CN108584352A - Rubber belt from deviating diagnostic system based on machine vision and method - Google Patents
Rubber belt from deviating diagnostic system based on machine vision and method Download PDFInfo
- Publication number
- CN108584352A CN108584352A CN201810623426.0A CN201810623426A CN108584352A CN 108584352 A CN108584352 A CN 108584352A CN 201810623426 A CN201810623426 A CN 201810623426A CN 108584352 A CN108584352 A CN 108584352A
- Authority
- CN
- China
- Prior art keywords
- adhesive tape
- scaling board
- picture
- belt
- scaling
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 239000002390 adhesive tape Substances 0.000 claims abstract description 142
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000012545 processing Methods 0.000 claims abstract description 10
- 238000012706 support-vector machine Methods 0.000 claims abstract description 10
- 238000003709 image segmentation Methods 0.000 claims abstract description 9
- 230000004438 eyesight Effects 0.000 claims abstract description 7
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 238000002405 diagnostic procedure Methods 0.000 claims description 4
- 238000003745 diagnosis Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 5
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000012512 characterization method Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 8
- 230000000694 effects Effects 0.000 description 3
- 239000003292 glue Substances 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000011897 real-time detection Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000007599 discharging Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G43/00—Control devices, e.g. for safety, warning or fault-correcting
- B65G43/02—Control devices, e.g. for safety, warning or fault-correcting detecting dangerous physical condition of load carriers, e.g. for interrupting the drive in the event of overheating
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G2203/00—Indexing code relating to control or detection of the articles or the load carriers during conveying
- B65G2203/02—Control or detection
- B65G2203/0266—Control or detection relating to the load carrier(s)
- B65G2203/0283—Position of the load carrier
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B65—CONVEYING; PACKING; STORING; HANDLING THIN OR FILAMENTARY MATERIAL
- B65G—TRANSPORT OR STORAGE DEVICES, e.g. CONVEYORS FOR LOADING OR TIPPING, SHOP CONVEYOR SYSTEMS OR PNEUMATIC TUBE CONVEYORS
- B65G2203/00—Indexing code relating to control or detection of the articles or the load carriers during conveying
- B65G2203/04—Detection means
- B65G2203/041—Camera
Landscapes
- Image Analysis (AREA)
Abstract
Rubber belt from deviating diagnostic system based on machine vision and method, system includes adhesive tape, the first scaling board, the second scaling board, high-definition camera and image processing module, first scaling board, the second scaling board are separately mounted to close to the both sides of adhesive tape lower surface, first scaling board, the size of the second scaling board and character are just the same, and there are apparent vision difference for the color of the first scaling board, the color of the second scaling board and adhesive tape.Method includes step:The scene picture run using high-definition camera real-time grasp shoot adhesive tape, feature samples are extracted with image segmentation, characterization extraction algorithm to the picture of candid photograph, feature samples are trained by support vector machines, are demarcated, complete the training study of tape movement scene, to judge adhesive tape whether sideslip, and further analysis adhesive tape deflection direction.Adhesive tape of the present invention itself is not required to do special designing, and scaling board is replaceable at any time, versatile, utmostly protects belt work life, accuracy of detection high.
Description
Technical field
The present invention relates to technical field of image processing, and in particular to a kind of rubber belt from deviating diagnostic system based on machine vision
And method.
Background technology
There is glue in operational process in the widely used grooved belt conveyor in the industries such as power station, coal mine and harbour
It is common failure with deviation phenomenon, in transmission process, is misaligned in case of discharging point or load imbalance, it will
Cause the adhesive tape of conveyer that random shift phenomenon occurs.When offset has exceeded every technic index, conveyor apparatus has
It is likely to occur adhesive tape derailing, the material in transport system is caused largely to come down in torrents, is worth expensive adhesive tape completeness damage, thus into
One step causes the disorderly closedown of a whole set of material-transporting system, device to be stopped work.
Rubber belt from deviating detection at present is mainly completed using large-scale control station unit matching, and system complex is of high cost, and traditional
Deviation switch detecting system be often unable to real-time judge sideslip state again, cause to rectify a deviation in time and pull adhesive tape, it is economical
It loses larger.The patent document of CN105197537B discloses a kind of rubber belt from deviating detecting system based on color detection and side
Method, detects rubber belt from deviating by the way of positioning colour band, color detecting sensor and programmable controller, but the detecting system
Middle positioning colour band is printed on conveyor belt, and during practical belt-conveying, widespread adoption is unable to without corresponding finished product,
Even if being positioned after long term wear using the printing of special facture technique but belt, colour band is unintelligible, and will also result in can not examine
It surveys or testing result is inaccurate.
