according to a distance criterion and an angle criterion, and connecting segment lines meeting allowable values of the distance criterion and the angle criterion at the same time; if the edge is obvious and the noise is extremely low, binarizing the edge image and thinning the edge image into a closed connected boundary diagram with single pixel width; directly detecting the target contour by utilizing the global characteristic of the image;

in a two-dimensional plane, a straight line passing through a point (x, y) can be expressed as:

yi＝axi+b

Wherein a is a slope and b is an intercept; in the polar coordinate system, ρ represents the normal distance of the straight line from the far point, θ represents the included angle between the normal line and the x axis, and this equation can be transformed into:

ρ＝xcosθ+ysinθ

The transformation is Hough transformation of the point (x, y) in the polar coordinate system; through Hough transform, a straight line in a rectangular coordinate system can be corresponding to a point in a polar coordinate system; converting the problem of detecting the straight line in the image space into the problem of detecting the point in the parameter space;

The preprocessing method is to adaptively correct the illumination non-uniform image by utilizing a two-dimensional gamma function;

The method for extracting the edge of the rock mass structural plane of the first image is to conduct derivation on the gray value of the first image;

The step of filtering the false edges of the binary image of the second image comprises the following steps:

An image description;

Removing branch points and corner points;

Filtering the noise of the non-structural surface;

The method for extracting the linear edge of the third image comprises the steps of adopting Hough transformation;

in the process of collecting a coal mine roadway heading head-on image, noise is generated due to the influence of non-uniform illumination and floating coal dust factors, and in order to reduce the influence of non-uniform illumination on the image, the quality of the image is improved, and the original image is enhanced;

preprocessing an image by using a self-adaptive correction algorithm of the illumination non-uniform image based on a two-dimensional gamma function;

for an image with uneven illumination, due to uneven distribution of illumination components in a scene, the brightness value of the image in a region with strong illumination in the image is enough or too strong, and the brightness value of the image in the region with weak illumination is insufficient;

Extracting illumination components in a scene by adopting a multi-scale Gaussian function, wherein the function expression is as follows:

wherein G is an illumination component, c is a scale factor, lambda is a normalization constant, x is an abscissa of an input image pixel, y is an ordinate of the input image pixel, and an estimated value of the illumination component can be obtained by convolving an original image with a Gaussian function, and the result is as follows:

I(x,y)＝F(x,y)G(x,y)

Wherein F (x, y) is an input image; i (x, y) is the estimated illumination component; and then, respectively extracting illumination components of the scene by utilizing a multi-scale Gaussian function, and weighting to finally obtain an estimated value of the illumination components, wherein the expression is as follows:

Wherein, I (x, y) is the illumination component value extracted and weighted by a plurality of Gaussian functions with different scales at the point (x, y); omega _i is the weighting coefficient of the illumination component extracted by the ith scale Gaussian function; i=1, 2,..n is the number of scales used, n=3, where the scale factor c has values of 15, 80 and 250, respectively; extracting illumination components of the image by using a 3-scale Gaussian function;

After the illumination component is extracted, an illumination non-uniformity correction function is constructed according to the distribution characteristic of the illumination component, correction processing is carried out on the illumination non-uniformity image, the brightness value of the area with over-illumination is reduced, and the brightness value of the area with over-illumination is improved; the parameters of the two-dimensional gamma function are adaptively adjusted by utilizing the distribution characteristics of illumination components of the image, so that the aim of improving the overall quality of the illumination non-uniform image is fulfilled; the two-dimensional gamma function expression is as follows:

Wherein 0 (x, Y) is a luminance value of the corrected output image, and Y is an index value for luminance enhancement; m is the luminance average of the illumination component;

Filtering the binary image pseudo edges of the second image to obtain a third image;

For the edge detection of the coal roadway tunneling head-on image, the result not only comprises complex and numerous structural surfaces, but also comprises non-structural surface noise; this is related to edge detection algorithms and parameter selection, special textures in the original image, and darker rock mass factors; for large images containing complex structural surfaces, non-structural surface filtering is required, and the method comprises the following steps:

Image description

For a binary image obtained by the detection of an edge detection operator, each edge curve is a pixel communication area; each communication area can be marked in Matlab, so that information of all edge curves is obtained, and the information of the edge curves is the length of a main shaft and the width of the main shaft; the structure is in a 'long line type' structure; the length-width ratio is used as an important index for distinguishing a long line structure from square or circular noise, which is an important basis for filtering background noise and extracting a fault structural surface;

r＝D/R

wherein D is the length of the main shaft of the communicating domain, R is the width of the main shaft of the communicating domain, and R is the length-width ratio;

clearing branch points and corner points

After the image is subjected to edge detection, branch points often exist in an image edge detection result; this prevents the noise of the unstructured surface of the bare rock in the binary image from being filtered by adopting a morphological method; in order to obtain more accurate trace parameters, branch points in the binary image obtained by edge detection need to be cleared;

For broken rock face images, linear targets formed after edge detection are mutually connected to form folding lines, and at the moment, the folding lines are split into two or more chain codes from corner points on the basis of corner point identification; the key point of the splitting is the detection of angular points; the corner points are high curvature points on the contour curves, and can estimate the curvature of the chain codes according to the chain code differences and judge the corner points;

Non-structural face noise filtering

Obtaining the length-width ratio and the area of all connected domains according to the selected mathematical characteristic attribute; according to the statistical rule, the ratio of the length of the main shaft of the 'long line type' structure to the width of the main shaft is more than 3; traversing the connected domain, and judging that the connected domain is of a long line structure if the ratio of the length of the main shaft to the width of the main shaft of the domain is more than 3 and the area of the connected domain is more than 10 pixels; otherwise, the pixels in the communication domain are set as foreground colors and used as non-structural surface noise for filtering;

the edge fitting module is used for extracting the linear edge of the third image to obtain a fourth image; if the edges are obvious and the noise is extremely low, the edge image can be binarized and thinned into a single-pixel wide closed connected boundary map; edge discontinuities due to noise, uneven illumination, and other effects due to introducing false brightness discontinuities, the resulting group of pixels rarely describe an edge in its entirety; combining edge pixels to a meaningful edge using a connection process immediately after using the edge detection algorithm; edge connections can be generally divided into two types: global edge connections and local edge connections.