CN-121999316-A - Quality detection method and device for track image

CN121999316ACN 121999316 ACN121999316 ACN 121999316ACN-121999316-A

Abstract

The invention provides a quality detection method and device for a track image, which comprises the steps of obtaining the track scene image, inputting the track scene image into a track quality detection model to obtain a track quality detection result, wherein the track quality detection result comprises a track fastener quality detection result and a track global quality detection result, the track quality detection model is obtained by training a deep neural network model based on a historical track scene image dataset, the track fastener quality detection result and the track global quality detection result are subjected to weighted fusion according to preset weights to obtain a track scene image quality evaluation result, and the track image quality detection method and device disclosed by the invention promotes the comprehensiveness and suitability of track image quality detection by considering the track fastener local detail and the track global scene quality evaluation, reduces manual intervention, and realizes intelligent and reliable track image quality detection.

Inventors

HE YU
HAN ZHI
ZHAO ZISHEN
FU QIANG
CHEN CHUNLEI
HAO JINFEI
HAN LUPING
WANG FUYIN
LIU KAI

Assignees

中国铁道科学研究院集团有限公司
中国铁道科学研究院集团有限公司基础设施检测研究所
北京铁科英迈技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260119

Claims (10)

1. A quality detection method of a track image, comprising: Acquiring a track scene image, wherein the track scene image comprises a rail clip and a track infrastructure; The method comprises the steps of inputting a track scene image into a track quality detection model to obtain a track quality detection result, wherein the track quality detection result comprises a track fastener quality detection result and a track global quality detection result, the track quality detection model is obtained by training a deep neural network model based on a historical track scene image dataset, and the historical track scene image dataset comprises a historical track scene image with quality grade marked based on a visual marking technology; And carrying out weighted fusion on the rail clip quality detection result and the rail global quality detection result according to preset weights to obtain a rail scene image quality evaluation result.
2. The method of claim 1, wherein the historical track scene image dataset is obtained as follows: Detecting a quality grade marking result of each historical track scene image after quality grade marking is carried out on the historical track scene images, wherein the historical track scene images are derived from data actually collected by a work inspection system; If a plurality of quality grades exist in one historical track scene image, calculating the difference degree between labeling results of every two quality grades; if the difference degree is higher than a preset threshold value, marking the track scene image as an abnormal sample and rechecking; And removing abnormal samples which do not pass the recheck to obtain a historical track scene image dataset.
3. The method of claim 2, wherein culling abnormal samples that fail review to obtain a historical track scene image dataset comprises: After abnormal samples which do not pass the rechecking are removed, the historical track scene images are divided according to a preset proportion to obtain a historical track scene image dataset, wherein the historical track scene image dataset comprises a training set, a verification set and a test set.
4. A method according to claim 1, wherein the rail clip quality measurement is determined by: Inputting the track scene image into a track quality detection model, and outputting a plurality of rail fastener quality grades and corresponding confidence degrees; And obtaining a rail clip quality detection result through weighted fusion according to the quality grades of the rail clips and the corresponding confidence degrees.
5. The method of claim 1, wherein the global quality of track detection result is determined by: inputting the track scene image into a track quality detection model, and outputting an initial global quality grade of the track scene image; obtaining response characteristics of the track scene image by an image characteristic extraction method; And correcting the initial global quality grade of the track scene image based on the response characteristics of the track scene image to obtain a track global quality detection result.
6. The method of claim 4, wherein the response characteristics of the track scene image include luminance variance, edge density, and spatial frequency.
7. A quality detection apparatus for a track image, comprising: The system comprises an image acquisition module, a storage module and a storage module, wherein the image acquisition module is used for acquiring a track scene image, and the track scene image comprises a rail clip and a track infrastructure; The quality detection module is used for inputting the track scene image into a track quality detection model to obtain a track quality detection result, wherein the track quality detection result comprises a track fastener quality detection result and a track global quality detection result, the track quality detection model is obtained by training a deep neural network model based on a historical track scene image dataset, and the historical track scene image dataset comprises a historical track scene image marked with quality grades based on a visual marking technology; And the quality evaluation module is used for carrying out weighted fusion on the rail fastener quality detection result and the track global quality detection result according to preset weights to obtain a track scene image quality evaluation result.
8. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1-6 when executing the computer program.
9. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a processor, implements the method of any of claims 1-6.
10. A computer program product, characterized in that the computer program product comprises a computer program which, when executed by a processor, implements the method of any of claims 1-6.

Description

Quality detection method and device for track image Technical Field The invention belongs to the technical field of railway service detection, and particularly relates to a quality detection method and device for a track image. Background With the continuous expansion of rail transit networks, the operation safety and operation and maintenance efficiency of rail equipment have become core points of concern in the fields of railway and urban rail transit. In order to ensure stable and safe running of the train, the track structure and key parts are required to be periodically subjected to inspection maintenance work, and the traditional track inspection mode depends on manual field detection and experience judgment, so that the operation efficiency is low, the labor cost is high, the operation is easily influenced by factors such as environmental illumination, personnel operation level and subjective judgment difference, and the standardization, the intellectualization and the precision of inspection evaluation are difficult to realize. In order to improve the inspection automation level, an image recognition-based track inspection system is developed, the system analyzes a track image by means of a machine vision technology, detection and quality judgment of key components are achieved, the limit of manual inspection is relieved to a certain extent, however, the conventional quality evaluation algorithm is multi-faced to natural images or general industrial scene design, the unique characteristics of the track inspection image cannot be fully adapted, and the reliability and practical application effect of the whole railway inspection system are affected. Disclosure of Invention The embodiment of the invention provides a quality detection method of a track image, which improves the comprehensiveness and suitability of the quality detection of the track image by considering the quality evaluation of the local detail of a rail clip and the global scene of the track, reduces the manual intervention and realizes the intelligent and reliable quality detection of the track image, and the quality detection method of the track image comprises the following steps: Acquiring a track scene image, wherein the track scene image comprises a rail clip and a track infrastructure; The method comprises the steps of inputting a track scene image into a track quality detection model to obtain a track quality detection result, wherein the track quality detection result comprises a track fastener quality detection result and a track global quality detection result, the track quality detection model is obtained by training a deep neural network model based on a historical track scene image dataset, and the historical track scene image dataset comprises a historical track scene image with quality grade marked based on a visual marking technology; And carrying out weighted fusion on the rail clip quality detection result and the rail global quality detection result according to preset weights to obtain a rail scene image quality evaluation result. The embodiment of the invention provides a quality detection device of a track image, which improves the comprehensiveness and suitability of the quality detection of the track image by considering the quality evaluation of the local detail of a rail clip and the global scene of the track, reduces the manual intervention and realizes the intelligent and reliable quality detection of the track image, and the quality detection device of the track image comprises: The system comprises an image acquisition module, a storage module and a storage module, wherein the image acquisition module is used for acquiring a track scene image, and the track scene image comprises a rail clip and a track infrastructure; The quality detection module is used for inputting the track scene image into a track quality detection model to obtain a track quality detection result, wherein the track quality detection result comprises a track fastener quality detection result and a track global quality detection result, the track quality detection model is obtained by training a deep neural network model based on a historical track scene image dataset, and the historical track scene image dataset comprises a historical track scene image marked with quality grades based on a visual marking technology; And the quality evaluation module is used for carrying out weighted fusion on the rail fastener quality detection result and the track global quality detection result according to preset weights to obtain a track scene image quality evaluation result. The embodiment of the invention also provides computer equipment, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the quality detection method of the track image when executing the computer program. The embodiment of the invention also provides a computer readable storage