CN102980896A - Method for detecting breakage of lugs of high-speed rail contact net suspension device - Google Patents

Abstract

The invention discloses a detection method adopting SIFT (Scale-Invariant Feature Transform) to detect a breakage fault of rotary double lugs. The method comprises the following steps of: firstly, carrying out local feature point matching on the image of a contact net supporting and suspending device and the image of a rotary double-lug template by utilizing an SIFT algorithm; secondly, screening successfully matched feature points by utilizing an improved RANSAC (Random Sample Consensus) algorithm and positioning the rotary double lugs; thirdly, extracting an upper boundary curve of the rotary double lugs by utilizing a boundary tracing method, and drawing a curvature curve of all points on the curve; and finally, carrying out difference treatment on the drawn curvature curve and the curvature curve under a normal condition at the corresponding point to judge whether the lugs have a breakage fault or not. The method provided by the invention can accurately identity the lug with the breakage fault in the image of a complex contact net suspending device, and greatly improves detection efficiency in comparison with an artificial screening method.

Description

High ferro overhead contact line device auricle fracture detection method

Technical field

The present invention relates to high-speed railway image detection field, particularly SIFT (Scale-Invariant Feature Transform, the conversion of yardstick invariant features) image invariant feature extraction, Feature Points Matching and screening, the methods such as auricle fracture defect identification.

Background technology

The rotation ears are positioned at the junction of high ferro contact net supporting and location device, are load parts important in the net-fault supporting construction, and safe train operation is played vital effect.In the actual operation of railway, often causes the auricle fracture defect because of train vibrations, cause contact net support device structure strength decreased, in the time of seriously even the danger that has steady arm to come off.Therefore be necessary rotation ears parts are detected, in time find and change trouble unit.

At present, China railways department mainly takes manual type that elements of contacting net is checked.This method efficient is lower, can not in time find fault, therefore is necessary to study automatic testing method.At present, on railway, use to some extent based on the Automatic Measurement Technique that image is processed, be mainly used in realizing the detection of contact net geometric parameter and pantograph wearing and tearing.Yet the fault detection method for the rotation ears not yet occurs at present.

Summary of the invention

The object of the present invention is to provide a kind of high ferro contact net rotation ears and the detection method of sheet fracture fault.The method is utilized convergent-divergent, rotation, brightness and the affine unchangeability of SIFT, can be from the contact net support and suspender image of different shooting environmental, different shooting angles, accurately identify and extract the rotation ears, carry out automatic decision to whether the auricle fracture defect occuring simultaneously.Its concrete means are:

A kind of high ferro contact net bracing or strutting arrangement auricle fracture detection method of utilizing yardstick invariant features conversion SIFT (Scale-Invariant Feature Transform), the CCD industrial camera that utilization is installed in the inspection vehicle roof supports contact net and the image that gathers is carried out local feature point coupling, cluster, location and fracture defect detection after gathering with the suspender image, may further comprise the steps:

A, CCD industrial camera and shot-light are installed in the inspection vehicle roof, the operation of inspection vehicle downline, the automatic shooting contact net supports and the suspender image when running into catenary mast, adopts the style of shooting that adds shot-light night to disturb to reduce daylight.

B, with A obtain contact net that imagery exploitation SIFT collects at the scene support with suspender image and rotation ears standard form image between realization local feature point mate;

C, with B obtain support with the suspender image in all unique points that the match is successful carry out cluster by space length, the image-region of the corresponding doubtful rotation ears of each cluster;

D, the random sampling unification algorism RANSAC (Random Sample Consensus) after utilize improving to the unique point in the described cluster to screening, judge in the corresponding image-region of this cluster and whether comprise the rotation ears, affine transformation matrix in calculation template image and contact net support and the suspender image between the rotation ears zone, realize the location of rotation ears in contact net support and suspender image, its algorithm flow is as follows:

A1: from described cluster, randomly draw 3 unique points, utilize the affine transformation matrix H between formula (1) calculating contact net support and suspender image and the template image ₁

H ₁＝BA ^T(AA ^T) ^-1 (1)

