CN113642534A - Mining equipment fault detection method and system based on artificial intelligence - Google Patents

Mining equipment fault detection method and system based on artificial intelligence Download PDF

Info

Publication number
CN113642534A
CN113642534A CN202111192892.6A CN202111192892A CN113642534A CN 113642534 A CN113642534 A CN 113642534A CN 202111192892 A CN202111192892 A CN 202111192892A CN 113642534 A CN113642534 A CN 113642534A
Authority
CN
China
Prior art keywords
roughness
image
pixel
descriptor
value
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202111192892.6A
Other languages
Chinese (zh)
Other versions
CN113642534B (en
Inventor
叶玮
王美容
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Haimen Heavy Mining Machinery Factory
Original Assignee
Haimen Heavy Mining Machinery Factory
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Haimen Heavy Mining Machinery Factory filed Critical Haimen Heavy Mining Machinery Factory
Priority to CN202111192892.6A priority Critical patent/CN113642534B/en
Publication of CN113642534A publication Critical patent/CN113642534A/en
Application granted granted Critical
Publication of CN113642534B publication Critical patent/CN113642534B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a mining equipment fault detection method and a mining equipment fault detection system based on artificial intelligence, wherein the method comprises the following steps: constructing a Gaussian pyramid based on the acquired mining equipment image; acquiring roughness description characteristics of each image in the Gaussian pyramid: for each pixel in an image, acquiring a roughness complexity and a roughness change descriptor of a window area with the pixel as a center based on the roughness of the pixel, and further acquiring a roughness descriptor of the window area; integrating the roughness descriptors of all the window regions to obtain the roughness description characteristics of the image; the roughness of the pixel is represented by utilizing the gray value, the saturation value, the brightness value and the heat value of the pixel position; and fusing the roughness description characteristics of the images in the Gaussian pyramid, and judging the wear degree of the mining equipment based on the fused roughness description characteristics. The invention can accurately evaluate the abrasion degree of the mining drill bit.

