A contact line rapid identification method and system based on parallel architecture

Abstract

This application discloses a method and system for rapid contact line identification based on a parallel architecture. The daytime working mode of the identification method includes: collecting a contact line image to obtain a known image; normalizing the one-dimensional data of the known image to obtain the mean of the one-dimensional data at the bull's-eye position; constructing a one-dimensional matching template for the contact line based on the mean; calculating the template matching results of the one-dimensional image data at the bull's-eye and the pixels above and below the bull's-eye position, with the contact line position being the maximum template matching result; the nighttime working mode includes: using one-dimensional template matching to identify the image heights of 30, H / 2, and H-30 at the start, center, and end positions, respectively, as Pos0, Pos1, and Pos2; determining the positional relationship between the center position and the start and end positions, and interpreting the contact line position. This method improves the accuracy and speed of rapid contact line detection, meeting the detection requirements for contact network geometric parameters at branch and anchor sections.

Description

Technical Field

[0001] The present invention belongs to the field of contact network monitoring and safety, and in particular relates to a contact line rapid identification method and system based on a parallel architecture. Background Art

[0002] With the rapid development of electrified railways, the catenary, a key component of railway electrification projects, is a specialized transmission line that supplies power to electric locomotives. As an open-air installation with no backup outdoor power supply, the catenary is subject to harsh weather conditions such as frost, rain, snow, wind, and sun. Damage to the catenary disrupts service, causing significant losses to railway transportation. Therefore, effective monitoring of the actual operating status of the catenary is essential to promptly detect faults.

[0003] Common contact wire inspection methods are categorized as contact and non-contact. Contact inspection methods require a large number of inspection devices, are complex, and are prone to causing additional wear and tear on the contact wire itself, impacting the transmission performance of the contact network. The mainstream form of non-contact measurement is image processing-based inspection, which analyzes and processes contact wire images to accurately identify contact wire problems. This method offers high accuracy, does not impact driving safety, and is highly real-time. Currently, commonly used contact network inspection equipment, such as the TYJJ-BX intelligent contact network inspection equipment developed by Beijing Precision Weiye Measurement and Control Technology, utilizes an image-based inspection mode, but requires manual alignment. The image-processing-based contact wire recognition algorithm is less intelligent, making it incapable of automated online inspection. The TGGT-DT-3 contact line detection uses a single-line laser radar to achieve real-time calculation of the contact line geometric parameters. Due to the limitations of the ranging accuracy of the laser radar itself, other similar methods such as those described in CN20494427U and CN204432429U generally have low intelligence and cannot meet the detection requirements of the contact network geometric parameters at the branch and anchor sections. Summary of the Invention

[0004] The purpose of the present invention is to provide a contact line rapid identification method based on a parallel architecture. The technical problem solved is to solve the problems of low intelligence level of contact line rapid identification, poor detection accuracy, and inability to meet the detection requirements of contact network geometric parameters at line branches and anchor sections.

[0005] The technical solution adopted in the present invention is:

[0006] A contact line fast identification method based on parallel architecture, including daytime working mode and nighttime working mode,

[0007] Daytime working mode includes the following steps:

[0008] Collect the contact line image to obtain a known image (W×H);

[0009] Normalize the one-dimensional data of the known image to obtain the mean value of the one-dimensional data at the bull's-eye position;

[0010] Construct a one-dimensional matching template for the contact line based on the mean value of the one-dimensional data at the bull's-eye position;

[0011] According to the one-dimensional matching template and the known image, calculate the template matching results of the one-dimensional image data at the bull's-eye and 3 - 10 pixels above and below the bull's-eye position. The maximum value of the template matching result is identified as the contact line position;

[0012] The night working mode includes the following steps:

[0013] Use one-dimensional template matching to respectively identify the template matching results of the known images where the contact lines are located at three different positions with image heights of 30±a, H / 2, and H - 30, where a∈[0,10]. Denote them as Pos0 - the recognition position of the contact line center (i0, H / 2), Pos1 - the recognition position of the contact line start end (i1, 30±a), and Pos2 - the recognition position of the contact line end (i2, H - 30±a);

[0014] Based on the positional relationship between the recognition position of the contact line center and the recognition positions of the contact line start end and the contact line end, judge the situation of the contact line position.

