CN108956614B - Mining steel wire rope dynamic flaw detection method and device based on machine vision - Google Patents
Mining steel wire rope dynamic flaw detection method and device based on machine vision Download PDFInfo
- Publication number
- CN108956614B CN108956614B CN201810432434.7A CN201810432434A CN108956614B CN 108956614 B CN108956614 B CN 108956614B CN 201810432434 A CN201810432434 A CN 201810432434A CN 108956614 B CN108956614 B CN 108956614B
- Authority
- CN
- China
- Prior art keywords
- steel wire
- wire rope
- image
- damage
- matrix
- 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.)
- Active
Links
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 175
- 239000010959 steel Substances 0.000 title claims abstract description 175
- 238000001514 detection method Methods 0.000 title claims abstract description 44
- 238000005065 mining Methods 0.000 title claims abstract description 20
- 239000011159 matrix material Substances 0.000 claims abstract description 72
- 238000013528 artificial neural network Methods 0.000 claims abstract description 39
- 238000012549 training Methods 0.000 claims abstract description 30
- 238000000034 method Methods 0.000 claims abstract description 24
- 238000012545 processing Methods 0.000 claims abstract description 23
- 230000007547 defect Effects 0.000 claims abstract description 16
- 238000005316 response function Methods 0.000 claims description 27
- 230000008859 change Effects 0.000 claims description 25
- 238000006073 displacement reaction Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 13
- 230000004044 response Effects 0.000 claims description 13
- 230000009466 transformation Effects 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000001914 filtration Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 7
- 238000013139 quantization Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 6
- 238000012935 Averaging Methods 0.000 claims description 4
- 230000001052 transient effect Effects 0.000 claims description 4
- 230000002950 deficient Effects 0.000 claims description 3
- 230000005284 excitation Effects 0.000 claims description 3
- 238000004458 analytical method Methods 0.000 description 6
- 230000011218 segmentation Effects 0.000 description 4
- 239000003245 coal Substances 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000001228 spectrum Methods 0.000 description 3
- 238000005299 abrasion Methods 0.000 description 2
- 238000003491 array Methods 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 239000012535 impurity Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 241001391944 Commicarpus scandens Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000005415 magnetization Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000013598 vector Substances 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analyzing Materials By The Use Of Ultrasonic Waves (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of image processing, and provides a dynamic flaw detection method and device for a mining steel wire rope based on machine vision. The method comprises the following steps: continuously shooting and recording the steel wire rope in dynamic motion through a camera to obtain a video signal; processing the video signal, and extracting a moving target to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope; obtaining a flexibility matrix of the steel wire rope to be detected; step S4: obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after the sample training; taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image; and processing the initial damage image of the steel wire rope, extracting features, and judging the surface defects based on the texture features by using a linear classifier. The invention solves the problem of image detection error, and has simple operation and low cost.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a dynamic flaw detection method and device for a mining steel wire rope based on machine vision.
Background
With the diversification of coal mining modes, various types of winches play an important role in inclined roadway transportation of coal mines. Personnel and equipment enter and exit a mine by virtue of a winch lifting system, a steel wire rope needs to bear heavy burden during working, the underground environment is special, potential safety hazards such as abrasion, cracks, corrosion and wire breakage of the steel wire rope easily occur in the steel wire rope in the long-term use process, the steel wire rope is easy to break when the steel wire rope is not replaced in time, and serious accidents are certainly caused. If the steel wire rope can be replaced in time before the steel wire rope is broken, serious accidents can be effectively avoided.
The steel wire rope is generally used for no more than two years, and the steel wire rope can be completely replaced regardless of whether the steel wire rope is damaged or not. The value of a steel wire rope used by a winch lifting system for a coal mine is generally about four to five million, and the mode of determining whether to replace the steel wire rope only by the service life is very waste on one hand, the safety cannot be ensured on the other hand, and sometimes, if the steel wire rope is not used correctly, the abnormal conditions of abrasion, breakage, distortion and deformation can occur when the steel wire rope is used for about one year.
The steel wire rope flaw detection comprises two categories of manual visual detection and electromagnetic detection. The manual visual inspection means that a special worker is equipped to observe whether the steel wire rope is damaged or not by naked eyes regularly. The method has the advantages of long detection time, high labor intensity, easy fatigue of detection personnel and low efficiency, greatly depends on the professional quality and working attitude of the working personnel, and has strong subjectivity, thus leading to high missed detection rate. The electromagnetic detection refers to magnetizing the steel wire rope to be detected and then detecting magnetic leakage. The method has the advantages of complex equipment and high cost, the signal is easy to be distorted by external interference, and the uneven magnetization can bring large errors to the detection result.
How to overcome the technical defects of the steel wire rope flaw detection system in the traditional technology is a technical problem to be solved urgently by technical personnel in the field.
Disclosure of Invention
The invention overcomes the defects of the prior art, and solves the technical problems that: the mining steel wire rope dynamic flaw detection method and device based on the machine vision are accurate in detection and simple in operation.