Invention content
The technical problem to be solved by the present invention is to detect above shortcomings for existing rubber belt from deviating, provide one kind
Rubber belt from deviating diagnostic system based on machine vision and method are realized in conjunction with picture depth learning method currently popular to glue
Real-time detection with sideslip, adhesive tape itself are not required to do special designing, and scaling board is replaceable at any time, versatile, at low cost, can be with
Real-time diagnosis adhesive tape whether sideslip, utmostly protect belt work life, accuracy of detection is high.
Used technical solution is the present invention to solve above-mentioned technical problem:
Rubber belt from deviating diagnostic system based on machine vision, including adhesive tape, the first scaling board, the second scaling board, high definition are taken the photograph
Camera and image processing module, first scaling board, the second scaling board are separately mounted to close to the both sides of adhesive tape lower surface,
First scaling board, the size of the second scaling board and character are just the same, the first scaling board, the color of the second scaling board and adhesive tape
There are apparent vision difference for color;The high-definition camera is mounted on above adhesive tape, and positioned at the first scaling board, the second calibration
The front or behind of plate so that the first scaling board, the second scaling board and adhesive tape occupy the center of high-definition camera shooting picture;Institute
It states image processing module with high-definition camera to be connected, for detecting in real time between the first scaling board, the second scaling board and adhesive tape
Position relationship, to judge adhesive tape whether sideslip.
By said program, light compensating lamp is also set up beside the high-definition camera (convenient for rather dark or work at night).
By said program, first scaling board and the second scaling board are replaced with using a complete scaling board, the mark
Fixed board is symmetrically arranged below adhesive tape.
The diagnostic method of the present invention also provides a kind of rubber belt from deviating diagnostic system based on machine vision, including walk as follows
Suddenly:
S1, below adhesive tape, it is complete that size and shape is installed respectively close to the adhesive tape both sides of (as possible close to) adhesive tape lower surface
The color of complete the first the same scaling board, the second scaling board, the first scaling board, the color of the second scaling board and adhesive tape exists obviously
Vision difference;
S2, above adhesive tape, the first scaling board, the second scaling board front or behind high-definition camera is installed so that the
One scaling board, the second scaling board and adhesive tape occupy the center of high-definition camera shooting picture;
S3, the scene picture run using high-definition camera real-time grasp shoot adhesive tape, picture of the image processing module to candid photograph
With image segmentation, the color form parameter information of extraction algorithm the first scaling board of extraction and the second scaling board is characterized as special
Sample is levied, feature samples are trained by support vector machines, are demarcated, completes the training study of tape movement scene, in real time
Detect the position relationship between the first scaling board, the second scaling board and adhesive tape, to judge adhesive tape whether sideslip, and further
Analyze the direction of adhesive tape deflection.
By said program, the step S3 is specifically included:
S31, sample collection
According to tape movement scene, adhesive tape is collected in different loads, the high definition scene picture of adhesive tape operation;
S32, image segmentation
Based on Mean Shift algorithms (mean shift algorithm is a kind of feature space clustering algorithm) to the scene graph of candid photograph
Piece is split, and marks off three blocks, respectively the first scaling board area, taped region, the second scaling board area;
S33, extraction is characterized
Using pairwise geometric histograms come the retrieval of extraction and profile to image shape feature, the method based on histogram
Several homogeneous regions are regarded with a join domain;
The image shape feature in the first scaling board area, taped region, the second scaling board area is extracted, variable is defined as follows
As feature samples:
Sidebdl:First scaling board left margin;
Sidebdr:Second scaling board right margin;
Sidebel:Adhesive tape left margin;
Sideber:Adhesive tape right margin;
XAL-BEL:First scaling board left margin SidebdlTo adhesive tape left margin SidebelPixel;
XBEL-BER:Adhesive tape left margin SidebelTo adhesive tape right margin SideberPixel;
XBR-BER:Adhesive tape right margin SideberWith the second scaling board right margin SidebdrPixel;
PA-BELT:First scaling board area and taped region pixel ratio, calculation formula are as follows:PA-BELT=XAL-BEL/XBEL-BER;
PB-BELT:Second scaling board area and taped region pixel ratio, calculation formula are as follows:PB-BELT=XBR-BER/XBEL-BER;
By characterizing extraction algorithm, whether just the first scaling board area, the second scaling board area and taped region pixel ratio are confirmed
Often:
In the case of adhesive tape normal operation, PA-BELTAnd PB-BELTValue variation only it is related with adhesive tape load, adhesive tape bear
In the case that load is more constant, PA-BELTAnd PB-BELTValue variation be metastable;
In adhesive tape operational process, if adhesive tape shifts, PA-BELTAnd PB-BELTValue significant change can occur, judge
Adhesive tape operation is abnormal;
S34, support vector machines training
Before starting to do training with support vector machines, front is collected and treated the feature samples (field of tape movement
Scape picture) it is divided into two classes, one kind is effective picture, i.e. adhesive tape the first scaling board, adhesive tape, under different loads or no-load condition
Two scaling board three position relationships are deviateing the picture (the normal picture of running orbit) in threshold range, and have sticked criterion
Label;Another kind of is invalid picture, i.e., adhesive tape under different loads or no-load condition adhesive tape traffic direction be biased to the first scaling board or
Adhesive tape traffic direction is biased to the picture (the abnormal picture of running orbit) of the second scaling board, and sticks invalid tag;Then from
A part is taken out in this two classes picture respectively to be used for doing supporting vector machine model training, then another part is used for doing test set,
Equally include effective picture and invalid picture in test set, to verify the trained classification effect of supporting vector machine model
Fruit.
