CN103034861B - The recognition methods of a kind of truck brake shoe breakdown and device - Google Patents
The recognition methods of a kind of truck brake shoe breakdown and device Download PDFInfo
- Publication number
- CN103034861B CN103034861B CN201210543988.7A CN201210543988A CN103034861B CN 103034861 B CN103034861 B CN 103034861B CN 201210543988 A CN201210543988 A CN 201210543988A CN 103034861 B CN103034861 B CN 103034861B
- Authority
- CN
- China
- Prior art keywords
- brake shoe
- lorry brake
- lorry
- feature
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses the recognition methods of a kind of truck brake shoe breakdown, including: from present image, extract the segmentation feature of three angles;Segmentation feature according to described three angles determines the lorry brake shoe characteristic area of present image;The feature of lorry brake shoe is extracted from the lorry brake shoe characteristic area of present image;Use support vector machine (SVM) algorithm that the feature calculation of lorry brake shoe draws the eigenvalue of present image, judge whether lorry brake shoe exists fault according to described eigenvalue and preset Fault Identification value.The invention also discloses the identification device of a kind of truck brake shoe breakdown, use the present invention to be avoided that manually-operated error, it is ensured that the discrimination of fault, Accident prevention occurs timely, thus ensures operation security.
Description
Technical field
The present invention relates to image processing field, particularly relate to recognition methods and the device of a kind of truck brake shoe breakdown.
Background technology
At present, lorry Block brake is the tread brake mode that domestic railway freight-car is commonly used, the lock on lorry brake shoe
Watt pricker is mainly used to prevent lorry brake shoe from coming off, and once brake shoe drill is lost, and in lorry running, lorry brake shoe is the most very likely
Come off, cause brake failure, if more seriously lorry brake shoe just comes off on rail, then can cause derailment accident.In recent years
Coming, the Ministry of Railways widelys popularize lorry operation troubles image dynamic detection system (TFDS) key position each to operating train and enters
Row imaging, and completed Fault Identification by artificial image browsing.This be there is poor efficiency by the mode manually carrying out Fault Identification, know
The problems such as rate is unstable, cannot meet the demand for development of safe train operation.
Visible, prior art uses TFDS carry out brake shoe breakdown identification, excessively rely on manual operation, therefore cannot ensure
The discrimination of fault, and then cannot occur by Accident prevention timely, thus operation security cannot be ensured.
Summary of the invention
In view of this, it is an object of the invention to provide recognition methods and the device of a kind of truck brake shoe breakdown, be avoided that
Manually-operated error, it is ensured that the discrimination of fault, Accident prevention occurs timely, thus ensures operation security.
For reaching above-mentioned purpose, the technical scheme is that and be achieved in that:
The invention provides the recognition methods of a kind of truck brake shoe breakdown, the method includes:
The segmentation feature of three angles is extracted from present image;
Segmentation feature according to described three angles determines the lorry brake shoe characteristic area of present image;
The feature of lorry brake shoe is extracted from the lorry brake shoe characteristic area of present image;
Use support vector machine (SVM, Support Vector Machine) algorithm that the feature calculation of lorry brake shoe is obtained
Go out the eigenvalue of present image, judge whether lorry brake shoe exists event according to described eigenvalue and preset Fault Identification value
Barrier.
In such scheme, the described segmentation feature extracting three angles from present image, including: from TDFS periodically
Extracting present image, the Gray Projection that present image carries out three angles obtains three drop shadow curves;To all drop shadow curves
It is filtered, using the maximum in each drop shadow curve filtered all as the segmentation feature of three angles of present image.
In such scheme, the described segmentation feature according to three angles determines the lorry brake shoe characteristic area of present image,
Including: use the marginal information of Canny operator extraction present image, according to the segmentation feature that angle is zero degree, determine lorry lock
The left margin of watt characteristic area and the coordinate figure of right margin;According to the segmentation feature that angle is-25 ° and 25 °, determine lorry brake shoe
The coboundary of characteristic area and the coordinate figure of lower boundary.
In such scheme, the described feature extracting lorry brake shoe from the lorry brake shoe characteristic area of present image, including:
Use the method estimated based on background area, the image in the lorry brake shoe characteristic area of present image is split and obtains lorry lock
Watt region binary picture;Use element marking method, from the binary picture of described lorry brake shoe region, extract largest connected district
Territory;Use Canny operator, the binary picture from largest connected region extracts lorry brake shoe edge contour;According to described
Lorry brake shoe edge contour extracts the feature of lorry brake shoe.
In such scheme, the feature of described lorry brake shoe includes: the smooth features value of lorry brake shoe edge contour, lorry lock
Watt convex and concave feature value of edge contour, the jaggedness of lorry brake shoe edge contour, lorry brake shoe edge contour solid by being worth, lorry
The compactness of brake shoe edge contour, the circle of lorry brake shoe edge contour, the length-width ratio of lorry brake shoe edge contour, lorry lock
The area of watt edge contour and the girth of lorry brake shoe edge contour.
In such scheme, before the described segmentation feature extracting three angles from present image, the method also includes: make
Positive and negative samples training is set up respectively with the lorry brake shoe drill image that there is fault with the lorry brake shoe drill image that there is not fault
Collection;The feature of all images in extraction positive and negative samples training set, nine Wei Te that composition positive and negative samples training set is corresponding respectively
Levying vector, the value that the positive and negative samples that uses SVM algorithm to calculate is corresponding, using value corresponding for negative sample training set as fault
Discre value.
