CN-121977457-A - Non-contact real-time monitoring method and system for thickness of train brake pad

CN121977457ACN 121977457 ACN121977457 ACN 121977457ACN-121977457-A

Abstract

The system comprises an image acquisition module, a trigger sensor and a data processing module, wherein the image acquisition module comprises a camera and a light source arranged on one side of the camera, and the data processing module comprises a target area coarse positioning module, a search area fine definition module, a candidate edge feature searching module, an edge scoring screening module, an edge fitting module and a thickness calculation output module. According to the invention, through the image acquisition module arranged beside the rail, when the train normally runs, the brake pad image is automatically shot, the abrasion thickness is accurately calculated, the non-contact detection is realized, the problems of missed detection risk, low efficiency, measurement error and the like existing in the traditional manual periodic detection are solved, so that the brake system is always in a safe state, the operation and maintenance cost is greatly reduced, the intelligent guarantee is provided for the safe operation of the train, the practicability is strong, and the popularization significance is strong.

Inventors

DONG HUI
GAN YU
PAN WENFENG
Xiao Aihui
WU GENGCAI
LIANG YINGCHANG
CHEN GANGFENG
CHEN RUNSONG
WU ZHENHUA
XU CHANGYUAN
Yin Sanyong
ZENG DEYAO

Assignees

东莞市诺丽科技股份有限公司

Dates

Publication Date: 20260505
Application Date: 20260119

Claims (9)

1. A non-contact real-time monitoring system for the thickness of a train brake pad is characterized by comprising an image acquisition module, a trigger sensor and a data processing module, wherein the trigger sensor is used for receiving train information and triggering the image acquisition module to start, the image acquisition module is used for acquiring images of the position of the brake pad and transmitting the images to the data processing module for processing and analyzing, the image acquisition module comprises a camera and a light source arranged on one side of the camera, the data processing module comprises a target area coarse positioning module, a search area precise definition module, a candidate edge feature search point module, an edge scoring screening module, an edge fitting module and a thickness calculation output module, the target area coarse positioning module is used for identifying and positioning a rough area of the brake pad in a picture acquired by the image acquisition module, the search area precise definition module is used for processing initial mask images output by the target area coarse positioning module and defining a precise search area, the candidate edge feature point module is used for extracting candidate edge feature points in a precise rectangular area, the edge scoring screening module is used for carrying out candidate edge feature points detected by the candidate edge feature point module and selecting two candidate edge feature points, and the edge points obtained by the candidate edge feature point module are directly fitted to two edge points which are calculated and belong to the two sharp edge fitting calculation module.
2. The system for non-contact real-time monitoring of train brake pad thickness according to claim 1, wherein the track is provided with a detection area, the two groups of image acquisition modules are respectively arranged on two sides of the track in the detection area, the directions of the two groups of image acquisition modules are different, the flash lamps of the two groups of image acquisition modules on the same side are adjacently arranged, and the camera is arranged on the outer side of the flash lamps.
3. The non-contact real-time monitoring system for train brake pad thickness of claim 2, wherein the trigger sensor is magnetic steel, and the magnetic steel is arranged on the inner side surface of the rail.
4. The non-contact real-time monitoring system for train brake pad thickness according to claim 2, wherein the trigger sensor is also provided with two groups, which are respectively arranged at the front side and the rear side of the train travelling direction in the detection area.
5. A non-contact real-time monitoring method for the thickness of a train brake pad is characterized by comprising the following steps: Step S1, collecting brake pad images, namely providing a train brake pad thickness non-contact real-time monitoring system according to any one of claims 1 to 4, wherein when a train enters a detection area, a trigger sensor detects the position of a wheel to generate a trigger signal, and a camera captures an image of a brake pad area; s2, coarsely positioning a target area, namely identifying and positioning the position of a brake pad by adopting Yolo deep learning semantic segmentation model or template matching; step S3, the search area is defined finely, namely morphological processing is carried out on the image of the identification positioning area and the processing of the minimum rectangular area operator is carried out, and the identified area is redefined as an accurate search area; S4, searching points by candidate edge features, namely calculating the gray value of each point in an accurate search area, carrying out first-order derivation on the difference value of two adjacent pixel points, and searching the maximum and minimum values of the derivative to find out key points of gray level and brightness variation of an image, namely the candidate edge feature points of the edge of a brake pad; Step S5, screening edge point scores, namely evaluating and screening the candidate edge feature points, wherein the first gray level comparison score is used for calculating the average gray level of foreground and background areas of each candidate point, the darker the background is, the brighter the foreground is, the higher the score is, the second gray level difference score is used for scoring the gray level difference value of the foreground and the background, the larger the difference value is, the higher the score is, and the third position deviation score is used for calculating the distance between the candidate point and a datum line, the closer the distance is, the higher the score is, wherein the datum line is a line marked on the correct edge position; S6, fitting edge points, namely fitting two discrete edge point sets which are screened by edge scoring and respectively belong to the upper edge and the lower edge of the brake pad into two line segments, wherein the two line segments respectively represent the upper edge and the lower edge of the brake pad; And S7, outputting thickness calculation, namely obtaining the thickness of the brake pad by calculating the result of multiplying the pixel value of the shortest distance between two line segments by the physical size value of each pixel calibrated by the camera, wherein the formula for calculating the residual thickness is as follows: The gate remaining uses the thickness = minimum distance pixel value of the upper edge line segment and the lower edge line segment X camera calibrated pixel physical size value.
6. The method for non-contact real-time monitoring of train brake pad thickness according to claim 1, wherein in step S5, an average gray scale is calculated for foreground and background of candidate edge feature points, wherein the foreground is a rectangular area above the point, the background is a rectangular area below the point, the rectangular area has a preset length, width and length, and then the calculated average gray scale is scored, and according to brightness features of the brake pad edge, the darker the background is, the higher the bright score of the front Jing Yue is.
7. The method of claim 1, wherein in step S2, yolo is a deep learning semantic segmentation model for identifying and pixel-level segmentation of a brake pad in an input image, and outputting an initial mask image where the brake pad is located.
8. The method for non-contact real-time monitoring of train brake pad thickness according to claim 1, wherein in step S2, the template matching mode is to match the acquired image with a standard template, calculate the similarity again and find out the area most matched with the template, so as to realize preliminary identification and positioning of the brake pad position.
9. The method for non-contact real-time monitoring of train brake pad thickness according to claim 1, wherein in step S4, the first-order derivative formula is: Wherein sigma is a smoothing parameter, and x is a difference value between adjacent pixels.