In order to solve the technical problems, the invention adopts the technical scheme that: a mining steel wire rope dynamic flaw detection method based on machine vision comprises the following steps:
step S1: erecting a camera in front of the steel wire rope, calibrating the actual distance represented by a unit pixel of the camera, acquiring distance calibration parameters, enabling the steel wire rope to move at a constant speed, and continuously photographing and recording the steel wire rope in dynamic motion through the camera to obtain a video signal;
step S2: processing the video signal, and extracting a moving target to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope;
step S3: according to the vibration track of each pixel point on the characteristic edge of the steel wire rope, obtaining a displacement space sequence matrix distributed along the length direction of the steel wire rope, performing discrete Fourier transform on each column of the matrix to obtain a frequency response function corresponding to each point on the steel wire rope, identifying inherent frequency according to the peak value of the frequency response function, obtaining the modal vibration mode of the steel wire rope according to the amplitude of the response frequency of the frequency response function of each point, and finally obtaining the flexibility matrix of the steel wire rope to be tested;
step S4: setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; according to the flexibility matrix of the steel wire rope to be detected, obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after sample training;
step S5: taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image;
step S6: according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; and then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge the surface defects based on the texture features.
In step S2, the processing of the video signal includes:
step S201: establishing a background model: establishing a background model by using a method of averaging multiple frames of images;
step S202: detection of the change area: carrying out difference processing on two continuous frames of images in a video sequence to determine a background area and a motion change area;
step S203: detecting a moving object: processing the current frame image, and detecting a moving object by only carrying out difference on the image in the change area and the background image;
step S204: target edge identification: extracting edge points with sub-pixel precision by adopting a 5-degree polynomial to fit an edge gray scale change curve; and processing a series of images to obtain the vibration track of each pixel point on the characteristic edge of the steel wire rope in a certain time period.
In step S4, when a training sample of the sample training BP neural network is set, each steel wire rope is divided into 8 units, and based on the difference in the wire breakage conditions, assuming that the stiffness thereof is respectively reduced by 25% and 40%, damage conditions of 16 degrees are obtained, natural frequencies and modal vibration patterns of vibrations of 16 kinds of damaged steel wire ropes are calculated, the first two-order frequencies and vibration patterns are taken, 16 sets of values are obtained, and GSL conversion is performed on the values, and the values are used as the training sample of the neural network.
In step S4, when setting a sample to train a BP neural network, selecting a Sigmoid function as an excitation function, selecting an error signal back propagation algorithm as a training algorithm, and updating the weight according to the formula:
wherein the first term represents the gradient of the error mean, the second term represents the transient term, the third term represents the random noise term, t represents the number of iterations,
represents the lth level weight for the tth iteration,
which is indicative of the delta error,
represents the output of the jth neuron of the lth layer of the kth training sample, η represents a learning factor, μ represents a transient constant,
representing a random noise term.
In step S6, the smoothness R is calculated by the formula: r is 1-1/(1+ sigma)2) (ii) a The formula for the calculation of entropy e is:
in the formula, σ represents a standard deviation of an image, and the calculation formula is:
l represents the gray level, ziRepresenting the gray value of each pixel point in the image area, m represents the average gray level,
p(zi) Representing the gray value z of each pixel point in the image areaiThe probability of (c).
In step S6, after the high-pass filtering and the imadjust-based gray-scale transformation are performed on the initial steel wire rope damage image, the image is further divided into smaller portions, the smoothness value and the entropy value of each portion are extracted, and each portion is classified and judged by a linear classifier.
In the step S6, the linear classifier includes a smoothness classification threshold T1 and an entropy classification threshold T2, and when the smoothness R > T1 and the entropy e > T2 of the image, it is determined that the surface is defective, and when the smoothness R < T1 and the entropy e < T2 of the image, it is determined that the surface is intact.
The invention also provides a mining steel wire rope dynamic flaw detection device based on machine vision, which comprises image acquisition equipment, a controller, a memory and a display, wherein the image acquisition equipment is used for acquiring images of the steel wire rope; the image acquisition equipment is used for acquiring a video signal of the steel wire rope; the controller is configured to execute the following program:
extracting a moving target from the video signal to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope;
obtaining a displacement space sequence matrix of the steel wire rope distributed along the rope length direction according to the vibration track, performing discrete Fourier transform on each column of the matrix to obtain a frequency response function corresponding to each point on the steel wire rope, identifying inherent frequency according to the peak value of the frequency response function, obtaining the modal vibration mode of the steel wire rope according to the amplitude of the response frequency of the frequency response function of each point, and finally obtaining the flexibility matrix of the steel wire rope;
setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; according to the flexibility matrix of the steel wire rope to be detected, obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after sample training;
taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image;
according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge surface defects based on texture features;
the memory is used for storing damaged steel wire rope pictures; the display equipment is used for displaying the damaged steel wire rope picture and marking the damage position and the damage degree.
Compared with the prior art, the invention has the following beneficial effects: the invention can carry out video photography on the state of the steel wire rope in real time in the normal production process, carry out background segmentation on the steel wire rope, provide the most complete characteristic data and extract a moving target, has accurate position and high speed, and can solve the problem of target solving failure caused by the change of the brightness and the gray level of the background color when a single-frame image is used for solving the edge to search the target. The method combines damage identification and image detection of a flexibility matrix, and solves the problem of image detection errors caused by the fact that spots formed by oil stains attached to the surface of a steel wire rope and other industrial impurity oil stains fall within the range of extracted strand images by using an image detection technology only by utilizing the influence of stress change of the steel wire rope caused by damage on flexibility. Meanwhile, complex equipment is not needed, the operation is simple, and the cost is low.