[0015] When the recognition position of the contact line center is not between the recognition positions of the contact line start end and the contact line end, that is, when i0>i1 and i0>i2, or i0<i1 and i0<i2, discard the contact line position recognized by the recognition position of the contact line center, calculate the recognition distances in the width direction between the start end, center, and end, which are Dis0 = i2 - i1, Dis1 = i0 - i1, Dis2 = i0 - i2, and make separate judgments according to the relationship between the recognition distances and the one-dimensional template width (model_width).

[0016] When Dis0<model_width, fit the straight line equation of the contact line y = kx + b. If the fitted straight line is a vertical straight line, the contact line position is: (x, y) = (0.5·(i1 + i2), H / 2). If the fitted straight line is not a vertical straight line, the actual contact line position is (x, y) = (1 / k(H / 2 - b), H / 2).

[0017] When Dis0>model_width, it includes the following situations:

[0018] When Dis1<model_width and Dis2<model_width, the actual contact line position is

[0019] When Dis1 < model_width and Dis2 > model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i1, H / 2), model_value(i0, H / 2));

[0020] When Dis1 > model_width and Dis2 < model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i2, H / 2), model_value(i0, H / 2));

[0021] When Dis1 > model_width and Dis2 > model_width, the actual position of the contact line is (x, y) = (min(d1, d2, d3), H / 2), where d1 = i0 - W / 2, d2 = i1 - W / 2, d3 = i2 - W / 2.

[0022] The central recognition position of the contact line is between the start-end recognition position and the end-end recognition position of the contact line. That is, when i2 > i0 > i1 or i2 < i0 < i1, it is judged separately according to the relationship between the recognition distance and the one-dimensional template width.

[0023] When Dis0 < model_width, the linear equation of the fitted contact line is y = kx + b. If the fitted line is a vertical line, the position of the contact line is: (x, y) = (0.5·(i1 + i2), H / 2). If the fitted line is not a vertical line, the actual position of the contact line is (x, y) = (1 / k(H / 2 - b), H / 2).

[0024] When Dis0 > model_width, the following situations are included:

[0025] When Dis1 < model_width and Dis2 < model_width, the actual position of the contact line is (x, y) = (i0, H / 2);

[0026] When Dis1 < model_width and Dis2 > model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i1, H / 2), model_value(i0, H / 2));

[0027] When Dis1 > model_width and Dis2 < model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i2, H / 2), model_value(i0, H / 2));

[0028] When Dis1>model_width and Dis2>model_width, the actual position of the contact line is (x,y)=(min(d1,d2,d3),H / 2), where d1=i0-W / 2, d2=i1-W / 2, and d3=i2-W / 2.

[0029] The contact line positions identified based on the three ends of the contact line, namely the starting end, the center and the end end, are used to set the contact line disturbance range to 5-10.

[0030] The one-dimensional template matching calculation formula is:

[0031]

[0032] Where pt(i, j) is the grayscale value of the pixel at the position on the image; model_width is the width of the one-dimensional template;

[0033]

[0034] A contact line rapid identification system based on a parallel architecture includes an image acquisition device for acquiring contact line images to obtain known images; an image processing device electrically connected to the image acquisition device, for receiving known images and identifying models and contact line images according to preset targets, and adopting a daytime working mode and a nighttime working mode to identify the contact line positions in the contact line images during the day and at night, respectively.

