Train surface foreign matter detection system based on image processing

Abstract

The invention discloses a train surface foreign matter detection system based on image processing. In the railway transportation industry, a train inspector inspects the appearance of a train entering and exiting a station in daily manners of visual inspection, manual photographing and the like, checks possible surface damage of the train in operation, notifies a maintainer to maintain the damage, and further checks whether foreign matters are abnormally suspended on the surface of the train or not. Foreign matters are prevented from flying up and being hung on the contact network in operation, and operation safety is avoided. The workload of manual inspection is huge, and the risk of missing inspection exists. According to the system, the image recognition technology is utilized, monitoring images installed on the rail side are intelligently analyzed, train surface losses or foreign matter possibly existing are found in time, an alarm is given to railway operation personnel, and the operation personnel make a processing decision. According to the system, the train appearance inspection workload is greatly reduced, the detection time is shortened, the manual labor intensity is reduced, and the railway transportation safety and the operation efficiency are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of train transportation safety, and particularly relates to a train surface foreign object detection system based on image processing. Background Art

[0002] Railway transportation production is the main artery of the national economy, and is an important means of transportation for improving the people's material and cultural living standards, meeting the people's travel needs, and strengthening national defense construction. Safety is the most basic requirement for railway transportation production. At present, with the rapid development of China's railways, especially high-speed railways, the safe operation of railways has become the focus and hot topic of social concern. Among the many factors affecting rail transit, the loosening of rail fasteners is a major safety hazard.

[0003] After a railway freight train is loaded with goods at a freight yard, it is towed by a shunting locomotive to a marshalling station for marshalling and departure. Before loading the goods, it is necessary to conduct a routine inspection of the vehicle appearance to check whether the train body has been damaged abnormally due to acceleration, deceleration, uphill and downhill, turning, vibration, or being hit by an external object during the previous operation, so as to repair the damage and eliminate potential safety hazards in railway production. After loading and before departure, it is necessary to conduct a routine inspection of the vehicle body appearance again to check whether there are any damages to the vehicle body appearance, such as doors and windows, due to collisions during the loading process, or whether the doors and windows are not closed, or whether there are abnormal conditions such as plastic bags and packaging boxes on the roof and side of the vehicle, which pose potential safety hazards to train operation. At present, the railway transportation department mainly uses manual inspection or the method of taking pictures with a hand-held photographing device and then manually checking. Under normal circumstances, the train is hundreds of meters to thousands of meters long, and the workload of routine inspection is huge. The train roof is much higher than the height of manual visual inspection. The inspection personnel use a hand-held photographing rod that is 5 to 7 meters long for shooting, which is difficult to operate, and there is a risk of the photographing rod hitting the vehicle window due to external force and causing damage to the vehicle, or there is a risk of personal safety due to the photographing rod touching the overhead contact line or other electrical equipment on the roof. Summary of the Invention

[0004] The problem to be solved by the present invention is to provide a train surface foreign object detection system based on image processing by applying video monitoring equipment combined with artificial intelligence image recognition technology in view of the above-mentioned defects of the existing train appearance inspection.

[0005] The technical solution adopted by the present invention to solve this technical problem is: Construct a train surface foreign object detection method and system based on image processing. The system is composed of a side image acquisition unit, a roof image acquisition unit, a data storage unit, an image recognition AI unit, an AI training unit, and a data display unit.

[0006] The AI training unit provides a one-stop model training service, including functional modules for dataset management, model training, model management, model verification, and model release. It provides a human-computer interaction interface for users. Users upload training samples (as shown in the figure) through the model training unit, annotate the samples, issue training instructions, start the system for model training, and publish the training results to the system.

[0007] The data storage unit mainly receives data from the AI training unit, the image acquisition unit, and the image recognition AI unit, stores various types of data in the system, and provides data support for training and computing.

[0008] The image acquisition unit consists of a side-of-car image acquisition unit and a roof-of-car image acquisition unit, which are installed on a gantry spanning the railway tracks and respectively acquire real-time video images of the side and roof of the train passing under the gantry.

[0009] The data display unit mainly displays the real-time images acquired by the video image acquisition unit and the detection results of the image AI unit, and issues an alarm for detected abnormalities. Detailed implementation manners

[0010] To make the purpose, technical solutions, and advantages of the present invention clearer, the following describes the technical solutions in the embodiments of the present invention clearly and completely in conjunction with the embodiments and the accompanying drawings. It should be noted that the described embodiments are some, but not all, embodiments of the present invention.

[0011] A train surface foreign object detection system based on image processing, as Figure 1 shown, includes the following units: The image acquisition unit mainly consists of a roof-of-car image acquisition unit and a side-of-car image acquisition unit. The side-of-car image acquisition unit real-time acquires the video image of the side of the train passing through the monitoring area, and the roof-of-car image acquisition unit acquires the video image of the top of the train passing through.

[0012] The data storage unit receives the data transmitted from other units and stores and manages it.

[0013] The model training unit provides a human-computer interaction interface for users. Users upload training samples through the model training unit, annotate the samples, issue training instructions, start the system for model training, and publish the training results to the system.

[0014] The image recognition AI unit receives the video stream acquired by the image acquisition unit, calls the knowledge base trained by the training unit, performs image recognition algorithm operations on the images in the video stream, feeds the calculation results back to the storage unit for storage, and at the same time publishes them to the data display unit to display the calculation results to users in real time.

[0015] The data display unit receives the calculation results of the video images collected by the image acquisition unit calculated by the image recognition AI unit, and displays them to the user in real time. It alarms for the abnormal information detected on the train surface during the calculation to prompt the operation and maintenance personnel to make handling. Description of the Drawings

[0016] Figure 1 Schematic diagram of the system architecture.

[0017] Figure 2 System flow chart.

[0018] Figure 3 Schematic diagram of the roof image sample.

[0019] Figure 4 Schematic diagram of the side image sample of the carriage.

[0020] Figure 5 Schematic diagram of the foreign object sample.

[0021] Figure 6 Schematic diagram of the foreign object sample.

Claims

1. A train surface foreign body detection system based on image processing, characterized in that: The image acquisition unit is mainly composed of a roof image acquisition unit and a side image acquisition unit. The side image acquisition unit acquires the side video images of the train passing through the monitoring area in real time, and the roof image acquisition unit acquires the top video images of the train passing through.

2. Data storage unit, which receives data transmitted from other units and stores and manages them.

3. The model training unit provides users with a human-computer interaction interface. Users upload training samples through the model training unit, label the samples, issue training instructions, start the system for model training, and publish the training results to the system.

4. The image recognition AI unit receives the video stream collected by the image acquisition unit, calls the knowledge base trained by the training unit, performs image recognition algorithm operations on the images in the video stream, and feeds the calculation results back to the storage unit for storage, and publishes them to the data display unit to display the calculation results to the user in real time.

5. The data display unit receives the calculation results of the video images collected by the image acquisition unit calculated by the image recognition AI unit, displays them to the user in real time, and issues an alarm for abnormal train surface information detected during the calculation, prompting the operation and maintenance personnel to take action.

6. The train surface foreign body detection system based on image processing according to claim 1 is characterized in that: The image captured by the vehicle-side image acquisition unit contains the car coding information. The image recognition AI unit identifies and marks the car number through the image captured by the vehicle-side image acquisition unit, and structures the unstructured video data collected by the image acquisition unit in combination with the timestamp, so as to facilitate system users to review historical video image data and identified car abnormality records.

7. The train surface foreign body detection system based on image processing according to claim 1 is characterized in that: The data display module provides a real-time display system video image window for the user, and alarms for abnormalities in the detected carriage appearance, and provides a review interface for the user.

Images