By said program, in the step S34, the characteristics of according to scene picture, the algorithm of supporting vector machine model training
The key factor of the Performance tuning of parameter is penalty factor and nuclear parameter r in SVM, and penalty factor is for determining sample data
The fiducial interval range of Learning machine is adjusted in subspace, optimal C is different in different data subspace, and nuclear parameter r
The implicit complexity for changing sample data Subspace Distribution of change.
By said program, first scaling board and the second scaling board use a complete scaling board, the scaling board left
The right side is arranged symmetrically in below adhesive tape.
Compared with prior art, beneficial effects of the present invention:
1, the present invention is by being arranged two scaling boards below adhesive tape, in conjunction with picture depth learning method currently popular,
Realize that the real-time detection to rubber belt from deviating, adhesive tape itself are not required to do special designing, scaling board is replaceable at any time, versatile, at
This is low, can with real-time diagnosis adhesive tape whether sideslip, utmostly protect belt work life;
It 2, can be very clear according to the first scaling board, the second scaling board and the taped region pixel ratio in feature samples data
Know that the direction of rubber belt from deviating, accuracy of detection are high.
Description of the drawings
Fig. 1 is that the present invention is based on the structural schematic diagrams of the rubber belt from deviating diagnostic system of machine vision;
Fig. 2 is the picture structure signal in the calibration area and taped region of the adhesive tape normal operation of high-definition camera of the present invention shooting
Figure;
Fig. 3 is the picture structure signal in calibration area and taped region that the adhesive tape of high-definition camera of the present invention shooting shifts
Figure;
The structural schematic diagram of first scaling board and taped region pixel ratio when Fig. 4 is adhesive tape normal operation of the embodiment of the present invention;
The structural schematic diagram of second scaling board and taped region pixel ratio when Fig. 5 is adhesive tape normal operation of the embodiment of the present invention;
Fig. 6 is the structural schematic diagram of the first scaling board and taped region pixel ratio when adhesive tape of the embodiment of the present invention shifts;
Fig. 7 is the structural schematic diagram of the second scaling board and taped region pixel ratio when adhesive tape of the embodiment of the present invention shifts;
Fig. 8 is the picture directory structure that the training of supporting vector machine model of the embodiment of the present invention preserves;
Fig. 9 is that preset parameter r, C influence schematic diagram as variable on SVM performances in SVM training of the embodiment of the present invention;
Figure 10 is that preset parameter C, r influence schematic diagram as variable on SVM performances in SVM training of the embodiment of the present invention;
Figure 11 is that parameter r, C equivalence situation influences schematic diagram to SVM performances in SVM training of the embodiment of the present invention;
In figure, the first scaling boards of 1-, the second scaling boards of 2-, 3- adhesive tapes, 4- high-definition cameras.
Specific implementation mode
Technical solution of the present invention is described in detail with reference to the accompanying drawings and examples.