[0035] In summary, this application has at least the following beneficial technical effects:

[0036] (1) The present invention adopts a parallel architecture design, and the contact line position is identified at three positions simultaneously, ensuring that the contact line position that best meets the requirements is selected as the actual contact line position in each identification. This makes up for the drawbacks of incorrect calculation of key parameters such as contact line guide height and pull-out value caused by low contact line position identification accuracy and poor stability in a single identification mode;

[0037] (2) The present invention adopts a three-end synchronous search mode of the starting, center and end contact lines in the night measurement mode, which can effectively solve the stable identification of the contact line suspension string, arm and positioning point position, and the identification is faster;

[0038] (3) The present invention sets the contact line disturbance range according to the contact line positions identified at the three ends of the contact line: the starting end, the center, and the end, and determines the disturbance direction by the specific disturbance value and position within the disturbance range, thereby determining the true position information of the contact line, and ensuring maximum coverage of the contact line position identification environment under the conditions of the contact line three-end identification position points. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1Flowchart for the contact line identification method under daytime working mode;

[0040] Figure 2 Flowchart for realizing the contact line identification method in night working mode;

[0041] Figure 3 The contact line identification results for different fixed-point measurement positions in daytime working mode;

[0042] Figure 4 These are the contact line identification results for different fixed-point measurement positions in night working mode. DETAILED DESCRIPTION

[0043] The present application is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0044] The embodiment of the present application discloses a method for rapid contact line identification based on a parallel architecture.

[0045] Reference Figure 1-2 , a contact line fast identification method based on parallel architecture, including fast identification methods in daytime working mode and nighttime working mode.

[0046] Daytime working mode:

[0047] Reference Figure 1 This is a flow chart of the contact line identification method in the daytime working mode. The contact line rapid identification in the daytime working mode includes the following steps:

[0048] Step 1: Given an image size of W×H, the grayscale value of the pixel at position (i, j) on the image is denoted as pt(i, j). The one-dimensional data of the image at the bull's-eye position (W / 2, H / 2) is normalized to obtain the mean ave of the one-dimensional data at the bull's-eye position. The calculation process can be expressed as:

[0049]

[0050] Step 2: Construct a one-dimensional matching template for the contact line. Set the width of the one-dimensional matching template, model_width, to 21, 28, or 35, and set the ratio of the rising width, center width, and falling width to 1:6:1.

[0051] Step 3: Traverse the entire width of the image, calculate the one-dimensional image data template matching result model_value at 3 pixels above and below the center of the target, and find the maximum value of the template matching result, which is identified as the contact line position. The one-dimensional template matching calculation formula can be expressed as:

[0052]

[0053] Reference Figure 1The template matching calculation results of the bull's eye position Pos'0, Pos'1 and Pos'2 at 3 pixels above and below the bull's eye position shown in can be expressed as:

[0054] Pos'0=model_value(i0,H / 2)

[0055] Pos'1=model_value(i1,H / 2+3)

[0056] Pos'2=model_value(i2,H / 2-3) (3)

[0058] After traversing the entire image width at positions j = H / 2, H / 2+3, and H / 2-3, we obtain the three contact line coordinates (i0, H / 2), (i1, H / 2+3), and (i2, H / 2-3) under different j values. The position with the maximum template matching result is the contact line identification position:

[0059] Pos'(x,y)=max(Pos'0,Pos'1,Pos'2), where 1 <i<w;

[0060] (x, y) is the actual contact line identification position, where the template matching value is the largest.

[0061] Night working mode:

[0062] The rapid contact line identification method in night mode is suitable for real-time contact line measurement under complex and changing conditions. During fixed-point measurement, both the contact line frame and the contact line are highlighted, making it difficult to determine the contact line's position. In practice, the image coordinates of the contact line's initial end, the center of the target, and the end must be measured to update the image coordinates of the contact line's center in real time.

