A method for detecting the tension of lifting wire rope based on visual vibration frequency recognition

Abstract

The present invention discloses a method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition, which processes the vibration image of the lifting wire rope collected by an orthogonal binocular camera, and calculates the tension of the lifting wire rope by the natural vibration frequency of the lifting wire rope, thereby achieving real-time detection of the tension of the lifting wire rope. The present invention improves the anti-interference performance of the algorithm by means of inclined rectangular fitting, and improves the detection accuracy of the algorithm by means of an absolute mean data bad point cleaning method based on differential theory, thereby achieving good detection effect.

Description

Technical Field

[0001] The invention relates to the field of lifting wire rope tension detection based on visual vibration frequency recognition, and in particular to a lifting wire rope tension detection method based on visual vibration frequency recognition. Background Art

[0002] During the operation of the mine hoisting system, the wire rope will inevitably vibrate, which will cause the tension of the wire rope to change. In severe cases, it will cause fatigue deformation or even breakage of the wire rope, posing a serious threat to the safe operation of the mine hoisting system. Therefore, how to accurately and efficiently detect the tension of the wire rope is of great significance for preventing mine accidents and is a scientific problem that needs to be solved urgently in the field of coal mine production.

[0003] However, the mine hoisting structure is complex and the tension detection of the hoisting system is difficult. The existing technology mostly uses sensors to detect the tension of the wire rope. These methods have some limitations, such as the equipment is difficult to install and the sensor will affect the operation of the wire rope to be tested, which is not conducive to the long-term detection of the wire rope tension.

[0004] Machine vision technology provides a new method to solve this problem, that is, to analyze the vibration image of the wire rope to be tested by computer to obtain its vibration parameters, and then calculate the tension value of the wire rope. The existing detection algorithm based on machine vision has poor anti-interference ability and is more sensitive to abnormal phenomena such as residual images and shadows; on the other hand, the existing method for cleaning bad points of wire rope vibration data also has limitations and cannot correctly screen out the hysteresis error in the corrected sample data, affecting the accuracy of the final detection result. Summary of the invention

[0005] The present invention aims to solve the above-mentioned problems and provides a lifting wire rope tension detection method based on visual vibration frequency recognition, which overcomes the problems brought about by the above-mentioned traditional detection methods, such as the equipment is difficult to install and is easy to affect the wire rope to be tested, thereby improving the efficiency and safety and reliability of detection, while improving the anti-interference ability of the wire rope image recognition algorithm and the accuracy of tension detection.

[0006] In order to achieve the above object, the present invention adopts the following technical solution:

[0007] The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition comprises the following steps:

[0008] Step A, collecting the vibration image of the lifting wire rope in real time, preprocessing the image, and obtaining a black and white image with good segmentation effect of the wire rope and the background;

[0009] Step B, extracting edge features and performing contour fitting based on the inclined rectangle method on the black and white image segmentation result obtained in step A, the specific steps are as follows:

[0010] B1, performing edge detection on the lifting wire rope target image obtained in step A;

[0011] B2, performs tilted rectangle fitting on the detected edge to eliminate the residual image error caused by the fast movement of the wire rope or the image acquisition equipment not meeting the use requirements; it also includes:

[0012] Step C, performing feature screening and geometric centroid positioning on the inclined rectangle fitted in step B to determine the time domain vibration data of the wire rope;

[0013] Step D, further processing the time domain vibration data of the wire rope obtained in step C, using the absolute mean method based on differential theory to filter the data bad points detected under abnormal working conditions to improve the accuracy of the data, specifically including the following sub-steps:

[0014] D1, determine the original array a to be processed o , and calculate its difference array a d ;

[0015] D2, define the filtering threshold T as the basis for filtering bad pixels;

[0016] D3, identify and correct bad data points;

[0017] Step E: Perform Fourier transform on the cleaned data to obtain its natural vibration frequency, and calculate the tension of the wire rope using the wire rope tension-vibration frequency formula.

[0018] Furthermore, the abnormal operating conditions include strong winds, lighting failure and human interference.

[0019] Furthermore, step A specifically includes the following sub-steps:

[0020] A1, select the ROI area based on the active area of ​​the wire rope in the image to reduce the total calculation amount of the algorithm;

[0021] A2, grayscale conversion of ROI area;

[0022] A3, performing adaptive threshold segmentation on the ROI grayscale image obtained in step A2;

[0023] A4, performing a Gaussian filter operation on the black and white image obtained in step A3 to reduce background interference;

[0024] A5, performing morphological processing on the filtering result obtained in step A4 to fill the internal holes in the wire rope target.