Description

Mining equipment fault detection method and system based on artificial intelligence
Technical Field
The invention relates to the field of artificial intelligence, in particular to a mining equipment fault detection method and system based on artificial intelligence.
Background
The importance of mining equipment in developing mineral resources is self evident, where the drill bit of the drilling equipment is the most susceptible to wear or failure. The categories of failures of drill bits vary, but of which the most important, also the most stressed part is the bit teeth. During the rotation of the drill bit, the drill bit teeth cut and grind the peripheral ore deposit, which is very easy to wear. Which over time affects the efficiency of the drilling rig. The existing detection of the abrasion fault of the drill bit still stays in an artificial observation and judgment stage, the analysis of the fault phenomenon has subjectivity, and misjudgment can be caused, so that the resource waste or the low mining efficiency is caused.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a mining equipment fault detection method and system based on artificial intelligence, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides a mining equipment fault detection method based on artificial intelligence, including the following specific steps:
constructing a Gaussian pyramid based on the acquired mining equipment image;
acquiring roughness description characteristics of each image in the Gaussian pyramid: for each pixel in an image, acquiring a roughness complexity and a roughness change descriptor of a window area with the pixel as a center based on the roughness of the pixel, and further acquiring a roughness descriptor of the window area; integrating the roughness descriptors of all the window regions to obtain the roughness description characteristics of the image; the roughness of the pixel is represented by utilizing the gray value, the saturation value, the brightness value and the heat value of the pixel position; the roughness complexity reflects the position distribution dispersion and the diversity of roughness values of some pixels in the window area, and the some pixels are pixels with different roughness from the central pixel in the window area;
And fusing the roughness description characteristics of the images in the Gaussian pyramid, and judging the wear degree of the mining equipment based on the fused roughness description characteristics.
Further, the gray value, the saturation value, the brightness value, and the heat value of the pixel position constitute a four-dimensional vector, which characterizes the roughness of the pixel.
Further, the rotation invariant circular LTP operator takes a pixel as a center, and an operator area obtained based on the radius of the operator is a window area corresponding to the pixel; and finishing coding according to a preset coding mode based on the roughness of the pixels in the window area to obtain a positive mode coding image and a negative mode coding image, further obtaining a positive mode characteristic value and a negative mode characteristic value corresponding to the central pixel point of the window area, wherein the sum of the positive mode characteristic value and the negative mode characteristic value represents the roughness complexity of the window area.
Further, the sum of difference vectors of the operator sampling points and the four-dimensional vectors of the central pixel points is a roughness change descriptor of the central pixel points;
for each operator sampling point, the sum of difference vectors of the operator sampling point and four-dimensional vectors of adjacent pixel points is a roughness change descriptor of the sampling point;
and the sum of the roughness change descriptor of the central pixel point and the roughness change descriptors of the sampling points is the roughness change descriptor of the window area.
Further, the roughness complexity of the window area and the roughness change descriptor are multiplied to obtain the roughness descriptor of the window area.
Further, an operator radius corresponding to the image is determined based on the size of the image and the degree of blurring.
Further, determining the weight of the image in the Gaussian pyramid based on the size and the fuzzy degree of the image, and performing weighted fusion on the roughness description features of the image in the Gaussian pyramid according to the weight.
Further, the roughness description features of the image are a roughness descriptor matrix obtained by integrating roughness descriptors of all window regions corresponding to the image.
And further, processing the fused roughness description characteristics by using a neural network to obtain the abrasion degree of the mining equipment.
In a second aspect, another embodiment of the present invention provides an artificial intelligence based mining equipment fault detection system, specifically comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which when executed by the processor implements the steps of an artificial intelligence based mining equipment fault detection method.