[0063] The contact line quick identification in night working mode includes the following steps:

[0064] Step 1: Use one-dimensional template matching to identify the image coordinates of the contact lines at three different positions, j = 30, H / 2, and H-30, respectively, for the initial end, center, and end positions. The template matching results are recorded as Pos0, Pos1, and Pos2 (Note: Pos1 here is the starting position of the contact line, and Pos2 is the end position of the contact line, which has a different meaning from the daytime working mode). It is expressed as:

[0065] Pos0(i0,H / 2)=model_value(i0,H / 2)

[0066] Pos1(i1,30)=model_value(i1,30)

[0067] Pos2(i2, H - 30) = model_value(i2, H - 30) (4)

[0069] Step 2: According to different situations between the image coordinates of the starting end, the center, and the ending end, determine whether the center position is between the image coordinates of the starting end and the ending end, and intelligently interpret the position of the contact wire:

[0070] 2.1. Case 1: If the contact wire recognized at the center position is at a position to the left or right of the image coordinates of the starting end and the ending end, and the recognized position Pos0 of the center point of the contact wire is not between the recognized position Pos1 of the starting end and the recognized position Pos2 of the ending end, that is, i0 > i1 and i0 > i2, or i0 < i1 and i0 < i2, discard the position of the contact wire recognized at the center position, and calculate the recognized distances in the width direction between the starting end, the center, and the ending end respectively according to the following situations, denoted as:

[0071] Dis0 = i2 - i1

[0072] Dis1 = i0 - i1

[0073] Dis2 = i0 - i2 (5)

[0075] 2.1.1. If the distance between the starting end and the ending end is less than the width of one template Dis0 < model_width (when comparing Dis0, Disl, Dis2 with model_width, model_width = one-dimensional template width + perturbation range, and the perturbation range in this embodiment is 5), fit the straight line equation of the contact wire y = kx + b, and judge whether the fitted straight line is a vertical straight line. If it is a vertical straight line, the position of the contact wire is: (x, y) = (0.5·(i1 + i2), H / 2). If the position of the contact wire is not a vertical straight line, then calculate the x value Pos = 1 / k(H / 2 - b) at y = H / 2 of the straight line, and the actual position of the contact wire is

[0076]

[0077] 2.1.2. If the distance between the starting end and the ending end is greater than the width of one template Dis0 > model_width, the following situations may occur:

[0078] 2.1.2a). If the distances between the starting end and the center, and between the ending end and the center position are both less than the width of one template (Dis1 < model_width and Dis2 < model_width), it means that all three points are near the contact wire, and both the starting end and the ending end are on the contact wire. The recognition result is accurate. Judge the image point closest to the center position as the position of the recognized contact wire, that is,

[0079]

[0080] 2.1.2 b). If the distance of the starting end from the center is relatively close (Dis1 < model_width and Dis2 > model_width), it indicates that the starting end and the center are on the contact line. According to Equation (2), calculate the matching degree model_value(i1, H / 2) at (i1, H / 2), and compare it with the matching degree model_value(i0, H / 2) at the center. The one with the higher matching degree is updated as the latest position of the contact line:

[0081] Pos(x, y) = max(model_value(i1, H / 2), model_value(i0, H / 2)) (7)

[0083] (x, y) is the actual recognition position of the contact line, where the template matching value is the largest;

[0084] 2.1.2 c). If the distance of the ending end from the center is relatively close (Dis1 > model_width and Dis2 < model_width), it indicates that the ending end and the center are on the contact line. According to Equation (2), calculate the matching degree model_value(i2, H / 2) at (i2, H / 2), and compare it with the matching degree model_value(i0, H / 2) at the center. The one with the higher matching degree is updated as the latest position of the contact line:

[0085] Pos(x, y) = max(model_value(i2, H / 2), model_value(i0, H / 2)) (8)

[0087] (x, y) is the actual recognition position of the contact line, where the template matching value is the largest;

[0088] 2.1.2 d). If the distances of both the starting end and the ending end positions from the center are greater than the width of one template (Dis1 > model_width and Dis2 > model_width), it indicates that any two of the starting end, the ending end, and the center are not on the contact line. Then calculate and judge the distances between the three contact line recognition positions and the abscissa of the bull's-eye as follows:

[0089] d1 = i0 - W / 2

[0090] d2 = i1 - W / 2

[0091] d3 = i2 - W / 2 (9)

[0093] The one with the closest distance is recognized as the position information of the contact line at the bull's-eye, (x, y) = (min(d1, d2, d3), H / 2).

