KR101707995B1 - A method of robust to the brightness change for detecting a dynamic declination of catenary - Google Patents

Abstract

The present invention relates to a method of detecting a dynamic deviation of a catenary, and more particularly, to a method of detecting a dynamic deviation of a catenary by acquiring an image through a measuring camera and a sensor disposed at the upper end of a railway, .
The present invention provides a method of detecting a pantograph and a dynamic deviation of a catenary of an input image by comparing the brightness of a reference image with a brightness of a reference image, comprising: obtaining a reference image with a different brightness condition; Determining a pantograph template representing a reference image to be compared with an input image in a set reference image; Detecting a pantograph and a catenary line from the pantograph template; And calculating dynamic excursions of the catenary and the pantograph. According to the present invention, there is no need for a separate illumination device for uniformly adjusting the brightness of the image, and there is an advantage of accurately detecting the dynamic deviation, and the advantage of reducing the calculation amount by utilizing the marker for reducing the detection area of the pantograph .

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for detecting dynamic deviation of a catenary,

The present invention relates to a method of detecting a dynamic deviation of a catenary, and more particularly, to a method of detecting a dynamic deviation of a catenary by acquiring an image through a measuring camera and a sensor disposed at the upper end of a railway, .

The electric cable is a facility to supply electric power to the railway car body stably. It is necessary to detect the deformation and damage of the electric cable by the vehicle or external factors and to keep it in the optimal condition according to the electric railway facilities regulation.

The horizontal distance from the center of the orbit of the catamaran is called the deviation. If the deviation of the catamaran is too large, the current collector deviates from the catamaran and causes an accident. Therefore, the deviation of the catenary line is defined by a certain limit. Usually, the deviation of the catenary line is within 250 mm from the center of the orbit perpendicular to the orbit. Since the current collecting device normally needs to collectively distribute the effective surface of the pantograph in order to even out the wear of the pantograph, the catenary line gives a right-left deviation from the center of the orbit. These catenary lines consist of a catenary that directly contacts an electric vehicle, supports supporting it, a feeder line and a return line. The electric cable is a part that supplies electric power to an electric vehicle, and its performance must be excellent because high voltage current is always flowing. Dynamic deviation, which indicates the alignment of the contact points between trains and electric lines in electric railways, is one of the important criteria for evaluating the safety performance of trains as one of criteria for judging whether or not a train can receive a stable current.

Korean Patent Laid-Open Publication No. 2014-0111712 (pantograph measurement system and pantograph measurement apparatus) acquires an image of a marker installed in a pantograph from a line sensor, generates a construction image from the acquired image, And the position of the pantograph is measured to estimate the position with respect to the height of the catenary.

However, the conventional image processing based pantograph measurement method can not obtain the dynamic deviation of the catenary because it only measures the position of the pantograph using the laser sensor.

Korean Patent Publication No. 2014-0111712

The present invention provides a method of increasing the accuracy of dynamic deviation measurement by detecting the position of a pantograph and the position of a catenary line without being influenced by brightness conditions by using a general camera capable of obtaining a continuous image.

According to another aspect of the present invention, there is provided a method of detecting a pantograph and a dynamic deviation of a cantilever of an input image by comparing the brightness of a reference image with a brightness of a reference image, Setting an image; Determining a pantograph template representing a reference image having a brightness most similar to an input image in a set reference image; Detecting a pantograph and a catenary line from the pantograph template; And calculating dynamic excursions of the catenary line and the pantograph, wherein the dynamic deviation of the input image is detected without being interfered with the brightness, and a method of detecting the dynamic deviation of the pantograph and the pantograph

Preferably, the reference image may be collected by class for different classes such as a normal situation, a background interference situation, an arc situation, a tunnel situation, and the like.

Preferably, if updating of the class is required due to the difference of the input image, the pantograph template of the input image may be compared with the average brightness of the class to reset the pantograph template serving as the basis of detection of the pantograph in the input image.

Preferably, the determining of the pantograph template may include calculating a feature vector of an input image by digitizing features of the input image such as brightness, variance, and correlation of the input image; And comparing the feature vector of the input image with the feature vector of the pantograph template to find the feature vector of the pantograph template that is the same as or the most similar to the feature vector of the input image to determine the target class and the target pantograph template.

Preferably, the detecting of the catenary includes simultaneously performing binarization and Hough transform of an image for digitizing an input image; And setting the brightness, height, width, area, and angle of the electric line, and comparing the binarized value with the pantograph template.

Preferably, the step of detecting the pantograph may utilize a template matching technique for extracting the brightness value of the input image and comparing the brightness value with the pantograph template.

Preferably, calculating the dynamic deviation comprises: detecting a contact surface corresponding to a horizontal plane where the catenary and the pantograph are in contact; Detecting a center point of the pantograph from the contact surface; And obtaining the intersection of the contact line and the contact line, and calculating the dynamic deviation of the intersection from the center point.

Preferably, the step of detecting the contact surface comprises: obtaining a straight line corresponding to the contact surface between the pantograph and the catenary using the Hough transform from the input image; And estimating a point where the straight line rises and comes into contact with the catenary as the contact surface.

Preferably, the step of detecting the contact surface comprises: detecting a plurality of markers linearly attached to the top of the pantograph to obtain a straight line; And estimating a point where the straight line rises and comes into contact with the catenary as the contact surface.

Preferably, the step of detecting the center point comprises the steps of: extracting the position of the marker attached in the middle among the markers linearly attached to the top of the pantograph; And estimating a point at which the marker attached at the middle is located as a center point.

According to the present invention, there is no need for a separate illumination device for uniformly adjusting the brightness of the image, and there is an advantage of accurately detecting the dynamic deviation.

Further, the present invention has an advantage of reducing the amount of calculation by utilizing a marker for reducing the detection area of the pantograph.

FIG. 1A is a block diagram of a method of detecting a dynamic deviation of a catenary without being affected by a brightness change according to the present invention.
FIG. 1B is a block diagram showing the detailed steps constituting step S10 in FIG. 1A;
2 shows a device for detecting a dynamic deviation of a catenary according to an embodiment of the present invention.
3 illustrates a process of calculating a feature vector according to an embodiment of the present invention and a step of classifying a reference image into classes based on a feature vector.
4 is a view showing a reference image according to an embodiment of the present invention divided into classes.
FIG. 5 is a diagram illustrating a process of updating and re-setting classes classified according to an embodiment of the present invention.
FIG. 6 shows a case where feature vectors obtained according to an embodiment of the present invention are different for each class.
FIG. 7 shows a detection unit detecting a pantograph in a pantograph template according to an embodiment of the present invention.
FIG. 8 shows the detection of a catenary according to the fusion of Hough transform and binarization according to the embodiment of the present invention.
9A shows a correctly detected pantograph image according to an embodiment of the present invention.
Figure 9b shows an incorrectly detected pantograph image according to an embodiment of the present invention.
10A shows a process of detecting a contact surface and a center point using an attachment marker according to an embodiment of the present invention.
10B shows an attachment marker according to an embodiment of the present invention.

Hereinafter, the present invention will be described in detail with reference to the accompanying drawings. However, the present invention is not limited to or limited by the exemplary embodiments. Like reference numerals in the drawings denote members performing substantially the same function.

Fig. 1 is a block diagram showing in sequence the entire steps of the present invention, and Fig. 2 shows a device 1 for detecting a dynamic deviation of a catenary according to an embodiment of the present invention. Referring to FIG. 2, the device 1 for detecting a dynamic deviation of a catenary includes a training unit 10, a detection unit 30, and a contact point calculation unit 50. The dynamic deviation detection device 1 of the catenary line can detect the dynamic deviation of the pantograph 70 and the catenary line 90 of the input image compared with the brightness of the reference image. The training unit 10 may acquire a plurality of reference images by classifying the similar types with different brightness conditions and acquire an optimal feature vector 101. [ The training unit 10 may compare the feature vector generated from the selected reference image with the feature vector obtained from the input image to determine a pantograph template 80 representing a reference image to be compared. The detection unit 30 can detect the pantograph 70 and the catenary line 90 existing in the background image from the pantograph template 80. [ The contact intersection point calculation unit 50 detects the contact surface 709 and the center point 707 from the pantograph template 80 and detects the dynamic of the catenary 90 and the pantograph 70 using the contact surface 709 and the center point 707. [ The deviation can be calculated. The device 1 for detecting the dynamic deviation of the catenary can detect the dynamic deviation of the input image without being interfered with the brightness.