Shown in referring to Fig.1, the present invention is based on the rubber belt from deviating diagnostic system of machine vision, including adhesive tape, the first scaling board,
Second scaling board, high-definition camera and image processing module, the first scaling board, the second scaling board are separately mounted to close to adhesive tape
The both sides (close to not contacting, facilitating observation, be also convenient for replacing) of lower surface, the first scaling board, the second scaling board size and property
Shape is just the same, and there are apparent vision difference, the first calibration for the color of the first scaling board, the color of the second scaling board and adhesive tape
Plate, the second scaling board color selected had with the article transported on adhesive tape and adhesive tape based on the first scaling board, the second scaling board
Bigger colour contrast is criterion;High-definition camera is mounted on above adhesive tape, and positioned at the first scaling board, second scaling board
Front or behind so that the first scaling board, the second scaling board and adhesive tape occupy the center of high-definition camera shooting picture;At image
Reason module is connected with high-definition camera, for detecting that the position between the first scaling board, the second scaling board and adhesive tape is closed in real time
System, to judge adhesive tape whether sideslip.
The present invention is based on the diagnostic methods of the rubber belt from deviating diagnostic system of machine vision, include the following steps:
S1, below adhesive tape 3, size and shape complete one is installed respectively close to 3 both sides of adhesive tape of 3 lower surface of adhesive tape as possible
The first scaling board 1, the second scaling board 2 of sample, the first scaling board 1, the second scaling board 2 color it is selected based on the first scaling board 1,
Second scaling board 2 has with the article transported on adhesive tape 3 and adhesive tape 3 subject to bigger colour contrast (apparent vision difference)
Then;
S2, above adhesive tape 3, the first scaling board 1, the second scaling board 2 front (dead astern) install high-definition camera
4 so that the first scaling board 1, the second scaling board 2 and adhesive tape 3 occupy the center that high-definition camera 4 shoots picture;It is rather dark or
When person works at night, 4 side of high-definition camera also sets up light compensating lamp;
S3, the scene picture run using 4 real-time grasp shoot adhesive tape 3 of high-definition camera, figure of the image processing module to candid photograph
The color form parameter information that piece extracts the first scaling board 1 and the second scaling board 2 with image segmentation, characterization extraction algorithm is made
It is characterized sample, feature samples are trained by support vector machines, are demarcated, completes the training study of 3 moving scene of adhesive tape,
Detect position relationship between the first scaling board 1, the second scaling board 2 and adhesive tape 3 in real time, to judge adhesive tape 3 whether sideslip,
And the direction of the deflection of adhesive tape 3 is further analyzed, it specifically includes:
S31, sample collection
According to 3 moving scene of adhesive tape, the high definition scene picture that adhesive tape 3 is run in different loads, adhesive tape 3 is collected,
The picture shot by high-definition camera 4 can significantly be told, and demarcate area and taped region, as shown in Figure 2;
S32, image segmentation
Based on Mean Shift algorithms (mean shift algorithm is a kind of feature space clustering algorithm) to the scene graph of candid photograph
Piece is split, and marks off three blocks, is the first scaling board area, taped region, the second scaling board area respectively;
Assuming that the image data of scene picture is { xij, i=1,2..., n, j=1,2...., m }, it is assumed that { kyK=1,
2 ... for the data vector collection on one group of image feature space, obtain the iteration mistake of the Mean Shift algorithms in image segmentation
Cheng Wei:
Wherein g (x) is the negative derivative of the core profile function of image feature space, it is assumed that M be in isolated area most
Small pixel number, then the image segmentation of Mean Shift algorithms be described as follows:
The first step:Select Epanechinov cores as kernel function;
Second step:For each point on image, its convergence point is calculated, z is denoted asij;
Third walks:In data set { zij, i=1,2..., n, j=1,2...., m } on carry out feature clustering, by coordinate space
Euclidean distance be less than hcAnd color space distance is less than hLData point gather for one kind, and remember that all feature class sets are:|cp
|P=1,2...q;
4th step:The class of j=1,2 ..., m, image data space are labeled as lij=p | xij∈cp};
5th step:Eliminate the class that element number is less than M;
S33, extraction is characterized
Using pairwise geometric histograms come the retrieval of extraction and profile to image shape feature, the method based on histogram
Several homogeneous regions are regarded with a join domain;Method based on histogram is suitable for describing in some given field
The global property of the pixel coordinate distribution in face, geometric histogram are defined as follows:
The function of given discrete function f and i-1 rank
In formula:f:X → υ, x ∈ χ, υ ∈ υ;