Description

Non-contact real-time monitoring method and system for thickness of train brake pad Technical Field The invention relates to the technical field of rail transit detection, in particular to a non-contact real-time monitoring method and system for train brake pad thickness. Background In a rail transit system, brake pads are used as core components of braking safety, and timely grasping of the abrasion state of the brake pads is important to train operation safety. However, the traditional brake pad overhaul mode has the defects that on one hand, the detection process depends on the shutdown and warehousing of a train and is limited by a fixed overhaul period, the real-time grasp of the abrasion state cannot be realized often for a plurality of weeks or even months, the undetected safety risk due to the overrun of abrasion exists in the detection interval period, on the other hand, the traditional manual measurement efficiency is low, the result is easily influenced by subjective factors, and a single person can only finish the detection tasks of a plurality of trains, and the detection tasks are high in long-term operation and maintenance investment and difficult to meet the detection requirements of high frequency and high consistency for vehicles with tens or even hundreds of scales. In recent years, a non-contact image detection technology (such as CN 116428995A, a train brake pad abrasion detection system and a detection method) gets rid of the limit of manual contact measurement to a certain extent, but has the following defects in practical application that continuous image capture is relied on, the requirements on hardware calculation power and environmental adaptability are high, an identification algorithm is easy to be interfered in a complex operation scene, the false alarm rate is high, the burden of secondary verification of operation and maintenance personnel is increased, and high-efficiency and reliable automatic detection cannot be truly realized. Disclosure of Invention Based on the above, it is necessary to provide a non-contact real-time monitoring method and system for train brake pad thickness, which aims at the defects in the prior art. The train brake pad thickness non-contact real-time monitoring system comprises an image acquisition module, a trigger sensor and a data processing module, wherein the trigger sensor is used for receiving train information and triggering the image acquisition module to start, the image acquisition module is used for acquiring images of brake pad positions and transmitting the images to the data processing module for processing analysis, the image acquisition module comprises a camera and a light source arranged on one side of the camera, the data processing module comprises a target area coarse positioning module, a search area precise definition module, a candidate edge feature search point module, an edge scoring screening module, an edge fitting module and a thickness calculation output module, the target area coarse positioning module is used for identifying and positioning a rough area of a brake pad in a picture acquired by the image acquisition module, the search area precise definition module is used for processing initial mask images output by the target area coarse positioning module and defining a precise search area, the candidate edge feature search point module is used for extracting candidate edge feature points in a precisely defined rectangular area, the edge scoring screening module is used for carrying out screening on candidate edge feature points detected by the candidate edge feature point module so as to select two edge candidate edge feature point modules, the edge feature point modules are directly fitted to two discrete edge score strips, and the edge score calculation and the two edge calculation and belong to two sharp edge calculation and fit between two sharp edge evaluation and straight line calculation and edge calculation and the two sharp edge calculation and are respectively fit to the two sharp edge calculation and the edge calculation module. Further, a detection area is arranged on the track, two groups of image acquisition modules are respectively arranged on two sides of the track of the detection area, the directions of the two groups of image acquisition modules are different, the flash lamps of the two groups of image acquisition modules on the same side are adjacently arranged, and the camera is arranged on the outer side of the flash lamp. Further, the triggering sensor is magnetic steel, and the magnetic steel is arranged on the inner side surface of the track. Further, the trigger sensors are also provided with two groups, and are respectively arranged at the front side and the rear side of the train travelling direction of the detection area. A non-contact real-time monitoring method for train brake pad thickness comprises the following steps: Step S1, collecting brake pad images, namely providing a train