The embodiment of the invention at least has the following beneficial effects: the method constructs a roughness concept, extracts the roughness complexity of the image and the direction and amplitude of roughness change in a multi-scale space, and then fuses the roughness description characteristics of the image, so that the extracted roughness description characteristics have scale invariance and rotation invariance, and further, the accurate evaluation of the wear degree of the mining drill bit is realized based on the roughness description characteristics of the image.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description is provided for the method and system for detecting faults of mining equipment based on artificial intelligence in accordance with the present invention, and the specific implementation, structure, features and effects thereof are described in detail below. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following application scenarios are taken as examples to illustrate the present invention:
the application scene is as follows: the embodiment detects the abrasion degree of the mining drill bit, particularly the mining tricone bit, acquires the image of the mining drill bit, processes the image of the mining drill bit and judges the abrasion degree of the mining drill bit.
One embodiment of the invention provides a mining equipment fault detection method based on artificial intelligence, which comprises the following steps:
And step S1, constructing a Gaussian pyramid based on the acquired mining equipment image.
Preferably, the mining drill bit image is preprocessed, specifically, the mining drill bit is made of metal, and the physical property of the metal determines that the surface reflection degree of the mining drill bit is sensitive to light source reaction, so that the condition of uneven illumination is easily generated in the RGB image of the mining drill bit, and in order to ensure the accuracy of detection of the abrasion degree of the drill bit in the subsequent steps, the embodiment adjusts the condition of uneven illumination in the mining drill bit image by using a Retinex algorithm, so as to obtain the RGB image of the mining drill bit after the uneven illumination is eliminated.
Constructing a Gaussian pyramid based on the preprocessed mining drill bit image, specifically:
because the Gaussian kernel is the only linear kernel for realizing scale transformation, the embodiment utilizes the Gaussian kernel to blur the image and then carries out point-spaced downsampling, thereby constructing scale spaces of different levels and expressing the characteristics of physiological visual perception: the near is big and the far is small; near sharp, far blurred. With this method, a scale space containing N levels can be constructed, each scale space containing k images of different degrees of blur.
Preferably, N is 4 and k is 5 in the embodiment, which represents 4 levels of the scale space, five images are in each level, that is, four sets of images are in the gaussian pyramid, and 5 levels are in each set of images, specifically, sp = { sp = { (sp) } 1，sp2，sp3，sp4}，sp1，sp2，sp3，sp4Respectively representing 4 levels, each level comprising 5 images numbered 0-4 spi={I0，I1，I2，I3，I4The value range of i is [1,4 ]]Wherein the hierarchy sp1Image I of (1)0Is a preprocessed raw mining drill bit RGB image. The images in one hierarchy have the same size and different fuzzy degrees, the images in different hierarchies have different sizes, and preferably, the first image in the next hierarchy is obtained by performing point-to-point downsampling on the last but one image in the previous hierarchy; wherein the size of the preprocessed original mining bit image is m × n, and the size of the image in the ith level is (1/4)i-1M n; since the scale space is discrete, in order to makeThe scale transition is smoother, each level utilizes different Gaussian kernel parameters sigma to obtain 5 images, the parameter is the bandwidth of the Gaussian kernel, namely the larger the sigma is, the larger the local influence range of the Gaussian kernel function is, and the stronger the fuzzy effect is; in each level, image IjCorresponding gaussian kernel parameter σj=2jJ has a value in the range of [0,4 ]]。
Thus, a gaussian pyramid is obtained.
Preferably, the weight of the image in the gaussian pyramid is determined based on the size and the blurring degree of the image for the subsequent fusion of the image features, for the image I in the ith level jThe weight W corresponding to the imagei,j=((1/4)i-1*m*n)/σjThe smaller the image size is, the farther the visual angle is in the simulated physiological vision, and the worse the details are, the lower the weight is, namely, the larger i is, the lower the weight of the hierarchical image is; the higher the gaussian kernel parameter value, the stronger the blurring effect and the worse the detail, the lower the image weight, i.e. the larger j, the lower the weight of the image. It should be noted that after determining the weights of the images in the gaussian pyramid, normalization processing needs to be performed on the acquired weights, and in the embodiment, the weights of all the images are normalized by using a Softmax function, that is, ownership is remapped to [0, 1]And the sum of all weights is 1; in particular, Wi,jAfter normalization, W' is obtainedi,j(ii) a The purpose of this step is to ensure that the change of different scale features is smoother, and also to provide a basis for the adaptive change of the subsequent LTP operator receptive field.
Step S2, acquiring roughness description characteristics of each image in the Gaussian pyramid: for each pixel in an image, acquiring a roughness complexity and a roughness change descriptor of a window area with the pixel as a center based on the roughness of the pixel, and further acquiring a roughness descriptor of the window area; integrating the roughness descriptors of all the window regions to obtain the roughness description characteristics of the image; the roughness of the pixel is represented by utilizing the gray value, the saturation value, the brightness value and the heat value of the pixel position; the roughness complexity reflects the position distribution dispersion and the diversity of roughness values of some pixels in the window area, and the some pixels are pixels with different roughness from the central pixel in the window area.