Present invention also offers the identification device of a kind of truck brake shoe breakdown, this device includes: characteristic extracting module and knowledge
Other module;Wherein,
Characteristic extracting module, for extracting the segmentation feature of three angles, according to described three angles from present image
Segmentation feature determine the lorry brake shoe characteristic area of present image, from the lorry brake shoe characteristic area of present image, extract goods
The feature of brake watt, is sent to identification module by the feature of described lorry brake shoe;
Identification module, for using the feature calculation of described lorry brake shoe that characteristic extracting module sent by SVM algorithm to obtain
Go out the eigenvalue of present image, judge whether lorry brake shoe exists event according to described eigenvalue and preset Fault Identification value
Barrier.
In such scheme, described characteristic extracting module, specifically for periodically extracting present image from the TDFS at place,
The Gray Projection that present image carries out three angles obtains three drop shadow curves;All drop shadow curves are filtered, will filter
The maximum in each drop shadow curve after ripple is all as the segmentation feature of three angles of present image.
In such scheme, described characteristic extracting module, specifically for using the edge letter of Canny operator extraction present image
Breath, according to the segmentation feature that angle is zero degree, determines left margin and the coordinate figure of right margin of lorry brake shoe characteristic area;According to
Angle is the segmentation feature of-25 ° and 25 °, determines coboundary and the coordinate figure of lower boundary of lorry brake shoe characteristic area.
In such scheme, described characteristic extracting module, the method estimated based on background area specifically for use, will be current
Image segmentation in the lorry brake shoe characteristic area of image obtains lorry brake shoe region binary picture;Use element marking method,
Largest connected region is extracted from the binary picture of described lorry brake shoe region;Use Canny operator, from largest connected region
Binary picture in extract lorry brake shoe edge contour;The spy of lorry brake shoe is extracted according to described lorry brake shoe edge contour
Levy.
In such scheme, described characteristic extracting module, specifically for extract lorry brake shoe edge contour smooth features value,
The convex and concave feature value of lorry brake shoe edge contour, the jaggedness of lorry brake shoe edge contour, consolidating of lorry brake shoe edge contour lean on
Value, the compactness of lorry brake shoe edge contour, the circle of lorry brake shoe edge contour, the length and width of lorry brake shoe edge contour
The girth of ratio, the area of lorry brake shoe edge contour and lorry brake shoe edge contour is as the feature of lorry brake shoe.
In such scheme, described identification module, it is also used for there is not the lorry brake shoe drill image of fault and there is event
The lorry brake shoe drill image of barrier sets up positive and negative samples training set respectively;Extract all images in positive and negative samples training set respectively
Feature, composition nine dimensional feature vectors corresponding to positive and negative samples training set, use the positive and negative samples pair that SVM algorithm calculates
The value answered, using value corresponding for negative sample training set as Fault Identification value.
The recognition methods of truck brake shoe breakdown provided by the present invention and device, can extract three automatically from present image
The segmentation feature of angle;Segmentation feature according to described three angles determines the lorry brake shoe characteristic area of present image;From working as
The lorry brake shoe characteristic area of front image extracts the feature of lorry brake shoe;And then according to the feature of described lorry brake shoe, and
Fault Identification condition determines whether lorry brake shoe exists fault.So, it becomes possible to avoid manually-operated error, it is ensured that fault
Discrimination, Accident prevention occurs timely, thus ensures operation security.
Accompanying drawing explanation
Fig. 1 is the recognition methods schematic flow sheet of the truck brake shoe breakdown of the present invention;
Fig. 2 is the identification device composition structural representation of the truck brake shoe breakdown of the present invention;
Fig. 3 is test result table.
Detailed description of the invention
The basic thought of the present invention is: extract the segmentation feature of three angles from present image;According to described three angles
The segmentation feature of degree determines the lorry brake shoe characteristic area of present image;Extract from the lorry brake shoe characteristic area of present image
The feature of lorry brake shoe;SVM algorithm is used the feature calculation of lorry brake shoe to draw the eigenvalue of present image, according to described spy
Value indicative and preset Fault Identification value judge whether lorry brake shoe exists fault.
Below in conjunction with the accompanying drawings and specific embodiment the present invention is further described in more detail.
The recognition methods of the truck brake shoe breakdown of the present invention, as it is shown in figure 1, comprise the following steps:
Step 101: extract the segmentation feature of three angles from present image.
Concrete, from TDFS, periodically extract present image, the Gray Projection that present image carries out three angles obtains
To three drop shadow curves;All drop shadow curves are filtered, using equal for the maximum in each drop shadow curve filtered as
The segmentation feature of three angles of present image;
Here, described three angles are-25 ° 0 ° and 25 ° of three angles set;
Described Gray Projection includes: be extended by present image, the image after being expanded;Image after extension is entered
Row projection calculates, and carries out discrete obtaining drop shadow curve to projecting calculated image.
Described being extended by present image as mending gray scale in image border is the pixel of zero, it is to avoid image limit during projection
The error that edge is caused, it is possible to use below equation calculates:
Wherein, described projection calculating can use formula:
Wherein, Assume to set up
Coordinate system, zero is o, if present image I is highly m, width is n,For projecting direction, two-dimensional image I will be projected to
Projection line Starting point is o,WithIt is mutually perpendicular to.R is any point on projection line, and p was r point, and withParallel straight line, l
For any point on it, (i, j) from projection lineVertical dimension, I (i, j) be image point (i, j) grey scale pixel value at place,
If projection line and x-axis angle theta are projecting direction angle, θ is-25 ° 0 ° or 25 °.
Described carry out discrete following formula to be used to calculate to projecting calculated image:
Wherein,θ is-25 ° 0 ° and 25 °,Currently
Image I is highly m, and width is n, δθR () represents image Gray Projection on the θ of projecting direction angle, it can reflect that image is being thrown
The grey scale change in shadow deflection θ direction.
Described all drop shadow curves are filtered using formula:
Step 102: determine the lorry brake shoe characteristic area of present image according to the segmentation feature of described three angles.