Drawings
FIG. 1 is a schematic flow chart of a dynamic flaw detection method for a mining steel wire rope based on machine vision according to the invention;
FIG. 2 is a Marxiweir mechanics model of a wire rope;
FIG. 3 is a flowchart of an algorithm for a BP neural network;
FIG. 4 is a flow chart of the linear classifier of the present invention;
fig. 5 is a schematic structural diagram of the mining steel wire rope dynamic flaw detection device based on machine vision.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments; all other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
As shown in fig. 1, the invention provides a dynamic flaw detection method for a mining steel wire rope based on machine vision, which comprises the following steps:
step S1: an industrial camera is erected in front of the steel wire rope, the actual distance represented by unit pixels of the industrial camera is calibrated, distance calibration parameters are obtained, the steel wire rope moves at a constant speed, and the steel wire rope in dynamic motion is continuously photographed and recorded by the camera to obtain video signals.
Step S2: and processing the video signal, and extracting the moving target to obtain the vibration track of each pixel point on the characteristic edge of the steel wire rope.
The video signal processing method comprises the following steps:
step S201: establishing a background model: and establishing a background model by using a method of averaging multiple frames of images. A background model is established by using a method of averaging multiple frames of images, and the formula is expressed as follows:
where N is the reconstructed image sequence frame number, BKFor the reconstructed background image, fkFor the k frame image, the value of each pixel point in the background image is the N frame image of the pixel pointLike the cumulative average of the gray levels.
Step S202: detection of the change area: and carrying out differential processing on two continuous frames of images in the video sequence to determine the area of the background and the area with changed motion.
The inter-frame difference method can detect a change region between two adjacent frames, wherein the change region actually comprises a region P covered by a moving object in a previous frame, and the region covered by the moving object in a current frame is the moving object Q. Let fk(i, j) and fk+1(i, j) are two continuous frames of images in the video sequence, wherein i, j represents the coordinate value of a pixel point, the two frames of images are subjected to differential processing, and the detection rule is as follows:
here, T is a threshold value for detection, and since the change region needs to be further processed with the background image to segment the moving object, the value of T does not need to be accurately selected here, and the adaptive range is wide, here, 15 is taken. B isk+1(i, j) represents a region determined as a background after the difference, Mk+1(i, j) represents a region determined to be a motion change after the difference.
Step S203: detecting a moving object: and processing the current frame image, and detecting the moving object by only carrying out difference on the image in the change area and the background image.
After a change area and a non-change area in an image are distinguished, for a current frame image, only the difference between the image in the change area and a background image is carried out to detect a moving object.
After the current frame image and the background image are differentiated, different thresholds are selected according to specific application occasions and image quality to carry out threshold segmentation, and the differential image is changed into a binary image. After threshold segmentation, a target can have small holes and burrs due to noise interference, and the holes and burrs are removed by a method of corroding and expanding a binary image in image morphology.
Step S204: target edge identification: extracting edge points with sub-pixel precision by adopting a 5-degree polynomial to fit an edge gray scale change curve; and processing a series of images to obtain the vibration track of each pixel point on the characteristic edge of the steel wire rope in a certain time period.
Each frame of image corresponding to different moments in the object vibration process can be represented as an m-row n-column matrix, the difference of the characteristic edge space positions in two adjacent frames of images reflects the vibration process of an actual object, and the difference of the characteristic edge space positions is calibrated and converted, so that the vibration track of the object at the corresponding moment can be obtained. The vibration track and the vibration characteristics of the structure in a certain time period can be obtained by processing a series of images, wherein the identification and the positioning of the characteristic edge are the key for realizing the method.
The edge position can be determined by the mathematical characterization of the gray values at the edge. The gray scale change of the image edge of the steel wire rope is an impulse-type edge, a classical operator such as a sobel operator is adopted to extract a boundary point set of two adjacent gray scale images, and only a boundary point set of an image on the right side of the steel wire rope is extracted. Obtaining two one-dimensional arrays containing n whole pixel edge positions and recording the two one-dimensional arrays as K1(n)、K2(n), expressed as follows:
K1(n)＝[k1(1),k1(2),…,k1(n)]； (3)
K2(n)＝[k2(1),k2(2),…,k2(n)]； (4)
however, the pixels of the gray level image are all integers, and obviously, the displacement precision is only an integer pixel level. Next, edge points of sub-pixel accuracy are extracted by a polynomial fitting method.
Fitting an edge gray level change curve by using a 5-degree polynomial:
I(z)＝(c0+c1z+c2z2+…+c5z5)； (5)
in the formula (5), I represents the gray value at the z point, and z represents the sub-pixel edgeThe position of the rim, c0,…,c5Representing the fitting polynomial coefficients.
Let K1(i)＝j,j∈[1,m]Indicating the number of edge groups K of the state 1 image1And (n), each frame of image can be represented as a matrix with m rows and n columns, and then (i, j) is the position of the edge. And 6 pixel points (including the edge) near the edge point (i, j) along the column direction and corresponding gray values are extracted. By introducing it into a polynomial function for solving the parameter c to be determined0,…,c5The following formula:
parameter c obtained from the above formula0,…,c5Bringing back i (z) ═ c0+c1z+c2z2+…+c5z5) And solving the second derivative of the position of the edge of the sub-pixel by making the second derivative be zero, wherein Z belongs to (j-1, j + 1).