In the formula, $A=\left(\begin{array}{cccc}{x}_1& x_2& K& x_N\\ y_1& y_2& K& y_N\\ 1& 1& K& 1\end{array}\right),$ $B=\left(\begin{array}{cccc}{x_1}^{\′}& {x_2}^{\′}& K& {x_N}^{\′}\\ {y_1}^{\′}& {y_2}^{\′}& K& {y_N}^{\′}\\ 1& 1& K& 1\end{array}\right),$ N is for calculating H ₁The time unique point used number, N=3, (x are arranged this moment _i, y _i), i ∈ [1, N] is the coordinate of unique point in contact net support and the suspender image, (x _i', y _i') be in the template image and (x _i, y _i) coordinate of the point that is complementary, with the S that counts in the optimum estimate _iValue set to 0;

A2: utilize (2) formula to calculate match point in template image of each point in this cluster (x ', y ') through H ₁The result who obtains after the determined affined transformation (x ", y ").

$\left(\begin{array}{c}{x}^{\′\′}\\ y^{\′\′}\\ 1\end{array}\right)=H_1\left(\begin{array}{c}{x}^{\′}\\ y^{\′}\\ 1\end{array}\right)---\left(2\right)$

If (x ", the distance of the point (x, y) in y ") and the cluster then is judged to be (x, y) an interior point less than T;

A3: judge H ₁In corresponding all point contact net support with the suspender image in relative position whether with template image in the relative position of the unique point that is complementary with it identical.If not identical, then think to comprise erroneous matching in the interior point, if identical, then judge H ₁Count in corresponding whether greater than the S that counts in the optimum estimate _iIf, greater than S _i, then with H ₁As the current optimum estimate of affine transformation matrix, and upgrade S _i

A4: return A1) continue operation, move to stipulated number after, circulation stops, if S _iValue be not 0, then think to comprise the rotation ears in the image-region of this cluster representative.The current optimum estimate H of affine transformation matrix ₁Namely as its final estimation H.Bring the coordinate on four summits of template image into formula (1), and make H ₁=H can calculate the region that rotates ears in contact net support and the suspender image;

E, from support with the suspender image extract rotation ears subimage, utilize the method based on the coboundary embroidery that the auricle fracture defect is detected, detection method comprises following steps:

E1: will from contact net support with the suspender image the rotation ears subimage that extracts carry out binaryzation, the rotation ears are separated with image background.From the subimage edge, utilize the method for boundary tracking to obtain successively the coordinate of each point on the curve of coboundary, form coboundary curvilinear coordinates sequence;

E2: the flexibility of k point is defined as it and k+50 inclination angle of putting line on the curve of coboundary, and its computing formula is as follows:

$c_k=\left\{\begin{array}{cc}\arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]>0)\\ \arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]+\mathrm{\&pi;}& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]<0)\end{array}\right.---\left(3\right)$

In the formula, c _kThe flexibility of k point on the curve of expression coboundary, x _k, y _k, x _K+50, y _K+50, represent on the curve of coboundary k point and k+50 the transverse and longitudinal coordinate of putting, c respectively _kUnit be radian, span is 0～π.

Calculate successively the flexibility of each point on the curve of coboundary, draw out the flexibility curve of each point on the curve of coboundary;

E3: above-mentioned flexibility curve and flexibility curve are under normal circumstances made difference at corresponding point position, if comprise the point greater than 0.5 in the difference curves, then think the auricle fracture defect has occured, send the fracture defect alarm signal.

Beneficial effect of the present invention is:

1, the present invention directly detects high ferro contact net rotation ears parts by image processing method, can obtain objective, testing result accurately, has reduced workload and the error of artificial cognition.

2, the present invention takes full advantage of convergent-divergent, rotation, brightness and the affine unchangeability of SIFT, and testing result is not subjected to the impact of shooting environmental and shooting angle, and the accuracy rate of Fault Identification is high.

Description of drawings

Fig. 1 is algorithm flow sketch of the present invention.

Fig. 2 is that the contact net that comprises the auricle fracture defect that ccd video camera photographs supports and the suspender image.

Fig. 3 is for being used for the rotation ears template image of local feature coupling.

Fig. 4 is the feature points clustering figure (Fig. 4 a, Fig. 4 b, Fig. 4 c represent respectively a cluster) that the match is successful.

The rotation ears subimage of Fig. 5 for from contact net support and suspender image, extracting.

Fig. 6 is the rotation ears coboundary curve that utilizes boundary tracking process to obtain.

Fig. 7 is the flexibility curve of the coboundary curve shown in Fig. 6.

Fig. 8 is the difference curves that the flexibility curve shown in Fig. 7 and flexibility curve under normal circumstances obtain when corresponding point position is done difference.

Fig. 9 is not for existing the rotation ears of auricle fracture defect, the flexibility difference curves of utilizing same method to obtain.

Embodiment:

Below in conjunction with accompanying drawing embodiments of the present invention are described in further detail.

Utilizing the SIFT algorithm that contact net shown in Figure 2 is supported carries out local feature point with suspender image and rotation ears template image shown in Figure 3 and mates.Matching result as shown in Figure 3.The unique point that the match is successful mainly concentrates on 3 zones, forms altogether 3 clusters, shown in Fig. 4 a, Fig. 4 b and Fig. 4 c.The large figure of left-half is contact net support and suspender image to be analyzed among the figure, and the little figure in the upper right corner is rotation ears template images, and dark green line represents the coupling of local feature point among two width of cloth figure.

Successively the cluster among Fig. 4 a, Fig. 4 b and Fig. 4 c is missed coupling with the RANSAC algorithm after improving and eliminate, S counts in the optimum estimate that the cluster among Fig. 4 b and Fig. 4 c finally obtains _iBe 0, so think and do not comprise the rotation ears in the image-region of their representatives.S counts in the optimum estimate that Fig. 4 a obtains after missing the coupling elimination _iBe not 0, so think and comprise the rotation ears in the image-region of its representative.With the final estimated result H of affine transformation matrix and template figure as four apex coordinates bring into formula (1) can realize rotating ears contact net support with the suspender image in the location.The subimage in rotation binaural localization zone as shown in Figure 5 in contact net support and the suspender image.

$\left(\begin{array}{c}{x}^{\&prime;\&prime;}\\ y^{\&prime;\&prime;}\\ 1\end{array}\right)=H\left(\begin{array}{c}{x}^{\&prime;}\\ y^{\&prime;}\\ 1\end{array}\right)---\left(1\right)$

Will from contact net support with the suspender image after the rotation ears subimage binaryzation that extracts, the rotation ears coboundary curve that utilizes boundary tracking process to obtain.The result as shown in Figure 6.Calculate the flexibility of each point on the curve of coboundary according to formula (2).Draw the flexibility curve as shown in Figure 7.

$c_k=\left\{\begin{array}{cc}\arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]>0)\\ \arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]+\mathrm{\&pi;}& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]<0)\end{array}\right.---\left(2\right)$

Flexibility curve among Fig. 7 and flexibility curve are under normal circumstances done difference in corresponding point, and the result as shown in Figure 8.As can be seen from Figure 8, difference curves comprise two greater than the peak value of threshold value 0.5, therefore judge that there is the auricle fracture defect in these rotation ears.

For the rotation ears that do not have the auricle fracture defect, utilize flexibility difference curves that same method obtains usually as shown in Figure 9, the value of having a few on the curve is all in threshold value below 0.5.

Claims (1)

1. high ferro contact net bracing or strutting arrangement auricle fracture detection method of utilizing yardstick invariant features conversion SIFT (Scale-Invariant Feature Transform), the CCD industrial camera that utilization is installed in the inspection vehicle roof supports contact net and the image that gathers is carried out local feature point coupling, cluster, location and fracture defect detection after gathering with the suspender image, may further comprise the steps:

A, CCD industrial camera and shot-light are installed in the inspection vehicle roof, the operation of inspection vehicle downline, the automatic shooting contact net supports and the suspender image when running into catenary mast, adopts the style of shooting that adds shot-light night to disturb to reduce daylight;

B, with A obtain contact net that imagery exploitation SIFT collects at the scene support with suspender image and rotation ears standard form image between realization local feature point mate;

C, with B obtain support with the suspender image in all unique points that the match is successful carry out cluster by space length, the image-region of the corresponding doubtful rotation ears of each cluster;