(a) Acquiring the roughness of the pixel:
preferably, the embodiment maps the concept of roughness to a high-dimensional space, extracts the trend of local roughness variation, and ensures that the extracted features can represent the roughness of the surface of the bit tooth as accurately as possible. In order to improve the characteristic performance and ensure that the characteristic performance contains more information, the roughness is characterized by combining with a thermal infrared image. When the drill bit equipment works, the temperature of the part, which is in contact with a mineral seam for cutting, of the drill bit is increased sharply, so that the higher the temperature is, the higher the wear degree is, and the lower the roughness is; and because the drill bit material is metal, the heat conductivity is stronger, and heat propagation efficiency is higher, probably the condition that the heat radiation scope is too wide appears, and the place heat that does not have the wearing and tearing is also higher promptly. Therefore, the heat signature needs to be combined with other signatures to more reasonably characterize the wear level of the bit teeth. After investigation, under the condition of uniform illumination, the lower the gray value is, the lower the reflection degree is, and the higher the roughness is; and the higher the saturation, the higher the roughness; the lower the luminance, the higher the roughness. And for infrared thermal imaging, the essence is a gray scale map, and the lower the heat, the lower the gray scale, the more slight the abrasion and the higher the roughness.
Thus, for each pixel, the grey value, saturation value, brightness value and heat value of the pixel position constitute a four-dimensional vector characterizing the roughness of the pixel, in particular: four-dimensional vector rough = [ GRAY, SAT, LUM, HEAT =]Wherein, in the step (A),
Figure DEST_PATH_IMAGE002
GRAY represents the influence of GRAY level on roughness, and has a value range of [1/255,1]Gray represents a gray value of a pixel; SAT = SAT, SAT representing the influence of saturation on roughness, and having a value in the range of [0,1]Sat represents the saturation of the pixel obtained based on the S channel in the HSV color space;
Figure DEST_PATH_IMAGE004
LUM represents the influence of luminance values on roughnessThe value range is [1/100,1 ]]Lum represents the luminance of the pixel obtained based on the L channel in the Lab color space;
Figure DEST_PATH_IMAGE006
HEAT represents the influence of HEAT value on roughness, and its value range is [1/255,1 ]]Heat represents a heat value of a pixel obtained based on the thermal infrared map; the thermal infrared map can be acquired in real time during the operation process or acquired with low delay just after the operation is finished.
It should be noted that when a gaussian pyramid is obtained, a gray-scale map gaussian pyramid, an S-channel map gaussian pyramid, an L-channel map gaussian pyramid, and a thermal infrared map gaussian pyramid are correspondingly required to be obtained, and the blur degree of images in the gaussian pyramids is different, which may cause the value of a pixel to change.
To eliminate the effect of the dimension, four components in a four-dimensional vector may be mapped between [0,255 ].
Because the obtained four-dimensional vector rough does not consider the correlation and the emphasis among the four components, the embodiment performs principal component analysis on the four components of the four-dimensional vector by using PCA to obtain the characteristic value corresponding to each component, and the characteristic value is used as the weight of each component after being normalized; the characteristic value of each component obtained by PCA is normalized and then respectively expressed as w1、w2、w3、w4Then, the representation mode of weighting each component based on the weight to obtain the final four-dimensional vector Rough is as follows: rough = [ w =1*GRAY，w2*SAT，w3*LUM，w4*HEAT]Rough is also known as a roughness vector. Compared with texture features, the roughness concept in the invention integrates several different kinds of information, and the roughness concept contains more information and can improve the robustness of the index roughness.
(b) For each pixel in an image, acquiring the roughness complexity of a window area with the pixel as a center, wherein the roughness complexity reflects the position distribution dispersion and the roughness value diversity of some pixels in the window area, and the roughness of some pixels is different from that of the center pixel in the window area. Counting the roughness values of some pixels in the window area to obtain a value set, wherein the values in the value set are different, and the more the number of values in the value set is, the more diversified the roughness values of the pixels in the window area are, the more the complexity value of the roughness of the window area is, and the larger the value of the roughness complexity of the window area is; furthermore, the more discrete the distribution of the positions of said certain pixels, the greater the value of the roughness complexity of the window area; the larger the value of the roughness complexity of the window area is, the more the surface of the mining drill bit is embodied as a hollow on the mining drill bit; specifically, the method for acquiring the roughness complexity of the window region comprises the following steps:
In one embodiment, the LTP operator may be a square or a circle, and preferably, a rotation-invariant circle LTP operator is used in the embodiment, and the roughness complexity of the window region is described by taking 8 sampling points as an example, in this embodiment, some of the pixels are pixels in the window region, whose roughness corresponds to a sampling point different from that of the central pixel:
the preset encoding mode of the rotation invariant circular LTP operator may be: if | Roughs|>|RoughcI, the operator sample point is coded as 1; if | Roughs|=|RoughcIf the operator sample point is coded as 0, ifs|<|RoughcI, the operator sample point is encoded as-1; | Roughs|、|RoughcAnd | respectively represents the modular length of the operator sampling point and the four-dimensional vector of the central pixel point.
In order to reduce the noise interference, the preset encoding mode of the rotation invariant circular LTP operator may also be: if | Roughs|≥|RoughcL + t, the operator sample point is coded as 1; if | | | Roughs|-|Roughc||<t, the operator sample point is coded as 0, if | Roughs|≤|RoughcL-t, the operator sample points are encoded as-1; | Roughs|、|RoughcI respectively represents the operator sampling point and the four-dimensional vector of the central pixel point, namely the modular length of the roughness vector;t is a threshold value, and the value of t is 5 in the example.
The rotation invariant circular LTP operator takes a pixel as a center, and an operator area obtained based on the radius of the operator is a window area corresponding to the pixel; based on the roughness of pixels in the window area, completing encoding according to a preset encoding mode to obtain a positive mode encoding image and a negative mode encoding image, further obtaining a positive mode characteristic value and a negative mode characteristic value corresponding to the central pixel point of the window area, wherein the sum of the positive mode characteristic value and the negative mode characteristic value represents the roughness complexity of the window area; specifically, the method comprises the following steps: after the coding is finished according to the preset coding mode, 1 is taken as a reference, all sampling points with the coding value smaller than 1 are coded to be 0 to obtain a positive mode coding graph, each sampling point is taken as a starting point, the binary codes in the positive mode coding graph are extracted clockwise to obtain 8 groups of binary codes, and the minimum value of 8 decimal characteristic values corresponding to the 8 groups of binary codes is a positive mode characteristic value LTP +(ii) a Similarly, taking-1 as a reference, all codes larger than-1 are 0, original-1 sampling points are recoded to be 1 to obtain a negative mode code graph, taking each sampling point as a starting point, clockwise extracting binary codes in the negative mode code graph, and taking the minimum value of 8 decimal characteristic values corresponding to 8 groups of binary codes as a negative mode characteristic value LTP-。
In another embodiment, the implementer may also use a rotation-invariant circular LBP operator, and the corresponding coding mode is: when | Roughs|≥|RoughcL + t or | Roughs|≤|RoughcWhen l-t, the operator sample point is coded as 1; when | | | Roughs|-|Roughc||<And when t is reached, the operator sampling points are coded to be 0 to obtain a binary code image, clockwise binary codes in the binary code image are extracted by taking each sampling point as a starting point respectively, 8 groups of binary codes can be obtained, the minimum value of 8 decimal characteristic values corresponding to the 8 groups of binary codes is an LBP characteristic value, and the LBP characteristic value represents the roughness complexity of a window area.
The roughness complexity of the window area obtained based on the two embodiments can reflect the position distribution dispersion and the diversity of roughness values of sampling points with different roughness and central pixels in the window area; in addition, in both of the above two embodiments, the roughness value of the central pixel is affected by noise and fluctuates when encoding is performed, so that the above two embodiments become insensitive to noise, that is, have strong anti-noise capability.
Preferably, in order to obtain more image detail information, the invention determines the operator radius corresponding to the image based on the size and the blurring degree of the image, that is, the radius of the rotation-invariant circular LTP operator or LBP operator corresponding to each image in the gaussian pyramid is different, that is, the size of the window region corresponding to the pixel of different images in the gaussian pyramid is different, specifically, for the image I in the ith leveljThe radius of the operator corresponding to the image is Ri，j=R0*W＇i,j，R0Is the first level sp1Middle image I0Radius of the corresponding operator, preferably R in the embodiment0Is taken to be 8 or 12, Wi,jIs Wi,jThe value obtained after normalization processing.
(c) For each pixel in an image, a descriptor of the roughness variation of the window region centered on the pixel is obtained:
the sum of the difference vectors of the operator sampling points and the four-dimensional vectors of the central pixel points is the roughness change descriptor of the central pixel points; specifically, the method comprises the following steps: directc=∑8 x=1(Roughx s-Roughc)，directcRoughx s represents a four-dimensional roughness vector of the sampling point of the x-th operator, and Rough is a roughness change descriptor of a central pixel pointcAnd representing the four-dimensional roughness vector of the central pixel point.