The training unit 10 may define different brightness conditions by matching a predetermined number of reference images 80 according to image acquisition conditions. The training unit 10 may calculate the feature vector 101 by calculating the brightness, variance, correlation, and the like of the reference image, and classify the reference image 80 into the class 103 based on the feature vector 101. The training unit 10 may acquire a plurality of similar images for defining the reference image 80 using cameras and sensors installed on the vehicle, and the obtained images may have different degrees of lightness, exposure values of the lenses, . For example, when the sunlight is reflected on the pantograph 70, a bright light image of the pantograph 70 is obtained, and in a cloudy situation, the pantograph 70 may appear relatively dark. In this case, there is a possibility that the detection performance of the pantograph 70 may be largely deteriorated. Therefore, a process of collecting the reference image 80 for each type should be performed first by defining a case where the acquired pantograph images are greatly different. The training unit 10 needs to collect the reference image for each defined type and to classify the class 103 in a situation where the feature vector 101 differs from the reference image of the same kind, SVM (Support Vector Machine). The support vector machine can store the feature information of the digitizable image such as the average of the brightness of the reference image 80, the contrast ratio, etc. in the form of the feature vector 101. The feature vector 101 stores the digitizable information in a vector form, and the stored information may differ substantially depending on the brightness condition.

FIG. 3 shows a process of calculating a feature vector 101 according to an embodiment of the present invention and a step of classifying a reference image into a class 103 based on a feature vector 101. FIG. Referring to FIG. 3, the training unit 10 can minimize an error in classifying the class 103 using a support vector machine. The training unit 10 may perform a linear classification process for performing training to classify reference images. The training unit 10 may extract features from a reference image including a plurality of pantographs 70 and catenary lines 90 obtained under various brightness conditions and quantify them according to the brightness condition. The training unit 10 may provide the quantified brightness condition and provide information so that the detection unit 30 can determine the pantograph template 80 according to the numerical value.

The training unit 10 can quantify the average of brightness, the brightness of the catenary line 90, and the contrast ratio with the pantograph 70 in the reference image, and store the feature vector 101 as a feature vector. The feature vector 101 can be digitized so as to represent one reference image. The training unit 10 can classify the reference image into classes 103 by referring to the feature vector 101.

4 is a view showing a reference image divided into classes 103 according to an embodiment of the present invention. Referring to FIG. 4, the support vector machine may calculate a brightness condition in a reference image, store the feature as a feature vector 101, and refer to the stored feature vector 101 to classify the reference image for each class 103. The support vector machine can classify a plurality of reference images measured by a camera into classes 103 such as bright, dark, tunnel interval, and background interference. However, the present invention is not limited to this, and the class 103 can be set to be subdivided according to the brightness condition. The training unit 10 may repeat the above-described classification process every time reference images are acquired. Since the training unit 10 can classify the class 103 through the feature vector 101, the process of calculating the feature vector 101 is important.

5 is a diagram illustrating a process of resetting a class 103 according to an embodiment of the present invention. Referring to FIG. 5, when the difference between the average brightness of the pantograph zone 104 and the average brightness of the class 103 among the updated feature vectors 101 exceeds the set value, the updated feature vector 101 is calculated The class 103 is reset. In an embodiment of the present invention, if the difference in average brightness between classes 103 is found and half or more of the difference is set as the reset threshold 105, when the brightness difference of the current image exceeds the reset threshold 105 The class 103 is reset.

Figure 6 shows a simplified two-dimensional illustration of a feature vector 101 for each class 103 according to an embodiment of the present invention. Dimensional characteristic vector 101 by plotting the average brightness of the pantograph area 104 and the pantograph lower area in each of the image obtained in the tunnel and the normal situation image in the background of the clear sky.

The training unit 10 of the present invention is capable of setting a determination boundary as to which class 103 the feature vector 101 of the class 103 corresponds to using the support vector machine algorithm. However, in the present invention, the method of determining the decision boundary of the feature vector 101 for the class 103 is not limited to the support vector machine.

The training unit 10 collects a reference image by selecting a class 103 such as a normal situation, a tunnel situation, a background interference situation, an arc occurrence situation, etc., which need to be classified; Constructing a feature vector (101) by digitizing features of the input image, such as brightness, correlation, and variance; A reset step for updating the class (103); Comparing a feature vector of an input image with a feature vector of a current class to determine a class 103 most suitable for an input image; And setting the pantograph template 80 corresponding to the determined class 103. The detection unit 30 can determine the position of the pantograph 70 using the reference pantograph image 80 determined by the training unit 10 and calculate the dynamic deviation of the catenary 90 based on the position of the pantograph.