D, the random sampling unification algorism RANSAC (Random Sample Consensus) after utilize improving to the unique point in the described cluster to screening, judge in the corresponding image-region of this cluster and whether comprise the rotation ears, affine transformation matrix in calculation template image and contact net support and the suspender image between the rotation ears zone, realize the location of rotation ears in contact net support and suspender image, its algorithm flow is as follows:

A1: from described cluster, randomly draw 3 unique points, utilize the affine transformation matrix H between formula (1) calculating contact net support and suspender image and the template image ₁

H ₁＝BA ^T(AA ^T) ^-1 (1)

In the formula, $A=\left(\begin{array}{cccc}{x}_1& x_2& K& x_N\\ y_1& y_2& K& y_N\\ 1& 1& K& 1\end{array}\right),$ $B=\left(\begin{array}{cccc}{x_1}^{\&prime;}& {x_2}^{\&prime;}& K& {x_N}^{\&prime;}\\ {y_1}^{\&prime;}& {y_2}^{\&prime;}& K& {y_N}^{\&prime;}\\ 1& 1& K& 1\end{array}\right),$ N is for calculating H ₁The time unique point used number, N=3, (x are arranged this moment _i, y _i), i ∈ [1, N] is the coordinate of unique point in contact net support and the suspender image, (x _i', y _i') be in the template image and (x _i, y _i) coordinate of the point that is complementary, with the S that counts in the optimum estimate _iValue set to 0;

A2: utilize (2) formula to calculate match point in template image of each point in this cluster (x ', y ') through H ₁The result who obtains after the determined affined transformation (x ", y ");

$\left(\begin{array}{c}{x}^{\&prime;\&prime;}\\ y^{\&prime;\&prime;}\\ 1\end{array}\right)=H_1\left(\begin{array}{c}{x}^{\&prime;}\\ y^{\&prime;}\\ 1\end{array}\right)---\left(2\right)$

If (x ", the distance of the point (x, y) in y ") and the cluster then is judged to be (x, y) an interior point less than T;

A3: judge H ₁In corresponding all point contact net support with the suspender image in relative position whether with template image in the relative position of the unique point that is complementary with it identical; If not identical, then think to comprise erroneous matching in the interior point, if identical, then judge H ₁Count in corresponding whether greater than the S that counts in the optimum estimate _iIf, greater than S _i, then with H ₁As the current optimum estimate of affine transformation matrix, and upgrade S _i

A4: return A1) continue operation, move to stipulated number after, circulation stops, if S _iValue be not 0, then think to comprise the rotation ears in the image-region of this cluster representative; The current optimum estimate H of affine transformation matrix ₁Namely as its final estimation H; Bring the coordinate on four summits of template image into formula (1), and make H ₁=H can calculate the region that rotates ears in contact net support and the suspender image;

E, from support with the suspender image extract rotation ears subimage, utilize the method based on the coboundary embroidery that the auricle fracture defect is detected, detection method comprises following steps:

E1: will from contact net support with the suspender image the rotation ears subimage that extracts carry out binaryzation, the rotation ears are separated with image background; From the subimage edge, utilize the method for boundary tracking to obtain successively the coordinate of each point on the curve of coboundary, form coboundary curvilinear coordinates sequence;

E2: the flexibility of k point is defined as it and k+50 inclination angle of putting line on the curve of coboundary, and its computing formula is as follows:

$c_k=\left\{\begin{array}{cc}\arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]>0)\\ \arctan \left[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}\right]+\mathrm{\&pi;}& \left(\mathrm{atrcan}\right[\frac{y_{k+50}-y_k}{x_{k+50}-x_k}]<0)\end{array}\right.---\left(3\right)$

In the formula, c _kThe flexibility of k point on the curve of expression coboundary, x _k, y _k, x _K+50, y _K+50, represent on the curve of coboundary k point and k+50 the transverse and longitudinal coordinate of putting, c respectively _kUnit be radian, span is 0～π;

Calculate successively the flexibility of each point on the curve of coboundary, draw out the flexibility curve of each point on the curve of coboundary;

E3: above-mentioned flexibility curve and flexibility curve are under normal circumstances made difference at corresponding point position, if comprise the point greater than 0.5 in the difference curves, then think the auricle fracture defect has occured, send the fracture defect alarm signal.