Performing column-by-column fitting and solving on the edge gray levels of the state 1 and state 2 images to obtain a series of sub-pixel edge positions, which are respectively marked as Z1(n) and Z2(n) is then Z1(n) and Z2The difference (n) is the obtained sub-pixel displacement, and is denoted as y (n). The method is used for processing each image of the steel wire rope in the vibration process frame by frame corresponding to the vibration displacement of the steel wire rope at the state 2, and the dynamic displacement of each point on the steel wire rope at each vibration moment can be obtained.
Processing each frame image of steel wire rope vibration by using a polynomial fitting gray intensity method to obtain the vibration track of each pixel point on the characteristic edge, and recording as y (t), wherein the displacement space sequence distributed along the length direction of the steel wire rope can be recorded as y1,y2,…,ynWhere each element is a function of time, then y (t) can be expressed in the form of a one-dimensional array, i.e.:
y(t)＝[y1(t),y2(t),…,yi(t),…,yn(t)],i＝1,2,…n； (7)
wherein y isi(t) represents a displacement function of a certain pixel point on the characteristic edge of the object vibration image with respect to time t.
Step S3: according to the vibration track of each pixel point on the characteristic edge of the steel wire rope, a displacement space sequence matrix distributed along the length direction of the steel wire rope is obtained, discrete Fourier transform is carried out on each column of the matrix, a frequency response function corresponding to each point on the steel wire rope is obtained, the inherent frequency is identified according to the peak value of the frequency response function, the modal vibration mode of the steel wire rope is obtained according to the amplitude of the response frequency of the frequency response function of each point, and finally the flexibility matrix of the steel wire rope to be tested is obtained.
Wherein, the displacement space sequence matrix that wire rope distributes along the rope length direction marks as y (t, i), and wherein i is 1,2, … n and expresses the pixel position, makes discrete fourier transform to each row of this matrix and can obtains the frequency response function H of each point on the corresponding wire rope, promptly:
H(ω,i)＝FFT(y(t,i))ω∈[0,1/△t]； (8)
wherein s images are collected within time t, and the corresponding time when each image is recorded is t1,t2,…,tsAnd the time interval is marked as Deltat, then: t is equal to t2-t1＝t3-t2＝…＝ts-ts-1(ii) a The data in the frequency response function H (ω, i) are complex numbers, which can be represented by amplitude and phase, or by real part and imaginary part, and are vectors of a complex plane, and the expression is:
H(ω)＝HR(ω)i+HI(ω)j； (9)
the amplitude-frequency characteristic curve is the relation curve between the amplitude and the frequency of the frequency response function, and the extreme point of the amplitude-frequency characteristic curve occurs at the natural frequency omega of the structure0Where, i.e. | H (ω)0) The absolute value is the maximum response amplitude of the displacement, so the natural frequency can be identified by the characteristic that the frequency response function has a peak value at the natural vibration frequency of the structure.
The mode shape can be obtained by combining the amplitudes of the response frequencies of the response spectrum of each point, and the amplitude of the response spectrum of each point with the maximum amplitude of the response spectrum is normalized, that is, when the frequency is the ith natural frequency, the amplitude of the frequency response function can be approximately expressed as:
in the formula phiri、Mi、ζiAre all constants determined by the ith order mode, phi1i……φNiRepresenting the i-order individual mode matrix. Therefore, the frequency response function is in direct proportion to the ith order mode of the structure, and the ratio of the peak value of the amplitude-frequency characteristic curve of each frequency response function at the mode frequency of a certain order can be approximately used as the mode shape of the order by neglecting the influence of the residual modes.
The mine hoist rope is actually a linear viscous, elastomeric, rather than rigid, body. We can simulate the properties of steel wire ropes using the maxwell model (c.maxwell model). The model is composed of a spring (an elastic model (H)) and a damper (a viscosity model (N)) in series, the structural schematic diagram is shown in FIG. 2, and the model symbols are expressed as: M-H-N. The stiffness and compliance matrix of a structure can be derived from its modal parameters, i.e. natural frequency and mode shape, knowing the system mass matrix M, the natural frequency ωi(i is 1 to n), the mode matrix phi, the stiffness matrix K and the compliance matrix F are respectively:
wherein Λ is a frequency matrix,
n is the degree of freedom.
The mode matrix Φ satisfies the regularization condition: phiTMΦ＝IΦTKΦ＝Λ。
As can be seen from equation (11), the stiffness matrix is inversely proportional to the square of the natural frequency, and therefore, in order to obtain an accurate stiffness matrix estimate, all modal parameters, or at least the higher order, must be measured. However, as can be seen from equation (12), the compliance matrix is inversely proportional to the square of the natural frequency, which means that the compliance matrix converges rapidly as the natural frequency increases, and thus an accurate estimate of the compliance matrix can be obtained from some modal parameters of lower order.
When a steel wire rope structure is cracked or locally damaged, a rigidity matrix of the steel wire rope structure is reduced, and the rigidity matrix is reduced due to the fact that a relation FK (where I is a unit matrix) exists between rigidity and flexibility, the flexibility matrix is increased definitely, if the flexibility is used as a sensitive parameter for representing structural damage, the physical significance of the steel wire rope structure is clear, and the steel wire rope structure is convenient to implement. Based on the characteristics of the compliance matrix, the change of the compliance matrix is used as a sensitive parameter for representing structural damage to diagnose the damage condition.
