Background
The steel wire rope core conveying belt is used in a large number in coal mine production, faults occur due to various reasons in use, if the faults cannot be detected and processed in time, safety accidents such as belt breakage or longitudinal tearing can be caused, and safety production is seriously influenced[1-3]. Therefore, each coal mine sets up a corresponding steel wire rope core conveying belt detection system.
At present, researchers have proposed effective methods such as eddy current detection, magnetic induction detection, shock wave detection, X-ray detection and the like besides manual detection for detection of steel wire rope core conveying belts, and have obtained certain application[4-9]. The X-ray detection method can acquire and display the internal steel wire rope core image of the steel wire rope core conveying belt in real time, can detect faults by using an image processing technology, and is simple to operate and high in detection speed, so that the detection device based on the method is valued by researchers and popular with conveying belt users. The inventor develops a nondestructive testing system of a steel wire rope core conveying belt based on X-ray in earlier stage, and researches an automatic detection method of the tensile failure of a joint of the steel wire rope core conveying belt[9,10]. In the use of the detection system, the user urgently requires that the faults except the joint stretching fault can be automatically detected on line, such as the faults of the conveyer belt breakage, the disconnection of a single steel wire rope core, the intersection of the ends of the steel wire rope cores and the like.
The X-ray image of the steel wire rope core conveying belt has strong periodic texture characteristics, and the texture is distributed in a regular stripe shape. In the aspect of regularity analysis of texture detection, although there are several effective methods, most of them are very sensitive to noise, the accuracy of detection results is not high, and in addition, the computation amount is often large, and therefore, the method is not suitable for on-line rapid detection. The invention can realize the fault detection of the steel wire rope core conveying belt on line, automatically and accurately.
Reference documents:
[1] mine steel rope core belt conveyor real-time working condition monitoring and fault diagnosis technology [ J ] coal science report, 2005, 30 (2): 245-250.
[2] Analysis of the cause of longitudinal tearing of a mining conveyor belt and prevention thereof [ J ] coal mine machinery, 2007, 28 (10): 182-183.
[3] Longitudinal tear monitoring method research [ J ] of university of mineral china, 2002, 31 (1): 49-52.
[4]Harrison A.A new technique for measuring loss of adhesion in conveyor belt splices[J].Australian J.Coal Mining Technology and Research,1983,(4):27-34.
[5]Sukhorukov V.Steel-cord conveyor belt NDT[J].International Journal of Materials&ProductTechnology,2006,27(3/4):236-246.
[6]Alport M,Basson J F,Padayachee T.Digital magnetic imaging of steel cord belts[J].BulkSolids Handling,2008,28(3):182-185.
[7] Design and implementation of a longitudinal anti-tear system of a heavy conveyor belt [ J ] mining machinery, 2006, 34 (6): 93-95.
[8]He Zhi-qiang,Wang Bai-yan,Gao Yu-lin.Research of algorithm for judgment of joint elongationof conveyer Belt[J].Journal of China University of Mining&Technology,1999,9(1):51-54.
[9]Li Xian-Guo,Miao Chang-Yun,Zhang Yan.An algorithm for selecting and stitching theconveyer belt joint images based on X-ray[A].2010International Conference on IntelligentComputation Technology and Automation[C].IEEE Computer Society,2010.vol.1,474-477.
[10] The method for automatically detecting the tensile failure of the joint of the steel wire rope core conveying belt by X-ray imaging comprises the following steps: china, 201010146881.X [ P ].2010-4-15.
Disclosure of Invention
The invention aims to provide an X-ray imaging on-line detection method for faults of a steel wire rope core conveying belt, which solves the problems that the current X-ray-based fault detection of the steel wire rope core conveying belt is poor in real-time performance, low in accuracy rate, mainly dependent on manual analysis and the like.
The technical scheme adopted by the invention is divided into two stages, namely a training stage and a detection stage.
The specific operation steps in the training phase are as follows:
step 1, reading a training image without fault information.
And 2, normalizing the input image, and enabling the pixel values of the image to be in the range of (0, 1) through linear transformation.