For each operator sampling point, the sum of difference vectors of the operator sampling point and four-dimensional vectors of adjacent pixel points is a roughness change descriptor of the sampling point; specifically, the method comprises the following steps: directx s = Σ 8 y =1(Roughy x-Roughx s), where directx s is a roughness change descriptor of the xth operator sampling point, Roughy x is a roughness vector of the yth neighborhood pixel point of the xth operator sampling point, and there are 8 neighborhood pixel points in one sampling point in the embodiment.
Roughness variation of center pixelThe sum of the descriptor and the roughness variation descriptor of each sampling point is the roughness variation descriptor of the window area, and specifically: direct = directcAnd +/-8 x =1directx s, wherein the direct is a roughness change descriptor of the window area, and the roughness change descriptor is a four-dimensional vector. The method for acquiring the roughness change descriptor considers the pixel-level information and the image-level information, so that the description of the direction and the amplitude of the roughness change in the window area is more accurate, the direction and the amplitude information are added for the subsequent task of judging the abrasion degree of the mining equipment, and the accuracy and the robustness in judging the abrasion degree are improved.
(d) Acquiring a roughness descriptor of the window area, and integrating the roughness descriptors of all the window areas to obtain the roughness description characteristics of the image:
for each window region, the roughness complexity and the roughness variation descriptor of that window region are multiplied to obtain a roughness descriptor of the window region, in particular graph = (LTP)++LTP-) Direct, graph represents a roughness descriptor of the window region, which is a four-dimensional vector.
For an image, acquiring a window area corresponding to each pixel in the image, and integrating roughness descriptors of all window areas to obtain a roughness description feature of the image, where the roughness description feature of the image is a roughness descriptor matrix obtained by integrating roughness descriptors of all window areas corresponding to the image, preferably, in the embodiment, the roughness descriptors of all window areas are integrated based on the positions of the pixels in the image to obtain a roughness descriptor matrix GRAPH, and the roughness descriptor matrix GRAPH is the roughness description feature of the image.
And step S3, fusing the roughness description characteristics of the images in the Gaussian pyramid, and judging the wear degree of the mining equipment based on the fused roughness description characteristics.
Because the sizes of the images in different levels in the Gaussian pyramid are different, and further the sizes of the roughness descriptor matrixes of the images in different levels are different, levels sp are required to be merged before fusion2，sp3，sp4The roughness descriptor matrix corresponding to the image in (1) is up-sampled, so as to sp2，sp3，sp4The roughness descriptor matrix and the level sp corresponding to the image in (1)1The roughness descriptor matrixes corresponding to the images in (1) have the same size, that is, the number of rows and columns is the same, and preferably, the upsampling is performed by using the mean difference value in the embodiment.
After upsampling, performing weighted fusion on the roughness description features of the images in the gaussian pyramid according to the weights of the images in the gaussian pyramid, specifically, G = ∑ 4 i =1 ∑ 4 j =0W ″i,j*GRAPHi,jG denotes roughness description characteristics after fusion, GRAPHi,jRepresenting images I in the ith leveljThe roughness descriptor matrix of (1).
Judging the wear degree of the mining equipment based on the fused roughness description characteristics, preferably, processing the fused roughness description characteristics G by using a neural network to obtain the wear degree of the mining equipment; in the embodiment, the abrasion degree corresponds to 6 abrasion grades 0-5, 0 represents that the mining drill bit is not abraded, 1-5 represents 5 abrasion grades, and the higher the abrasion grade is, the more serious the abrasion is; correspondingly, the training data set used for training the neural network CNN includes roughness description features corresponding to the mining drill bit images of 6 wear grades after fusion, and labels are respectively corresponding wear grades and are trained by adopting a cross entropy loss function. And processing the input fused roughness description characteristics G by the neural network, outputting the probability that the roughness description characteristics belong to each wear level, wherein the wear level corresponding to the maximum probability value is the wear degree of the mining drill bit in the mining drill bit image.
Based on the same inventive concept as the above method embodiments, an embodiment of the present invention provides an artificial intelligence based mining equipment fault detection system, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which computer program, when executed by the processor, implements the steps of an artificial intelligence based mining equipment fault detection method.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (9)