[0025] Furthermore, step C specifically includes the following sub-steps:

[0026] C1, defines three characteristic parameters of the inclined rectangle in step B2: compactness, vertex height, and aspect ratio;

[0027] C2, combining the clipping logic of the ROI area to define the confidence intervals of the three feature parameters in C1, and screen out the undesired target fitting rectangles outside the interval to eliminate interference caused by background shadows, oil stains or scratches;

[0028] C3, determining the geometric centroid of the target rectangle screened in step C2 as the position of the wire rope in the current frame, and then finally determining the time domain vibration data of the wire rope by collecting the centroid frame by frame.

[0029] Furthermore, the step B1 includes the following sub-steps:

[0030] B11, adding white edges with a width of two pixels to the upper and lower edges of the black and white image obtained in step A, so that there is a closed contour in the image that can completely cover all edges of the wire rope target;

[0031] B12, using the Canny edge operator to extract the edge of the target area of ​​the lifting wire rope;

[0032] Step B2 includes the following sub-steps:

[0033] B21, for the target edge extracted in step B1, use the inclined circumscribed rectangle method to fit the contour. The fitting principle of the inclined rectangle is: select the point with the largest y value in the edge of the wire rope as the lower vertex Point 0 of the circumscribed rectangle, use it as the vertex to make two rays L1 and L2 in the direction parallel to the x-axis, rotate L1 clockwise with point Point 0 as the center, and stop when L1 touches the edge of the wire rope to be measured. Use the intersection point Point 1 of the edge and L1 at this time as the second vertex of the rectangle, so as to determine the first side of the rectangle. Then make a perpendicular line to this side, and the angle between it and L2 is θ2, and then determine point Point3 in this direction, so that the distance from Point 0 to Point 3 is the diameter d1 of the wire rope, thereby completing the establishment of the inclined rectangle.

[0034] Furthermore, the specific steps of step C1 are as follows:

[0035] C11, calculate the aspect ratio F1 of the inclined rectangle obtained in step B2, the calculation formula is:

[0036]

[0037] In the formula, a is the number of pixels occupied by the length of the inclined rectangle, and b is the number of pixels occupied by the width of the inclined rectangle;

[0038] C12, calculate the compactness F2 of the inclined rectangle obtained in step B2, the calculation formula is:

[0039]

[0040] In the formula, S1 is the total number of pixels occupied by the tilted rectangle, and S2 is the total number of pixels in the ROI area;

[0041] C13, calculate the vertex height F3 of the inclined rectangle obtained in step B2, the calculation formula is:

[0042]

[0043] Where h1 is the ordinate value of the highest vertex of the inclined rectangle, and h2 is the number of pixels of the height of the ROI area;

[0044] The specific steps of step C2 are as follows:

[0045] C21, set the confidence interval Q1 of the aspect ratio F1 of the tilted rectangle to [1,5],

[0046] The confidence interval Q2 of the compactness F2 of the inclined rectangle is [0.05, 0.7], and the confidence interval Q3 of the vertex height F3 of the inclined rectangle is [0, 0.1]; if the characteristic parameters of the inclined fitting rectangle are outside the confidence interval, it will be excluded as an interfering contour;

[0047] The specific steps of step C3 are as follows:

[0048] C31, record the geometric centroid of the screened wire rope target contour, record the coordinate P1 of the centroid in the ROI area as the tracking basis, and convert it to the original image coordinate P2. The values ​​of the horizontal and vertical coordinates of the original image coordinate P2 and the values ​​of the horizontal and vertical coordinates of the centroid in the ROI area P1 are as follows:

[0049]

[0050]

[0051] In the formula, k1 is the ordinate of the upper left corner when intercepting the ROI area;

[0052] x P1 is the abscissa value of the centroid P1,

[0053] x P2 is the horizontal coordinate value of the tracking point P2,

[0054] y P1 is the ordinate value of the centroid P1,

[0055] y P2 is the ordinate value of the tracking point P2;

[0056] C32, initialize an original data array a0, add the tracking points P2 obtained in step C31 according to the order of the image frames to the original data array a0 one by one, and complete the information recording of the tracking target within one cycle for subsequent data processing.

[0057] Furthermore, the specific steps of step D1 are as follows:

[0058] D11, add the wire rope vibration displacement detected in step C3 to the original data array a in the order of frame numbers. o Then reorder them into a according to the rule of small to large values. o [P o 1,P o 2,P o 3,…,P o n], where P o 1≤P o 2≤P o 3≤…≤P o n;

[0059] D12, then calculate array a o The difference array a d , which is calculated by the following formula:

[0060] P d i=P o (i+1)-P o i

[0061] Where P d i represents the difference array a d The value of the element with sequence number i in ,

[0062] P o i represents the original array a o The value of the element with sequence number i in ,

[0063] Next, calculate the difference array a d The absolute mean of The calculation method can be expressed by the following formula:

[0064]

[0065] Where n is the difference array a d The number of elements in

[0066] The specific steps of step D2 are as follows:

[0067] D21, the definition method of the filtering threshold T is:

[0068]

[0069] In the formula, k is an empirical coefficient, which takes a value between 4 and 5. Different k values ​​are selected for different wire rope vibration conditions according to the dispersion degree of different parts of the data.

[0070] The degree of data dispersion is determined by the standard deviation σ of the original data before it is sorted by size. To calculate the standard deviation σ, we need to first determine the array a o The residual array a e , a e The calculation method of each element value in is determined by the following formula:

[0071]

[0072] Where Pe i Represents the residual array a e The value of the element with sequence number i in;

[0073] Calculate the standard deviation σ as:

[0074]

[0075] Divide the original data before sorting into N parts evenly, and calculate the standard deviation of each part σ1, σ2, σ3, …, σ in the same way. N , and then determine the empirical coefficient k of each part as:

[0076]

[0077] The specific steps of step D3 are as follows:

[0078] D31, after determining the k value, check the difference array a according to the filtering threshold T d In the problem data, if array a d An element P d m satisfies:

[0079]

[0080] It is considered that the corresponding original data array a o P o m、P o (m+1) There is at least one bad data point between the two points;

[0081] Next, P o m、P o (m+1) two points are used to determine whether they are bad pixels. If point P o m satisfies:

[0082] |[P o mxD o (m-1)]-[Po (m-1)-P o (m-2)]|>σ

[0083] It is believed that P o m is the data bad point, for P o The judgment method for (m+1) is the same;

[0084] D32, correct the bad data points identified after judgment, the correction method is:

[0085] P o w=P o (w-1)+P d (w-1)

[0086] Where P o w is the data bad pixel,

[0087] P o (w-1) is array a o The value of the element with sequence number w-1 in the

[0088] P d (w-1) is array a d The value of the element with sequence number w-1;

[0089] If the local data standard deviation σ i If it is approximately 0, it is considered that this part of the data has a hysteresis error. The algorithm will alert this part of the data and remove it from the sample to avoid affecting the subsequent parameter calculation results.

[0090] The corrected array values ​​are restored to the original acquisition sequence, thus completing the data cleaning of the wire rope lateral vibration data.

[0091] Beneficial effects:

[0092] Compared with the prior art, the present invention adopts the above technical solution and has the following technical effects:

[0093] First point: The present invention can effectively reduce the detection error caused by shooting afterimage. This is because the present invention adopts the inclined rectangle fitting method to perform contour fitting and positioning of the wire rope, and completes the positioning of the wire rope without involving the wire rope afterimage generation area in the image.

[0094] Second point: The present invention avoids the false detection caused by oil stains, scratches in the detection background or background shadows caused by uneven lighting. This is because the present invention screens out the background interference targets in the image through the feature parameter screening of inclined rectangle fitting, and only retains the target contour of the wire rope to be tested.

[0095] The third point: The present invention has a better bad pixel filtering effect for wire rope vibration data, because the present invention adopts an absolute mean filtering method based on differential theory. The advantages of this method are: first, this method sets different filtering thresholds according to the fluctuation degree of wire rope vibration data, and has a better filtering effect for the characteristics of different fluctuation degrees of wire rope vibration data in different service stages; then, this method can filter the potential bad pixels in wire rope vibration data whose absolute values ​​differ greatly from those of its adjacent points, but whose absolute values ​​themselves are small, based on the idea of ​​differential method. Therefore, the bad pixel filtering method of the present invention can effectively identify the diffusion, concealment and hysteresis errors in wire rope vibration data, and has better applicability for the characteristics of non-normal distribution and strong randomness of wire rope vibration data. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] Figure 1 It is the algorithm flow chart of the present invention.

[0097] Figure 2 This is the basic principle diagram of the inclined rectangle fitting method.

[0098] Figure 3 It is a comparison chart of the detection effects of the traditional fitting algorithm and the inclined rectangle fitting method when the shooting afterimage situation occurs and when the shooting afterimage situation does not occur.

[0099] Figure 4 This is a picture of a wire rope with shadows and the image segmentation effect.

[0100] Figure 5 It is a diagram showing the false detection phenomenon of the traditional algorithm and the screening result of the algorithm of the present invention.

[0101] Figure 6 It is a flow chart of the absolute mean correction method based on difference theory. DETAILED DESCRIPTION

[0102] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings:

[0103] like Figure 1 As shown, the present invention provides a method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition, including four parts: image acquisition and preprocessing, target edge extraction and inclined rectangle fitting, feature screening and centroid positioning, data processing and result output. The specific implementation steps are as follows:

[0104] A1, select the ROI area based on the active area of ​​the wire rope in the image to reduce the total calculation amount of the algorithm;

[0105] A2, grayscale conversion of ROI area;

[0106] A3, performing adaptive threshold segmentation on the ROI grayscale image obtained in step A2;

[0107] A4, performing a Gaussian filter operation on the black and white image obtained in step A3 to reduce background interference;

[0108] A5, performing morphological processing on the filtering result obtained in step A4 to fill the internal holes in the wire rope target;

[0109] Step B, extracting edge features and performing contour fitting based on the inclined rectangle method on the image segmentation result obtained in step A, the specific steps are as follows:

[0110] B1, performing edge detection on the lifting wire rope target image obtained in step A;

[0111] B2, perform tilted rectangle fitting on the detected edge to eliminate the residual image error caused by the fast movement of the wire rope or the image acquisition equipment not meeting the use requirements;

[0112] Step C, performing feature screening and geometric centroid positioning on the inclined rectangle fitted in step B, specifically includes the following steps:

[0113] C1, defines three characteristic parameters of the inclined rectangle in step B2: compactness, vertex height, and aspect ratio;

[0114] C2, combining the clipping logic of the ROI area to define the confidence intervals of the three feature parameters in C1, and screen out the undesired target fitting rectangles outside the interval to eliminate interference caused by background shadows, oil stains or scratches;

[0115] C3, determining the geometric centroid of the target rectangle screened in step C2 as the position of the wire rope in the current frame, and then finally determining the lateral vibration displacement curve of the wire rope by collecting the centroid frame by frame.

[0116] Step D, performing data processing on the lateral vibration parameters of the steel wire rope obtained in step D to calculate the tension of the steel wire rope, specifically comprises the following steps:

[0117] Step D, further process the detected data, use the absolute mean method based on differential theory to filter out the bad points of the data detected under abnormal working conditions (strong wind, lighting failure, human interference, etc.), and improve the accuracy of the data. The process of this method is as follows: Figure 6 shown.

[0118] D1, determine the original array a to be processed o , and calculate its difference array a d ;

[0119] D2, define the filtering threshold T as the basis for filtering bad pixels;

[0120] D3, identify and correct bad data points;

[0121] Step E, further processing the cleaned data to obtain the wire rope tension, the specific steps are as follows:

[0122] E1, perform Fourier transform on the data to obtain the natural frequency of wire rope vibration;

[0123] E2, calculate the value of the tension that the wire rope bears based on its natural frequency of vibration.

[0124] The specific steps of step A2 are:

[0125] A21, calculate the gray value of the i-th pixel of the image:

[0126] Gray i =0.299R i +0.587G i +0.114B i

[0127] Where R i , G i , B i is the value of the red, green, and blue channels of the i-th pixel in the three-channel color image. Then, the grayscale value is assigned to this pixel and the entire image is traversed to complete the grayscale image conversion.

[0128] The specific steps of step A3 are:

[0129] A31. Calculate the grayscale histogram of the current image input image to determine a mean value M, and use the M value as the dividing line to divide the histogram into two parts, MA and MB, and define the number of pixels occupied by the MA part as PA and the number of pixels occupied by the MB part as PB, and define it through the inter-class variance:

[0130] ICV=PA·(MA-M) 2 +PB·(MB-M) 2

[0131] In the formula, ICV is the inter-class variance, and the threshold M that makes it reach the maximum value is the optimal threshold;

[0132] A32. Use the threshold M as the dividing line to divide the image into two parts: the target and the background.

[0133] The specific steps of step A4 are:

[0134] A41. Define the one-dimensional discrete zero-mean Gaussian filter function g(x) as:

[0135]

[0136] Where σ is the Gaussian distribution parameter, which determines the width of the Gaussian function. The present invention uses a two-dimensional discrete function to smooth the image. First, the image is convolved with the above-mentioned one-dimensional Gaussian function, and then convolved with a one-dimensional Gaussian function that is the same but perpendicular to the convolution result, to obtain a two-dimensional Gaussian function:

[0137]

[0138] A42. Obtain the weights at discrete points by discretizing the two-dimensional Gaussian function. For example, to obtain a filter template of size (2n+1)×(2n+1), a matrix A[2n+1,2n+1] is established with the pixel at the center of the template as the central element. The value of A[i,j] in the matrix is ​​determined by the following formula:

[0139]

[0140] Then the calculated values ​​of each point are normalized to obtain the weight V of each point. The calculation process is:

[0141]

[0142] After that, the weighted average of each pixel in the image can be performed to achieve a smoothing filtering effect.

[0143] The specific steps of step B1 are as follows:

[0144] B11, adding white edges with a width of two pixels to the upper and lower edges of the black and white image obtained in step A, so that there is a closed contour in the image that can completely cover all edges of the wire rope target;

[0145] B12, using the Canny edge operator to extract the edge of the target area of ​​the lifting wire rope, the present invention uses a relatively simple convolution operator in the Canny algorithm, which is expressed as follows:

[0146]

[0147] Where the first-order partial derivative matrices P and Q in the x-direction and y-direction, the gradient amplitude M and direction θ are expressed as:

[0148] P[i,j]=(f[i,j+1]-f[i,j]+f[i+1,j+1]-f[i+1,j]) / 2

[0149] Q[i,j]=(f[i,j]-f[i+1,j]+f[i,j+1]-f[i+1,j+1]) / 2

[0150]

[0151] θ[i,j]=tan -1 (Q[i,j] / P[i,j])

[0152] Further operations are performed by solving the matrix to obtain the boundary image of the lifting wire rope target.

[0153] The specific steps of step B2 are:

[0154] B21, for the target edge extracted in step B1, use the inclined circumscribed rectangle method to fit the contour. The fitting principle of the inclined rectangle is: select the point with the largest y value in the edge of the wire rope as the lower vertex Point 0 of the circumscribed rectangle, use it as the vertex to make two rays L1 and L2 in the direction parallel to the x-axis, rotate L1 clockwise with point Point 0 as the center, and stop when L1 touches the edge of the wire rope to be measured. Use the intersection point Point 1 of the edge and L1 at this time as the second vertex of the rectangle, so as to determine the first side of the rectangle. Then make a perpendicular line to this side, and the angle between it and L2 is θ2, and then determine point Point3 in this direction, so that the distance from Point 0 to Point 3 is the diameter d1 of the wire rope, thereby completing the establishment of the inclined rectangle.

[0155] The specific steps of step C1 are as follows:

[0156] C11, calculate the aspect ratio F1 of the inclined rectangle obtained in step B2, the calculation formula is:

[0157]

[0158] Where a is the number of pixels occupied by the length of the inclined rectangle, and b is the number of pixels occupied by the width of the inclined rectangle;

[0159] C12, calculate the compactness F2 of the inclined rectangle obtained in step B2, the calculation formula is:

[0160]

[0161] Where S1 is the total number of pixels occupied by the tilted rectangle, and S2 is the total number of pixels in the ROI area;

[0162] C13, calculate the vertex height F3 of the inclined rectangle obtained in step B2, the calculation formula is:

[0163]

[0164] Where h1 is the ordinate value of the highest vertex of the inclined rectangle, and S2 is the number of pixels of the height of the ROI area;

[0165] The specific steps of step C2 are as follows:

[0166] C21, set the confidence interval Q1 of F1 to [1,5], the confidence interval Q2 of F2 to [0.05,0.7], and the confidence interval Q3 of F3 to [0,0.1]; if the characteristic parameters of the inclined fitting rectangle are outside the confidence interval, it will be excluded as an interfering contour.

[0167] The specific steps of step C3 are as follows:

[0168] C31, record the geometric centroid of the screened wire rope target contour, record the coordinate P1 of the centroid in the ROI area as the tracking basis, and convert it to the original image coordinate P2. The values ​​of the horizontal and vertical coordinates of P2 and P1 are as follows:

[0169]

[0170]

[0171] Where k1 is the ordinate of the upper left corner when intercepting the ROI area; x P1 is the horizontal coordinate value of the centroid P1, x P2 is the horizontal coordinate value of the tracking point P2, y P1 is the ordinate value of the centroid P1, y P2 is the ordinate value of the tracking point P2;

[0172] C32, initialize an integer array a1, add the tracking points P2 obtained in step C31 according to the sequence of video frames to a1 one by one, complete the information recording of the tracking target for subsequent data processing.

[0173] The specific steps of step D1 are as follows:

[0174] D11, add the wire rope vibration displacement detected in C3 to the original data array a in the order of frame numbers. o Then reorder them into a according to the rule of small to large values. o [P o 1,P o 2,P o 3,…,P o n], where P o 1≤P o 2≤P o 3≤…≤P o n;

[0175] D12, then calculate array a o The difference array a d , which can be calculated by the following formula:

[0176] P d i=P o (i+1)-Po i

[0177] Where P d i represents the difference array a d The value of the element with sequence number i in P o i represents the original array a o Next, calculate the difference array a d The absolute mean of The calculation method can be expressed by the following formula:

[0178]

[0179] The specific steps of step D2 are as follows:

[0180] D21, the definition method of the filtering threshold T is:

[0181]

[0182] Where k is an empirical coefficient, which usually takes a value between 4 and 5. This method selects different k values ​​for different wire rope vibration conditions according to the degree of dispersion of different parts of the data.

[0183] The degree of data dispersion is determined by the standard deviation σ of the original data before it is sorted by size. To calculate the standard deviation σ, we need to first determine the array a o The residual array a e , a e The calculation method of each element value in can be determined by the following formula:

[0184]

[0185] Where Pe i Represents the residual array a e The value of the element with index i in .

[0186] Furthermore, the standard deviation σ can be calculated as:

[0187]

[0188] Divide the original data before sorting into N parts evenly, and calculate the standard deviation of each part σ1, σ2, σ3, …, σ in the same way. N , and then determine the empirical coefficient k of each part as:

[0189]

[0190] The specific steps of step D3 are as follows:

[0191] After D31 determines the k value, the difference array a can be checked according to the filter threshold T.d In the problem data, if array a d An element P d m satisfies:

[0192]

[0193] It is considered that the corresponding original data array a o P o m、P o There is at least one bad data point between the two points (m+1).

[0194] Next, we need to analyze P o m、P o (m+1) two points are used to determine whether they are bad pixels. o m satisfies:

[0195] |[P o mxD o (m-1)]-[P o (m-1)-P o (m-2)]|>σ

[0196] It is believed that P o m is a bad pixel. o The judgment method for (m+1) is similar.

[0197] D32, correct the bad data points identified after judgment. The correction method is:

[0198] P o w=P o (w-1)+P d (w-1)

[0199] Where P o w is the data bad point, through its original data array a o The previous value in the sequence and the difference array a d The correction is done by adding the corresponding elements in .

[0200] In addition, if the local data standard deviation σ i If it is approximately 0, it is considered that this part of the data has a hysteresis error. The algorithm will alert this part of the data and remove it from the sample to avoid affecting the subsequent parameter calculation results.

[0201] The corrected array values ​​are restored to the original acquisition sequence, thus completing the data cleaning of the wire rope lateral vibration data.

[0202] The specific steps of step E1 are as follows:

[0203] E11, perform Fourier transform on the time domain function y(t) of the transverse vibration of the wire rope:

[0204]

[0205] Where Y(ω) is the frequency domain signal function obtained after transformation, and the peak value of the function waveform is calculated to obtain the nth order vibration natural frequency ω of the wire rope n ;

[0206] The specific steps of step E2 are as follows:

[0207] E21, according to the natural frequency ω of the wire rope vibration obtained in step E11 n Calculate the tension:

[0208]

[0209] Where T s is the tension of the wire rope, ρ s is the wire rope density, l is the wire rope length. The wire rope tension test result can be obtained through this calculation formula.

[0210] like Figure 2 The figure shows the basic principle of the inclined rectangle fitting method. First, select the point with the largest y value on the edge of the wire rope as the lower vertex Point 0 of the circumscribed rectangle, and make two rays L1 and L2 with it as the vertex in the direction parallel to the x-axis. Rotate L1 clockwise with Point 0 as the center. Stop when L1 touches the edge of the wire rope to be measured. Take the intersection of the edge and L1 Point 1 as the second vertex of the rectangle, thus determining the first side of the rectangle. Then make a perpendicular line to this side, with an angle of θ2 with L2, and finally determine Point 3 in this direction, so that the distance from Point 0 to Point 3 is the diameter d1 of the wire rope, thus completing the establishment of the inclined rectangle.

[0211] When the afterimage phenomenon occurs during shooting, the computer will mistakenly regard the afterimage as part of the wire rope target, thus causing detection errors. However, since the afterimage must appear in the opposite direction of the wire rope's movement, the inclined rectangle fitting method can be used to determine the circumscribed rectangle of the wire rope contour without touching the area where the afterimage may occur, and then the wire rope to be tested is located by the geometric centroid of the rectangle to reduce the detection error caused by the afterimage phenomenon.

[0212] like Figure 3Shown is a comparison of the detection effects of the traditional fitting algorithm and the tilted rectangle fitting method when there is a shooting afterimage and when there is no shooting afterimage. The light color in the figure is the fitting effect of the traditional method, and the dark color is the fitting effect of the tilted rectangle method proposed in the present invention. Figure 1 is a comparison of the detection effects in the case of high resolution (1224×2448) and shooting afterimage; Figure 2 is a comparison of the detection effects in the case of high resolution (1224×2448) and no shooting afterimage; Figure 3 is a comparison of the detection effects in the case of low resolution (480×640) and shooting afterimage; Figure 4 is a comparison of the detection effects in the case of low resolution (480×640) and no shooting afterimage. Figure 3 It can be seen that under different resolutions, the positioning results of the two fitting methods are basically consistent when there is no afterimage phenomenon, but there is an obvious positioning deviation when there is afterimage phenomenon. This shows that the inclined rectangle fitting method plays an important role in suppressing the positioning deviation caused by the afterimage phenomenon.