Step 3, normalizing the normalized image, and subtracting the pixel mean value of the normalized image
All the pixel values are at
Within the range.
Step 4, line scanning, calculating LRB (Light regulated Band, bright ruled Band, which measures bright defects by using regularity of texture image) and DRB (Dark regulated Band, which measures Dark defects by using regularity of texture image) parameter values of each pixel point of each line, wherein when the LRB parameter values are calculated on a certain line of the image, the definition formula is as follows
When calculating DRB parameter value, the formula is
Wherein the moving average
And standard deviation of
Is expressed as
Where n is an integer representing the dimension of the repeated elements in the input image, i.e. the texture period, x
ijIs the gray value of pixel points in the ith row and the jth column after the image X with the size of p multiplied by q is normalized, and r is more than or equal to 0
1≤r
n≤q-1,r
n=r
1+n-1,r
1∈[0,q-n],r
n∈[n-1,q-1]And is and
and 5, scanning lines and calculating the mean value of LRB parameter values of each line. For example, for a training image of size p × q, the mean of the LRB parameter values for row i is
Wherein
The LRB parameter values of the (i, j) pixel points obtained by calculation in the row direction are used. And calculating the mean value of the DRB values of each row in the same way.
Step 6, line scanning, calculating two thresholds of LRB parameter
And
and two thresholds for DRB parameters
And
during calculation, the maximum deviation amount of the mean value of the LRB parameter values and the DRB parameter values corresponding to the positive deviation and the negative deviation of the LRB parameter values and the DRB parameter values of the training images is used as the reference of a threshold value, and an adjusting parameter is added for self-adaptive adjustment and used as the threshold value. For example, for a training image of size p × q, the LRB parameter values for row i are biased forward from the maximum value of the mean of the LRB parameter values for that row
The forward deviation of the LRB parameter values obtained by all the training rows is as the maximum value
In practice, the maximum value of the forward deviation of the LRB parameter value used is
<math>
<mrow>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<msub>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
</msub>
<mo>×</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>α</mi>
<mo>)</mo>
</mrow>
</mrow>
</math>
Wherein alpha is an adjusting parameter, alpha is more than or equal to 0 and is defaulted to be 1, and the maximum value of the negative deviation of the LRB parameter value is
<math>
<mrow>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<msub>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
</msub>
<mo>×</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>α</mi>
<mo>)</mo>
</mrow>
</mrow>
</math>
For obtaining DRB parameter values in the same wayTwo thresholds, i.e. maximum of the DRB parameter value in the forward direction
And maximum value of negative deviation of DRB parameter value
In the training process, the mean value of the LRB and DRB parameter values is independently calculated for each row, and the maximum value of the mean value of the corresponding row with positive deviation and negative deviation is calculated respectively, so that the method is equivalent to training with all rows respectively, which is equivalent to increasing the number of training samples and reducing the accidental influence of training sample selection.
The specific operation steps in the detection stage are as follows:
step 1, reading an image to be detected.
And 2, normalizing the image read in the step 1, and enabling each pixel value of the image to be in a (0, 1) range through linear transformation.
Step 3, normalizing the normalized image obtained in the step 2, and subtracting the pixel mean value of the normalized image
All the pixel values are at
Within the range.
In particular, the pixel mean subtracted here
Is the normalized pixel mean value of the image to be detected
So as to adapt to different changes of the image to be detected and keep the image to be detected and the training image independent in operation.
And 4, scanning the normalized image obtained in the step 3, and calculating LRB and DRB parameter values of each pixel point in each line.
