CN115861825B - 2C detection method based on image recognition - Google Patents
2C detection method based on image recognition Download PDFInfo
- Publication number
- CN115861825B CN115861825B CN202310167299.9A CN202310167299A CN115861825B CN 115861825 B CN115861825 B CN 115861825B CN 202310167299 A CN202310167299 A CN 202310167299A CN 115861825 B CN115861825 B CN 115861825B
- Authority
- CN
- China
- Prior art keywords
- image
- abnormal
- image data
- train
- point
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Train Traffic Observation, Control, And Security (AREA)
Abstract
The invention discloses a 2C detection method based on image recognition, which comprises the following steps: collecting panorama of a contact network suspension and high-speed railway line on a train in real time by using a collecting device; processing the acquired image data and judging the abnormality in the image data; marking an abnormality in the image data, recording a time point when the abnormal image data is acquired, judging the position of the marking point based on the time point, and recording the longitude and latitude of the position; and selecting all the abnormal point images to form a two-dimensional relation graph, taking longitude and latitude as a relation axis of the two-dimensional graph, displaying the positions of the abnormal points, and displaying the positions on a terminal. The method is based on an image recognition technology, the acquired image is recognized, the image is detected based on a machine learning algorithm, an abnormal image is judged, the time point of the abnormal image is recorded, the abnormal position in the image is judged according to a positioning method and a calculation method, the longitude and latitude are used as the relation axis of a two-dimensional graph, the position of the abnormal point is displayed, and the abnormal position is displayed on a terminal.
Description
Technical Field
The invention relates to the field of detection, in particular to a 2C detection method based on image recognition.
Background
The utility model provides a contact net safety inspection device (2C for short), which is a safety monitoring device for guiding the operation and maintenance of the contact net by temporarily erecting portable equipment in a cab of an operating motor train unit (or locomotive), and carrying out video acquisition on the state of the contact net and the external environment.
The method has the main functions of monitoring whether the contact net equipment has obvious disconnection, deviation and other abnormal conditions;
whether bird nests, tree harm and other surrounding environmental factors possibly endangering the power supply safety of the overhead line system exist or not;
there are intrusion restrictions, obstacles that hinder the running of the rolling stock, and the like.
However, the existing network security inspection method (2C detection method for short) basically collects video images through a network security inspection device, and then inspects the video images manually through inspection personnel to judge the state of the contact network;
although the accuracy rate of the method is higher, the efficiency is lower, and the video of one train needs to be checked for a long time, so that the method is not consistent with the actual situation;
meanwhile, the detection workers look over the video for a long time, visual fatigue is easy to occur, judgment accuracy is further affected, and inspection effect is further affected.
In the prior art, there are technical means for image monitoring by using 2C:
prior art 1 (CN 111598855 a) discloses a 2C equipment high-speed rail overhead line system hanger defect detection method based on deep learning and migration learning; specifically disclosed is (1) inputting a photographed high-resolution image; (2) Sending the input image into a target detection network to obtain a prediction result, and filtering out detection frames with overlarge overlapping parts in all detection results by using a non-maximum suppression algorithm; (3) performing coordinate matching on the detected normal part; (4) positioning the missed detection part according to the matching result; (5) Classifying the positioned missed detection part by using a classification network; (6) However, the prior art needs to monitor based on deep learning and transfer learning, the algorithm has higher requirements on hardware computing capacity, higher equipment cost, and the problems of slower calculation and insufficient real-time monitoring capacity, and the deep learning needs to perform model training, so that the initial precision is too low.
Prior art 2 (CN 105501248A) discloses a railway line inspection system; the data acquisition subsystem is arranged on the unmanned aerial vehicle, acquires field data of the railway line, and sends the field data to the data and fault processing subsystem for processing, wherein the field data comprises radar scanning data and field image data of the railway line. However, the prior art does not relate to a technical means for processing an image to improve detection accuracy, specifically, in an actual scene, due to the problems of frost, rain, snow and dust along a railway, the image accuracy of the unmanned aerial vehicle line inspection detection means is seriously reduced, so that the detection accuracy is insufficient, accurate image data cannot be provided, and the processing of the original image data cannot be completed under the conditions of simplicity and high efficiency.
Prior art 3 (CN 104299260 a) discloses a three-dimensional reconstruction method of a catenary based on point cloud registration of SIFT and LBP; the method specifically discloses a method for carrying out feature description on key points by using an image local binary pattern (LBP: local Binary Patterns) of a uniform pattern to obtain feature description vectors of the key points; then, the inter-vector distance is used as a similarity judgment measure between key points, and the corresponding relation between point clouds under different view angles is determined according to the similarity judgment measure; and finally, performing point cloud rough registration and nearest point iteration (ICP: iterative Closest Point) fine registration to obtain the complete three-dimensional point cloud data of the contact net part to be rebuilt. The technical means realizes detection of the contact network through three-dimensional modeling, and also has the problems of overlarge calculated amount, unnecessary modeling data and the like, and particularly, the method is only suitable for 'large inspection', namely, after normal modeling is completed, the method needs to be modeled again in large inspection, and is compared with normal modeling, so that abnormal points can be found, and real-time monitoring cannot be carried out.
Prior art 4 (CN 104567708A) discloses a device and method for detecting full-section high-speed dynamic health of a tunnel based on active panoramic vision; specifically, the RGB color space of the panorama is converted into the HIS color space, then 1.2 times of average brightness on an imaging plane is used as a threshold value for extracting red laser projection points, and in order to obtain the accurate position of the laser projection lines, the invention adopts a Gaussian approximation method to extract the central position of the laser projection lines. However, the method requires Gaussian transformation, has high requirements on constraint conditions and position relations, causes daily vibration along the track, causes parameter distortion in the Gaussian transformation, and reduces calculation accuracy.
Disclosure of Invention
The invention aims to: A2C detection method based on image recognition is provided to solve the above problems existing in the prior art.
The technical scheme is as follows: a 2C detection method based on image recognition, comprising:
collecting panorama of a contact network suspension and high-speed railway line on a train in real time by using a collecting device, and recording a time point in real time in the process;
processing the acquired image data and judging the abnormality in the image data;
marking an abnormality in the image data, recording a time point when the abnormal image data is acquired, judging the position of the marking point based on the time point, and recording the longitude and latitude of the position;
and selecting all the abnormal point images to form a two-dimensional relation graph, taking longitude and latitude as a relation axis of the two-dimensional graph, displaying the positions of the abnormal points, and displaying the positions on a terminal.
When displayed, two modes can be included;
mode one:
selecting and only checking abnormal points, at this time, all pictures without abnormality can be checked, and the abnormal position and abnormal state can be checked more rapidly;
mode two:
all pictures are displayed one by one, and the mode can be used for checking leakage and repairing defects to a certain extent, but the speed is different from that of the mode one.
The method is based on an image recognition technology, the acquired image is recognized, the image is detected based on a machine learning algorithm, an abnormal image is judged, the time point of the abnormal image is recorded, the abnormal position in the image is judged according to a positioning method and a calculation method, the longitude and latitude are used as the relation axis of a two-dimensional graph, the position of the abnormal point is displayed, and the abnormal position is displayed on a terminal. The method comprises the steps of carrying out a first treatment on the surface of the
Meanwhile, the image is subjected to pre-processing, the image is subjected to gray scale speech processing, and the image is subjected to denoising, so that the processed image is more convenient for detection of a machine learning algorithm, the traditional macroscopic recognition process of detection personnel is replaced, the inspection time is greatly shortened, and the machine learning algorithm is increasingly accurate in detection along with the increase of acquired data, so that the inspection accuracy is further compensated.
In a further embodiment, determining the location of the marker point based on the time point includes a positioning method and a calculation method;
the positioning method comprises the following steps:
based on a time point corresponding to an abnormal point in the image data, backtracking the position of the train at the time point, and further judging the abnormal position in the image data;
by designing the positioning method, the method can be adopted in areas with strong signals and the like, however, when the signals are weak and the nearby interference is strong, the position can be obtained by selecting a calculation method.
The calculation method comprises the following steps:
based on the running speed of the train in each time period, the running distance of the train on the track is calculated by combining the abnormal time existing in the image data, and the abnormal position existing in the image data is judged on the track.
Because the track of the high-speed rail train is fixed, the position with abnormality can be known only by calculating the travel distance;
the starting station can be started, and the starting station can be started according to the intermediate station;
when the starting station starts, the final position can be obtained by directly combining the running data of the train in the time period with the sampling time of the abnormal picture;
the starting time of the intermediate station is the time point of the train starting from the intermediate station.
In a further embodiment, determining whether the two positions obtained by the positioning method and the calculation method overlap;
if yes, the overlapping position is determined to be an abnormal position;
if not, calculating the linear distance between the two positions obtained by the positioning method and the calculation method, and judging whether the linear distance is smaller than a set value;
if yes, the position of the positioning method is determined to be an abnormal position;
and if not, determining the position of the calculation method as an abnormal position.
The set point may be dependent on the local signal conditions.
By setting the double positioning mode, the error of the abnormal position in the image is avoided, and the labor efficiency is increased to a certain extent.
In a further embodiment, the positioning system comprises GPS;
the acquisition equipment comprises a lens and a CCD camera matched with the lens.
In a further embodiment, the terminal comprises a mobile phone, a computer and a display screen.
In a further embodiment, the image data processing includes:
s1, carrying out gray processing on an image, and intercepting an acquired video image into a picture state before processing;
s11, converting the color image from an RGB color model to a CIE Lab color model;
s12, obtaining the perceived brightness of an original color image in a CIE Lab color space, and correcting the perceived brightness of the image by utilizing chromaticity components a and b to realize the preliminary graying of the color image so as to obtain a preliminary gray image; wherein a represents red and green, and b represents yellow and blue;
s13, performing contourlet decomposition on the color image in a CIE Lab color model to obtain 1 low-frequency component sub-band image and a plurality of high-frequency component sub-band images;
s14, carrying out contour wave decomposition on the preliminary gray level image in the same scale as the color image;
s15, calculating the ratio of the chromaticity contrast between each sub-band of the color image and the primary gray image, and adding the chromaticity information of the color image into the primary gray image according to the proportion to obtain a final gray image with enhanced local contrast;
s2, denoising the image subjected to the graying treatment by using a mean value filtering method;
by selecting the template during filtering, the pixel values of each point in the image are replaced by the average of the pixel values of all points in the template;
the mean value filtering formula is:
in the above-mentioned method, the step of,representing an original image containing noise, i.e. a graying processed image +.>Representing the image obtained after mean filtering, < >>Representation dot->A set of pixels in the template that is centered, +.>Representing the template size;
s3, detecting the image after denoising treatment based on a machine learning algorithm, and judging an abnormal image;
s31, classifying according to the abnormal types, and displaying in a two-dimensional relation chart by using different colors according to the classification;
the maintenance is red, the obstacle removal is yellow, the dredging is green, so that a detector can visually check the abnormal type, and maintenance personnel can be rapidly arranged to perform maintenance, obstacle removal, dredging and other works.
S311, classifying maintenance when the train contact net equipment has poor contact; for example, monitoring whether the contact net equipment has obvious disconnection, deviation and other abnormal conditions;
s312, classifying the overhead line system power supply safety as obstacle removal when the overhead line system power supply safety is influenced by surrounding environment factors; for example, there are no bird nest, tree harm and other surrounding environmental factors which may endanger the power supply safety of the overhead line system;
s313, classifying as dredging when the train track encounters an obstacle. Examples of the obstacle include a non-invasive limit, an obstacle that hinders the running of a rolling stock;
s4, marking the abnormal image.
By classifying the abnormal states, the inspection personnel can quickly locate the abnormal states and the abnormal positions.
The beneficial effects are that:
1. the invention discloses a 2C detection method based on image recognition, which is characterized in that an acquired image is recognized based on an image recognition technology, an abnormal image is detected based on a machine learning algorithm, the time point of the abnormal image is recorded, the abnormal position in the image is judged according to a positioning method and an algorithm, the longitude and latitude are used as the relation axis of a two-dimensional graph, the abnormal position is displayed, and the abnormal position is displayed on a terminal.
2. The invention carries out the pre-processing to the image, carries out the gray-scale speech processing to the image, and simultaneously carries out the denoising to the image, so that the processed image is more convenient for the detection of a machine learning algorithm, replaces the traditional naked eye identification process of detection personnel, greatly shortens the inspection time, and along with the increase of the acquired data, the detection of the machine learning algorithm is more accurate, thereby compensating the inspection accuracy.
3. According to the invention, based on CIE Lab color space, the original image is subjected to contourlet decomposition to obtain 1 low-frequency component and a plurality of high-frequency component sub-band images, so that the contour shape of the image can be rapidly and efficiently determined on the premise of not carrying out complex transformation or artificial intelligence algorithm training on the original image, and further whether the contour of the image has abnormal conditions such as deformation, falling and the like can be judged; in the image processing stage, the algorithm does not relate to common modeling and training processes in the intelligent algorithm, does not need operations such as Gaussian transformation and the like, further avoids the influence of error parameters on processing results, and particularly has higher efficiency and accuracy for image judgment under the weather conditions such as rain, snow, wind, frost, sand, dust, haze and the like for rail line equipment in open areas.
4. The invention processes the preprocessed image by the machine learning algorithm, can rapidly judge the abnormal image, in particular, the invention avoids using intelligent algorithms such as artificial intelligence, machine learning and the like in the preprocessing stage of the image data, can avoid introducing data errors, can determine the specific machine learning algorithm according to actual needs in the abnormal judging stage, and can judge in a background server, thereby having higher real-time performance and accuracy.
5. According to the method, all the abnormal point images are selected to form the two-dimensional relation graph, the longitude and latitude are used as the relation axis of the two-dimensional graph, the positions of the abnormal points are displayed and displayed on the terminal, the abnormal areas can be displayed quickly, the processing is convenient, the areas easy to occur are judged based on the historical abnormal data, targeted inspection or reinforcement can be performed, the safety rate of the whole track line is improved, and unnecessary inspection is reduced.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention.
FIG. 2 is a schematic diagram of the collection and processing procedure of the present invention.
Fig. 3 is a schematic illustration of the marking process of the present invention.
Fig. 4 is a schematic illustration of the present invention.
Detailed Description
The present application relates to a 2C detection method based on image recognition, and is explained in detail below by way of specific embodiments.
A 2C detection method based on image recognition, as shown in fig. 1, includes:
collecting panorama of a contact network suspension and high-speed railway line on a train in real time by using a collecting device, and recording a time point in real time in the process;
processing the acquired image data and judging the abnormality in the image data;
marking an abnormality in the image data, recording a time point when the abnormal image data is acquired, judging the position of the marking point based on the time point, and recording the longitude and latitude of the position;
and selecting all the abnormal point images to form a two-dimensional relation graph, taking longitude and latitude as a relation axis of the two-dimensional graph, displaying the positions of the abnormal points, and displaying the positions on a terminal.
When displayed, two modes can be included;
mode one:
selecting and only checking abnormal points, at this time, all pictures without abnormality can be checked, and the abnormal position and abnormal state can be checked more rapidly;
mode two:
all pictures are displayed one by one, and the mode can be used for checking leakage and repairing defects to a certain extent, but the speed is different from that of the mode one.
The method is based on an image recognition technology, the acquired image is recognized, the image is detected based on a machine learning algorithm, an abnormal image is judged, the time point of the abnormal image is recorded, the abnormal position in the image is judged according to a positioning method and a calculation method, the longitude and latitude are used as the relation axis of a two-dimensional graph, the position of the abnormal point is displayed, and the abnormal position is displayed on a terminal. The method comprises the steps of carrying out a first treatment on the surface of the
Meanwhile, the image is subjected to pre-processing, the image is subjected to gray scale speech processing, and the image is subjected to denoising, so that the processed image is more convenient for detection of a machine learning algorithm, the traditional macroscopic recognition process of detection personnel is replaced, the inspection time is greatly shortened, and the machine learning algorithm is increasingly accurate in detection along with the increase of acquired data, so that the inspection accuracy is further compensated.
Judging the position of the mark point based on the time point comprises a positioning method and a calculation method, as shown in figure 3;
the positioning method comprises the following steps:
based on a time point corresponding to an abnormal point in the image data, backtracking the position of the train at the time point, and further judging the abnormal position in the image data;
by designing the positioning method, the method can be adopted in areas with strong signals and the like, however, when the signals are weak and the nearby interference is strong, the position can be obtained by selecting a calculation method.
The calculation method comprises the following steps:
based on the running speed of the train in each time period, the running distance of the train on the track is calculated by combining the abnormal time existing in the image data, and the abnormal position existing in the image data is judged on the track.
Because the track of the high-speed rail train is fixed, the position with abnormality can be known only by calculating the travel distance;
the starting station can be started, and the starting station can be started according to the intermediate station;
when the starting station starts, the final position can be obtained by directly combining the running data of the train in the time period with the sampling time of the abnormal picture;
the starting time of the intermediate station is the time point of the train starting from the intermediate station.
Judging whether two positions obtained by a positioning method and an calculating method are overlapped or not;
if yes, the overlapping position is determined to be an abnormal position;
if not, calculating the linear distance between the two positions obtained by the positioning method and the calculation method, and judging whether the linear distance is smaller than a set value;
if yes, the position of the positioning method is determined to be an abnormal position;
and if not, determining the position of the calculation method as an abnormal position.
The set point may be dependent on the local signal conditions.
By setting the double positioning mode, the error of the abnormal position in the image is avoided, and the labor efficiency is increased to a certain extent.
The positioning system comprises a GPS;
the acquisition equipment comprises a lens and a CCD camera matched with the lens.
The terminal comprises a mobile phone, a computer and a display screen.
The image data processing is as shown in fig. 2, and includes:
s1, carrying out gray processing on an image, and intercepting an acquired video image into a picture state before processing;
s11, converting the color image from an RGB color model to a CIE Lab color model;
s12, obtaining the perceived brightness of an original color image in a CIE Lab color space, and correcting the perceived brightness of the image by utilizing chromaticity components a and b to realize the preliminary graying of the color image so as to obtain a preliminary gray image;
s13, in a CIE Lab color model, performing contourlet decomposition on an input color image to obtain 1 low-frequency component and a plurality of high-frequency component sub-band images;
s14, carrying out contour wave decomposition on the preliminary gray level image in the same scale as the color image;
s15, calculating the ratio of the chromaticity contrast between each sub-band of the color image and the preliminary gray image, and adding the chromaticity information of the color image into the preliminary gray image according to the proportion to obtain a final gray image with enhanced local contrast;
s2, denoising the image subjected to the graying treatment by using a mean value filtering method;
by selecting the template during filtering, the pixel values of each point in the image are replaced by the average of the pixel values of all points in the template;
the mean value filtering formula is:
in the above-mentioned method, the step of,representing an original image containing noise, i.e. a graying processed image +.>Representing the image obtained after mean filtering, < >>Representation dot->A set of pixels in the template that is centered, +.>Representing the template size;
s3, detecting the image after denoising treatment based on a machine learning algorithm, and judging an abnormal image;
s31, classifying according to the abnormal types, and displaying in a two-dimensional relation diagram by using different colors according to the classification, as shown in FIG. 4;
the maintenance is red, the obstacle removal is yellow, the dredging is green, so that a detector can visually check the abnormal type, and maintenance personnel can be rapidly arranged to perform maintenance, obstacle removal, dredging and other works.
S311, classifying maintenance when the train contact net equipment has poor contact; for example, monitoring whether the contact net equipment has obvious disconnection, deviation and other abnormal conditions;
s312, classifying the overhead line system power supply safety as obstacle removal when the overhead line system power supply safety is influenced by surrounding environment factors; for example, there are no bird nest, tree harm and other surrounding environmental factors which may endanger the power supply safety of the overhead line system;
s313, classifying as dredging when the train track encounters an obstacle. Examples of the obstacle include a non-invasive limit, an obstacle that hinders the running of a rolling stock;
s4, marking the abnormal image.
By classifying the abnormal states, the inspection personnel can quickly locate the abnormal states and the abnormal positions.
Description of working principle: the method comprises the steps of carrying out real-time acquisition on panorama of a contact network suspension and a high-speed railway line by using a CCD camera on a train, and recording a time point in real time in the process;
processing the acquired image data and judging the abnormality in the image data;
carrying out graying treatment on the image, and intercepting the acquired video image into a picture state before the treatment;
converting the color image from the RGB color model to the CIE Lab color model;
obtaining the perceived brightness of an original color image in a CIE Lab color space, and correcting the perceived brightness of the image by utilizing chromaticity components a and b to realize the preliminary graying of the color image so as to obtain a preliminary gray image;
in a CIE Lab color model, performing contourlet decomposition on the color image to obtain 1 low-frequency component and a plurality of high-frequency component sub-band images;
performing contour wave decomposition on the preliminary gray level image in the same scale as the color image;
calculating the ratio of the chromaticity contrast between each sub-band of the input color image and the preliminary gray image, and adding the chromaticity information of the color image into the preliminary gray image according to the ratio to obtain a final gray image with enhanced local contrast;
denoising the image signal by an average filtering method;
by selecting the template during filtering, the pixel values of each point in the image are replaced by the average of the pixel values of all points in the template;
detecting the processed image based on a machine learning algorithm, and judging an abnormal image;
classifying according to the abnormal types, and displaying in a two-dimensional relation chart by using different colors according to the classification;
the maintenance is red, the obstacle removal is yellow, the dredging is green, so that a detector can visually check the abnormal type, and maintenance personnel can be rapidly arranged to perform maintenance, obstacle removal, dredging and other works.
When the train contact net equipment has poor contact, classifying the train contact net equipment as maintenance; for example, monitoring whether the contact net equipment has obvious disconnection, deviation and other abnormal conditions;
classifying the contact net power supply safety as obstacle elimination when the contact net power supply safety is influenced by surrounding environment factors; for example, there are no bird nest, tree harm and other surrounding environmental factors which may endanger the power supply safety of the overhead line system;
when the train track encounters an obstacle, it is classified as unblocked. Examples of the obstacle include a non-invasive limit, an obstacle that hinders the running of a rolling stock;
marking the abnormal image;
marking an abnormality in the image data, recording a time point when the abnormal image data is acquired, judging the position of the marking point based on the time point, and recording the longitude and latitude of the position;
based on a time point corresponding to an abnormal point in the image data, backtracking the position of the train at the time point, and further judging the abnormal position in the image data;
based on the running speed of the train in each time period, the running distance of the train on the track is calculated by combining the abnormal time existing in the image data, and the abnormal position existing in the image data is judged on the track.
Judging whether two positions obtained by a positioning method and an calculating method are overlapped or not;
if yes, the overlapping position is determined to be an abnormal position;
if not, calculating the linear distance between the two positions obtained by the positioning method and the calculation method, and judging whether the linear distance is smaller than a set value;
if yes, the position of the positioning method is determined to be an abnormal position;
and if not, determining the position of the calculation method as an abnormal position.
And selecting all the abnormal point images to form a two-dimensional relation graph, taking longitude and latitude as a relation axis of the two-dimensional graph, displaying the positions of the abnormal points, and displaying the positions on a terminal.
At this time, the patrol workers can conduct personnel maintenance according to the corresponding classification and the position.
The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the specific details of the above embodiments, and various equivalent changes can be made to the technical solutions of the present invention within the scope of the technical concept of the present invention, and these equivalent changes all fall within the scope of the present invention.
Claims (6)
1. A 2C detection method based on image recognition, comprising:
collecting panorama of a contact network suspension and high-speed railway line on a train in real time by using a collecting device, and recording a time point in real time;
processing the acquired image data and judging the abnormality in the image data;
the processing of the acquired image data comprises:
s1, carrying out gray processing on an image, and intercepting an acquired video image into a picture state before processing;
the S1 further includes:
s11, converting the color image from an RGB color model to a CIE Lab color model;
s12, obtaining the perceived brightness of an original color image in a CIE Lab color space, and correcting the perceived brightness of the image by utilizing chromaticity components a and b to realize the preliminary graying of the color image so as to obtain a preliminary gray image; wherein a represents red and green, and b represents yellow and blue;
s13, performing contourlet decomposition on the color image in a CIE Lab color model to obtain 1 low-frequency component sub-band image and a plurality of high-frequency component sub-band images;
s14, carrying out contour wave decomposition on the preliminary gray level image in the same scale as the color image;
s15, calculating the ratio of the chromaticity contrast between each sub-band of the color image and the primary gray image, and adding the chromaticity information of the color image into the primary gray image according to the proportion to obtain a final gray image with enhanced local contrast;
s2, denoising the image subjected to the graying treatment by using a mean value filtering method;
the S2 comprises the steps that a template is selected in the filtering process, and the pixel value of each point in the image is replaced by the average value of the pixel values of all points in the template;
the filtering formula of the mean filtering method is as follows:
in the above, the->Representing an original image containing noise, i.e. a graying processed image +.>Representing the image obtained after mean filtering, < >>Representation dot->A set of pixels in the template that is centered, +.>For aggregate sign +.>Representing the template size;
s3, detecting the image after denoising treatment based on a machine learning algorithm, and judging an abnormal image;
s4, marking the abnormal image;
marking an abnormality in the image data, recording a time point when the abnormal image data is acquired, judging the position of the marking point by a positioning system based on the time point, and recording the longitude and latitude of the position;
and selecting all the abnormal point images to form a two-dimensional relation graph, taking longitude and latitude as a relation axis of the two-dimensional graph, displaying the positions of the abnormal points, and displaying the positions on a terminal.
2. The 2C detection method based on image recognition according to claim 1, wherein: judging the position of the mark point based on the time point, wherein the position comprises a positioning method and a calculation method;
the positioning method comprises the following steps:
based on a time point corresponding to an abnormal point in the image data, backtracking the position of the train at the time point, and further judging the abnormal position in the image data;
the calculation method comprises the following steps:
based on the running speed of the train in each time period, the running distance of the train on the track is calculated by combining the abnormal time existing in the image data, and the abnormal position existing in the image data is judged on the track.
3. The 2C detection method based on image recognition according to claim 2, wherein: judging whether two positions obtained by a positioning method and an calculating method are overlapped or not;
if yes, the overlapping position is determined to be an abnormal position;
if not, calculating the linear distance between the two positions obtained by the positioning method and the calculation method, and judging whether the linear distance is smaller than a set value;
if yes, the position of the positioning method is determined to be an abnormal position;
and if not, determining the position of the calculation method as an abnormal position.
4. The 2C detection method based on image recognition according to claim 1, wherein: the positioning system comprises a GPS; the acquisition equipment comprises a lens and a CCD camera matched with the lens.
5. The 2C detection method based on image recognition according to claim 1, wherein: the terminal comprises a mobile phone, a computer and a display screen.
6. The 2C detection method based on image recognition according to claim 5, wherein: the step S3 further includes:
s31, classifying according to the abnormal types, and displaying in a two-dimensional relation chart by using different colors according to the classification;
s311, classifying maintenance when the train contact net equipment has poor contact;
s312, classifying the train contact net power supply safety as obstacle removal when the train contact net power supply safety is influenced by surrounding environment factors;
s313, classifying the train rail as dredging when the train rail has an obstacle.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310167299.9A CN115861825B (en) | 2023-02-27 | 2023-02-27 | 2C detection method based on image recognition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310167299.9A CN115861825B (en) | 2023-02-27 | 2023-02-27 | 2C detection method based on image recognition |
Publications (2)
Publication Number | Publication Date |
---|---|
CN115861825A CN115861825A (en) | 2023-03-28 |
CN115861825B true CN115861825B (en) | 2023-04-25 |
Family
ID=85658995
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310167299.9A Active CN115861825B (en) | 2023-02-27 | 2023-02-27 | 2C detection method based on image recognition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115861825B (en) |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112330646A (en) * | 2020-11-12 | 2021-02-05 | 南京优视智能科技有限公司 | Motor car bottom abnormity detection method based on two-dimensional image |
CN115100562A (en) * | 2022-06-13 | 2022-09-23 | 国网安徽省电力有限公司安庆供电公司 | Intelligent monitoring system and method for equipment line based on video image and deep learning |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018122589A1 (en) * | 2016-12-30 | 2018-07-05 | 同济大学 | Method for detecting degree of development of asphalt pavement fracture based on infrared thermal image analysis |
-
2023
- 2023-02-27 CN CN202310167299.9A patent/CN115861825B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112330646A (en) * | 2020-11-12 | 2021-02-05 | 南京优视智能科技有限公司 | Motor car bottom abnormity detection method based on two-dimensional image |
CN115100562A (en) * | 2022-06-13 | 2022-09-23 | 国网安徽省电力有限公司安庆供电公司 | Intelligent monitoring system and method for equipment line based on video image and deep learning |
Also Published As
Publication number | Publication date |
---|---|
CN115861825A (en) | 2023-03-28 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110261436B (en) | Rail fault detection method and system based on infrared thermal imaging and computer vision | |
WO2020199538A1 (en) | Bridge key component disease early-warning system and method based on image monitoring data | |
CN105373135B (en) | A kind of method and system of aircraft docking guidance and plane type recognition based on machine vision | |
JP6904614B2 (en) | Object detection device, prediction model creation device, object detection method and program | |
CN108230344B (en) | Automatic identification method for tunnel water leakage diseases | |
KR20200013148A (en) | Method, system and computer program for providing defect analysis service of concrete structure | |
CN108198417B (en) | A kind of road cruising inspection system based on unmanned plane | |
CN108109385A (en) | A kind of vehicle identification of power transmission line external force damage prevention and hazardous act judgement system and method | |
CN110046584B (en) | Road crack detection device and detection method based on unmanned aerial vehicle inspection | |
CN113947731B (en) | Foreign matter identification method and system based on contact net safety inspection | |
CN112819094A (en) | Target detection and identification method based on structural similarity measurement | |
CN112528979B (en) | Transformer substation inspection robot obstacle distinguishing method and system | |
CN109489724A (en) | A kind of tunnel safe train operation environment comprehensive detection device and detection method | |
CN111768417B (en) | Railway wagon overrun detection method based on monocular vision 3D reconstruction technology | |
CN111667655A (en) | Infrared image-based high-speed railway safety area intrusion alarm device and method | |
CN113688817A (en) | Instrument identification method and system for automatic inspection | |
CN113788051A (en) | Train on-station running state monitoring and analyzing system | |
CN112633722A (en) | Vehicle-mounted road safety risk assessment system and method | |
CN112508911A (en) | Rail joint touch net suspension support component crack detection system based on inspection robot and detection method thereof | |
CN112288682A (en) | Electric power equipment defect positioning method based on image registration | |
CN117237597A (en) | Data processing terminal based on Beidou satellite data and AI graph fusion | |
CN113673614B (en) | Metro tunnel foreign matter intrusion detection device and method based on machine vision | |
CN115861825B (en) | 2C detection method based on image recognition | |
CN112115737B (en) | Vehicle orientation determining method and device and vehicle-mounted terminal | |
Saini et al. | Feature-based template matching for joggled fishplate detection in railroad track with drone images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |