Data processing method for track foreign matter recognition

Abstract

The invention discloses a data processing method for track foreign matter identification, and particularly relates to the technical field of track traffic safety monitoring, and the method comprises the following steps: S1, information collection and processing, S2, track area extraction, S3, depth information acquisition, S4, dynamic background modeling and updating, S5, safe driving frame determination, S6, foreign matter detection and identification, and S7, danger degree evaluation. S8, intelligent alarm and feedback mechanism; and S9, result output and recording. According to the method, accurate recognition and efficient processing of the track foreign matter are achieved through cooperation of all the steps, track area extraction and depth information acquisition are combined, the detection range is accurately delimited, the three-dimensional space position is defined, the detection precision is improved, the foreign matter is accurately detected through foreign matter detection and recognition and deep learning, and the detection efficiency is improved. The processing priority is determined in cooperation with danger degree evaluation, information is transmitted in time in combination with an intelligent alarm and feedback mechanism, and the accuracy, reliability and timeliness of track foreign matter recognition are greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of rail transit safety monitoring, and in particular to a data processing method for identifying foreign objects on a track. Background Art

[0002] In today's rapidly developing rail transit landscape, ensuring train safety is paramount. Track foreign object identification, a key component of ensuring operational safety, places extremely high demands on data processing methods. Existing data processing methods for track foreign object identification suffer from numerous drawbacks. Some rely on simple image recognition algorithms, enabling only preliminary assessments of obvious foreign objects. Other data processing methods lack the ability to integrate multi-source data, resulting in incomplete information utilization.

[0003] Therefore, it is urgent to design a data processing method for track foreign object identification that can integrate multi-source data, efficiently process complex scene data, meet real-time requirements and make full use of historical data. Summary of the Invention

[0004] The main purpose of the present invention is to provide a data processing method for identifying foreign objects on tracks, which can effectively solve the problems mentioned in the background technology.

[0005] To achieve the above object, the technical solution adopted by the present invention is: A data processing method for identifying foreign objects on a track comprises the following steps: S1, information collection and processing: using a high-definition camera array unit to collect real-time images of the track area, and then processing the collected images; S2, track area extraction: Based on the images collected and processed in S1, the track area is identified and extracted through a hybrid algorithm; S3, depth information acquisition: After completing the acquisition and preprocessing of the track area image in S1 and the accurate extraction of the track area in S2, the image coordinates and depth coordinates are established for the processed image through relevant algorithms and mathematical models to obtain the depth information of the track area; S4, dynamic background modeling and updating: The Gaussian mixture model algorithm and the kernel density estimation algorithm are used to dynamically model the background of the track area, update the background model parameters in real time, adapt to changes in ambient lighting and shadows, and establish a background update threshold mechanism; S5, safe driving frame determination: based on the depth information of the track area, the safe driving frame is determined according to the track design specifications and safety standards; S6, foreign object detection and identification: The images collected and processed in S1 are input into a foreign object detection model based on deep learning. The model's recognition capability is used to determine whether there is a foreign object within the safe driving box. Based on the judgment result, if a foreign object is present, the next step is carried out. If no foreign object is present, the system indicates normal operation. S7, Danger level assessment: Based on the judgment results in S6, the danger level of the foreign object is assessed according to factors such as the location, size, and intersection ratio with the track, and a dynamic danger level assessment mechanism is established; S8, intelligent alarm and feedback mechanism: When a high-risk foreign object is detected, the intelligent alarm mechanism is automatically triggered and the alarm information is recorded, and the detailed situation of the foreign object is recorded and fed back; S9, result output and recording: output the results of foreign body detection and store the detection results and alarm information in the database for subsequent query and analysis.

[0006] Preferably, in S1, the collected image is preprocessed using a filtering algorithm, and the preprocessing operation includes but is not limited to denoising, contrast enhancement, and defogging.

[0007] Preferably, in S2, the hybrid algorithm includes but is not limited to grayscale conversion, edge detection, and Hough transform.

[0008] Preferably, in S4, when the background pixel change exceeds a threshold, the background model is updated, and the background difference method is used to detect the foreground target, and the noise and holes in the detection result are eliminated in combination with morphological processing to accurately separate the moving object.

[0009] Preferably, in S5, the detection range is within a safe driving frame.

[0010] Preferably, in S7, the foreign bodies are classified into three levels: high risk, medium risk and low risk, and priority recommendations are provided for subsequent processing, with high risk foreign bodies being processed first.

[0011] Preferably, in S7, the dynamic assessment mechanism updates the foreign object hazard level in real time according to the train operation plan and the changes in foreign objects.

[0012] Preferably, in S8, the intelligent alarm mechanism includes but is not limited to sound and light alarms, SMS notifications, and system pop-up windows to issue alarms to relevant personnel.

[0013] Preferably, in S9, the results of the foreign body detection include but are not limited to the location, type, and danger level of the foreign body.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention provides high-quality image data for subsequent steps through information collection and processing, reducing environmental interference; combines track area extraction and depth information acquisition to accurately define the detection range and clarify the three-dimensional spatial position, thereby improving detection accuracy; dynamic background modeling and updating eliminate dynamic interference and reduce false detection rates; safe driving frame determination further narrows the detection range; similar foreign object detection and identification uses deep learning to accurately detect foreign objects; risk assessment clarifies processing priorities; intelligent alarm and feedback mechanisms promptly transmit information; and result output and recording facilitate data management and decision-making.

[0015] 2. The present invention realizes the precise identification and efficient handling of foreign objects on the track through the coordinated efforts of various steps, which not only greatly improves the accuracy, reliability and timeliness of foreign object identification on the track, but also provides a scientific basis for train scheduling and track maintenance, effectively ensuring the safe operation of rail transit and reducing the risk of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0017] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0018] Example 1, see Figure 1 , a data processing method for identifying foreign objects on a track, comprising the following steps: S1, Information Collection and Processing: A multi-view HD camera array is deployed to form a three-dimensional monitoring network on both sides of the track and above it, ensuring no blind spots. The camera frame rate is set to at least 30 frames per second to ensure continuous capture of dynamic targets and capture dynamic images at a high frame rate. The collected images are filtered to remove noise and adjust brightness and contrast to enhance image clarity and legibility. Establish an environmental parameter monitoring subsystem to collect data such as light intensity, temperature and humidity, and atmospheric particulate matter concentration in real time, and automatically adjust image acquisition and preprocessing strategies according to environmental parameters.

[0019] Ensure the acquisition of high-quality track image data, provide clear and complete visual information for subsequent processing, ensure the accuracy of foreign object detection, reduce the impact of environmental factors (such as weather and lighting) on images, and provide clearer input data for subsequent image analysis and foreign object detection.

[0020] The above filtering algorithm is used to pre-process the collected images, including denoising, contrast enhancement, defogging and other operations. The original image data is input according to: The filtered image can be obtained, which removes noise, enhances contrast, and improves image clarity and recognizability.

[0021] in, G(x,y): the value of the Gaussian filter at position (x,y); σ: standard deviation of Gaussian filtering, which controls the smoothness of the filter; x and y: coordinates of the image pixel.

[0022] S2, Track Area Extraction: First, the color image is converted into a grayscale image. The Canny edge detection algorithm is used to extract image edge information. The Hough transform is used to identify track straight line features. The track curve is accurately fitted through an iterative optimization algorithm to determine the track boundary and position. The track area can be accurately divided. A track template library is established to store standard track models of different track types (such as high-speed rail, subway, and light rail). Template matching technology is used to assist in track area identification, reduce the interference of non-track areas on foreign object detection, and improve detection efficiency and accuracy. When the Canny edge detection algorithm is used above, the input data is a grayscale image, according to: Image edge information can be obtained for subsequent track feature extraction.

[0023] in, M: edge strength; θ: edge direction; Gx and Gy: are the gradients of the image in the a and y directions respectively.

[0024] When using Hough transform to identify track straight line features, the input data is the image after edge detection, according to: The straight line characteristics of the track can be derived to determine the boundaries and positions of the track.

[0025] in, ρ: the vertical distance from the origin to the line; θ: angle between the line and the axis; x and y: coordinates of the edge point in the image.

[0026] S3, Depth Information Acquisition: After completing the acquisition and preprocessing of the track area image in S1 and the precise extraction of the track area in S2, the processed image is used to establish image coordinates and depth coordinates using relevant algorithms and mathematical models to obtain the depth information of the track area. This depth information determines the three-dimensional spatial position of the track area, providing a basis for the three-dimensional location and size determination of foreign objects, further improving detection accuracy.

[0027] S4, dynamic background modeling and updating: The Gaussian mixture model algorithm and the kernel density estimation algorithm are used to dynamically model the background of the track area. The background model parameters are updated in real time to adapt to changes in ambient lighting and shadows. A background update threshold mechanism is established. When the background pixel change exceeds the threshold, the background model update is triggered to ensure that the background model always matches the actual environment. The background difference method is used to detect foreground targets, and morphological processing is combined to eliminate noise and holes in the detection results to accurately separate moving objects.