And 5, calculating the mean value of the LRB and DRB parameter values of each row according to the result obtained in the step 4. Because the difference between the mean values of the LRB parameter values and the DRB parameter values in the X-ray image of the non-joint area of the steel wire rope core conveying belt is not large, and the size of the fault area is much smaller than that of the whole image, in order to reduce the influence of the LRB or DRB parameter values of the fault area on the line where the fault is positioned on the mean value of the line, the mean value of the mean values of the LRB and DRB parameter values of all lines of the detected image is used as the mean value for actually judging all lines, and for the image to be detected with the size of p multiplied by q, the mean value for actually judging and using the LRB is used as the mean
And mean value used for DRB actual discrimination
Are respectively calculated as
<math>
<mrow>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>Σ</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>p</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>i</mi>
</msubsup>
</mrow>
<mi>p</mi>
</mfrac>
</mrow>
</math>
<math>
<mrow>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>Σ</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>p</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>i</mi>
</msubsup>
</mrow>
<mi>p</mi>
</mfrac>
</mrow>
</math>
Wherein,
is the mean of the LRB parameter values of row i;
is the mean of the DRB parameter values of row i.
Step 6, using the result of step 5 and two thresholds of LRB parameter value obtained in training phase
And two thresholds for DRB parameter values
Line scanning, performing threshold segmentation on the original image, if the (i, j) pixel point
<math>
<mrow>
<msubsup>
<mi>L</mi>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>∉</mo>
<mfenced open='[' close=']'>
<mtable>
<mtr>
<mtd>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>,</mo>
</mtd>
<mtd>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</math> Or
<math>
<mrow>
<msubsup>
<mi>D</mi>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>∉</mo>
<mfenced open='[' close=']'>
<mtable>
<mtr>
<mtd>
<msubsup>
<mi>d</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>,</mo>
</mtd>
<mtd>
<msubsup>
<mi>d</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
</mrow>
</math> Making the gray value of the pixel point be white, otherwise setting the gray value to be black, wherein
In lineThe LRB parameter values of the (i, j) pixel points obtained by calculation in the direction,
and calculating the DRB parameter value of the (i, j) pixel point in the row direction.
And 7, counting white pixel points according to the result of thresholding division in the step 6, and when the number of the white pixel points exceeds the set number, considering the image as an image containing fault information.
And 8, reporting the detection result according to the judgment of the step 7.
The invention has the advantages that the practical application shows that the invention can automatically and accurately realize the on-line detection of the faults of the steel wire rope core conveyer belt breakage, the single steel wire rope core breakage, the steel wire rope core end crossing and the like, and has good use value in industrial application. The technology of the invention has the following advantages:
(1) the method has the advantages of high automation degree, simple use and few parameters needing to be set, and the detection result is not very sensitive to the value of n (n is the texture period of the input image).
(2) In the training process, each row and each column are processed independently, which is equivalent to increase the number of training samples, reduces the accidental influence of the selection of the training samples, and has high accuracy of fault detection.
(3) The method can detect fine faults, has high image contrast, can well highlight the fault area, and the detected fault area is closer to the appearance of the fault, thereby being beneficial to further segmentation and fault identification of the fault area.
(4) The method has small calculation amount and is suitable for the on-line detection of the fault of the steel wire rope core conveying belt.
(5) The texture regularity analysis-based method can also be applied to the online detection of other articles with regular texture characteristics, such as fabrics, wallpaper, ceramics and the like.
Detailed Description
The following detailed description of the embodiments of the invention refers to the accompanying drawings and examples.
The method provided by the invention is divided into two stages when being implemented specifically, namely, firstly, a threshold value required for discrimination is acquired through a training stage, and an operation flow chart of the on-line detection training stage of the fault of the steel wire rope core conveying belt is shown as an attached figure 1; then, the detection of the image to be detected is realized in the detection stage, and the operation flow chart of the on-line detection stage of the fault of the steel wire rope core conveying belt is shown in the attached figure 2.
The steel wire rope cores in the steel wire rope core conveying belt are longitudinally arranged according to a certain distance, and the X-ray image has strong texture characteristics, namely the gray scale of the image is regularly changed, and the texture is regularly distributed in a stripe shape (vertical stripe shape). In order to reduce the complexity of implementation and improve the execution efficiency, the fault on-line detection of the steel wire rope core conveying belt only needs to be carried out in the row direction in the actual operation.
The method is used for detecting three typical faults of common conveyor belt breakage of the steel wire rope core conveyor belt, single steel wire rope core breakage, steel wire rope core end intersection and the like. The detection database comprises 2577X-ray images of four steel wire rope core conveyor belts with large differences of four coal mines, wherein the X-ray images comprise 8 conveyor belt fracture images (wherein a certain conveyor belt fracture image is shown in a figure 3 (a)), 33 single steel wire rope core fracture images (wherein a certain single steel wire rope core fracture image is shown in a figure 3 (b)), and 22 steel wire rope core end crossing images (wherein a certain steel wire rope core end crossing image is shown in a figure 3 (c)). In order to enhance the display effect, the image shown in fig. 3 is a partial image cut from the original complete image and containing the fault region with a size of 160 × 160 pixels. And when each conveying belt is detected, taking the first image without fault information as a training image. n is 12, alpha is a default value of 1, and the statistical threshold of the white pixel point is 5.
The operation flow of the training phase is as follows:
step 1, reading a first training image without fault information in a database of a current conveying belt.
And 2, normalizing the input image, and enabling the pixel values of the image to be in the range of (0, 1) through linear transformation.
Step 3, normalizing the normalized image, and subtracting the pixel mean value of the normalized image
All the pixel values are at
Within the range.
Step 4, line scanning, calculating LRB (Light regulated Band, bright ruled Band, which measures bright defects by using regularity of texture image) and DRB (Dark regulated Band, which measures Dark defects by using regularity of texture image) parameter values of each pixel point of each line, wherein when the LRB parameter values are calculated on a certain line of the image, the definition formula is as follows
When calculating DRB parameter value, the formula is
Wherein the moving average
And standard deviation of
Is expressed as
Where n is an integer representing the dimension of the repeated elements in the input image, i.e. the texture period, x
ijIs the gray value of pixel points in the ith row and the jth column after the image X with the size of p multiplied by q is normalized, and r is more than or equal to 0
1≤r
n≤q-1,r
n=r
1+n-1,r
1∈[0,q-n],r
n∈[n-1,q-1]And is and
and 5, scanning lines and calculating the mean value of LRB parameter values of each line. For example, for a training image of size p × q, the mean of the LRB parameter values for row i is
Wherein
The LRB parameter values of the (i, j) pixel points obtained by calculation in the row direction are used. And calculating the mean value of the DRB values of each row in the same way.
Step 6, line scanning, calculating two thresholds of LRB parameter
And
and two thresholds for DRB parameters
And
for a training image of size p × q, the LRB parameter values for row i deviate positively from the maximum value of the mean of the LRB parameter values for that row
The forward deviation of the LRB parameter values obtained by all the training rows is as the maximum value
In practice, the maximum value of the forward deviation of the LRB parameter value used is
<math>
<mrow>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<msub>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
</msub>
<mo>×</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>α</mi>
<mo>)</mo>
</mrow>
</mrow>
</math>
Wherein, alpha is an adjusting parameter, alpha is 1, and the maximum value of the negative deviation of the LRB parameter value is
<math>
<mrow>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<msub>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
</msub>
<mo>×</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>+</mo>
<mi>α</mi>
<mo>)</mo>
</mrow>
</mrow>
</math>
Two threshold values of DRB parameter value, namely the maximum value of the positive deviation of DRB parameter value
And maximum value of negative deviation of DRB parameter value
Four thresholds are established by the training phase
And then, the detection stage can be entered, and the fault on-line detection of the steel wire rope core conveying belt is realized.
The operation flow of the detection stage is as follows:
step 1, reading an image to be detected, wherein the image to be detected may be a fault image or a non-fault image.
And 2, normalizing the image read in the step 1, and enabling each pixel value of the image to be in a (0, 1) range through linear transformation.
Step 3, normalizing the normalized image obtained in the step 2, and subtracting the pixel mean value of the normalized image
All the pixel values are at
Within the range.
And 4, scanning the normalized image obtained in the step 3, and calculating LRB and DRB parameter values of each pixel point in each line.
And 5, calculating the mean value of the LRB and DRB parameter values of each row according to the result obtained in the step 4. For the image to be detected with the size of p multiplied by q, the average value used by LRB actual discrimination
And mean value used for DRB actual discrimination
Are respectively calculated as
<math>
<mrow>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>Σ</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>p</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>i</mi>
</msubsup>
</mrow>
<mi>p</mi>
</mfrac>
</mrow>
</math>
<math>
<mrow>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>=</mo>
<mfrac>
<mrow>
<munderover>
<mi>Σ</mi>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>0</mn>
</mrow>
<mrow>
<mi>p</mi>
<mo>-</mo>
<mn>1</mn>
</mrow>
</munderover>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>i</mi>
</msubsup>
</mrow>
<mi>p</mi>
</mfrac>
</mrow>
</math>
Wherein,
is the mean of the LRB parameter values of row i;
is the mean of the DRB parameter values of row i.
Step 6, using the result of step 5 and two thresholds of LRB parameter value obtained in training phase
And two thresholds for DRB parameter values
Line scanning, performing threshold segmentation on the original image, if the (i, j) pixel point
<math>
<mrow>
<msubsup>
<mi>L</mi>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>u</mi>
<mrow>
<mi>L</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>∉</mo>
<mfenced open='[' close=']'>
<mtable>
<mtr>
<mtd>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>,</mo>
</mtd>
<mtd>
<msubsup>
<mi>l</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</math> Or
<math>
<mrow>
<msubsup>
<mi>D</mi>
<mi>r</mi>
<mrow>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
</mrow>
</msubsup>
<mo>-</mo>
<msubsup>
<mi>u</mi>
<mrow>
<mi>D</mi>
<mo>,</mo>
<mi>r</mi>
</mrow>
<mi>use</mi>
</msubsup>
<mo>∉</mo>
<mfenced open='[' close=']'>
<mtable>
<mtr>
<mtd>
<msubsup>
<mi>d</mi>
<mrow>
<mi>max</mi>
<mo>-</mo>
</mrow>
<mi>use</mi>
</msubsup>
<mo>,</mo>
</mtd>
<mtd>
<msubsup>
<mi>d</mi>
<mrow>
<mi>max</mi>
<mo>+</mo>
</mrow>
<mi>use</mi>
</msubsup>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
</mrow>
</math> Making the gray value of the pixel point be white, otherwise setting the gray value to be black, wherein
For the LRB parameter values of the (i, j) pixel points calculated in the row direction,
and calculating the DRB parameter value of the (i, j) pixel point in the row direction.
And 7, counting white pixel points according to the result of thresholding and dividing in the step 6, and when the number of the white pixel points exceeds a set threshold value, determining that the image is an image containing fault information.
And 8, reporting the detection result according to the judgment of the step 7.
Fig. 4 is a detection result image of the fault image of the steel wire rope core conveyer belt shown in fig. 3 (in order to enhance the display effect, fig. 4 is a local image cut from the result image and corresponding to the fault area shown in fig. 3 and having a size of 160 × 160 pixels), from which it can be seen that the method of the present invention can detect a fine fault (as shown in fig. 3(b), that is, a single steel wire rope core disconnection fault caused by corrosion or manufacturing process, etc., because the disconnection degree is low, the gray value has no large change, and the position thereof is close to the edge of the image, it is generally difficult for human eyes to identify it, but the method of the present invention can still correctly identify it), the image contrast is high, and the fault area can be well highlighted, and the detected fault area is closer to the appearance of the fault, thereby being beneficial to further segmentation and fault identification of the fault area.
For four conveyer belt images of the detection database, the three types of faults contained in the images are successfully detected, and the accuracy rate is 100%.
The detection stage program provided by the invention is written by C # language, and runs on a PC with Intel Core 2Duo CPU T58702.0GHz main frequency and 1GB memory, the detection time of an X-ray image of a 600X 960 steel wire Core conveying belt is 630ms, which is less than the acquisition time of 750ms, so the method is suitable for online detection. The method can also be applied to the online detection of other articles with regular texture characteristics, such as fabrics, wallpaper, ceramics and the like.
Practical application shows that the invention can realize the on-line, automatic and accurate detection of faults of the steel wire rope core conveyer belt breakage, single steel wire rope core breakage, steel wire rope core end crossing and the like, and has good use value in industrial application.