1. A mining equipment fault detection method based on artificial intelligence is characterized by comprising the following steps:
constructing a Gaussian pyramid based on the acquired mining equipment image;
acquiring roughness description characteristics of each image in the Gaussian pyramid: for each pixel in an image, acquiring a roughness complexity and a roughness change descriptor of a window area with the pixel as a center based on the roughness of the pixel, and further acquiring a roughness descriptor of the window area; integrating the roughness descriptors of all the window regions to obtain the roughness description characteristics of the image; the roughness of the pixel is represented by utilizing the gray value, the saturation value, the brightness value and the heat value of the pixel position; the roughness complexity reflects the position distribution dispersion and the diversity of roughness values of some pixels in the window area, and the some pixels are pixels with different roughness from the central pixel in the window area;
And fusing the roughness description characteristics of the images in the Gaussian pyramid, and judging the wear degree of the mining equipment based on the fused roughness description characteristics.
2. The method of claim 1, wherein the gray value, the saturation value, the luminance value, and the heat value of a pixel location constitute a four-dimensional vector that characterizes the roughness of the pixel.
3. The method of claim 2, wherein the roughness complexity of the window region is obtained by:
the rotation invariant circular LTP operator takes a pixel as a center, and an operator area obtained based on the radius of the operator is a window area corresponding to the pixel; and finishing coding according to a preset coding mode based on the roughness of the pixels in the window area to obtain a positive mode coding image and a negative mode coding image, further obtaining a positive mode characteristic value and a negative mode characteristic value corresponding to the central pixel point of the window area, wherein the sum of the positive mode characteristic value and the negative mode characteristic value represents the roughness complexity of the window area.
4. A method as claimed in claim 3, characterized in that the acquisition of the roughness variation descriptor of the window region is embodied as:
the sum of the difference vectors of the operator sampling points and the four-dimensional vectors of the central pixel points is the roughness change descriptor of the central pixel points;
For each operator sampling point, the sum of difference vectors of the operator sampling point and four-dimensional vectors of adjacent pixel points is a roughness change descriptor of the sampling point;
and the sum of the roughness change descriptor of the central pixel point and the roughness change descriptors of the sampling points is the roughness change descriptor of the window area.
5. The method of claim 4, wherein the roughness complexity and the roughness variation descriptor of a window region are multiplied to obtain a roughness descriptor vector for the window region.
6. The method of claim 5, wherein the operator radius corresponding to the image is determined based on the size of the image and the degree of blur.
7. The method of claim 6, wherein the weight of the image in the Gaussian pyramid is determined based on the size and the degree of blurring, and the roughness description features of the image in the Gaussian pyramid are subjected to weighted fusion according to the weight.
8. The method of claim 7, wherein the roughness-describing features of the image are a roughness-descriptor matrix obtained by integrating roughness descriptors of all window regions corresponding to the image.
9. The method of claim 8, wherein the fused roughness descriptive features are processed using a neural network to derive a degree of wear of the mining equipment.
An artificial intelligence based mining equipment fault detection system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program, when executed by the processor, implements the steps of the method according to any one of claims 1-9.
CN202111192892.6A 2021-10-13 2021-10-13 Mining equipment fault detection method and system based on artificial intelligence Active CN113642534B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111192892.6A CN113642534B (en) 2021-10-13 2021-10-13 Mining equipment fault detection method and system based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111192892.6A CN113642534B (en) 2021-10-13 2021-10-13 Mining equipment fault detection method and system based on artificial intelligence

