CN-122024209-A - Railway ground signal image characteristic analysis and recognition method and system

CN122024209ACN 122024209 ACN122024209 ACN 122024209ACN-122024209-A

Abstract

The application relates to the technical field of railway signal identification, provides a railway ground signal image feature analysis and identification method and a railway ground signal image feature analysis and identification system, and solves the problem that a railway operation vehicle is low in ground signal identification accuracy in a complex environment. The method comprises the steps of collecting visible light and thermal infrared images of a ground signal machine, obtaining mileage coordinate information and type information of the visible light and thermal infrared images, determining a target area of a signal lamp in the images based on the images, calculating predicted mileage coordinates corresponding to the target area according to real-time positioning data, matching the predicted mileage coordinates with the mileage coordinate information, determining the type of the target signal machine by combining the type information, inputting a feature vector sequence and the type corresponding to the target area in continuous multi-frame images into a state recognition model, comparing and determining color types and signal states, combining the coordinates, the color types and the signal states of the target area into recognition results, and packaging the recognition results and sending the recognition results to a cab display terminal. The application improves the accuracy of the recognition of the ground signals of the railway operation vehicle in a complex environment.

Inventors

TIAN XINYING
WANG XIAOQIN
WANG SHUAI
LIU JIAXING
LIU AIWU
DONG LIGUO
SUN ZHANHU
WU CHUANCHUAN
ZHANG YAKE
YU HAIFENG
LIANG WEI
SHEN PENGYU