Fig. 7 shows the detection unit 30 detecting the pantograph 70 in the pantograph template 80 according to the embodiment of the present invention. Referring to FIG. 7, the detection unit 30 can detect the pantograph 70 from the pantograph template 80 determined by the training unit 10. The detection unit 30 can extract the pantograph 70 from the pantograph template 80 using the following equation (1).

[Equation 1]

(I: input image, T: pantograph template image, w: input vector,

X: coordinate in the pantograph template image, p: linear transformation parameter of the input vector)

The detection unit 30 can determine the optimal pantograph template 80 through the identification process and perform the linear transformation w (x, p) of the input image. The detection unit 30 can obtain a point having the highest degree of similarity in the pantograph template 80 through this calculation.

FIG. 8 shows a manner of detecting the catenary line 90 according to the fusion of the Hough transform, which is a method of detecting a straight line, and the binarization, which is a method of distinguishing a background from a foreground. Referring to FIG. 8, the detection unit 30 performs binarization on an input image. Limiting the brightness, height, width, area, and angle of the catenary line 90 and comparing the binarized values to the pantograph template 80. [ The detection unit 30 may utilize a template matching technique for extracting the brightness of the input image and comparing the brightness of the input image with the pantograph template 80 using Equation 1 when the pantograph 70 and the catenary 90 are detected.

9A shows an accurately detected pantograph image 70 according to an embodiment of the present invention. FIG. 9B shows an incorrectly detected pantograph image 70 according to an embodiment of the present invention. Referring to FIG. 9A, when the brightness condition is similar, the detection unit 30 can detect the actual position (green) and the detected position (red) in a substantially identical manner. In this case, it is possible to accurately calculate the dynamic deviation of the catenary 90. Referring to FIG. 9B, when the lightness condition is not met, the detection unit 30 may detect the actual position (green) and the detected position (red) differently, and may incorrectly calculate the dynamic deviation. Therefore, it can be seen that the reference pantograph template 80 must be accurately set in order to find the correct pantograph 70 and the caten grid 90. [

The contact intersection point calculation section 50 can calculate the dynamic deviation of the catenary line 90 using the detected pantograph 70 and the catenary line 90. The contact intersection point calculation unit 50 detects a contact surface 709 corresponding to a horizontal plane where the detected catenary line 90 and the pantograph 70 contact; Detecting a center point (707) of the pantograph (70) from the contact surface (709); And an intersection of the catenary line 90 and the contact surface 709. The contact intersection point calculation section 50 can calculate the dynamic deviation of the intersection point from the center point 707. [

The step of detecting the contact surface 709 in the contact intersection point calculation unit 50 includes the steps of obtaining a straight line of the pantograph 70 from the input image using Hough transform; And estimating as a contact surface 709 a point where the straight line rises and contacts the catenary 90. [ That is, when the pantograph 70 is brought into close contact with the catenary 90 suspended in the air using air pressure, if it finds the intersection of the horizontal plane of the pantograph 70 and the catenary 90 through the binarized image, Similar data can be obtained. In this case, a gap may be generated between the pantagraph 70 and the caten grid 90 due to the vibration, but the difference between the pantograph 70 and the caten grid 90 is very small, so that the approximation can be included within the error range. The contact intersection point calculation section 50 can calculate the dynamic deviation of the catenary line 90 by detecting the contact surface 709, the center point 707 and the intersection point and calculating the distance from the center point 707 to the intersection point. The accurate value can be measured by substituting the proportional equation using the length of the pantograph 70 at the detected intersection and converting the distance detected in pixel unit in mm.

Although measurement using the dynamic deviation detection device 1 of the catenary line can be performed as described above, there is a method of simplifying the detection of the patagraph using the attachment marker.

10A shows a process of detecting a contact surface 709 and a center point 707 using an attachment marker according to an embodiment of the present invention. 10B shows an attachment marker according to an embodiment of the present invention. Referring to FIG. 10A, a separate marker as shown in FIG. 10B can be attached to the pantograph 70. In this case, the detection of the pantograph 70 and the method of detecting the contact surface 709 can be simplified. The detection unit 30 must find a detection region corresponding to the image size of the pantograph 70 from the entire image as described above. However, when the marker is attached, only the portion with the marker is detected, so that the calculation amount can be greatly reduced.

