Intelligent identification and risk assessment method and system for light and hard floating objects along railway

Abstract

The invention provides an intelligent identification and risk assessment method and system for light and hard floating objects along a railway, and belongs to the field of railway safety monitoring. According to the method, unmanned aerial vehicle aerial photography and fixed monitoring image data are acquired through a multi-source data acquisition system, a floating object is identified and segmented through an improved MaskDINO instance segmentation model, actual parameters are calculated by means of perspective transformation and conversion to a physical coordinate system, the intrusion probability is calculated through a risk assessment model by fusing multiple factors such as real-time meteorology and the like, and the risk assessment accuracy is improved. And generating graded early warning and pushing to management personnel. The system comprises a multi-source data acquisition module, a pre-processing module, a recognition module, a coordinate conversion module, a risk assessment module and an early warning release module, an edge and cloud computing collaborative architecture is adopted, a mobile terminal application is matched, and full-process intelligent monitoring and closed-loop management are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of railway safety monitoring, in particular to a railway line light and hard floating object intelligent identification and risk assessment method and system. BACKGROUND

[0002] Railway is the core national transportation infrastructure, and its operation safety is directly related to the safety of people's life and property and the stable operation of national economy. Light and hard floating objects along the railway are easily swept into the railway safety limit in windy weather, causing serious safety hazards to railway operation and threatening train driving safety and transportation order. These floating objects are diverse, including light objects such as plastic film, dust screen and advertising cloth, and hard components such as color steel tile, engineering protective net and temporary building materials, which have wide distribution range, are greatly affected by the environment and are difficult to monitor.

[0003] The current detection of environmental safety hazards along the railway is still mainly based on manual patrol, which is not only low in efficiency, but also has a large number of blind areas, making it difficult to meet the needs of continuous monitoring in long distances and complex terrain. Although the existing image recognition-based solutions have been applied, there are obvious shortcomings: the recognition accuracy is poor, and it is difficult to accurately identify various floating objects in complex backgrounds and variable lighting conditions; there is a lack of scientific and effective risk assessment mechanism, which cannot provide accurate decision support for management departments combined with multi-dimensional information; the data source is single, making it difficult to achieve continuous monitoring of all sections without dead angles; the dynamic monitoring capability is insufficient, which cannot effectively predict the motion trajectory and state change of the floating objects, making it difficult to prevent risks in advance. SUMMARY

[0004] In view of the above technical deficiencies, the purpose of the present application is to provide a railway line light and hard floating object intelligent identification and risk assessment method and system to solve the problems of low recognition accuracy and lack of effective risk assessment in the prior art.

[0005] To achieve the above purpose, the technical scheme adopted by the present application is as follows:

[0006] A railway line light and hard floating object intelligent identification and risk assessment method, comprising the following steps:

[0007] S1, acquiring image data along the railway by a multi-source data acquisition system, including unmanned aerial vehicle aerial image and fixed monitoring device image;

[0008] S2, using an improved MaskDINO instance segmentation model to identify and segment the light and hard floating objects in the image;

[0009] S3, mapping the identified floating objects from the image coordinate system to the physical coordinate system combined with the perspective transformation matrix, calculating their actual physical size and distance from the railway safety limit;

[0010] S4, acquiring real-time weather data, fusing floating object type, physical characteristics, distance information, and calculating the probability of floating object intrusion into the safety limit of the railway using a risk assessment model;

[0011] S5, generating graded warning information according to the risk assessment results and pushing it to relevant management personnel through a warning release system.

[0012] Preferably, the multi-source data acquisition system adopts a heterogeneous image data fusion strategy, including fusion analysis of image data obtained from different angles and different times in the same area.

[0013] Preferably, the improved MaskDINO instance segmentation model fuses weather data as auxiliary input features and performs cross-modal fusion of weather conditions and visual features through a multi-head attention mechanism.

[0014] Preferably, the risk assessment model is a multi-factor weighted model, considering factors including floating object type, floating object size, distance between floating object and railway safety limit, environmental wind speed and direction, and historical intrusion records.

[0015] Preferably, in step S1, the unmanned aerial vehicle autonomously plans the inspection path, including dynamically generating an optimal inspection path based on the geometric characteristics of the railway line, the historical frequency of floating object occurrence, and real-time weather conditions.

[0016] A railway line light and hard floating object intelligent identification and risk assessment system, comprising:

[0017] A multi-source data acquisition module including an unmanned aerial vehicle shooting unit and fixed monitoring equipment;

[0018] A data preprocessing module for denoising, enhancing, and standardizing the collected image data;

[0019] A floating object identification module using an improved MaskDINO instance segmentation model for light and hard floating object identification and segmentation;

[0020] A coordinate conversion module for converting image coordinates to physical coordinates through perspective transformation;

[0021] A risk assessment module for fusing multi-source information to assess floating object risk levels;

[0022] A warning release module for generating warning information according to risk levels and pushing it to management personnel.

[0023] Preferably, it also includes a mobile terminal application; the mobile terminal application is used to receive warning information, view floating object details, and submit processing feedback.

[0024] Preferably, the system adopts an edge computing and cloud computing collaborative architecture, and the identification algorithm is partially deployed on the edge device for real-time processing, and the model training and risk assessment are deployed on the cloud server.

[0025] Compared with the prior art, the present application has the following advantages:

[0026] The multi-source data acquisition and heterogeneous fusion strategy, combined with the unmanned aerial vehicle autonomous dynamic inspection path planning and continuous monitoring of fixed monitoring equipment, solves the problem of single traditional monitoring data source and insufficient continuous monitoring of the whole section, and improves the monitoring coverage and data integrity.

[0027] The improved MaskDINO instance segmentation model fuses meteorological data as auxiliary features, enhances the recognition ability in complex environment and variable light through cross-modal fusion and optimization of network structure, and significantly improves the accuracy of light and hard floating object identification and segmentation.

[0028] The perspective transformation realizes accurate mapping of image coordinates to physical coordinates, can accurately obtain the actual size and safety distance of the floating object, and provides reliable data support for risk assessment, solving the problem of inaccurate physical parameter estimation of traditional methods.

[0029] The multi-factor fusion risk assessment model comprehensively considers the characteristics of floating objects, environmental conditions and historical data, realizes scientific division of risk levels, provides accurate decision basis for management departments, and makes up for the defects of the prior art that lack effective risk assessment.

[0030] The edge computing and cloud computing collaborative architecture ensures real-time processing of the identification algorithm and efficient operation of model training and risk assessment, and the mobile terminal application realizes the closed-loop management of early warning receiving, detail viewing and processing feedback, improving the emergency disposal efficiency and management standardization. BRIEF DESCRIPTION OF DRAWINGS

[0031] Figure 1 The system structure diagram of the present application;

[0032] Figure 2 The flowchart of the method of the present application;

[0033] Figure 3 The structure diagram of the MaskDINO instance segmentation model in the present application;

[0034] Figure 4 The structure diagram of the risk assessment model in the present application. DETAILED DESCRIPTION

[0035] The present application will be further described below in conjunction with the drawings.

[0036] A railway line light and hard floating object intelligent identification and risk assessment method, comprising the following steps:

[0037] S1. Acquire image data along the railway line through a multi-source data acquisition system, including aerial images taken by drones and images from fixed monitoring equipment;

[0038] S2. Use the improved MaskDINO instance segmentation model to identify and segment light, rigid floating objects in the image;

[0039] S3. Combine the perspective transformation matrix to map the identified floating objects from the image coordinate system to the physical coordinate system, and calculate their actual physical size and distance from the railway safety clearance.

[0040] S4. Obtain real-time meteorological data, integrate information on floating object type, physical characteristics, and distance, and use a risk assessment model to calculate the probability of floating objects intruding into the railway safety clearance.

[0041] S5. Generate tiered early warning information based on the risk assessment results and push it to relevant management personnel through the early warning release system.

[0042] In this embodiment, the multi-source data acquisition system adopts a heterogeneous image data fusion strategy, which includes fusing and analyzing image data acquired from different angles and at different times in the same area.

[0043] In this embodiment, the improved MaskDINO instance segmentation model incorporates meteorological data as auxiliary input features and uses a multi-head attention mechanism to perform cross-modal fusion of meteorological conditions and visual features.

[0044] In this embodiment, the risk assessment model is a multi-factor weighted model, which considers factors including: floating object type, floating object size, distance between the floating object and the railway safety clearance, ambient wind speed and direction, and historical intrusion records.

[0045] In this embodiment, in step S1, the autonomous inspection path planning of the UAV includes: dynamically generating the optimal inspection path based on the geometric characteristics of the railway line, the frequency of historical floating objects, and real-time weather conditions.