Assignees

中铁电气化铁路运营管理有限公司

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The railway ground signal image characteristic analysis and identification method is characterized by comprising the following steps: collecting visible light image data and thermal infrared image data of a ground signal machine, and acquiring mileage coordinate information and signal machine type information of the ground signal machine along a railway; determining a target area of a signal lamp in an image based on the visible light image data and the thermal infrared image data; calculating predicted mileage coordinates of a ground signal machine corresponding to the target area along a railway according to real-time positioning data of the operation vehicle, matching the predicted mileage coordinates with the mileage coordinate information, and determining the type of the target signal machine corresponding to the target area by combining the signal machine type information; Inputting a corresponding characteristic vector sequence of the target region in continuous multi-frame images and the type of the target signal machine into a state identification model to call a corresponding signal state change rule, and comparing the change rule of the characteristic vector sequence with the signal state change rule to determine the color type and the signal state of the signal machine; and combining the coordinates of the target area, the color types and the signal states into an identification result, packaging according to the preset railway communication requirement, and sending the identification result to a cab display terminal.
2. The method according to claim 1, wherein the step of inputting the feature vector sequence corresponding to the target area in the continuous multi-frame images and the target signal machine type into a state recognition model to retrieve a corresponding signal state change rule, and comparing the change rule of the feature vector sequence with the signal state change rule to determine the color class and the signal state of the signal machine comprises: Arranging the feature vectors corresponding to the target areas in the continuous multi-frame images according to the acquisition time sequence to form a feature vector sequence; Inputting the characteristic vector sequence and the type of the target signal machine into a state recognition model, and performing time sequence coding on the characteristic vector sequence through a characteristic coding module of the state recognition model to generate a time sequence characteristic diagram, wherein the time sequence characteristic diagram is used for representing characteristic change trend among continuous frames; the rule matching module of the state identification model is used for calling a corresponding color change time sequence rule from a preset signal state rule base according to the type of the target signal machine; Performing inter-frame difference detection on the time sequence feature map through a time sequence analysis module of the state identification model to identify the frame position of the color feature value in the feature vector sequence, and generating a jump time interval set according to the acquisition time interval between adjacent jump frames by a time interval calculation unit built in the time sequence analysis module; And comparing the color jump sequence corresponding to the jump frame position, each time interval in the jump time interval set with the color appearance sequence and the duration time range specified in the color change time sequence rule item by item through a state judgment module of the state identification model, and generating the color category and the signal state of the annunciator in the current frame image according to the comparison result.
3. The method for identifying the image features of the railway ground signal according to claim 2, wherein the time sequence analysis module passing through the state identification model performs inter-frame difference detection on the time sequence feature map to identify the frame position where the color feature value hops in the feature vector sequence, and the time interval calculation unit built in the time sequence analysis module generates a hop time interval set according to the acquisition time interval between adjacent hop frames, and the method comprises the steps of: Performing differential operation between adjacent frames on each feature vector arranged according to time sequence in the time sequence feature map through a differential calculation sub-module of the time sequence analysis module to generate an inter-frame differential value sequence; Comparing each difference value in the inter-frame difference value sequence with a preset jump detection threshold value through a threshold value comparison sub-module of the time sequence analysis module, marking the image acquisition time of the next frame corresponding to the difference value as a jump frame position of which the color characteristic value is jumped when the difference value is larger than the jump detection threshold value, and recording a frame index corresponding to the jump frame position, wherein the jump detection threshold value is set according to the color change characteristic of the railway signal machine; Through a jump frame screening sub-module of the time sequence analysis module, all marked jump frame positions are subjected to redundancy elimination, and only the first jump frame position in a plurality of continuous jump frame positions is reserved as an effective jump frame position by combining with the upper limit of the color switching times of a railway signal machine in a fixed period, so that a simplified jump frame position list is obtained; Calculating, by a time interval calculating unit of the timing analysis module, a time difference value between each adjacent hopping frame according to image acquisition time stamps corresponding to two adjacent hopping frame positions in the reduced hopping frame position list, and taking the time difference value as a hopping time interval; And according to the standard range of each color duration of the railway signal machine, the time interval calculation unit performs validity screening on the jump time intervals, eliminates time intervals exceeding the standard range, arranges the residual jump time intervals according to the sequence of jump occurrence, and generates a jump time interval set.
4. The method for analyzing and identifying the characteristics of the railway ground signal image according to claim 1, wherein the calculating the predicted mileage coordinates of the ground signal corresponding to the target area along the railway according to the real-time positioning data of the working vehicle, matching the predicted mileage coordinates with the mileage coordinate information, and determining the type of the target signal corresponding to the target area by combining the signal type information comprises: Extracting the current global positioning system coordinates of the operation vehicle from the real-time positioning data, and converting the global positioning system coordinates into railway mileage values of the operation vehicle according to a preset railway line geographic information table; According to the pixel coordinates of the target area in the image, combining the internal parameters and the external installation parameters of the image acquisition equipment, calculating the distance and azimuth angle of the ground signal machine corresponding to the target area relative to the operation vehicle, and according to the distance, the azimuth angle and the railway mileage value of the operation vehicle at present, calculating the predicted mileage coordinates of the ground signal machine corresponding to the target area along the railway; Reading mileage coordinate information and signal type information of all ground signal machines from a preset signal machine position database, inputting the predicted mileage coordinates