The step of detecting the contact surface 709 in the detection unit 30 includes the steps of obtaining a straight line from the marker attached to the top of the pantograph 70; And estimating as a contact surface 709 a point at which the straight line is raised to contact the catenary 90. [ The step of detecting the center point 707 in the detection unit 30 includes the steps of extracting a marker attached in the middle among the markers linearly attached to the top of the pantograph 70; And estimating a point at which the marker attached in the middle is located as a center point 707, thereby reducing the amount of computation.

In the embodiment of the present invention, three markers are attached to the pantograph 70. The first marker 701 can be attached to the left of the pantograph 70, the second marker 703 can be attached to the middle, and the third marker 705 can be attached to the right. The detection unit 30 can linearly detect the three points detected when three markers are installed horizontally, and estimate the position at which the contact is made as the contact surface 709. [ The detection unit 30 can estimate the center point 707 where the marker located in the middle is located. Therefore, the process of estimating the contact surface 709 and the center point 707 can be simplified by extracting the marker, thereby reducing the amount of computation.

While the present invention has been particularly shown and described with reference to exemplary embodiments thereof, it is to be understood that the invention is not limited to the disclosed exemplary embodiments. will be. Therefore, the scope of the present invention should not be limited to the above-described embodiments, but should be determined by all changes or modifications derived from the scope of the appended claims and equivalents of the following claims.

1: Dynamic deviation detection device of catenary line
10: training part 30: detection part
50: contact point calculation unit
70: Pantograph 80: Pantograph template
90: electric line 101: characteristic vector
103: Class 104: Pantograph Area
701: first marker 703: second marker
705: Third marker
707: Center point 709: Contact surface

Claims (10)

A method for detecting a dynamic deviation of a pantograph and an electric line of an input image by comparing with a brightness of a reference image,
Acquiring the reference image with a different brightness condition, and setting the reference image by matching the brightness condition;
Determining a pantograph template representing a reference image to be compared with the input image in the set reference image;
Detecting the pantograph and the electric line from the pantograph template; And
And calculating dynamic excursions of said catenary and said pantograph. ≪ Desc / Clms Page number 15 >
The method according to claim 1,
Wherein the step of setting the reference image comprises:
And collecting the reference images for different classes of normal conditions, background interference conditions, arc conditions, and tunnel conditions on a class-by-class basis.
3. The method of claim 2,
When the class needs to be updated due to the difference of the input image,
And comparing the average brightness of the class with the pantograph region of the input image to reset the pantograph template serving as a reference for detecting the pantograph in the input image.
4. The method according to any one of claims 1 to 3,
Wherein determining the pantograph template comprises:
Calculating a feature vector of the input image by digitizing brightness, variance, and correlation characteristics of the input image in each region; And
Comparing the feature vector of the input image with a feature vector of the pantograph template to determine a feature vector of the pantograph template that is the same as or closest to the feature vector of the input image, and determining a target class and a target pantograph template, And detecting the dynamic deviation of the pantograph.
4. The method according to any one of claims 1 to 3,
The step of detecting the electric line includes:
Simultaneously performing binary conversion and Hough transform of an image digitizing the input image; And
Setting the brightness, height, width, area and angle of the electric line, and comparing the binarized value with the pantograph template.
6. The method of claim 5,
Wherein detecting the pantograph comprises:
Wherein a template matching technique is used for extracting brightness values of the input image and comparing the extracted brightness values with the pantograph template.
The method according to claim 1,
Wherein calculating the dynamic deviation comprises:
Detecting a contact surface corresponding to a horizontal plane where the catenary and the pantograph contact;
Detecting a center point of the pantograph from the contact surface; And
And obtaining an intersection between the electric line and the contact surface,
And calculating the dynamic deviation of the intersection from the center point.
8. The method of claim 7,
Wherein the step of detecting the contact surface comprises:
Obtaining a straight line corresponding to a contact surface between the pantograph and the electric line using the Hough transform from the input image; And
And raising the straight line to estimate a point of contact with the catenary as a contact surface.
9. The method according to claim 7 or 8,
Wherein the step of detecting the contact surface comprises:
Detecting a plurality of markers linearly attached to the top of the pantograph to obtain a straight line; And
And estimating, as a contact surface, a point where the straight line rises and comes into contact with the electric line, to detect a dynamic deviation of the pantograph.
10. The method of claim 9,
The step of detecting the center-
Extracting a position of a marker attached in the middle among a plurality of markers linearly arranged on an upper end of the pantograph; And
And estimating a point at which the marker attached at the middle is located as the center point.

Images