Particularly as follows: use the marginal information of Canny operator extraction present image, according to the segmentation feature that angle is zero degree,
Determine left margin and the coordinate figure of right margin of lorry brake shoe characteristic area;According to the segmentation feature that angle is-25 ° and 25 °, really
The coboundary of order brake watt characteristic area and the coordinate figure of lower boundary.
Here, described according to the segmentation feature that angle is zero degree, determine left margin and the right of lorry brake shoe characteristic area
The coordinate figure on boundary includes: using angle is that the marginal information of present image is divided into two subgraphs in left and right by the segmentation feature of zero degree
Picture, calculates the grey scale pixel value sum of two subimages respectively, compares grey scale pixel value sum big of said two subimage
Little, if the grey scale pixel value sum of left side subimage is relatively big, then the coordinate figure of the right margin of lorry brake shoe characteristic area is equal to zero
Degree segmentation feature abscissa value, deduct preset right margin distance value, the coordinate of the left margin of lorry brake shoe characteristic area
Value deducts preset width value equal to the coordinate figure of its right margin;Otherwise, the coordinate figure of the left margin of lorry brake shoe characteristic area
Equal to the abscissa value of the segmentation feature of zero degree, add preset left margin distance value, the right margin of lorry brake shoe characteristic area
Coordinate figure adds preset width value equal to the coordinate figure of its left margin;
Such as, it is assumed that the marginal information of present image is I, angle is that the segmentation of zero degree is characterized as l in figure2, will currently scheme
Marginal information I of picture is divided into two subimages in left and right and is respectively I1And I2;Calculate two subimage I respectively1And I2Pixel ash
Angle value sum edge1 and edge2;If edge1 > edge2, then Brigth=l2-h1, Bleft=Brigth-N;As edge1 < edge2
Time, Bleft=l2+h2, Bright=Bleft+N;Wherein, l2For the abscissa value of the segmentation feature of zero degree, h1For preset right margin
Distance value, h2For preset left margin distance value, N is preset width value;
Wherein, Wherein I1(i j) is subimage I1Point (i,
J) grey scale pixel value at place, M1, N1For I1Height and width;I2(i j) is subimage I2Point (i, j) grey scale pixel value at place,
M2, N2For subimage I2Height and width.
Described according to the segmentation feature that angle is-25 ° and 25 °, determine that the coboundary of lorry brake shoe characteristic area is with following
The coordinate figure on boundary, including: set angle straight line l corresponding to the segmentation feature of-25 °1With image coboundary intersection point abscissa value
For top1, lower boundary intersection point abscissa value is bot1;Angle is the straight line l corresponding to segmentation feature of 25 °3With image coboundary
Intersection point abscissa value is top3, and lower boundary intersection point abscissa value is bot3, calculating ratio ratio=| (top1-top3)/
(bot1-bot3)|;As ratio < 1, lorry brake shoe drill region is positioned at the tip position of present image;As ratio > 1,
Lorry brake shoe drill region is positioned at the bottom position of present image;According to the frame for movement feature in lorry brake shoe drill region, according to pine
Relaxation principle extracts the coordinate figure of coboundary and lower boundary.Wherein, described lax principle is prior art, is used for ensureing selected goods
Brake watt characteristic area can completely represent lorry brake shoe, and its implementation does not repeats.
Step 103: extract the feature of lorry brake shoe from the lorry brake shoe characteristic area of present image.
Concrete, use the method estimated based on background area, by the figure in the lorry brake shoe characteristic area of present image
As segmentation obtains lorry brake shoe region binary picture;Use element marking method, from described lorry brake shoe region binary picture
The largest connected region of middle extraction;Use Canny operator, the binary picture from largest connected region extracts lorry brake shoe limit
Edge profile;The feature of lorry brake shoe is extracted according to described lorry brake shoe edge contour.
Here, the method that described background area is estimated is prior art, does not repeats;Element marking method is existing skill
Art, does not repeats;Canny operator is prior art, does not repeats.
The feature of described lorry brake shoe includes: the smooth features value of lorry brake shoe edge contour, lorry brake shoe edge contour
Convex and concave feature value, the jaggedness of lorry brake shoe edge contour, lorry brake shoe edge contour solid by being worth, lorry brake shoe edge wheel
The compactness of exterior feature, the circle of lorry brake shoe edge contour, the length-width ratio of lorry brake shoe edge contour, lorry brake shoe edge contour
Area and the girth of lorry brake shoe edge contour.
Wherein, the calculating of the smooth features value of described lorry brake shoe edge contour includes: calculate lorry brake shoe edge
The central point of profile, is expressed as polar coordinate by lorry brake shoe edge contour centered by described central point;It is expressed as pole to described
The lorry brake shoe edge contour of coordinate form carries out discretization calculating, obtains the lorry brake shoe edge contour of discretization;To described
After the lorry brake shoe edge contour of discretization is normalized calculating, it is ranked up according to preset rule, then obtains after sequence
To result carry out wavelet decomposition and obtain wavelet coefficient, use wavelet coefficient to calculate smooth features value;
The described central point calculating lorry brake shoe edge contour may include that hypothesis lorry brake shoe edge contour be x, then with
Relative arc length s is parameter, then lorry brake shoe edge contour is represented by:Wherein 0≤s≤1,For lorry
Brake shoe edge contour;Then the central point computing formula of lorry brake shoe edge contour is
Wherein 0≤s≤1.
Described centered by central point, lorry brake shoe edge contour be expressed as polar coordinate and can be expressed as:
Described the described lorry brake shoe edge contour being expressed as polar form is carried out discretization calculate can be to obtain
Polar coordinate (r (the s of each discrete point on profilei), θ (si)), whereinI=0,1 ... N-1, N are discrete point number.