Publications (2)

Publication Number Publication Date
CN113642534A true CN113642534A (en) 2021-11-12
CN113642534B CN113642534B (en) 2022-02-18

Family

ID=78426614

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111192892.6A Active CN113642534B (en) 2021-10-13 2021-10-13 Mining equipment fault detection method and system based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN113642534B (en)

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104281994A (en) * 2014-10-22 2015-01-14 南京大学 Three-dimensional grid watermark embedding and detecting method based on local roughness analysis
CN110728677A (en) * 2019-07-22 2020-01-24 中南大学 Texture roughness defining method based on sliding window algorithm

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104281994A (en) * 2014-10-22 2015-01-14 南京大学 Three-dimensional grid watermark embedding and detecting method based on local roughness analysis
CN110728677A (en) * 2019-07-22 2020-01-24 中南大学 Texture roughness defining method based on sliding window algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
XU GUANG-MEI等: "A New Metric Algorithm of Image Textural Coarseness", 《2010 INTERNATIONAL CONFERENCE ON COMPUTER APPLICATION AND SYSTEM MODELING》 *
徐赵飞: "花粉图像鉴别特征提取算法的研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 *

Also Published As

Publication number Publication date
CN113642534B (en) 2022-02-18

Similar Documents

Publication Publication Date Title
CN111950649B (en) Attention mechanism and capsule network-based low-illumination image classification method
Zamri et al. Tree species classification based on image analysis using Improved-Basic Gray Level Aura Matrix
Laishram et al. A novel minimal distortion-based edge adaptive image steganography scheme using local complexity: (BEASS)
CN113128271A (en) Counterfeit detection of face images
CN111179202B (en) Single image defogging enhancement method and system based on generation countermeasure network
CN116682068B (en) Oil well sand prevention operation construction monitoring method and system thereof
CN115619743A (en) Construction method and application of OLED novel display device surface defect detection model
CN114882599A (en) Off-line handwritten signature segmentation system and method based on double-branch neural network
Freitas et al. On the application LBP texture descriptors and its variants for no-reference image quality assessment
Liu et al. Blind image blur metric based on orientation-aware local patterns
JP2024505766A (en) End-to-end watermarking system
Han et al. Low contrast image enhancement using convolutional neural network with simple reflection model
CN113642534B (en) Mining equipment fault detection method and system based on artificial intelligence
CN117422696A (en) Belt wear state detection method based on improved YOLOv8-Efficient Net
Yang et al. EHNQ: Subjective and objective quality evaluation of enhanced night-time images
CN112967227A (en) Automatic diabetic retinopathy evaluation system based on focus perception modeling
JP4795737B2 (en) Face detection method, apparatus, and program
CN115272378A (en) Character image segmentation method based on characteristic contour
JP4202692B2 (en) Image processing method and apparatus
Yuan et al. RM-IQA: A new no-reference image quality assessment framework based on range mapping method
CN114463379A (en) Dynamic capturing method and device for video key points
Mahmood Defocus Blur Segmentation Using Genetic Programming and Adaptive Threshold.
Xiao et al. Blind Quality Metric via Measurement of Contrast, Texture, and Colour in Night-Time Scenario.
Raghavan et al. Performance Evaluation of Weighted Entropy Based Fusion Technique for Face Recognition with Different Pre-processing Techniques
Costa et al. Enhancement of weakly illuminated images using CNN and retinex theory

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant