CN-121998973-A - Track sleeper falls piece disease detecting system based on PCA technique

CN121998973ACN 121998973 ACN121998973 ACN 121998973ACN-121998973-A

Abstract

The invention discloses a rail sleeper block falling disease detection system based on PCA technology, which relates to the technical field of block falling disease detection, wherein an image acquisition module is used for acquiring an original sleeper image as an image to be detected, a sleeper block falling image analysis module is used for extracting a data dictionary by acquiring a normal sleeper image set, reducing a vitamin data model through a PCA algorithm and combining projection transformation of the image to be detected to construct a reference image, preprocessing the image to be detected, performing difference between the preprocessed image to be detected and the reference image to obtain a comparison image, and judging a block falling disease area through threshold segmentation and communication area calculation and labeling an output result. The invention realizes high-efficiency and high-precision identification of sleeper block falling disease detection and effectively adapts to the scene of complex surface texture and high-dimensional data redundancy of the sleeper.

Inventors

LIN QING
LIU CHENG
HU CHENGKAI
Lan Yayi
MAO HONGJUN
LUO CHAOYANG

Assignees

成都精工华耀科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260407

Claims (10)

1. The utility model provides a track sleeper falls piece disease detecting system based on PCA technique which characterized in that includes: the image acquisition module is fixed on the track defect detection device and is used for acquiring an original sleeper image serving as an sleeper image to be detected; The sleeper block dropping image analysis module is used for realizing reference image construction, sleeper image processing to be detected, image comparison and disease judgment algorithm execution; The reference image construction algorithm comprises the steps of collecting a normal sleeper image set, extracting a data dictionary from the normal sleeper image set by adopting a preset method, performing data dimension reduction and data model generation on the data dictionary by adopting a PCA algorithm, and performing projection transformation based on a sleeper image to be detected to obtain a reference image which has correlation with the sleeper image to be detected; the sleeper image processing algorithm to be detected comprises the following steps: preprocessing the sleeper image to be detected; The image comparison algorithm comprises the steps of carrying out difference absolute value operation on the preprocessed sleeper image to be detected and the reference image to obtain a comparison image; The disease judgment algorithm comprises the steps of carrying out threshold segmentation on the comparison image, calculating the area of a connected region in the segmented image, judging the region with the area exceeding a set threshold value as a block-falling disease region, marking the corresponding region and outputting a disease detection result.
2. The system for detecting the defect of the falling blocks of the track sleeper based on the PCA technology as in claim 1, wherein the preset method is an SC coding method.
3. The system for detecting track sleeper block defect based on the PCA technology according to claim 1, wherein performing data dimension reduction and data model generation on the data dictionary by a PCA algorithm comprises: the sample data of the normal sleeper image set is subjected to decentration, and global offset interference is eliminated; Based on the sample data after the decentralization, calculating to obtain a covariance matrix describing the linear correlation of the pixel dimension; performing eigenvalue decomposition on the covariance matrix to obtain a corresponding eigenvalue set and eigenvector set; Sorting the characteristic values from large to small based on the characteristic value set, and screening the characteristic values in the sorted characteristic value set; selecting K eigenvectors corresponding to the K largest eigenvalues from the screened eigenvalue set, and forming a new eigenvector matrix by taking all the selected eigenvectors as row vectors; and reconstructing the original image based on the new feature vector matrix to obtain the reference image.
4. A track sleeper fall block defect detection system based on PCA technology as described in claim 3 wherein screening feature values in the ordered feature value set comprises: The characteristic values are filtered and determined based on the accumulated contribution rate, and the accumulated contribution rate is calculated according to the following formula: as the cumulative contribution rate of the feature values, As a result of the sorted feature values, In order to obtain the dimension of the principal component after screening, Is the original data dimension.
5. The system for detecting track sleeper blocking disease based on PCA as defined in claim 4, wherein the screening of the eigenvalues based upon cumulative contribution determination comprises: and for a preset contribution rate threshold, reserving a corresponding characteristic value only when the accumulated contribution rate is not smaller than the preset contribution rate threshold.
6. The system for detecting rail sleeper blocking disease based on PCA technology as in claim 1, wherein preprocessing the sleeper image to be detected comprises filtering the sleeper image to be detected with Gaussian filtering.
7. The system for detecting the defect of the falling blocks of the rail sleeper based on the PCA technology according to claim 1, wherein the step of performing difference absolute value operation on the preprocessed sleeper image to be detected and the reference image to obtain a comparison image comprises the following steps: In order to compare the images of the images, For the sleeper image to be detected, Is a reference image.
8. The railroad tie blocking disease detection system based on PCA technology of claim 1, wherein thresholding the comparison image comprises: at coordinates for a thresholded binarized image Pixel values at; For comparing the image at the coordinates Pixel values at; a preset pixel gray threshold value; When (when) When the area is judged to be a normal area; When (when) And judging the suspected block disease area.
9. The railroad tie blocking disease detection system based on PCA technology of claim 8, wherein calculating an area of the connected region in the segmented image, determining a region whose area exceeds a set threshold as a blocking disease region comprises: And judging the suspected block-falling disease area with the pixel area exceeding a set threshold as the block-falling disease area.
10. The system for detecting the defect of the falling blocks of the track sleeper based on the PCA technology as in claim 1, wherein the steps of marking the corresponding area and outputting the defect detection result comprise: After disease analysis is completed, a feedback result image with a block-falling area outline mark is output, and a corresponding detection result is synchronously output to be 0; And the detection result is 0, which indicates that the sleeper in the corresponding area has the block dropping disease, and the detection result is 1, which indicates that the sleeper in the corresponding area has no block dropping disease.

Description

Track sleeper falls piece disease detecting system based on PCA technique Technical Field The invention relates to the technical field of block dropping disease detection, in particular to a track block dropping disease detection system based on a PCA technology. Background The sleeper is used as a core bearing component of a railway track, directly bears the weight of a train and the track, and the structural integrity (especially avoiding block dropping diseases) is a key for guaranteeing the safe transportation of the railway. With the rapid expansion of urban rail transit and railway networks, the demand for sleeper maintenance detection is increasingly urgent. The current sleeper block falling detection still uses manual inspection as a main mode, and part of schemes design image acquisition devices such as laser and industrial cameras, but still need to manually judge detection results after acquisition. The manual inspection is limited by environmental factors such as weather and illumination, has the problems of low efficiency, strong subjectivity, high false detection rate of missed detection and large labor intensity, is difficult to adapt to the requirement of large-scale and high-frequency track maintenance, and cannot guarantee the consistency and reliability of detection. In addition, although the machine vision technology is widely applied in the field of track detection, the surface texture of the sleeper is complex, the sleeper is easy to be interfered by dust, water stain, slight scratch and the like after being put into use, the surface condition is changed frequently, sleeper images are high-dimensional data, contain a large amount of redundant information and also can directly influence the identification precision of the falling block diseases, and the conventional vision detection method is difficult to accurately extract core characteristics and cannot be directly transferred and applied to the falling block detection of the sleeper due to the reasons. In summary, the automation, high precision and high efficiency requirements of sleeper detection are difficult to be met in the existing detection technology, and a corresponding technical scheme needs to be provided for solving the problem. Disclosure of Invention In view of the above, the application provides a track sleeper block falling disease detection system based on PCA technology, so as to solve the defects existing in the prior art. The first aspect of the application provides a track sleeper block falling disease detection system based on PCA technology, comprising: the image acquisition module is fixed on the track defect detection device and is used for acquiring an original sleeper image serving as an sleeper image to be detected; The track disease detection device comprises an electric bus, a data acquisition module, a system control module and a power supply module, wherein the electric bus is used for providing a running carrier, the data acquisition module is a calculation platform and is used for acquiring imaging data from the image acquisition module and transmitting the imaging data to a sleeper block falling image analysis module, the system control module is an embedded calculation platform and comprises a function of calculating the real-time running speed of the electric bus, outputting corresponding pulse signals to the image acquisition module at the real-time running speed and controlling the image acquisition module to acquire images at a set frequency, and the power supply module is a battery pack or an external power supply interface and is used for supplying power for all the modules; The sleeper block dropping image analysis module is used for realizing reference image construction, sleeper image processing to be detected, image comparison and disease judgment algorithm execution; The reference image construction algorithm comprises the steps of collecting a normal sleeper image set, extracting a data dictionary from the normal sleeper image set by adopting a preset method, performing data dimension reduction and data model generation on the data dictionary by adopting a PCA algorithm, and performing projection transformation based on a sleeper image to be detected to obtain a reference image which has correlation with the sleeper image to be detected; the sleeper image processing algorithm to be detected comprises the following steps: preprocessing the sleeper image to be detected; The image comparison algorithm comprises the steps of carrying out difference absolute value operation on the preprocessed sleeper image to be detected and the reference image to obtain a comparison image; The disease judgment algorithm comprises the steps of carrying out threshold segmentation on the comparison image, calculating the area of a connected region in the segmented image, judging the region with the area exceeding a set threshold value as a block-falling disease region, marking the corresponding region and outp