Described normalization calculates and can use formula:
Described it is ranked up may include that note θ ' (s according to preset rulei) it is θ ' (i), r ' (si) it is r ' (i), to r '
I () is resequenced: R (i)=r ' (η (i)) so that rightThere is θ (η (i)) < θ (η (j));
The described wavelet decomposition that carries out the result that obtains after sequence obtains wavelet coefficient and may include that
Described use wavelet coefficient calculates smooth features value: E=| | d (i) | |2。
The formula used that calculates of the convex and concave feature value of described lorry brake shoe edge contour is:Its
In, c (n), n=1, the sequence of points of 2 N lorry brake shoe edge contours, c (n)=| | c (n)-c0| |, n=1,2 ...,
N,
The rectangular degree computing formula of described lorry brake shoe edge contour can be:Wherein A0It it is the face of profile
Long-pending, AMERBeing the area of the minimum enclosed rectangle of profile, acquisition methods is prior art, does not repeats.
Admittedly the property computing formula of leaning on of described lorry brake shoe edge contour can be:Wherein, A represents the face of profile
Long-pending, CA is the area of its minimal convex polygon, and acquisition methods is prior art, does not repeats.
The compactness computing formula of described lorry brake shoe edge contour can be:Wherein, P represents the week of profile
Long, A represents that area, acquisition methods are prior art, does not repeats.
The circle computing formula of described lorry brake shoe edge contour can be:
Wherein,μRIt it is the average distance from regional barycenter to boundary point.δRIt it is the distance mean square deviation from regional barycenter to boundary point.
The width of the minimum enclosed rectangle that length-width ratio is lorry brake shoe edge contour of described lorry brake shoe edge contour and length
Ratio;
The area of described lorry brake shoe edge contour can be: the number of pixels in the range of lorry brake shoe edge contour is carried out
Statistics obtains.
The girth of described lorry brake shoe edge contour can be: the boundary length of lorry brake shoe edge contour.
Step 104: use SVM algorithm the feature calculation of lorry brake shoe to draw the eigenvalue of present image, according to described
Eigenvalue and preset Fault Identification value judge whether lorry brake shoe exists fault.
Here, the determination method of described Fault Identification value is: uses and there is not the lorry brake shoe image of fault and there is event
The lorry brake shoe image of barrier sets up positive and negative samples training set respectively;Extract all images in positive and negative samples training set respectively
Feature, nine dimensional feature vectors that composition positive and negative samples training set is corresponding, the positive and negative samples using SVM algorithm to calculate is corresponding
Value, using value corresponding for negative sample training set as Fault Identification value.Wherein, described SVM algorithm be embodied as prior art, this
In do not repeat.
Described according to eigenvalue and preset Fault Identification value judges whether lorry brake shoe exists fault and include: if current
The eigenvalue of image is equal to Fault Identification value, it is determined that lorry brake shoe breaks down;Otherwise, it determines there is not event in lorry brake shoe
Barrier.
As in figure 2 it is shown, the invention provides the identification device of a kind of truck brake shoe breakdown, this device includes: feature extraction
Module 21 and identification module 22;Wherein,
Characteristic extracting module 21, for extracting the segmentation feature of three angles, according to described three angles from present image
The segmentation feature of degree determines the lorry brake shoe characteristic area of present image, extracts from the lorry brake shoe characteristic area of present image
The feature of lorry brake shoe, is sent to identification module 22 by the feature of described lorry brake shoe;
Identification module 22, based on the feature using described lorry brake shoe that characteristic extracting module 21 sent by SVM algorithm
Calculate the eigenvalue drawing present image, judge whether lorry brake shoe exists according to described eigenvalue and preset Fault Identification value
Fault.
Described characteristic extracting module 21, specifically for periodically extracting present image, to current figure from the TDFS at place
Gray Projection as carrying out three angles obtains three drop shadow curves, is filtered all drop shadow curves, by filtered respectively
Maximum in individual drop shadow curve is all as the segmentation feature of three angles of present image.Wherein, described three angles are for setting
-25 ° 0 ° and 25 ° of three angles.
Described characteristic extracting module 21, the image specifically for present image is extended, after being expanded;Will extension
After image carry out projection and calculate, carry out discrete obtaining drop shadow curve to projecting calculated image.
Described characteristic extracting module 21, is the pixel of zero specifically for mending gray scale in image border, it is to avoid scheme during projection
The error caused as edge, is extended present image, it is possible to use below equation calculates:
Described characteristic extracting module 21, calculates specifically for using below equation to carry out projection:
Wherein, Assume to set up
Coordinate system, zero is o, if present image I is highly m, width is n,For projecting direction, two-dimensional image I will be projected to
Projection line Starting point is o,WithIt is mutually perpendicular to.R is any point on projection line, and p was r point, and withParallel straight line, l
For any point on it, (i, j) from projection lineVertical dimension, I (i, j) be image point (i, j) grey scale pixel value at place, if
Projection line and x-axis angle theta are projecting direction angle, and θ is-25 ° 0 ° or 25 °.
Described characteristic extracting module 21, specifically for using following formula to carry out discrete to projecting calculated image:
Wherein,θ is-25 ° 0 ° and 25 °,Currently
Image I is highly m, and width is n, δθR () represents image Gray Projection on the θ of projecting direction angle, it can reflect that image is being thrown
The grey scale change in shadow deflection θ direction.
Described characteristic extracting module 21, specifically for using below equation that all drop shadow curves are filtered:
Described characteristic extracting module 21, specifically for using the marginal information of Canny operator extraction present image, according to angle
Degree is the segmentation feature of zero degree, determines left margin and the coordinate figure of right margin of lorry brake shoe characteristic area;According to angle be-
The segmentation feature of 25 ° and 25 °, determines coboundary and the coordinate figure of lower boundary of lorry brake shoe characteristic area.