and the mileage coordinate information into a hash-code-based quick matching model, and respectively carrying out hash function mapping on the predicted mileage coordinates and the mileage coordinate information by the hash-code-based quick matching model to generate a predicted mileage hash value and each mileage coordinate hash value; Sorting the similarity by calculating the Hamming distance between the predicted mileage hash value and each mileage coordinate hash value, and screening out the signal machine type information corresponding to the mileage coordinate information with the minimum Hamming distance as a preliminary matching result; Inputting mileage coordinate information and annunciator type information of the ground annunciator corresponding to the preliminary matching result into a dynamic tracker based on Kalman filtering, and performing iterative updating and error correction on a predicted mileage coordinate of the ground annunciator corresponding to the preliminary matching result by the dynamic tracker based on Kalman filtering to generate a corrected target mileage coordinate; And reading corresponding signal type information from a preset signal position database according to the corrected target mileage coordinates, and taking the corresponding signal type information as a target signal type corresponding to the target area.
5. The railway ground signal image characteristic analysis and recognition method according to claim 1, further comprising, after being transmitted to the cab display terminal: Acquiring current running state parameters and real-time distance between the working vehicle and a front annunciator, wherein the running state parameters comprise real-time vehicle speed, train pipe pressure and position of a driver handle, and the real-time distance is calculated according to coordinates of the target area and real-time positioning data of the working vehicle; inputting the signal state and the running state parameters in the identification result into a control rule base based on fuzzy logic, and carrying out fuzzification processing and fuzzy reasoning on the current signal state and the running state parameters through the control rule base based on the fuzzy logic to generate a security level label at the current moment; Arranging the identification results, the running state parameters, the real-time distance and the security level labels at a plurality of continuous moments according to a time sequence to form running state time sequence data, and inputting the running state time sequence data into a long-short-period memory network, wherein the long-short-period memory network predicts the relative position change trend and the possibility of signal state change between the working vehicle and a front annunciator at the next moment according to the historical change rule in the running state time sequence data; triggering an early warning instruction when the long-period memory network predicts that signal state switching occurs within a preset time threshold or the operation vehicle possibly passes over a annunciator; And generating secondary early warning information containing early warning levels according to the early warning instructions, packaging the secondary early warning information according to preset railway communication requirements, sending the packaged secondary early warning information to a cab display terminal for highlighting, and simultaneously playing corresponding early warning voice through vehicle-mounted sound equipment.
6. The method for analyzing and identifying the image features of the railway ground signal according to claim 1, wherein the step of combining the coordinates of the target area, the color class and the signal state into an identification result, and transmitting the identification result to a cab display terminal after packaging according to a preset railway communication requirement comprises the steps of: Converting the coordinates of the target area in the image into actual position coordinates under a railway coordinate system, and combining the actual position coordinates, the color categories and the signal states into a recognition result; Packaging the identification result according to a preset railway communication requirement to generate a communication data frame; The communication data frame is sent to a cab display terminal through a vehicle-mounted network in the operation vehicle, the cab display terminal analyzes the communication data frame, the position of a signal machine corresponding to the actual position coordinate is marked on an electronic map in an icon mode, the color type and the signal state are displayed in a specified area of a screen, and corresponding voice prompt information is generated according to the signal state and played through vehicle-mounted sound equipment.
7. The method of claim 1, wherein determining a target area of a signal in an image based on the visible light image data and the thermal infrared image data comprises: according to the pre-stored joint calibration parameters, performing spatial coordinate system registration on the visible light image data and the thermal infrared image data to obtain registered multi-mode image data; Analyzing the thermal infrared image data in the registered multi-mode image data by adopting a region segmentation algorithm to extract a candidate signal region, and mapping the coordinates of the candidate signal region to the visible light image data in the registered multi-mode image data; And verifying the candidate signal areas in the visible light image data, removing the interference areas according to the verification result, and taking the image areas corresponding to the residual candidate signal areas as target areas of the signal lamps in the images.
8. A railway ground signal image feature analysis and recognition system, comprising: the acquisition module is used for acquiring visible light image data and thermal infrared image data of the ground annunciator and acquiring mileage coordinate information and annunciator type information of the ground annunciator along a railway; The determining module is used for determining a target area of the signal lamp in the image based on the visible light image data and the thermal infrared image data; The calculation module is used for calculating the predicted mileage coordinates of the ground annunciators corresponding to the target areas along the railway according to the real-time positioning data of the operation vehicle, matching the predicted mileage coordinates with the mileage coordinate information, and determining the types of the target annunciators corresponding to the target areas by combining the annunciator type information; The input module is used for inputting the corresponding characteristic vector sequence of the target area in the continuous multi-frame images and the type of the target signal machine into a state recognition model so as to call a corresponding signal state change rule, comparing the change rule of the characteristic vector sequence with the signal state change rule and determining the color type and the signal state of the signal machine; And the combination module is used for combining the coordinates of the target area, the color types and the signal states into a recognition result, packaging the recognition result according to the preset railway communication requirement and then transmitting the recognition result to a cab display terminal.
9. An electronic device, comprising: A memory for storing a computer program; A processor for implementing the steps of the railway ground signal image feature analysis and identification method according to any one of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium, wherein a computer program is stored in the computer readable storage medium, and when executed by a processor, the computer program is capable of implementing the railway ground signal image feature analysis and identification method according to any one of claims 1 to 7.