[0213] like Figure 4 The image segmentation result is shown when there is background shadow interference in the captured image. Since the background shadow interference is obvious and has a large area, it cannot be filtered out by image preprocessing.

[0214] like Figure 5 The left side shows the false detection phenomenon of the traditional method under the interference of background shadows. This is because the traditional method will also locate the background interference as the wire rope; the right side of the figure shows the detection effect of the present invention after filtering with rectangular feature parameters, correctly filtering out the interference contour and retaining only the target contour.

[0215] The specific implementation modes of the present invention are described above in conjunction with the accompanying drawings, but these descriptions cannot be understood as limiting the scope of the present invention. Any changes based on the claims of the present invention are within the protection scope of the present invention.

Claims

1. A method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition comprises the following steps: Step A, collecting the vibration image of the lifting wire rope in real time, preprocessing the image, and obtaining a black and white image with good segmentation effect of the wire rope and the background; Step B, extracting edge features and performing contour fitting based on the inclined rectangle method on the black and white image segmentation result obtained in step A, the specific steps are as follows: B1, performing edge detection on the lifting wire rope target image obtained in step A5; B2, performing tilted rectangle fitting on the detected edge to eliminate the error of the shooting afterimage; characterized in that it also includes: Step C, performing feature screening and geometric centroid positioning on the inclined rectangle fitted in step B to determine the time domain vibration data of the wire rope; Step D, further processing the time domain vibration data of the wire rope obtained in step C, using the absolute mean method based on differential theory to filter the bad points of the data detected under abnormal working conditions, and improving the accuracy of the data, specifically includes the following sub-steps: D1, determine the original array a to be processed o , and calculate its difference array a d ; D2, define the filtering threshold T as the basis for filtering bad pixels; D3, identify and correct bad data points; Step E, performing Fourier transform on the cleaned data to obtain its natural vibration frequency, and calculating the tension of the wire rope in combination with the wire rope tension-vibration frequency formula; The step B1 comprises the following sub-steps: B11, adding white edges with a width of two pixels to the upper and lower edges of the black and white image obtained in step A, so that there is a closed contour in the image that can completely cover all edges of the wire rope target; B12, using the Canny edge operator to extract the edge of the target area of ​​the lifting wire rope; Step B2 includes the following sub-steps: B21, for the target edge extracted in step B1, the contour is fitted by means of an inclined circumscribed rectangle, specifically: select the point with the largest y value in the edge of the wire rope as the lower vertex Point 0 of the circumscribed rectangle, make two rays L1 and L2 with it as the vertex in the direction parallel to the x-axis, rotate L1 clockwise with point Point 0 as the center, stop when L1 touches the edge of the wire rope to be measured, and use the intersection point Point 1 of the edge and L1 at this time as the second vertex of the rectangle, thus determining the first side of the rectangle; then make a perpendicular line to this side, with an angle of θ2 with L2, and then determine point Point 3 in this direction, so that the distance from Point 0 to Point 3 is the diameter d1 of the wire rope, thereby completing the establishment of the inclined rectangle.

2. The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition according to claim 1 is characterized in that: The abnormal operating conditions include strong winds, lighting failure and human interference.

3. The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition according to claim 1 is characterized in that: Step A specifically includes the following sub-steps: A1, select the ROI area based on the active area of ​​the wire rope in the image to reduce the total calculation amount of the algorithm; A2, grayscale conversion of ROI area; A3, performing adaptive threshold segmentation on the ROI grayscale image obtained in step A2; A4, performing a Gaussian filter operation on the black and white image obtained in step A3 to reduce background interference; A5, performing morphological processing on the filtering result obtained in step A4 to fill the internal holes in the wire rope target.

4. The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition according to claim 1 is characterized in that: Step C specifically includes the following sub-steps: C1, defines three characteristic parameters of the inclined rectangle in step B2: compactness, vertex height, and aspect ratio; C2, combining the clipping logic of the ROI area to define the confidence intervals of the three feature parameters in C1, and screen out the undesired target fitting rectangles outside the interval to eliminate interference caused by background shadows, oil stains or scratches; C3, determining the geometric centroid of the target rectangle screened in step C2 as the position of the wire rope in the current frame, and then finally determining the time domain vibration data of the wire rope by collecting the centroid frame by frame.

5. The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition according to claim 4 is characterized in that: The specific steps of step C1 are as follows: C11, calculate the aspect ratio F1 of the inclined rectangle obtained in step B2, the calculation formula is: In the formula, a is the number of pixels occupied by the length of the inclined rectangle, and b is the number of pixels occupied by the width of the inclined rectangle; C12, calculate the compactness F2 of the inclined rectangle obtained in step B2, the calculation formula is: In the formula, S1 is the total number of pixels occupied by the tilted rectangle, and S2 is the total number of pixels in the ROI area; C13, calculate the vertex height F3 of the inclined rectangle obtained in step B2, the calculation formula is: Where h1 is the ordinate value of the highest vertex of the inclined rectangle, and h2 is the number of pixels of the height of the ROI area; The specific steps of step C2 are as follows: C21, set the confidence interval Q1 of the aspect ratio F1 of the tilted rectangle to [1,5], The confidence interval Q2 of the compactness F2 of the inclined rectangle is [0.05, 0.7], and the confidence interval Q3 of the vertex height F3 of the inclined rectangle is [0, 0.1]; if the characteristic parameters of the inclined fitting rectangle are outside the confidence interval, it will be excluded as an interfering contour; The specific steps of step C3 are as follows: C31, record the geometric centroid of the screened wire rope target contour, record the coordinate P1 of the centroid in the ROI area as the tracking basis, and convert it to the original image coordinate P2. The values ​​of the horizontal and vertical coordinates of the original image coordinate P2 and the values ​​of the horizontal and vertical coordinates of the centroid in the ROI area P1 are as follows: In the formula, k1 is the ordinate of the upper left corner when intercepting the ROI area; P1 is the horizontal coordinate value of the centroid P1, x P2 is the horizontal coordinate value of the tracking point P2, y P1 is the ordinate value of the centroid P1, y P2 is the ordinate value of the tracking point P2; C32, initialize an original data array a0, add the tracking points P2 obtained in step C31 according to the order of the image frames to the original data array a0 one by one, and complete the information recording of the tracking target within one cycle for subsequent data processing.

6. The method for detecting the tension of a lifting wire rope based on visual vibration frequency recognition according to claim 1, characterized in that: The specific steps of step D1 are as follows: D11, add the wire rope vibration displacement detected in step C3 to the original data array a in the order of frame numbers. o Then reorder them into a according to the rule of small to large values. o [P o 1,P o 2,P o 3,…,P o n], where P o 1≤P o 2≤P o 3≤…≤P o n; D12, then calculate array a o The difference array a d , which is calculated by the following formula: P d i=P o (i+1)-P o i Where P d i represents the difference array a d The value of the element with sequence number i in , P o i represents the original array a o The value of the element with sequence number i in , Next, calculate the difference array a d The absolute mean of The calculation method can be expressed by the following formula: Where n is the difference array a d The number of elements in The specific steps of step D2 are as follows: D21, the definition method of the filtering threshold T is: In the formula, k is an empirical coefficient, which takes a value between 4 and 5. Different k values ​​are selected for different wire rope vibration conditions according to the dispersion degree of different parts of the data. The degree of data dispersion is determined by the standard deviation σ of the original data before it is sorted by size. To calculate the standard deviation σ, we need to first determine the array a o The residual array a e , a e The calculation method of each element value in is determined by the following formula: Where Pe i Represents the residual array a e The value of the element with sequence number i in; Calculate the standard deviation σ as: Divide the original data before sorting into N parts evenly, and calculate the standard deviation of each part σ1, σ2, σ3, …, σ in the same way. N , and then determine the empirical coefficient k of each part as: The specific steps of step D3 are as follows: D31, after determining the k value, check the difference array a according to the filtering threshold T d In the problem data, if array a d An element P d m satisfies: It is considered that the corresponding original data array a o P o m、P o (m+1) There is at least one bad data point between the two points; Next, P o m、P o (m+1) two points are used to determine whether they are bad pixels. If point P o m satisfies: |[P o mP o (m-1)]-[P o (m-1)-P o (m-2)]|>σ It is believed that P o m is the data bad point, for P o The judgment method for (m+1) is the same; D32, correct the bad data points identified after judgment, the correction method is: P o w=P o (w-1)+P d (w-1) Where P o w is the data bad pixel, P o (w-1) is array a o The value of the element with sequence number w-1 in the P d (w-1) is array a d The value of the element with sequence number w-1; If the local data standard deviation σ i If it is approximately 0, it is considered that this part of the data has a hysteresis error. The algorithm will alert this part of the data and remove it from the sample to avoid affecting the subsequent parameter calculation results. The corrected array values ​​are restored to the original acquisition sequence, thus completing the data cleaning of the wire rope lateral vibration data.

Images