Described characteristic extracting module 21, believes the edge of present image specifically for the segmentation feature using angle to be zero degree
Breath is divided into two subimages in left and right, calculates the grey scale pixel value sum of two subimages respectively, compares said two subimage
The size of grey scale pixel value sum, if the grey scale pixel value sum of left side subimage is relatively big, then lorry brake shoe characteristic area
The coordinate figure of right margin is equal to the abscissa value of the segmentation feature of zero degree, deducts preset right margin distance value, and lorry brake shoe is special
The coordinate figure of the left margin levying region deducts preset width value equal to the coordinate figure of its right margin;Otherwise, lorry brake shoe feature
The coordinate figure of the left margin in region is equal to the abscissa value of the segmentation feature of zero degree, adds preset left margin distance value, lorry lock
The coordinate figure of the right margin of watt characteristic area equal to the coordinate figure of its left margin plus preset width value;
Such as, it is assumed that the marginal information of present image is I, angle is that the segmentation of zero degree is characterized as l in figure2, will currently scheme
Marginal information I of picture is divided into two subimages in left and right and is respectively I1And I2;Calculate two subimage I respectively1And I2Pixel ash
Angle value sum edge1 and edge2;If edge1 > edge2, then Brigth=l2-h1, Bleft=Brigth-N;As edge1 < edge2
Time, Bleft=l2+h2, Bright=Bleft+N;Wherein, l2For the abscissa value of the segmentation feature of zero degree, h1For preset right margin
Distance value, h2For preset left margin distance value, N is preset width value;
Wherein, Wherein I1(i j) is subimage I1Point (i,
J) grey scale pixel value at place, M1, N1For I1Height and width;I2(i j) is subimage I2Point (i, j) grey scale pixel value at place,
M2, N2For subimage I2Height and width.
Described characteristic extracting module 21, specifically for setting angle straight line l corresponding to the segmentation feature of-25 °1With image
Coboundary intersection point abscissa value is top1, and lower boundary intersection point abscissa value is bot1;Angle is corresponding to the segmentation feature of 25 °
Straight line l3Being top3 with image coboundary intersection point abscissa value, lower boundary intersection point abscissa value is bot3, calculating ratio ratio=
|(top1-top3)/(bot1-bot3)|;As ratio < 1, lorry brake shoe drill region is positioned at the tip position of present image;
As ratio > 1, lorry brake shoe drill region is positioned at the bottom position of present image;Machinery knot according to lorry brake shoe drill region
Structure feature, extracts the coordinate figure of coboundary and lower boundary according to lax principle.
Described characteristic extracting module 21, the method estimated based on background area specifically for use, by the goods of present image
Image segmentation in brake watt characteristic area obtains lorry brake shoe region binary picture;Use element marking method, from described goods
Brake watt region binary picture extracts largest connected region;Use Canny operator, the binary system from largest connected region
Image extracts lorry brake shoe edge contour;The feature of lorry brake shoe is extracted according to described lorry brake shoe edge contour.
Here, the method that described background area is estimated is prior art, does not repeats;Element marking method is existing skill
Art, does not repeats;Canny operator is prior art, does not repeats.
Described characteristic extracting module 21, specifically for calculating the central point of lorry brake shoe edge contour, in described
Centered by heart point, lorry brake shoe edge contour is expressed as polar coordinate;To the described lorry brake shoe edge being expressed as polar form
Profile carries out discretization calculating, obtains the lorry brake shoe edge contour of discretization;Lorry brake shoe edge wheel to described discretization
After exterior feature is normalized calculating, it is ranked up according to preset rule, then the result that obtains after sequence is carried out wavelet decomposition obtains
To wavelet coefficient, wavelet coefficient is used to calculate smooth features value;
Wherein, described in calculate the central point of lorry brake shoe edge contour and may include that hypothesis lorry brake shoe edge contour is x,
Then with relative arc length s as parameter, then lorry brake shoe edge contour is represented by:Wherein 0≤s≤1,For goods
Brake watt edge contour;Then the central point computing formula of lorry brake shoe edge contour is
Wherein 0≤s≤1.
Described centered by central point, lorry brake shoe edge contour be expressed as polar coordinate and can be expressed as:
Described the described lorry brake shoe edge contour being expressed as polar form is carried out discretization calculate can be to obtain
Polar coordinate (r (the s of each discrete point on profilei), θ (si)), whereinI=0,1 ... N-1, N are discrete point number.
Described normalization calculates and can use formula:
Described it is ranked up may include that note θ ' (s according to preset rulei) it is θ ' (i), r ' (si) it is r ' (i), to r '
I () is resequenced: R (i)=r ' (η (i)) so that rightThere is θ (η (i)) < θ (η (j));
The described wavelet decomposition that carries out the result that obtains after sequence obtains wavelet coefficient and may include that
Described use wavelet coefficient calculates smooth features value: E=| | d (i) | |2。
The formula used that calculates of the convex and concave feature value of described lorry brake shoe edge contour is:Its
In, c (n), n=1, the sequence of points of 2 N lorry brake shoe edge contours, c ' (n)=| | c (n)-c0| |, n=1,
2, N,
The rectangular degree computing formula of described lorry brake shoe edge contour can be:Wherein A0It it is the face of profile
Long-pending, AMERIt it is the area of the minimum enclosed rectangle of profile.
Admittedly the property computing formula of leaning on of described lorry brake shoe edge contour can be:Wherein, A represents the face of profile
Long-pending, CA is the area of its minimal convex polygon.
The compactness computing formula of described lorry brake shoe edge contour can be:Wherein, P represents the week of profile
Long, A represents area.
The circle computing formula of described lorry brake shoe edge contour can be:
Wherein,μRIt it is the average distance from regional barycenter to boundary point.
δRIt it is the distance mean square deviation from regional barycenter to boundary point.
The width of the minimum enclosed rectangle that length-width ratio is lorry brake shoe edge contour of described lorry brake shoe edge contour and length
Ratio;
The area of described lorry brake shoe edge contour can be: the number of pixels in the range of lorry brake shoe edge contour is carried out
Statistics obtains.
The girth of described lorry brake shoe edge contour can be: the boundary length of lorry brake shoe edge contour.
Described identification module 22, specifically for preserving Fault Identification value.
Described identification module 22, if the eigenvalue specifically for present image is equal to Fault Identification value, it is determined that lorry lock
Watt break down;Otherwise, it determines lorry brake shoe does not break down.
Described identification module 22, is also used for there is not the lorry brake shoe image of fault and there is the lorry brake shoe of fault
Image sets up positive and negative samples training set respectively;The feature of all images in extraction positive and negative samples training set respectively, just forming,
Nine dimensional feature vectors that negative sample training set is corresponding, use the value that the positive and negative samples that calculates of SVM algorithm is corresponding, by negative sample
Value corresponding to this training set is as Fault Identification value.
Said apparatus can be installed in the management equipment of TDFS as logic module.
The method and device that the application of the invention provides, truck brake shoe breakdown sample image 75, non-event are chosen in test
Barrier sample image 75 is set up sample set and is carried out classifier training, for testing the adaptability of recognizer, uses cross validation
(cross validation is called again cross validation to thought, and first by training sample v decile, a copy of it is used as test set, additionally v-1 part
As training set.Rotation successively, until a test set all made by every part of sample, has i.e. carried out v training and the process of prediction.
Therefore, the accuracy rate of cross validation is the meansigma methods of v time) carry out Fault Identification test, make 10 experiments altogether, take experimental result
Meansigma methods be final recognition result as shown in Figure 3.Visible, recognition methods and the dress of the truck brake shoe breakdown of present invention offer are provided
Postponing, recognition time is less than two seconds, hence it is evident that be less than manually being distinguished by image, and loss, false drop rate are the lowest.
The above, only presently preferred embodiments of the present invention, it is not intended to limit protection scope of the present invention.
Claims (6)
1. the recognition methods of a truck brake shoe breakdown, it is characterised in that the method includes:
The segmentation feature of three angles is extracted from present image;
Segmentation feature according to described three angles determines the lorry brake shoe characteristic area of present image;
The feature of lorry brake shoe is extracted from the lorry brake shoe characteristic area of present image;
Support vector machines algorithm is used the feature calculation of lorry brake shoe to draw the eigenvalue of present image, according to described spy
Value indicative and preset Fault Identification value judge whether lorry brake shoe exists fault;
Wherein, before the described segmentation feature extracting three angles from present image, the method also includes: uses and there is not event
The lorry brake shoe image of barrier and the lorry brake shoe image that there is fault set up positive and negative samples training set respectively;Extract positive and negative respectively
The feature of all images that sample training is concentrated, nine dimensional feature vectors that composition positive and negative samples training set is corresponding, use SVM to calculate
The value that positive and negative samples that method calculates is corresponding, using value corresponding for negative sample training set as Fault Identification value;
Wherein, the described segmentation feature extracting three angles from present image, including: from fault picture dynamic detection system
Periodically extracting present image in TDFS, the Gray Projection that present image carries out three angles obtains three drop shadow curves;Right
All drop shadow curves are filtered, using the maximum in each drop shadow curve filtered all as three angles of present image
Segmentation feature;
Described Gray Projection includes: be extended by present image, the image after being expanded;Image after extension is thrown
Shadow calculates, and carries out discrete obtaining drop shadow curve to projecting calculated image;
The described segmentation feature according to three angles determines the lorry brake shoe characteristic area of present image, including: use Canny to calculate
Son extracts the marginal information of present image, according to the segmentation feature that angle is zero degree, determines the left side of lorry brake shoe characteristic area
Boundary and the coordinate figure of right margin;According to the segmentation feature that angle is-25 ° and 25 °, determine the coboundary of lorry brake shoe characteristic area
Coordinate figure with lower boundary;
Described according to the segmentation feature that angle is zero degree, determine left margin and the coordinate figure of right margin of lorry brake shoe characteristic area
Including: using angle is that the marginal information of present image is divided into two subimages in left and right by the segmentation feature of zero degree, counts respectively
Calculate the grey scale pixel value sum of two subimages, compare the size of the grey scale pixel value sum of said two subimage, if the left side
The grey scale pixel value sum of subimage is relatively big, then the coordinate figure of the right margin of lorry brake shoe characteristic area is special equal to the segmentation of zero degree
The abscissa value levied, deducting preset right margin distance value, the coordinate figure of the left margin of lorry brake shoe characteristic area is right equal to it
The coordinate figure on border deducts preset width value;Otherwise, the coordinate figure of the left margin of lorry brake shoe characteristic area is equal to zero degree
Splitting the abscissa value of feature, add preset left margin distance value, the coordinate figure of the right margin of lorry brake shoe characteristic area is equal to
The coordinate figure of its left margin is plus preset width value;
Described according to the segmentation feature that angle is-25 ° and 25 °, determine the coboundary of lorry brake shoe characteristic area and lower boundary
Coordinate figure, including: set angle straight line l corresponding to the segmentation feature of-25 °1With image coboundary intersection point abscissa value it is
Top1, lower boundary intersection point abscissa value is bot1;Angle is the straight line l corresponding to segmentation feature of 25 °3Hand over image coboundary
Point abscissa value is top3, and lower boundary intersection point abscissa value is bot3, calculating ratio ratio=| (top1-top3)/(bot1-
bot3)|;As ratio < 1, lorry brake shoe drill region is positioned at the tip position of present image;As ratio > 1, lorry lock
Watt pricker region is positioned at the bottom position of present image;According to the frame for movement feature in lorry brake shoe drill region, according to lax principle
Extract the coordinate figure of coboundary and lower boundary.