[0046] A smart identification and risk assessment system for lightweight, rigid floating objects along railway lines includes:

[0047] Multi-source data acquisition module, including drone shooting unit and fixed monitoring equipment;

[0048] The data preprocessing module performs noise reduction, enhancement, and standardization on the acquired image data;

[0049] The floating object recognition module uses an improved MaskDINO instance segmentation model to identify and segment light and rigid floating objects;

[0050] The coordinate transformation module converts image coordinates into physical coordinates through perspective transformation;

[0051] The risk assessment module integrates multi-source information to assess the risk level of floating objects;

[0052] The early warning release module generates early warning information based on the risk level and pushes it to administrators.

[0053] In this embodiment, a mobile terminal application is also included; the mobile terminal application is used to receive early warning information, view details of floating objects, and submit processing feedback.

[0054] In this embodiment, the system adopts a collaborative architecture of edge computing and cloud computing. The identification algorithm is deployed on edge devices to achieve real-time processing, while model training and risk assessment are deployed on cloud servers.

[0055] Example 1

[0056] like Figure 1 As shown, the system of the present invention includes a multi-source data acquisition module, a data preprocessing module, a floating object identification module, a coordinate transformation module, a risk assessment module, and an early warning release module.

[0057] The multi-source data acquisition module consists of fixed monitoring equipment deployed along the railway line and inspection drones. The fixed monitoring equipment is responsible for continuous monitoring of key sections, while the drones collect data from multiple angles in key areas according to preset paths. The drone inspection paths are dynamically optimized based on the geometric characteristics of the railway line, the frequency of historical floating objects, and real-time weather conditions.

[0058] The data preprocessing module performs noise reduction, enhancement, and standardization on the acquired image data, including illumination normalization and contrast enhancement, to improve the accuracy of subsequent recognition algorithms.

[0059] The floating object identification module employs an improved MaskDINO instance segmentation model. For example... Figure 3 As shown, the model has been improved based on the original MaskDINO as follows:

[0060] (1) Using the Swin Transformer as the backbone network, its hierarchical attention and shift window mechanism can better capture the local details and global context of floating objects;

[0061] (2) Introduce Feature Pyramid Network (FPN) to enhance the detection capability of small floating objects;

[0062] (3) Real-time meteorological data is used as an auxiliary input and fused with visual features through a cross-modal attention mechanism;

[0063] (4) Adopt data augmentation strategies tailored to the characteristics of the railway environment, including simulating rain and snow weather, different lighting conditions, etc.

[0064] The coordinate transformation module maps the identified floating objects from the image coordinate system to the physical coordinate system through perspective transformation. Specifically, this includes:

[0065] (1) Camera calibration: Place a calibration board in the monitoring area and calculate the camera's intrinsic and extrinsic parameters;

[0066] (2) Calculation of perspective transformation matrix: Calculate the perspective transformation matrix under the current viewpoint based on the calibration results;

[0067] (3) Coordinate transformation: Convert the pixel coordinates of the floating object into physical coordinates;

[0068] (4) Calculation of physical parameters: Based on the transformed coordinates, calculate the actual size of the floating object and the distance from the railway safety clearance.

[0069] The risk assessment module uses a multi-factor weighted model, such as Figure 4 As shown. This model comprehensively considers the following factors:

[0070] (1) Risk coefficient of floating object type: determined based on the characteristics of floating object such as material and shape;

[0071] (2) Risk factor of floating object size: the larger the size, the higher the risk;

[0072] (3) Distance risk coefficient: The closer the distance to the railway safety clearance, the higher the risk;

[0073] (4) Meteorological risk coefficient: mainly considering wind speed, the higher the wind speed, the higher the risk;

[0074] (5) Historical data risk coefficient: Consider the frequency of past accidents at this location.

[0075] The formula for calculating the overall risk value is:

[0076] ;

[0077] Where T, S, D, W, and H represent the risk coefficients of each factor, and α, β, γ, δ, and ε are the weights of each factor, obtained through machine learning training on historical data.

[0078] Based on the comprehensive risk value R, the risk is divided into four levels:

[0079] Low risk (R<0.3): Continue to monitor;

[0080] Medium risk (0.3≤R<0.6): Notify inspection personnel to pay attention;

[0081] High risk (0.6≤R<0.8): Send an early warning message;

[0082] Extremely high risk (R≥0.8): Take immediate action.

[0083] The early warning module generates tiered early warning information based on risk assessment results and pushes it to relevant management personnel via SMS, mobile applications, and other means. Management personnel can view details of floating objects and submit handling feedback through mobile devices, forming a closed loop for safety management.