[0028] It effectively eliminates dynamic interference such as lighting changes and shadow movement, and the background modeling accuracy reaches over 98%, reducing false detections caused by background changes. The dynamic background update mechanism enables the system to quickly adapt to environmental changes, reducing the false alarm rate caused by environmental interference by 40%, and improving the reliability of foreign object detection.

[0029] The parameter update formula of the above Gaussian mixture model can be expressed as: in, μ k and ∑ k are the mean and covariance matrices of the kth Gaussian component respectively; : The posterior probability that the sample belongs to the Gaussian component; N: total number of samples.

[0030] When the background difference method is used to detect foreground targets, the input data is the current frame image and the background model image. According to: The foreground target image, that is, the detected moving object, can be obtained.

[0031] in, F(t): foreground image (i.e., the difference between the current frame and the background model); I(t): current frame image; B(t): background model image.

[0032] When using morphological processing, the input data is the foreground target image, according to: The image after morphological processing can be obtained, which eliminates noise and holes and accurately separates moving objects.

[0033] in, E(B,S): image after corrosion; B: input image; S: Structural element.

[0034] S5, Determine the safe travel frame: Based on the depth information of the track area, the size and position of the safe travel frame are determined according to track design specifications and safety standards. The safe travel frame is appropriately extended to the sides and top of the track to reserve a safe buffer area. In addition, factors such as track slope and curvature are taken into account, and the safe travel frame is spatially transformed to ensure that the safe travel area can be accurately delineated under different track terrains. A dynamic adjustment mechanism for the safe travel frame is established, and the size and position of the safe travel frame are adjusted in real time based on factors such as train speed and track maintenance status. The demarcation of the safe driving frame narrows the foreign object detection range to the actual effective area, avoiding the mistaken identification of objects outside the track as foreign objects, further narrowing the detection range, improving the targetedness and reliability of detection, and reducing the false alarm rate.

[0035] S6, foreign object detection and identification: The preprocessed images from S1 are input into a deep learning-based foreign object detection model (such as the improved YOLOv8 model). First, the improved YOLOv8 model undergoes transfer learning. This model is trained and optimized using a dataset containing a large number of track foreign object samples to improve its ability to identify track foreign objects. The model's identification capabilities are then used to determine whether there are foreign objects within the safe driving box. If so, the system proceeds to the next step. If not, the system indicates normal operation and the YOLOv8 model continues to process other input images. Utilizing the powerful recognition capabilities of deep learning models, foreign objects on the track can be accurately detected and their location, category, and other information can be output.

[0036] When using the YOLOv8 model to process images, input the pre-processed image in step S1, according to: The specific location, category and other information of the foreign matter that can be detected.

[0037] in, L: total loss; λ coord ,λ noobj ,λ cls : weight coefficient; x i 、yi 、w i 、h i : The center coordinates, width and height of the ground truth bounding box; : The center coordinates, width and height of the bounding box.

[0038] S7, Hazard Level Assessment: Based on the judgment results in S6, a foreign object hazard level assessment index system is established. The hazard level of foreign objects is assessed by comprehensively considering factors such as the location of the foreign object (such as whether it is located at the center of the track or close to the track edge), size (relative size to the track), motion state (stationary, moving), and intersection ratio with the track. A dynamic hazard level assessment mechanism is established, and the foreign object hazard level is updated in real time based on the train operation plan and changes in foreign objects. Foreign objects are classified into three levels: high-risk, medium-risk, and low-risk, and priority recommendations are provided for subsequent processing. The dynamic assessment mechanism ensures the timeliness and accuracy of risk level assessment, provides a reliable decision-making basis for train scheduling and track maintenance, gives priority to handling high-risk foreign objects, and improves the efficiency of track safety maintenance.

[0039] When evaluating according to the dynamic risk assessment mechanism, the input data is the location, size, intersection and union ratio of the foreign object with the track, etc., according to: The risk level of the foreign body can be determined (high risk, medium risk, low risk).

[0040] in, D: Danger level of foreign matter; P: the position weight of the foreign object (e.g., whether it is located at the center of the track); S: size weight of the foreign body (relative size to the track); I: intersection-combination ratio of foreign matter and track; w1,w2,w3: weight coefficients, adjusted according to actual conditions.

[0041] S8, intelligent alarm and feedback mechanism: When a high-risk foreign object is detected, the intelligent alarm mechanism is automatically triggered to alert relevant personnel through sound and light alarms, SMS notifications, system pop-up windows, etc. In addition, an alarm information management system is established to record information such as alarm time, foreign object location, category, and hazard level, and to classify, count, and analyze the detection results and alarm information, and store them in a database for subsequent query and analysis. It can also provide an alarm feedback function, and relevant personnel can confirm and process the alarm information through the system and feedback the processing results.

[0042] S9, Result Output and Recording: Develop an intuitive and friendly system interface to display foreign body detection results in the form of maps, charts, etc., including information such as the location, category, and hazard level of the foreign body. At the same time, the recorded data can be used for subsequent statistical analysis and system optimization. The detection results and alarm information are stored in the database in real time. A data index and query mechanism is established to facilitate rapid query and retrieval of historical data. The data in the database is regularly backed up and archived to ensure data security and integrity.

[0043] The visual output of results provides track maintenance personnel with intuitive and clear detection information. The perfect data storage and management mechanism provides rich data resources for track safety analysis and system optimization, supporting track safety decision-making based on big data.

[0044] During the operation of this embodiment, the various steps work together to achieve accurate identification and efficient processing of foreign objects on the track. Information collection and processing provide high-quality image data for subsequent steps, reducing environmental interference. Track area extraction and depth information acquisition are combined to accurately define the detection range and clarify the three-dimensional spatial position, improving detection accuracy. Dynamic background modeling and updating eliminate dynamic interference and reduce false detection rates. Safe driving frame determination further narrows the detection range. Similarly, foreign object detection and identification utilize deep learning to accurately detect foreign objects. Danger level assessment clarifies processing priorities. Intelligent alarm and feedback mechanisms deliver timely information. Result output and recording facilitate data management and decision-making. This not only significantly improves the accuracy, reliability, and timeliness of foreign object identification on the track, but also provides a scientific basis for train scheduling and track maintenance, effectively ensuring the safe operation of rail transit and reducing the risk of accidents.

[0045] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A data processing method for identifying foreign objects on a track, characterized in that: The following steps are involved: S1, information collection and processing: using a high-definition camera array unit to collect real-time images of the track area, and then processing the collected images; S2, track area extraction: Based on the images collected and processed in S1, the track area is identified and extracted through a hybrid algorithm; S3, depth information acquisition: After completing the acquisition and preprocessing of the track area image in S1 and the accurate extraction of the track area in S2, the image coordinates and depth coordinates are established for the processed image through relevant algorithms and mathematical models to obtain the depth information of the track area; S4, dynamic background modeling and updating: The Gaussian mixture model algorithm and the kernel density estimation algorithm are used to dynamically model the background of the track area, update the background model parameters in real time, adapt to changes in ambient lighting and shadows, and establish a background update threshold mechanism; S5, safe driving frame determination: based on the depth information of the track area, the safe driving frame is determined according to the track design specifications and safety standards; S6, foreign object detection and identification: The images collected and processed in S1 are input into a foreign object detection model based on deep learning. The model's recognition capability is used to determine whether there is a foreign object within the safe driving box. Based on the judgment result, if a foreign object is present, the next step is carried out. If no foreign object is present, the system indicates normal operation. S7, Danger level assessment: Based on the judgment results in S6, the danger level of the foreign object is assessed according to factors such as the location, size, and intersection ratio with the track, and a dynamic danger level assessment mechanism is established; S8, intelligent alarm and feedback mechanism: When a high-risk foreign object is detected, the intelligent alarm mechanism is automatically triggered and the alarm information is recorded, and the detailed situation of the foreign object is recorded and fed back; S9, result output and recording: output the results of foreign body detection and store the detection results and alarm information in the database for subsequent query and analysis.

2. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S1, the collected image is preprocessed using a filtering algorithm, and the preprocessing operation includes but is not limited to denoising, contrast enhancement, and defogging.

3. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S2, the hybrid algorithm includes but is not limited to grayscale conversion, edge detection, and Hough transform.

4. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S4, when the background pixel change exceeds a threshold, the background model is updated, and the foreground target is detected using the background difference method. The noise and holes in the detection result are eliminated in combination with morphological processing to accurately separate the moving object.

5. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S5, the detection range is within the safe driving frame.

6. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S7, the foreign bodies are classified into three levels: high risk, medium risk, and low risk, and priority recommendations are provided for subsequent processing, with high-risk foreign bodies being processed first.

7. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S7, the dynamic assessment mechanism updates the foreign object hazard level in real time according to the train operation plan and the changes in foreign objects.

8. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S8, the intelligent alarm mechanism includes but is not limited to sound and light alarms, SMS notifications, and system pop-up windows to issue alarms to relevant personnel.

9. The data processing method for identifying foreign objects on a track according to claim 1, characterized in that: In S9, the results of the foreign body detection include but are not limited to the location, type, and danger level of the foreign body.

Images