In order to be distinguished with original spatial histogram, replace these used here as geometric histogram (geogram)
The tuple that i-1 rank functions are constituted is κ rank geometric histograms Gk(ν) is defined as the tuple that k-1 rank functions are constituted, they are all
The derived function κ rank geometric histograms for being multiplied by function f are described as:
In formula:
<.>A function that integer part is obtained from given real number is indicated, in order to calculateIntroduce 00→1;
Zeroth order geometric histogram is equivalent to the perimeter in some feature distribution region, hereafter, is replaced using this title of homogeneous region
The distributed areas of given feature;
The image shape feature in the first scaling board area, taped region, the second scaling board area is extracted, variable is defined as follows
As feature samples (these variables embody the color form parameter information of the first scaling board 1 and the second scaling board 2):
Sidebdl:First scaling board left margin;
Sidebdr:Second scaling board right margin;
Sidebel:Adhesive tape left margin;
Sideber:Adhesive tape right margin;
XAL-BEL:First scaling board left margin SidebdlTo adhesive tape left margin SidebelPixel;
XBEL-BER:Adhesive tape left margin SidebelTo adhesive tape right margin SideberPixel;
XBR-BER:Adhesive tape right margin SideberWith the second scaling board right margin SidebdrPixel;
PA-BELT:First scaling board area and taped region pixel ratio, calculation formula are as follows:PA-BELT=XAL-BEL/XBEL-BER;
PB-BELT:Second scaling board area and taped region pixel ratio, calculation formula are as follows:PB-BELT=XBR-BER/XBEL-BER;
By characterizing extraction algorithm, whether just the first scaling board area, the second scaling board area and taped region pixel ratio are confirmed
Often:
In the case of 3 normal operation of adhesive tape, PA-BELTAnd PB-BELTValue variation only it is related with the load of adhesive tape 3, in glue
In the case that the load of band 3 is more constant, PA-BELTAnd PB-BELTValue variation be it is metastable, as shown in Figure 4, Figure 5;
In 3 operational process of adhesive tape, if adhesive tape 3 shifts, PA-BELTAnd PB-BELTValue significant change can occur, such as
Shown in Fig. 6, Fig. 7, transmission adhesive tape 3 is deviated to the second scaling board 2, it is abnormal to judge that adhesive tape 3 is run, and can basis
PA-BELTAnd PB-BELTThe very clear direction for knowing 3 sideslip of adhesive tape of curve graph;
S34, support vector machines training
Before starting to do training with support vector machines, front is collected and treated the feature samples (field that adhesive tape 3 moves
Scape picture) it is divided into two classes, one kind is effective picture, i.e. adhesive tape 3 first scaling board 1, adhesive tape under different loads or no-load condition
3,2 three's position relationship of the second scaling board is deviateing the picture (the normal picture of running orbit) in threshold range, and has sticked
Criterion label;Another kind of is invalid picture, i.e. 3 traffic direction of adhesive tape under different loads or no-load condition of adhesive tape 3 is biased to the first mark
Fixed board 1 or 3 traffic direction of adhesive tape are biased to the picture (the abnormal picture of running orbit) of the second scaling board 2, and stick no criterion
Label;Then it takes out a part respectively from this two classes picture to be used for doing supporting vector machine model training, then another part is used for
Test set is done, equally includes effective picture and invalid picture in this test set, to verify supporting vector machine model instruction
The classifying quality perfected.
Bibliographic structure as shown in Figure 8 is built up, effective picture and invalid picture is stored, separately includes for trained
Data and for test verification data, this process is exactly label application process.Picture after the good class of sample point is extracted
Characteristic set is just added in SVM training algorithms after feature extraction and is trained by the operation of feature.It is the characteristics of according to scene, right
The Performance tuning of SVM training algorithm parameters, it is found that penalty factor and r in SVM are to influence SVM from the analysis of SVM principles
The key factor of performance.The effect of parameter C is to determine the fiducial interval range that Learning machine is adjusted in data subspace, different numbers
It is different according to C optimal in subspace, and the change of nuclear parameter r actually impliedly changes mapping function to change sample
The complexity of data Subspace Distribution, the i.e. maximum VC dimensions of linear classification, also just determine that the minimum that linear classification reaches is missed
Difference.Following two figure is that the model trained according to fixed C and r certain Graph One factor therein goes out to what test sample was predicted
Properties curve reflects influences of the parameter C and r to SVM performances respectively.
From Fig. 9 it may be seen that in fixed r, C is as variable, after being trained to model, influences of the C to SVM performances
Situation.Clearly when C is increasing, after reaching 12 or more, assessment that model is either still integrated from discrimination, recall ratio
Score has all reached optimal, as C up increases again, curve then regional stability, even without variation.
Figure 10 is fixed penalty factor, and parameter r influences SVM performances, it may be seen that when r values are gradually increasing
SVM model performances are optimal when general 7 to 11 section, increase again then as r values, and the performance of SVM declines therewith,
Finally tend to be steady after 20.
It from Figure 11 it may be seen that in the case of C and r equivalences, gradually increases, the property of SVM in 14 to 22 section
It can be optimal, so the further value range for reducing SVM parameters C and r, can be that final value makes reference, when
Optimization picture treatment effect and be also particularly important to the optimization of feature extraction while so obtaining optimal C and r for training
's.
First scaling board 1 and the second scaling board 2 can also use a complete scaling board, the symmetrical cloth of the scaling board
It sets below adhesive tape 3.
Obviously, the above embodiment is merely an example for clearly illustrating the present invention, and is not to the present invention
The restriction of embodiment.For those of ordinary skill in the art, spirit under this invention is extended out aobvious and easy
The variation or variation seen are still in the protection scope of this invention.
Claims (6)
1. the Direction Deviation in Belt Conveyer diagnostic system based on machine vision, which is characterized in that including adhesive tape, the first calibration
Plate, the second scaling board, high-definition camera and image processing module, first scaling board, the second scaling board are separately mounted to
Both sides close to adhesive tape lower surface, the first scaling board, the size of the second scaling board and character are just the same, the first scaling board, the
There are apparent vision difference for the color of two scaling boards and the color of adhesive tape;The high-definition camera is mounted on above adhesive tape, and
Positioned at the front or behind of the first scaling board, the second scaling board so that the first scaling board, the second scaling board and adhesive tape occupy high definition
Video camera shoots the center of picture;Described image processing module is connected with high-definition camera, for detecting the first calibration in real time
Position relationship between plate, the second scaling board and adhesive tape, to judge adhesive tape whether sideslip.
2. the rubber belt from deviating diagnostic system according to claim 1 based on machine vision, which is characterized in that the high definition is taken the photograph
Light compensating lamp is also set up beside camera.
3. the rubber belt from deviating diagnostic system according to claim 1 based on machine vision, which is characterized in that first mark
Fixed board and the second scaling board are replaced with using a complete scaling board, which is symmetrically arranged below adhesive tape.
4. a kind of diagnosis side of rubber belt from deviating diagnostic system of 1~3 any one of them of the claims based on machine vision
Method, which is characterized in that include the following steps:
S1, below adhesive tape, size and shape duplicate the is respectively symmetrically installed close to the adhesive tape both sides of adhesive tape lower surface
One scaling board, the second scaling board, the first scaling board, the second scaling board color with the color of adhesive tape ontology, there are apparent visions
Difference;
S2, above adhesive tape, the first scaling board, the second scaling board front or behind install high-definition camera so that first mark
Fixed board, the second scaling board and adhesive tape occupy the center of high-definition camera shooting picture;
S3, the scene picture run using high-definition camera real-time grasp shoot adhesive tape, image processing module use the picture of candid photograph
Image segmentation characterizes the color form parameter information of extraction algorithm the first scaling board of extraction and the second scaling board as feature sample
This, is trained feature samples by support vector machines, demarcates, and the training study of tape movement scene is completed, in real time
Detect the position relationship between the first scaling board, the second scaling board and adhesive tape, to judge adhesive tape whether sideslip, and further
Analyze the direction of adhesive tape deflection.