After the structure is damaged, (F)A+△F)(KA+. DELTA K) ═ I; the above formula is developed and finished to obtain: -FE△K＝(FE-FA)KA(ii) a Let E be-FEΔ K, then E ═ FE-FA)KA. Wherein,
Δ F, Δ K are the variation of the compliance matrix and stiffness matrix, respectively, I is the identity matrix, FA，FECompliance matrices, K, before and after breakage of the structure, respectivelyAAs a stiffness matrix before structural failure, phiiTo satisfy the ith order mode, ω, of the regularization conditioniIs the ith order frequency, and n is the degree of freedom.
If only the freedom degree of the perturbation direction is considered, the order frequency and the mode shape are changed from E to FEDelta K may result in a matrix El. Is provided witheiIs a matrix ElCorresponding row element d ofijThe maximum of the absolute values, i.e.:ei＝max|dijand | i and j are the degrees of freedom of the perturbation direction.
Thus, from the above analysis, it can be seen that the compliance matrix of a structure can be derived from its modal parameters, i.e., natural frequency and mode shape. Through the processing calculation of the pictures, after the natural frequency and the vibration mode of the steel wire rope are obtained, the flexibility matrix of the steel wire rope can be obtained.
Step S4: setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; and obtaining the damage position and the damage degree of the steel wire rope to be detected by utilizing the BP neural network after the sample training according to the flexibility matrix of the steel wire rope to be detected.
The BP neural network is composed of a three-layer network structure, the first layer is an input layer, the second layer is a hidden layer, the last layer is an output layer, when the output of signals does not reach the preset learning purpose, training is carried out through reverse propagation, when reverse transmission is carried out, error signals return in sequence along the path of the initial neurons, the weight of connection between the neurons is continuously modified when each layer returns, and finally the reverse propagation is completed. The BP neural network algorithm flowchart is shown in fig. 3, p represents a sample, and t represents the number of iterations. The upper case P is the number of samples and the lower case P is the representation of the P-th sample. First, let t equal to 1, calculate the first iteration. And assigning an initial value 1 to P, if the lower case P is less than the total sample number of the upper cases P, automatically adding 1 to the lower case P, and continuously circulating until the lower case P is not less than the upper cases P. And if the iteration times t do not reach the standard, adding one to the iteration times t, calculating the second iteration, and continuously and circularly performing the iterative calculation until the standard is reached, and stopping the calculation. The ultimate goal of training is to keep the error signal within a specified reasonable range based on repeated iterations of the forward and backward propagation processes. In this embodiment, when the sample is set to train the BP neural network, a Sigmoid function is selected as an excitation function, and an error signal back propagation algorithm is selected as a training algorithm. Suppose that the transmission function and the output of the jth neuron at the L-1 layer of the p-th training sample are respectively
And
then:
in formula (13), E is the overall error;
is the weight;
is a delta error.
In the formula (14), m represents the mth neuron, and the log-tangent function is a functional form of Sigmoid function, and the basic form is
Weight value
The formula of the updating process is as follows:
in the formula (15), t represents the number of iterations,
which is indicative of the delta error,
represents the output of the jth neuron at the lth layer for the pth training sample,
represents the L-th layer weight of the t-th iteration, mu represents that the instantaneous constant is 0.8, eta represents that the learning factor is 0.1,
representing a random noise term.
In addition, the training samples are obtained as follows: dividing each steel wire rope into 8 units, respectively reducing the rigidity by 25% and 40% according to different wire breakage conditions, thus obtaining 16 degrees of breakage conditions, calculating the vibration frequency and vibration mode of the 16 breakage steel wire ropes, taking the first two-order frequency and vibration mode to obtain 16 groups of values, carrying out GSL transformation (namely an orthogonal space method, being capable of correcting the localization phenomenon of data points point by point and enabling the data points to be uniformly distributed in a data space), and taking the 16 groups of values as input parameters of a neural network. The learning factor η is determined to be 0.1.
The final output results of the BP neural network are shown in table 1.
TABLE 1BP neural network Final output
Step S5: and taking the damage position unit judged by the BP neural network as an ROI area to be extracted from the whole image and storing the damage position unit into a new steel wire rope damage initial image.
Step S6: according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; and then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge the surface defects based on the texture features.
And extracting the characteristics which can reflect the texture properties of the defects and are relatively stable from the preprocessed image, and using the characteristics as a basis for identifying the defects. The selection of the defect features of the present embodiment mainly depends on the feature analysis and experimental analysis results of the defects. The criterion for selecting the features is that the same feature values are most obviously different, the features are ensured to have larger mutual independence, and finally smoothness and entropy are selected for analysis.
Wherein the smoothness is calculated by the formula:
R＝1-1/(1+σ2)； (16)
in equation (16), σ represents a standard deviation of the image, and the calculation formula is:
l represents the gray level, ziRepresenting the gray value of each pixel point in the image area, m represents the average gray level,
p(zi) Representing the gray value z of each pixel point in the image areaiThe probability of (c). The roughness of the texture of an image can be described using smoothness R, with low smoothness textures being less variable in gray scale than high smoothness textures, i.e. the lower the smoothness, the flatter the image.
The formula for calculating the entropy is:
in the formula, p (z)i) The gray value of each pixel point in the image area is ziThe probability of (d); l is a gray scale. The randomness of the image texture can be described by using the entropy e, and in the steel wire rope image, the entropy e of the steel wire rope with a defect on the surface is larger than the entropy value of the steel wire rope with a defect, which represents that the randomness of the steel wire rope with the defect is larger.
In addition, in view of a certain correlation among the features of the steel wire rope image, in order to reduce the complexity of the recognition algorithm, the embodiment utilizes a linear classifier in the determination, and the form of the discriminant function of the linear classifier and the decision criterion are shown as follows:
g(x)＝WTX+w0； (18)
wherein
n is the dimension of the feature space.
If it is not
If g (x) is 0, the classification is rejected.
The state of the wire rope is discriminated using the above-described texture features. The characteristic threshold value is obtained by experimental analysis and characteristic analysis results. In the determination process, in order to increase the accuracy of the detection result, the image of the steel wire rope is further divided into smaller parts, and the smoothness value and the entropy value of each part are extracted and compared with the classification threshold value. FIG. 4 is a schematic diagram of a linear classifier, wherein T is1，T2Classification thresholds for smoothness and entropy, respectively. When the smoothness R > T1 and the entropy e > T2 of the image, the surface is judged to be defective, and when the smoothness R < T1 and the entropy e < T2 of the image, the surface is judged to be intact.
After the judgment is finished, the detection result and the damage picture are classified and filed finally, so that the detection result and the damage picture can be conveniently checked and rechecked by workers.
In addition, as shown in fig. 5, the invention also provides a machine vision-based dynamic flaw detection device for the mining steel wire rope, which comprises an image acquisition device 1, a controller 2, a memory 3 and a display 4; the image acquisition equipment 1 is used for acquiring a video signal of the steel wire rope 5; the controller 2 is configured to execute the following program:
extracting a moving target from the video signal to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope;
obtaining a displacement space sequence matrix of the steel wire rope distributed along the rope length direction according to the vibration track, performing discrete Fourier transform on each column of the matrix to obtain a frequency response function corresponding to each point on the steel wire rope, identifying inherent frequency according to the peak value of the frequency response function, obtaining the modal vibration mode of the steel wire rope according to the amplitude of the response frequency of the frequency response function of each point, and finally obtaining the flexibility matrix of the steel wire rope;
setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; according to the flexibility matrix of the steel wire rope to be detected, obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after sample training;
taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image;
according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge surface defects based on texture features;
the memory 3 is used for storing damaged steel wire rope pictures;
the display device 4 is used for displaying the damaged steel wire rope picture and marking the damage position and the damage degree.
The invention can carry out video photography on the state of the steel wire rope in real time in the normal production process and carry out background segmentation on the steel wire rope, the image processing method can provide the most complete characteristic data and extract a moving target, the position is accurate, the speed is high, and the problem of target solving failure caused by the brightness and gray level change of background color when a single-frame image is used for solving the edge to search for the target can be solved. The method combines damage identification and image detection of a flexibility matrix, and solves the problem of image detection errors caused by the fact that spots formed by oil stains attached to the surface of a steel wire rope and other industrial impurity oil stains fall within the range of extracted strand images by using an image detection technology only by utilizing the influence of stress change of the steel wire rope caused by damage on flexibility. Meanwhile, complex equipment is not needed, the operation is simple, and the cost is low.
Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.
Claims (8)
1. A mining steel wire rope dynamic flaw detection method based on machine vision is characterized by comprising the following steps:
step S1: erecting a camera in front of the steel wire rope, calibrating the actual distance represented by a unit pixel of the camera, acquiring distance calibration parameters, enabling the steel wire rope to move at a constant speed, and continuously photographing and recording the steel wire rope in dynamic motion through the camera to obtain a video signal;
step S2: processing the video signal, and extracting a moving target to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope;
step S3: according to the vibration track of each pixel point on the characteristic edge of the steel wire rope, obtaining a displacement space sequence matrix distributed along the length direction of the steel wire rope, performing discrete Fourier transform on each column of the matrix to obtain a result as a frequency response function corresponding to each point on the steel wire rope, identifying inherent frequency according to the peak value of the frequency response function, obtaining the modal vibration mode of the steel wire rope according to the amplitude value of the response frequency of the frequency response function of each point, and finally obtaining the flexibility matrix of the steel wire rope to be tested; wherein, the modal shape is the combination of the amplitude of the response frequency corresponding to the frequency response function of each pixel point, and the computational formula of the compliance matrix is as follows: f ═ phi Λ-1ΦTWherein F represents a compliance matrix, phi represents a mode shape matrix, and Lambda represents a natural frequency matrix;
step S4: setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; according to the flexibility matrix of the steel wire rope to be detected, obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after sample training;
step S5: taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image;
step S6: according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; and then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge the surface defects based on the texture features.
2. The mining steel wire rope dynamic flaw detection method based on machine vision according to claim 1, wherein in the step S2, the step of processing the video signal is as follows:
step S201: establishing a background model: establishing a background model by using a method of averaging multiple frames of images;
step S202: detection of the change area: carrying out difference processing on two continuous frames of images in a video sequence to determine a background area and a motion change area;
step S203: detecting a moving object: processing the current frame image, and detecting a moving object by only carrying out difference on the image in the change area and the background image;
step S204: target edge identification: extracting edge points with sub-pixel precision by adopting a 5-degree polynomial to fit an edge gray scale change curve; and processing a series of images to obtain the vibration track of each pixel point on the characteristic edge of the steel wire rope in a certain time period.
3. The mining steel wire rope dynamic flaw detection method based on machine vision according to claim 1, characterized in that, in step S4, when a training sample of a sample training BP neural network is set, each steel wire rope is divided into 8 units, the rigidity is assumed to be respectively reduced by 25% and 40% according to different wire breakage conditions, 16 damage conditions are obtained, natural frequencies and modal vibration patterns of 16 kinds of damaged steel wire rope vibrations are calculated, the first two-order frequencies and vibration patterns are taken to obtain 16 groups of values, and the values are subjected to GSL transformation to be used as the training sample of the neural network.
4. The mining steel wire rope dynamic flaw detection method based on machine vision according to claim 1, characterized in that in step S4, when a sample training BP neural network is set, a Sigmoid function is selected as an excitation function, an error signal back propagation algorithm is selected as a training algorithm, and a weight updating process formula is as follows:
wherein the first term represents the gradient of the error mean, the second term represents the transient term, the third term represents the random noise term, t represents the number of iterations,
represents the lth level weight for the tth iteration,
which is indicative of the delta error,
represents the output of the jth neuron of the lth layer of the kth training sample, η represents a learning factor, μ represents a transient constant,
representing a random noise term.
5. The mining steel wire rope dynamic flaw detection method based on machine vision according to claim 1, characterized in that in the step S6, a calculation formula of smoothness R is as follows: r is 1-1/(1+ sigma)2) (ii) a The formula for the calculation of entropy e is:
in the formula, σ represents a standard deviation of an image, and the calculation formula is:
l represents the gray level, ziRepresenting the gray value of each pixel point in the image area, m represents the average gray level,
p(zi) Representing the gray value z of each pixel point in the image areaiThe probability of (c).
6. The mining steel wire rope dynamic flaw detection method based on machine vision according to claim 1, characterized in that in step S6, after high-pass filtering and imadjust-based gray scale transformation are performed on an initial steel wire rope damage image, the image is further divided into smaller parts, a smoothness value and an entropy value of each part are extracted, and each part is classified and judged through a linear classifier.
7. The machine vision-based mining steel wire rope dynamic flaw detection method of claim 1, wherein in the step S6, the linear classifier comprises a smoothness classification threshold T1 and an entropy classification threshold T2, when the smoothness R of the image is greater than T1 and the entropy e is greater than T2, the surface is judged to be defective, and when the smoothness R of the image is less than T1 and the entropy e is less than T2, the surface is judged to be intact.
8. A mining steel wire rope dynamic flaw detection device based on machine vision is characterized by comprising image acquisition equipment, a controller, a memory and a display;
the image acquisition equipment is used for acquiring a video signal of the steel wire rope;
the controller is configured to execute the following program:
extracting a moving target from the video signal to obtain a vibration track of each pixel point on the characteristic edge of the steel wire rope;
obtaining the distribution of the wire ropes along the length direction of the ropes according to the vibration trajectoryDisplacing the spatial sequence matrix, performing discrete Fourier transform on each column of the matrix to obtain a result as a frequency response function corresponding to each point on the steel wire rope, identifying inherent frequency according to the peak value of the frequency response function, obtaining the modal shape of the steel wire rope according to the amplitude of the response frequency of the frequency response function of each point, and finally obtaining the flexibility matrix of the steel wire rope; wherein, the modal shape is the combination of the amplitude of the response frequency corresponding to the frequency response function of each pixel point, and the computational formula of the compliance matrix is as follows: f ═ phi Λ-1ΦTWherein F represents a compliance matrix, phi represents a mode shape matrix, and Lambda represents a natural frequency matrix;
setting a BP (back propagation) neural network, setting input parameters of the BP neural network as a flexibility matrix of the steel wire ropes, setting output quantity as a structure damage unit and a damage degree quantization value, dividing each steel wire rope into a plurality of units, setting a plurality of rigidity reduction values for each unit, obtaining flexibility matrixes of various damaged steel wire ropes, and taking the flexibility matrixes as training samples to train; according to the flexibility matrix of the steel wire rope to be detected, obtaining the damage position and the damage degree of the steel wire rope to be detected by using the BP neural network after sample training;
taking the damage position unit judged by the BP neural network as an ROI area, extracting from the whole image and storing into a new steel wire rope damage initial image;
according to a Retinex principle, firstly, performing high-pass filtering on an initial steel wire rope damage image, and then performing grey level transformation based on imadjust; then, extracting smoothness R and entropy e of the processed image, and utilizing a linear classifier to judge surface defects based on texture features;
the memory is used for storing damaged steel wire rope pictures;
the display equipment is used for displaying the damaged steel wire rope picture and marking the damage position and the damage degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810432434.7A CN108956614B (en) | 2018-05-08 | 2018-05-08 | Mining steel wire rope dynamic flaw detection method and device based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810432434.7A CN108956614B (en) | 2018-05-08 | 2018-05-08 | Mining steel wire rope dynamic flaw detection method and device based on machine vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108956614A CN108956614A (en) | 2018-12-07 |
| CN108956614B true CN108956614B (en) | 2020-12-29 |
Family
ID=64499692
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810432434.7A Active CN108956614B (en) | 2018-05-08 | 2018-05-08 | Mining steel wire rope dynamic flaw detection method and device based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108956614B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109613117B (en) * | 2018-12-19 | 2021-11-09 | 广州广电计量检测股份有限公司 | Method and device for obtaining performance parameters of vibration flaw detector |
| CN109859170B (en) * | 2019-01-04 | 2023-04-18 | 中国矿业大学 | LBP (local binary pattern) feature-based intelligent monitoring method and system for surface damage of steel wire rope |
| CN110617873B (en) * | 2019-04-26 | 2022-01-14 | 深圳市豪视智能科技有限公司 | Method for detecting vibration of cable and related product |
| CN110517266B (en) * | 2019-04-26 | 2022-06-24 | 深圳市豪视智能科技有限公司 | Rope vibration detection method and related device |
| CN111324949B (en) * | 2020-02-10 | 2022-09-20 | 大连理工大学 | Engineering structure flexibility recognition method considering noise influence |
| CN113567460A (en) * | 2020-04-29 | 2021-10-29 | 中国石油化工股份有限公司 | Automatic detection device and detection method for broken filaments in drafting process based on image recognition |
| CN111796710B (en) * | 2020-06-02 | 2023-05-23 | 南京信息工程大学 | Method for reproducing image contour features on touch screen |
| CN112270658A (en) * | 2020-07-13 | 2021-01-26 | 安徽机电职业技术学院 | Elevator steel wire rope detection method based on machine vision |
| CN112085787B (en) * | 2020-07-20 | 2024-04-23 | 中国矿业大学 | Method for measuring space vibration of hoisting steel wire rope based on monocular vision |
| CN111862083B (en) * | 2020-07-31 | 2023-09-19 | 中国矿业大学 | Visual-electromagnetic detection-based steel wire rope state comprehensive monitoring system and method |
| CN111983973B (en) * | 2020-08-11 | 2021-08-06 | 重庆见芒信息技术咨询服务有限公司 | Cast workpiece edge deburring processing method based on template matching |
| CN112307932B (en) * | 2020-10-27 | 2023-02-17 | 上海交通大学 | Parameterized full-field visual vibration modal decomposition method |
| CN114541197B (en) * | 2021-12-31 | 2024-04-05 | 上海天链轨道交通检测技术有限公司 | Dynamic measurement method, device and system for track settlement and offset |
| CN114445404A (en) * | 2022-04-06 | 2022-05-06 | 湖南大学 | Automatic structural vibration response identification method and system based on sub-pixel edge detection |
| CN115147450B (en) * | 2022-09-05 | 2023-02-03 | 中印云端(深圳)科技有限公司 | Moving target detection method and detection device based on motion frame difference image |
| CN118654733B (en) * | 2024-08-19 | 2024-10-25 | 山东济宁运河煤矿有限责任公司 | Anchor cable winding and tensioning safety monitoring system and method |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20020065637A1 (en) * | 2000-11-30 | 2002-05-30 | Thompson John S. | Method and apparatus for simulating the measurement of a part without using a physical measurement system |
| CN100593716C (en) * | 2007-11-09 | 2010-03-10 | 无锡东望科技有限公司 | On-line detecting method of machine vision system for printed calico flaw |
| CN104280406A (en) * | 2014-09-16 | 2015-01-14 | 中国科学院广州能源研究所 | Machine vision system for detecting surface defects of copper part |
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107618533A (en) * | 2017-09-29 | 2018-01-23 | 兰州交通大学 | A kind of machine vision detection device and method of the discrete defect of Rail Surface |
| CN107764839B (en) * | 2017-10-12 | 2020-05-05 | 中南大学 | Machine vision-based steel wire rope surface defect online detection method and device |
-
2018
- 2018-05-08 CN CN201810432434.7A patent/CN108956614B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN108956614A (en) | 2018-12-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108956614B (en) | Mining steel wire rope dynamic flaw detection method and device based on machine vision | |
| CN109870461B (en) | Electronic components quality detection system | |
| CN114937055B (en) | Image self-adaptive segmentation method and system based on artificial intelligence | |
| KR102256181B1 (en) | Method of inspecting and evaluating coating state of steel structure and system for the same | |
| WO2021000524A1 (en) | Hole protection cap detection method and apparatus, computer device and storage medium | |
| CN111507976B (en) | Defect detection method and system based on multi-angle imaging | |
| CN109472822A (en) | Dimension of object measurement method based on depth image processing | |
| CN108242054A (en) | A kind of steel plate defect detection method, device, equipment and server | |
| CN105654121A (en) | Complex jacquard fabric defect detection method based on deep learning | |
| CN107369155A (en) | A kind of cloth surface defect detection method and its system based on machine vision | |
| CN109444169A (en) | A kind of bearing defect detection method and system | |
| CN114066857A (en) | Infrared image quality evaluation method and device, electronic equipment and readable storage medium | |
| CN103389310B (en) | Online sub-pixel optical component damage detection method based on radiation calibration | |
| CN111539927B (en) | Detection method of automobile plastic assembly fastening buckle missing detection device | |
| CN104680541A (en) | Remote sensing image quality evaluation method based on phase congruency | |
| CN111415339A (en) | Image defect detection method for complex texture industrial product | |
| CN109584206B (en) | Method for synthesizing training sample of neural network in part surface flaw detection | |
| CN112508950B (en) | Anomaly detection method and device | |
| CN111238927A (en) | Fatigue durability evaluation method and device, electronic equipment and computer readable medium | |
| CN113554645B (en) | Industrial anomaly detection method and device based on WGAN | |
| Choudhary et al. | Determination of rate of degradation of iron plates due to rust using image processing | |
| KR102189951B1 (en) | System and method for inspection of ship painting condition using image analysis | |
| CN115855961B (en) | Distribution box fault detection method used in operation | |
| CN116758149A (en) | Bridge structure displacement detection method based on unmanned aerial vehicle system | |
| CN114677428B (en) | Power transmission line icing thickness detection method based on unmanned aerial vehicle image processing |
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 |