Method the most according to claim 1, it is characterised in that described carry from the lorry brake shoe characteristic area of present image
The feature of picking brake watt, including: use the method estimated based on background area, by the lorry brake shoe characteristic area of present image
In image segmentation obtain lorry brake shoe region binary picture;Use element marking method, enter from described lorry brake shoe region two
The imaged largest connected region of middle extraction;Use Canny operator, the binary picture from largest connected region extracts lorry
Brake shoe edge contour;The feature of lorry brake shoe is extracted according to described lorry brake shoe edge contour.
Method the most according to claim 2, it is characterised in that the feature of described lorry brake shoe includes: lorry brake shoe edge
The smooth features value of profile, the convex and concave feature value of lorry brake shoe edge contour, the jaggedness of lorry brake shoe edge contour, lorry lock
Watt edge contour solid by value, the compactness of lorry brake shoe edge contour, the circle of lorry brake shoe edge contour, lorry brake shoe
The length-width ratio of edge contour, the area of lorry brake shoe edge contour and the girth of lorry brake shoe edge contour.
4. the identification device of a truck brake shoe breakdown, it is characterised in that this device includes: characteristic extracting module and identification mould
Block;Wherein,
Characteristic extracting module, for extracting the segmentation feature of three angles, dividing according to described three angles from present image
Cut feature and determine the lorry brake shoe characteristic area of present image, from the lorry brake shoe characteristic area of present image, extract lorry lock
Watt feature, the feature of described lorry brake shoe is sent to identification module;
Identification module, works as using the feature calculation of described lorry brake shoe that characteristic extracting module sent by SVM algorithm to draw
According to described eigenvalue and preset Fault Identification value, the eigenvalue of front image, judges whether lorry brake shoe exists fault;
Described identification module, is also used for the lorry brake shoe image that there is not fault and divides with the lorry brake shoe image that there is fault
Do not set up positive and negative samples training set;The feature of all images in extraction positive and negative samples training set, forms positive and negative samples respectively
Nine dimensional feature vectors that training set is corresponding, use the value that the positive and negative samples that calculates of SVM algorithm is corresponding, are trained by negative sample
The value of collection correspondence is as Fault Identification value;
Described characteristic extracting module, specifically for periodically extracting present image from the TDFS at place, is carried out present image
The Gray Projection of three angles obtains three drop shadow curves;All drop shadow curves are filtered, by each projection filtered
Maximum in curve is all as the segmentation feature of three angles of present image;Described Gray Projection includes: entered by present image
Row extension, the image after being expanded;Will extension after image carry out projection calculate, to project calculated image carry out from
Dissipate and obtain drop shadow curve;
Described characteristic extracting module, is specifically also used for the marginal information of Canny operator extraction present image, according to angle is
The segmentation feature of zero degree, determines left margin and the coordinate figure of right margin of lorry brake shoe characteristic area;According to angle be-25 ° and
The segmentation feature of 25 °, determines coboundary and the coordinate figure of lower boundary of lorry brake shoe characteristic area;
Described according to the segmentation feature that angle is zero degree, determine left margin and the coordinate figure of right margin of lorry brake shoe characteristic area
Including: using angle is that the marginal information of present image is divided into two subimages in left and right by the segmentation feature of zero degree, counts respectively
Calculate the grey scale pixel value sum of two subimages, compare the size of the grey scale pixel value sum of said two subimage, if the left side
The grey scale pixel value sum of subimage is relatively big, then the coordinate figure of the right margin of lorry brake shoe characteristic area is special equal to the segmentation of zero degree
The abscissa value levied, deducting preset right margin distance value, the coordinate figure of the left margin of lorry brake shoe characteristic area is right equal to it
The coordinate figure on border deducts preset width value;Otherwise, the coordinate figure of the left margin of lorry brake shoe characteristic area is equal to zero degree
Splitting the abscissa value of feature, add preset left margin distance value, the coordinate figure of the right margin of lorry brake shoe characteristic area is equal to
The coordinate figure of its left margin is plus preset width value;
Described according to the segmentation feature that angle is-25 ° and 25 °, determine the coboundary of lorry brake shoe characteristic area and lower boundary
Coordinate figure, including: set angle straight line l corresponding to the segmentation feature of-25 °1With image coboundary intersection point abscissa value it is
Top1, lower boundary intersection point abscissa value is bot1;Angle is the straight line l corresponding to segmentation feature of 25 °3Hand over image coboundary
Point abscissa value is top3, and lower boundary intersection point abscissa value is bot3, calculating ratio ratio=| (top1-top3)/(bot1-
bot3)|;As ratio < 1, lorry brake shoe drill region is positioned at the tip position of present image;As ratio > 1, lorry lock
Watt pricker region is positioned at the bottom position of present image;According to the frame for movement feature in lorry brake shoe drill region, according to lax principle
Extract the coordinate figure of coboundary and lower boundary.
Device the most according to claim 4, it is characterised in that described characteristic extracting module, specifically for using based on the back of the body
The method that scape area is estimated, obtains lorry brake shoe region two by the image segmentation in the lorry brake shoe characteristic area of present image and enters
Imaged;Use element marking method, from the binary picture of described lorry brake shoe region, extract largest connected region;Use
Canny operator, extracts lorry brake shoe edge contour in the binary picture from largest connected region;According to described lorry brake shoe
Edge contour extracts the feature of lorry brake shoe.
Device the most according to claim 4, it is characterised in that described characteristic extracting module, specifically for extracting lorry lock
Watt smooth features value of edge contour, the convex and concave feature value of lorry brake shoe edge contour, the jaggedness of lorry brake shoe edge contour,
Consolidating by value, the compactness of lorry brake shoe edge contour, the circle of lorry brake shoe edge contour, goods of lorry brake shoe edge contour
The girth of the length-width ratio of brake watt edge contour, the area of lorry brake shoe edge contour and lorry brake shoe edge contour is as lorry
The feature of brake shoe.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210543988.7A CN103034861B (en) | 2012-12-14 | 2012-12-14 | The recognition methods of a kind of truck brake shoe breakdown and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201210543988.7A CN103034861B (en) | 2012-12-14 | 2012-12-14 | The recognition methods of a kind of truck brake shoe breakdown and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103034861A CN103034861A (en) | 2013-04-10 |
CN103034861B true CN103034861B (en) | 2016-12-21 |
Family
ID=48021740
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201210543988.7A Expired - Fee Related CN103034861B (en) | 2012-12-14 | 2012-12-14 | The recognition methods of a kind of truck brake shoe breakdown and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103034861B (en) |
Families Citing this family (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103499584B (en) * | 2013-10-16 | 2016-02-17 | 北京航空航天大学 | Railway wagon hand brake chain bar loses the automatic testing method of fault |
CN104268588B (en) * | 2014-06-19 | 2018-02-27 | 江苏大学 | Railway wagon brake shoe pricker loses the automatic testing method of failure |
CN106778740A (en) * | 2016-12-06 | 2017-05-31 | 北京航空航天大学 | A kind of TFDS non-faulting image detecting methods based on deep learning |
CN107392313B (en) * | 2017-06-12 | 2020-09-29 | 五邑大学 | Steel rail identification method based on deep learning |
CN109377511B (en) * | 2018-08-30 | 2021-09-03 | 西安电子科技大学 | Moving target tracking method based on sample combination and depth detection network |
CN109632007B (en) * | 2019-01-17 | 2020-12-04 | 北京理工大学 | Edge point extraction method and gear high-precision vision measurement system |
CN111080605A (en) * | 2019-12-12 | 2020-04-28 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying railway wagon manual brake shaft chain falling fault image |
CN112197715B (en) * | 2020-10-27 | 2022-07-08 | 上海市特种设备监督检验技术研究院 | Elevator brake wheel and brake shoe gap detection method based on image recognition |
CN112699794B (en) * | 2020-12-29 | 2021-09-14 | 哈尔滨市科佳通用机电股份有限公司 | Method for identifying dislocation fault images of middle rubber and upper and lower floor plates of wagon axle box rubber pad |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101339669A (en) * | 2008-07-29 | 2009-01-07 | 上海师范大学 | Three-dimensional human face modelling approach based on front side image |
CN101699470A (en) * | 2009-10-30 | 2010-04-28 | 华南理工大学 | Extracting method for smiling face identification on picture of human face |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2003036438A (en) * | 2001-07-25 | 2003-02-07 | Minolta Co Ltd | Program for specifying red-eye in image, recording medium, image processor and method for specifying red- eye |
KR100453943B1 (en) * | 2001-12-03 | 2004-10-20 | 주식회사 세넥스테크놀로지 | Iris image processing recognizing method and system for personal identification |
-
2012
- 2012-12-14 CN CN201210543988.7A patent/CN103034861B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101339669A (en) * | 2008-07-29 | 2009-01-07 | 上海师范大学 | Three-dimensional human face modelling approach based on front side image |
CN101699470A (en) * | 2009-10-30 | 2010-04-28 | 华南理工大学 | Extracting method for smiling face identification on picture of human face |
Non-Patent Citations (2)
Title |
---|
基于计算机视觉的列车闸瓦检测方法;杨雪荣 等;《内燃机车》;20090630;第43页第3段至第44页第4段 * |
货车典型故障图像识别算法研究;戴鹏;《中国博士学位论文全文数据库》;20110515;第2.2.1节,2.4节,3.5.1节和3.5.2节 * |
Also Published As
Publication number | Publication date |
---|---|
CN103034861A (en) | 2013-04-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103034861B (en) | The recognition methods of a kind of truck brake shoe breakdown and device | |
CN106529424B (en) | A kind of logo detection recognition method and system based on selective search algorithm | |
CN106156780B (en) | The method of wrong report is excluded on track in foreign body intrusion identification | |
CN107025432B (en) | A kind of efficient lane detection tracking and system | |
CN104268596B (en) | A kind of Car license recognition device and its detection method of license plate and system | |
CN104392239B (en) | A kind of licence plate recognition method and system | |
CN100463000C (en) | Human eye state detection method based on cascade classification and hough circle transform | |
CN105550654B (en) | Bullet train image capturing system, real-time license number detection system and method | |
CN105335743A (en) | Vehicle license plate recognition method | |
CN109637151A (en) | A kind of recognition methods that highway Emergency Vehicle Lane is driven against traffic regulations | |
CN105404857A (en) | Infrared-based night intelligent vehicle front pedestrian detection method | |
CN104881661B (en) | Vehicle checking method based on structural similarity | |
CN105891220B (en) | A kind of pavement marker line defect detection device and its detection method | |
DE102015205309A1 (en) | Stable detection of windshields via the location of marking points | |
Guo et al. | Image-based seat belt detection | |
CN104318582A (en) | Detection method for bad state of rotating double-lug component pin of high-speed rail contact network on basis of image invariance target positioning | |
CN103903018A (en) | Method and system for positioning license plate in complex scene | |
CN107273802A (en) | A kind of detection method and device of railroad train brake shoe drill ring failure | |
CN105976392B (en) | Vehicle tyre detection method and device based on maximum output probability | |
CN107273852A (en) | Escalator floor plates object and passenger behavior detection algorithm based on machine vision | |
CN106485694A (en) | A kind of high ferro contact net double-jacket tube connector six-sided nut based on cascade classifier comes off defective mode detection method | |
CN103020596A (en) | Method for identifying abnormal human behaviors in power production based on block model | |
CN103632427A (en) | Gate cracking protection method and gate control system | |
CN107909009A (en) | A kind of obstacle detection method and device based on road surface study | |
CN108009574A (en) | A kind of rail clip detection method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20161221 Termination date: 20211214 |