[0084] Example 2

[0085] The method of this invention is applied to a section of a high-speed railway, and the specific implementation process is as follows:

[0086] (1) The system discovered a plastic film accumulation 100 meters along the railway line through fixed monitoring equipment, and identified it as a lightweight floating object through the improved MaskDINO instance segmentation model;

[0087] (2) The actual size is calculated to be approximately 2m × 3m through perspective transformation, and it is 80 meters away from the railway safety clearance.

[0088] (3) Obtain real-time meteorological data; the wind speed is level 6 (a strong wind).

[0089] (4) Calculate based on the risk assessment model:

[0090] Risk factor T for floating debris type: 0.8;

[0091] Size risk factor S: 0.6;

[0092] Distance risk coefficient D: 0.7;

[0093] Meteorological condition risk coefficient W: 0.9;

[0094] Historical data risk coefficient H: 0.8;

[0095] The overall risk value R = 0.755;

[0096] (5) If the system determines that the risk level is high, it will automatically generate an early warning message and send it to the maintenance section management personnel;

[0097] (6) The management personnel dispatched personnel to clean up the plastic film and reported the results of the handling through a mobile terminal;

[0098] (7) The system updates the database and records the entire process information of this event.

[0099] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features, and such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent identification and risk assessment of lightweight, rigid floating objects along railway lines, characterized in that, Includes the following steps: S1. Acquire image data along the railway line through a multi-source data acquisition system, including aerial images taken by drones and images from fixed monitoring equipment; S2. Use the improved MaskDINO instance segmentation model to identify and segment light, rigid floating objects in the image; S3. Combine the perspective transformation matrix to map the identified floating objects from the image coordinate system to the physical coordinate system, and calculate their actual physical size and distance from the railway safety clearance. S4. Obtain real-time meteorological data, integrate information on floating object type, physical characteristics, and distance, and use a risk assessment model to calculate the probability of floating objects intruding into the railway safety clearance. S5. Generate tiered early warning information based on the risk assessment results and push it to relevant management personnel through the early warning release system.

2. The intelligent identification and risk assessment method for lightweight and rigid floating objects along railway lines as described in claim 1, characterized in that, The multi-source data acquisition system adopts a heterogeneous image data fusion strategy, which includes fusing and analyzing image data acquired from different angles and at different times in the same area.

3. The intelligent identification and risk assessment method for lightweight and rigid floating objects along railway lines as described in claim 1, characterized in that, The improved MaskDINO instance segmentation model incorporates meteorological data as auxiliary input features and uses a multi-head attention mechanism to fuse meteorological conditions and visual features across modalities.

4. The intelligent identification and risk assessment method for lightweight and rigid floating objects along railway lines as described in claim 1, characterized in that, The risk assessment model is a multi-factor weighted model, which considers factors including: floating object type, floating object size, distance between the floating object and the railway safety clearance, ambient wind speed and direction, and historical intrusion records.

5. The intelligent identification and risk assessment method for lightweight and rigid floating objects along railway lines as described in claim 1, characterized in that, In step S1, the UAV autonomous inspection path planning includes: dynamically generating the optimal inspection path based on the geometric characteristics of the railway line, the frequency of historical floating objects, and real-time weather conditions.

6. A smart identification and risk assessment system for lightweight, rigid floating objects along railway lines, used to implement the method described in any one of claims 1 to 5, characterized in that, include: Multi-source data acquisition module, including drone shooting unit and fixed monitoring equipment; The data preprocessing module performs noise reduction, enhancement, and standardization on the acquired image data; The floating object recognition module uses an improved MaskDINO instance segmentation model to identify and segment light and rigid floating objects; The coordinate transformation module converts image coordinates into physical coordinates through perspective transformation; The risk assessment module integrates multi-source information to assess the risk level of floating objects; The early warning release module generates early warning information based on the risk level and pushes it to administrators.

7. The intelligent identification and risk assessment system for lightweight, rigid floating objects along railway lines as described in claim 6, characterized in that, It also includes a mobile terminal application; the mobile terminal application is used to receive early warning information, view details of floating objects, and submit processing feedback.

8. The intelligent identification and risk assessment system for lightweight, rigid floating objects along railway lines as described in claim 6, characterized in that, The system adopts a collaborative architecture of edge computing and cloud computing. The identification algorithm is deployed on edge devices for real-time processing, while model training and risk assessment are deployed on cloud servers.

Images