Description

Railway ground signal image characteristic analysis and recognition method and system Technical Field The application relates to the technical field of railway signal identification, in particular to a railway ground signal image characteristic analysis and identification method and system. Background The railway ground signal image characteristic analysis and identification method is mainly applied to a signal confirmation scene of a railway contact net operation vehicle in the shunting and interval operation process, and along with the development of railway transportation to an intelligent direction and the improvement of all-weather operation requirements, the identification of the ground signal machine state by an image processing technology is an important technical direction for improving the operation safety. The current common railway signal image recognition method mainly adopts a single visible light camera to collect the image of a signal machine, extracts the color characteristics and the shape characteristics of a signal lamp through an image processing algorithm to recognize, can obtain a better recognition effect in daytime or under good illumination conditions, and can also be combined with the fixed position information of the signal machine along the railway to carry out auxiliary judgment on the recognition result, so that the method has been applied to part of railway operation scenes. However, in an actual railway running environment, the signal identification process is easily affected by illumination changes such as day and night alternation, rain, snow, haze and the like and weather factors, so that a single visible light image is difficult to accurately reflect the real state of a signal lamp under a low-visibility condition, meanwhile, the color change rule of a signal machine in a continuous time period is directly related to driving safety, and the signal machine is difficult to adapt to complex and changeable site conditions by judging only by means of static features of a single frame image, so that the technical problem of insufficient identification reliability of railway ground signals in the complex environment exists in the prior art. Disclosure of Invention The application provides a railway ground signal image feature analysis and identification method and system, which are used for solving the problem of low accuracy in identifying ground signals in a complex environment of a railway operation vehicle in the prior art. In order to solve the technical problems, in a first aspect, the present application provides a method for analyzing and identifying features of a railway ground signal image, including: collecting visible light image data and thermal infrared image data of a ground signal machine, and acquiring mileage coordinate information and signal machine type information of the ground signal machine along a railway; determining a target area of a signal lamp in an image based on the visible light image data and the thermal infrared image data; calculating predicted mileage coordinates of a ground signal machine corresponding to the target area along a railway according to real-time positioning data of the operation vehicle, matching the predicted mileage coordinates with the mileage coordinate information, and determining the type of the target signal machine corresponding to the target area by combining the signal machine type information; Inputting a corresponding characteristic vector sequence of the target region in continuous multi-frame images and the type of the target signal machine into a state identification model to call a corresponding signal state change rule, and comparing the change rule of the characteristic vector sequence with the signal state change rule to determine the color type and the signal state of the signal machine; and combining the coordinates of the target area, the color types and the signal states into an identification result, packaging according to the preset railway communication requirement, and sending the identification result to a cab display terminal. In a second aspect, the present application provides a railway ground signal image feature analysis and recognition system, comprising: the acquisition module is used for acquiring visible light image data and thermal infrared image data of the ground annunciator and acquiring mileage coordinate information and annunciator type information of the ground annunciator along a railway; The determining module is used for determining a target area of the signal lamp in the image based on the visible light image data and the thermal infrared image data; The calculation module is used for calculating the predicted mileage coordinates of the ground annunciators corresponding to the target areas along the railway according to the real-time positioning data of the operation vehicle, matching the predicted mileage coordinates with the mileage coordinate information, and determining the ty