[0094] 2.2. Case 2: If the recognized contact line at the central position is between the image coordinates of the start and end, that is, i2 > i0 > i1, or i2 < i0 < i1, recalculate according to the following cases:

[0095] 2.2.1. If the distance between the start and the end is less than the width of one template Dis0 < model_width, fit the linear equation of the contact line y = kx + b, and calculate the position information of the contact line at the bull's-eye, the same as 2.1.1;

[0096] 2.2.2. If the distance between the start and the end is greater than the width of one template Dis0 > model_width, the following situations may occur:

[0097] 2.2.2a). If the distances from the start to the center and from the end to the center are both less than the width of one template (Dis1 < model_width and Dis2 < model_width), it means that all three points are on the contact line, the recognition result of the central contact line is accurate, and the position of the contact line at the bull's-eye is not updated, (x, y) = (i0, H / 2);

[0098] 2.2.2b). If the distance from the start to the center is relatively close (Dis1 < model_width and Dis2 > model_width), judge the matching degree model_value(i1, H / 2) at the position (i1, H / 2) of the start, and compare it with the matching degree model_value(i0, H / 2) at the center. The one with the higher matching degree is updated as the latest position of the contact line, and the calculation method is the same as 2.1.2b).;

[0099] 2.2.2c). If the distance from the end to the center is relatively close (Dis1 > model_width and Dis2 < model_width), it means that the end and the center are on the contact line. Calculate the matching degree model_value(i2, H / 2) at the position (i2, H / 2) according to formula (2), and compare it with the matching degree model_value(i0, H / 2) at the center. The one with the higher matching degree is updated as the latest position of the contact line, and the calculation method is the same as 2.1.2c);

[0100] 2.2.2d). If the distance between the start and end positions and the center is greater than the width of a template (Dis1>model_width and Dis2>model_width), it means that any two of the start, end, and center positions are not on the contact line. Then determine the distance between the three contact line identification positions and the center of the target. The one with the closest distance is identified as the contact line position information at the center of the target. The calculation method is the same as 2.1.2d).

[0101] The practical effect of this method can be illustrated by the following experiment

[0102] (1) Test experiment implementation conditions:

[0103] The operating environment is vs2015 (win32), and it relies on the database opencv3.4.5

[0104] High-resolution camera (MAKO G-223B), lens focal length 60mm

[0105] (2) Test experiment implementation content and results

[0106] Test 1: Figure 3 In the daytime working mode, according to the method for rapid contact line identification based on a parallel architecture proposed by the present invention, the contact line identification result obtained by testing takes 3ms.

[0107] Test 2: Figure 4 In the night working mode, according to the present invention, a fast contact line identification method based on a parallel architecture is proposed. The contact line identification result obtained by testing takes 5ms.

[0108] Figure 3 and Figure 4 The test results show that regardless of the night working mode or the day working mode, the contact line rapid identification method based on the parallel architecture proposed in the present invention can achieve stable and rapid identification.

[0109] The above is only a specific example of the present invention and does not constitute any limitation to the present invention. Obviously, after understanding the content and principles of the present invention, professionals in this field may make various modifications and changes in form and details without departing from the principles and structure of the present invention. However, these modifications and changes based on the ideas of the present invention are still within the scope of protection of the claims of the present invention.

[0110] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A contact line rapid identification method based on a parallel architecture, characterized by: It includes a daytime working mode and a nighttime working mode. The daytime working mode includes: Collect the image of the contact wire to obtain a known image of W×H. Normalize the one-dimensional data of the known image to obtain the mean value of the one-dimensional data at the bull's-eye position. Construct a one-dimensional matching template for the contact wire based on the mean value of the one-dimensional data at the bull's-eye position. According to the one-dimensional matching template and the known image, calculate the template matching results of the one-dimensional image data at the bull's-eye and 3 - 10 pixels above and below the bull's-eye position. The maximum value of the template matching results is identified as the position of the contact wire. The rapid recognition of the contact wire in the daytime working mode includes the following steps: Step 1: The size of the known image is W×H. The gray value of the pixel at the position (i, j) on the image is represented as pt(i, j). Normalize the one-dimensional data of the image at the bull's-eye position (W / 2, H / 2) to obtain the mean value ave of the one-dimensional data at the bull's-eye position. Step 2: Construct a one-dimensional matching template for the contact wire. The width model_width of the one-dimensional matching template is set to 21, 28, or 35, and the ratios of the rising width, the central width, and the falling width are set to 1:6:

1. The calculation formula of the one-dimensional matching template is: Among them, pt(i, j) is the grayscale value of the pixel at the position on the image; model_width is the width of the one-dimensional matching template; ; Step 3: Traverse the entire width of the image, calculate the template matching results model_value of the one-dimensional image data at 3 pixels above and below the bull's-eye position, and find the maximum value of the template matching results, which is identified as the position of the contact wire. After traversing the entire width of the image at the positions j = H / 2, H / 2 + 3, and H / 2 - 3 respectively, three contact wire coordinate values (i0, H / 2), (i1, H / 2 + 3), and (i2, H / 2 - 3) at different j values are obtained. The position with the maximum template matching result is the contact wire recognition position. Pos'(x, y)=max(Pos'0, Pos'1, Pos'2), where 1 < i < w; (x, y) is the actual contact wire recognition position, and the template matching value is the largest at this position. The nighttime working mode includes: Use the one-dimensional matching template to respectively identify the template matching results of the known images where the contact wires are located at n different positions with different image heights. According to the positional relationship between different recognition positions of the contact wires, judge the position situation of the contact wires, where n is a positive integer not less than 3. The rapid recognition of the contact wire in the nighttime working mode includes the following steps: Step 1: Use one-dimensional matching templates to identify the image coordinates of the contact lines at three different positions, j = 30, H / 2, and H-30, respectively, and record them as Pos0, Pos1, and Pos2. , where Pos0 is the contact line center identification position (i0, H / 2), Pos1 is the contact line start identification position (i1, 30 a), Pos2-contact line end identification position (i2,H-30 a); Step 2: According to different situations between the image coordinates of the start end, the center, and the end, judge whether the center position is between the image coordinates of the start end and the end, and judge the position situation of the contact wire: Step 2.

1. Situation 1: If the contact wire recognized at the center position is at a position to the left or right of the image coordinates of the start end and the end, and the recognition position Pos0 of the contact wire center point is not between the recognition position Pos1 of the start end and the recognition position Pos2 of the end, discard the contact wire position recognized at the center position. Calculate the recognition distances in the width direction between the start end, the center, and the end respectively according to the following situations, which are: Dis0 = i2 - i1 Dis1 = i0 - i1 Dis2 = i0 - i2 Make separate judgments according to the relationship between the recognition distance and the width model_width of the one-dimensional matching template. Step 2.1.

1. If the distance between the start end and the end end is less than the width of one template, Dis0 < model_width, fit the straight line equation of the contact line y = kx + b. If the fitted straight line is a vertical straight line, the position of the contact line is: (x, y) = (0.5·(i1 + i2), H / 2). If the fitted straight line is not a vertical straight line, the actual position of the contact line is (x, y) = (1 / k(H / 2 - b), H / 2); Step 2.1.