5. the diagnostic method of the rubber belt from deviating diagnostic system according to claim 1 based on machine vision, which is characterized in that
The step S3 is specifically included:
S31, sample collection
According to tape movement scene, adhesive tape is collected in different loads, the high definition scene picture of adhesive tape operation;
S32, image segmentation
The scene picture of candid photograph is split based on Mean Shift algorithms, marks off three blocks, the respectively first calibration
Plate area, taped region, the second scaling board area;
S33, extraction is characterized
Using pairwise geometric histograms come the retrieval of extraction and profile to image shape feature, if the method handle based on histogram
Dry homogeneous region is regarded with a join domain;
The image shape feature in the first scaling board area, taped region, the second scaling board area is extracted, variable conduct is defined as follows
Feature samples:
Sidebdl:First scaling board left margin;
Sidebdr:Second scaling board right margin;
Sidebel:Adhesive tape left margin;
Sideber:Adhesive tape right margin;
XAL-BEL:First scaling board left margin SidebdlTo adhesive tape left margin SidebelPixel;
XBEL-BER:Adhesive tape left margin SidebelTo adhesive tape right margin SideberPixel;
XBR-BER:Adhesive tape right margin SideberWith the second scaling board right margin SidebdrPixel;
PA-BELT:First scaling board area and taped region pixel ratio, calculation formula are as follows:PA-BELT=XAL-BEL/XBEL-BER;
PB-BELT:Second scaling board area and taped region pixel ratio, calculation formula are as follows:PB-BELT=XBR-BER/XBEL-BER;
By characterizing extraction algorithm, confirm whether the first scaling board area, the second scaling board area and taped region pixel ratio are normal:
In the case of adhesive tape normal operation, PA-BELTAnd PB-BELTValue variation only it is related with adhesive tape load, in adhesive tape duty factor
In the case of more constant, PA-BELTAnd PB-BELTValue variation be metastable;
In adhesive tape operational process, if adhesive tape shifts, PA-BELTAnd PB-BELTValue can occur significant change, judge adhesive tape
It runs abnormal;
S34, support vector machines training
Before starting to do training with support vector machines, front is collected and treated that feature samples are divided into two classes, one kind is that have
Imitate picture, i.e., adhesive tape under different loads or no-load condition the first scaling board, adhesive tape, second scaling board three's position relationship inclined
From the picture in threshold range, and stick effective label;Another kind of is invalid picture, i.e., adhesive tape is in different loads or no-load condition
Lower adhesive tape traffic direction is biased to the first scaling board or adhesive tape traffic direction is biased to the picture of the second scaling board, and sticks no criterion
Label;Then it takes out a part respectively from this two classes picture to be used for doing supporting vector machine model training, then another part is used for
Test set is done, includes equally effective picture and invalid picture in test set, is trained to verify supporting vector machine model
Classifying quality.
6. the diagnostic method of the rubber belt from deviating diagnostic system according to claim 5 based on machine vision, which is characterized in that
In the step S34, the characteristics of according to scene picture, the key of the Performance tuning of the algorithm parameter of supporting vector machine model training
Factor is penalty factor and nuclear parameter r in SVM, and penalty factor adjusts Learning machine for determining in sample data subspace
Fiducial interval range, optimal C is different in different data subspace, and the change of nuclear parameter r is implicit changes sample number
According to the complexity of Subspace Distribution.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810623426.0A CN108584352B (en) | 2018-06-15 | 2018-06-15 | Machine vision-based adhesive tape deviation diagnosis system and method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810623426.0A CN108584352B (en) | 2018-06-15 | 2018-06-15 | Machine vision-based adhesive tape deviation diagnosis system and method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108584352A true CN108584352A (en) | 2018-09-28 |
CN108584352B CN108584352B (en) | 2024-02-13 |
Family
ID=63628074
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810623426.0A Active CN108584352B (en) | 2018-06-15 | 2018-06-15 | Machine vision-based adhesive tape deviation diagnosis system and method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108584352B (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110759037A (en) * | 2019-11-22 | 2020-02-07 | 江苏东方赛云电子科技有限公司 | IPC-based conveyer belt deviation detection method |
CN111369538A (en) * | 2020-03-05 | 2020-07-03 | 北京科技大学 | Conveyor belt deviation detection method based on deep convolutional neural network |
CN112193763A (en) * | 2020-10-29 | 2021-01-08 | 河南理工大学 | Belt deviation detection system of belt conveyor |
CN112209053A (en) * | 2020-09-21 | 2021-01-12 | 山东科技大学 | Belt conveyor running state detection system based on line laser |
CN113076911A (en) * | 2021-04-16 | 2021-07-06 | 河南六米电子科技有限公司 | AGV ribbon guide track wear degree detection method and system based on artificial intelligence |