2. If the distance between the start end and the end end is greater than the width of one template, Dis0 > model_width, analyze the distances between the start end and the center and between the end end and the center, and determine the actual recognition position of the contact line; Including the following cases: When Dis1 < model_width and Dis2 < model_width, the actual position of the contact wire is x = , y = H / 2; When Dis1 < model_width and Dis2 > model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i1, H / 2), model_value(i0, H / 2)); When Dis1 > model_width and Dis2 < model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i2, H / 2), model_value(i0, H / 2)); When Dis1 > model_width and Dis2 > model_width, the actual position of the contact line is (x, y) = (min(d1, d2, d3), H / 2), where d1 = i0 - W / 2, d2 = i1 - W / 2, d3 = i2 - W / 2; Step 2.

2. Case 2: If the identified position of the center of the contact wire is between the identified position of the start end of the contact wire and the identified position of the end of the contact wire, that is, i2 > i0 > i 1、 or when i2 < i0 < i1, make separate judgments according to the relationship between the identified distance and the width of the one-dimensional matching template: Step 2.2.

1. If the distance between the start end and the end end is less than the width of one template, Dis0 < model_width, fit the straight line equation of the contact line. If the fitted straight line is a vertical straight line, the position of the contact line is: (x, y) = (0.5·(i1 + i2), H / 2). If the fitted straight line is not a vertical straight line, the actual position of the contact line is (x, y) = (1 / k(H / 2 - b), H / 2); Step 2.2.

2. If the distance between the start end and the end end is greater than the width of one template, Dis0 > model_width, analyze the distances between the start end and the center and between the end end and the center, and determine the actual recognition position of the contact line; Including the following cases: When Dis1 < model_width and Dis2 < model_width, the actual position of the contact line is (x, y) = (i0, H / 2); When Dis1 < model_width and Dis2 > model_width, the actual position of the contact line is Pos(x, y) = max(model_value(i1, H / 2), model_value(i0, H / 2)); When Dis1 > model_width and Dis2 < model_width, the actual position of the contact wire is Pos(x, y) = max(model_value(i2, H / 2), model_value(i0, H / 2)); When Dis1 > model_width and Dis2 > model_width, the actual position of the contact wire is (x, y) = (min(d1, d2, d3), H / 2), where d1 = i0 - W / 2, d2 = i1 - W / 2, and d3 = i2 - W / 2.

2. The method for rapid contact line identification based on a parallel architecture according to claim 1, characterized in that: According to the contact wire positions identified at the three ends of the starting end, the center, and the ending end of the contact wire, the contact wire perturbation range is set to 5 - 10.

3. A contact line rapid identification system based on a parallel architecture that implements the contact line rapid identification method based on a parallel architecture as described in any one of claims 1-2, characterized in that: It includes an image acquisition device for acquiring contact wire images to obtain known images; An image processing device, electrically connected to the image acquisition device, for receiving the known images and identifying the contact wire in the contact wire images according to a preset target recognition model, and respectively identifying the contact wire positions in the contact wire images during daytime and nighttime using the daytime working mode and the nighttime working mode.

4. The contact line rapid identification system based on parallel architecture according to claim 3, characterized in that: The daytime working mode includes: Acquiring contact wire images to obtain known images W×H; Performing normalization processing on the one-dimensional data of the known images to obtain the mean value of the one-dimensional data at the bull's-eye position; Constructing a one-dimensional matching template for the contact wire according to the mean value of the one-dimensional data at the bull's-eye position; According to the one-dimensional matching template and the known images, calculating the template matching results of the one-dimensional image data at the bull's-eye and at 3 - 10 pixels above and below the bull's-eye position, and identifying the maximum value of the template matching results as the contact wire position; The nighttime working mode includes: Using the one-dimensional matching template to respectively identify the template matching results of the known images where the contact wires are located at n different positions with different image heights, and judging the contact wire position situation based on the positional relationship between different identified positions of the contact wire, where n is a positive integer not less than 3.

Images