CN113191394A (en) * | 2021-04-07 | 2021-07-30 | 国电汉川发电有限公司 | Machine vision-based conveyor belt deviation diagnosis method and system |
CN113343834A (en) * | 2021-06-02 | 2021-09-03 | 华电邹县发电有限公司 | Belt deviation diagnosis method based on machine vision and laser line |
CN114655655A (en) * | 2022-03-09 | 2022-06-24 | 南京北路软件技术有限公司 | Conveyor belt deviation detection method based on UNet network |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102602681A (en) * | 2012-01-13 | 2012-07-25 | 天津工业大学 | Machine vision based online deviation fault detecting method for conveying belts |
JP2015173410A (en) * | 2014-03-12 | 2015-10-01 | 地方独立行政法人青森県産業技術センター | image processing system, image processing method and image processing program |
-
2018
- 2018-06-15 CN CN201810623426.0A patent/CN108584352B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102602681A (en) * | 2012-01-13 | 2012-07-25 | 天津工业大学 | Machine vision based online deviation fault detecting method for conveying belts |
JP2015173410A (en) * | 2014-03-12 | 2015-10-01 | 地方独立行政法人青森県産業技術センター | image processing system, image processing method and image processing program |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110759037A (en) * | 2019-11-22 | 2020-02-07 | 江苏东方赛云电子科技有限公司 | IPC-based conveyer belt deviation detection method |
CN111369538A (en) * | 2020-03-05 | 2020-07-03 | 北京科技大学 | Conveyor belt deviation detection method based on deep convolutional neural network |
CN111369538B (en) * | 2020-03-05 | 2023-07-04 | 北京科技大学 | Conveyor belt deviation detection method based on deep convolutional neural network |
CN112209053A (en) * | 2020-09-21 | 2021-01-12 | 山东科技大学 | Belt conveyor running state detection system based on line laser |
CN112193763A (en) * | 2020-10-29 | 2021-01-08 | 河南理工大学 | Belt deviation detection system of belt conveyor |
CN113191394A (en) * | 2021-04-07 | 2021-07-30 | 国电汉川发电有限公司 | Machine vision-based conveyor belt deviation diagnosis method and system |
CN113076911A (en) * | 2021-04-16 | 2021-07-06 | 河南六米电子科技有限公司 | AGV ribbon guide track wear degree detection method and system based on artificial intelligence |
CN113076911B (en) * | 2021-04-16 | 2023-04-14 | 天津万事达物流装备有限公司 | AGV ribbon guide track wear degree detection method and system based on artificial intelligence |
CN113343834A (en) * | 2021-06-02 | 2021-09-03 | 华电邹县发电有限公司 | Belt deviation diagnosis method based on machine vision and laser line |
CN114655655A (en) * | 2022-03-09 | 2022-06-24 | 南京北路软件技术有限公司 | Conveyor belt deviation detection method based on UNet network |
Also Published As
Publication number | Publication date |
---|---|
CN108584352B (en) | 2024-02-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108584352A (en) | Rubber belt from deviating diagnostic system based on machine vision and method | |
CN107123131B (en) | Moving target detection method based on deep learning | |
CN103279765B (en) | Steel wire rope surface damage detection method based on images match | |
CN108109137A (en) | The Machine Vision Inspecting System and method of vehicle part | |
CN103914708B (en) | Food kind detection method based on machine vision and system | |
Farazi et al. | A monocular vision system for playing soccer in low color information environments | |
CN104809481A (en) | Natural scene text detection method based on adaptive color clustering | |
CN105426828A (en) | Face detection method, face detection device and face detection system | |
CN107886131A (en) | One kind is based on convolutional neural networks detection circuit board element polarity method and apparatus | |
CN109344842A (en) | A kind of pedestrian's recognition methods again based on semantic region expression | |
CN105083912A (en) | Conveyer belt deviation detecting method based on image identification | |
CN103325122A (en) | Pedestrian retrieval method based on bidirectional sequencing | |
Wang et al. | End-to-end scene text recognition in videos based on multi frame tracking | |
CN109918971A (en) | Number detection method and device in monitor video | |
CN107330360A (en) | A kind of pedestrian's clothing colour recognition, pedestrian retrieval method and device | |
CN106548131A (en) | A kind of workmen's safety helmet real-time detection method based on pedestrian detection | |
CN105654054A (en) | Semi-supervised neighbor propagation learning and multi-visual dictionary model-based intelligent video analysis method | |
Tsintotas et al. | DOSeqSLAM: Dynamic on-line sequence based loop closure detection algorithm for SLAM | |
CN103854278A (en) | Printed circuit board image registration method based on shape context of mass center of communicated region | |
CN109165612B (en) | Pedestrian re-identification method based on depth feature and bidirectional KNN sequencing optimization | |
CN106447694A (en) | Video badminton motion detection and tracking method | |
Leordeanu et al. | Unsupervised learning of object features from video sequences | |
CN113343834A (en) | Belt deviation diagnosis method based on machine vision and laser line | |
CN105512631A (en) | Violence and horror video detection method based on MoSIFT and CSD features | |
CN208843186U (en) | Rubber belt from deviating diagnostic system based on machine vision |
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 |