LiDAR-based Train Detection System

Neuvition LiDAR-based Train Detection System is a solution for train compartment number recognition and unbalanced loading detection.

1. Work Process of Neuvition LiDAR-based Train Detection System

Based on highly reliable and safe software and hardware platform, Neuvition LiDAR-based Train Detection System makes an innovative combination of LiDAR + high-speed camera video sensor and achieves effective coverage of point cloud collection on a large scale.

Using high-definition camera image data, through heterogeneous data fusion, relying on the powerful artificial intelligence deep learning algorithm of edge calculator, the whole detection system can realize train compartment number recognition, train unbalanced loading detection, and compartment volume ratio calculation, etc.

2. Main Workflow of Neuvition LiDAR-based Train Detection System 

Neuvition LiDAR-based Train Detection System monitors the real-time status of the railway through the camera and Neuvition LiDAR. When the camera detects that the train enters the system detection area, the system automatically activates the LiDAR measurement system and the high-speed camera enters the working state. Then LiDAR is responsible for collecting train contour information in real time. The high-speed camera system uses intelligent learning algorithms to automatically record the image data of the train compartment number and train number. With the collected data, the system can acquire the structured data through a complete set of mature algorithms independently developed by Neuvition, and then is uploaded to the software control center for data comparison, recording, and archiving.

1. Detect train speed: collect high-definition point cloud data and video information through two 3D LiDARs and built-in cameras, to obtain real-time train speed;

2. Point cloud-video fusion: The edge calculator receives the point cloud data + video information of the LiDAR, and synthesizes the color point cloud data through the point cloud video fusion algorithm;

3. Point cloud registration: 3D reconstruction of the 3D model of each train through the point cloud registration algorithm independently developed by Neuvition, based on the vehicle speed information and test point cloud data;

4. Compartment number recognition: Use LiDAR to accurately locate the train, high-speed camera system collects the compartment number image data, and then train compartment number can be accurately identified according to the neural network recognition algorithm;

5. Unbalanced loading detection: By analyzing the 3D model data of the train compartment, the length, width, and height data, as well as the distribution of cargo loading of the compartment, can be detected in real time, also, the deflection angle of the compartment can be obtained in real time;

6. Intelligent analysis: the loading quality of the train(whether there is foreign matter, whether it exceeds the limit), the closed state of the train compartment door, etc.;

7. Compartment volume ratio calculation (optional): For open-top cars, LiDAR can obtain the volume of cargo inside the vehicle in real time;

The Advantages of Neuvition LiDAR-based Train Detection System

1. High reliability: LiDAR works independently of ambient light, and can work normally at night and under low-light conditions. With Neuvition’s intelligent and advanced algorithms, it guarantees the performance of LiDAR in rain, snow, and fog;

2. High precision: at a detection distance of 60m, it can achieve a ranging accuracy of ≤1cm, and obtain real-time vehicle posture, tilt angle, and 3D information of the cargo inside the vehicle;

3. High resolution: based on MEMS technology, Titan M1 solid-state LiDAR has a maximum resolution of 0.03°, ensuring effective coverage of point cloud information in the compartment;

Technologies Used in Neuvition LiDAR-based Train Detection System:

1. LiDAR Technology

Neuvition Titan M1-A has a vertical 700 line, can collect track environment point cloud data based on 1550nm laser echo signal measurement technology. Combined with the point cloud registration algorithm, a high-precision train 3D model can be generated, and then the compartment posture and volume ratio can be calculated through the algorithm. Neuvition LiDAR-based Train Detection System sends the detected train data to the system host, and the system can perform calculation, analysis, and storage of the data remotely. It is mainly used for unbalanced loading detection, compartment number identification, and volume ratio detection of freight train systems.

 Neuvition LiDAR has a high resolution, high detection accuracy, and accurate echo intensity. At the same time, it takes into account the angle coverage and angular resolution in the pitch direction, and achieves the following effects:

–Effectively resist the interference of ambient light intensity on detection;

–The vertical field of view takes into account the coverage and grid resolution, and the horizontal angular resolution is up to 0.03°;

–Industrialized vehicle design, has obtained third-party certification test reports that comply with the motion, vibration, electromagnetic interference, temperature, and humidity environment of the train platform;

The effect of the detection scene is as follows:

2. Data Fusion Processing Technology

The data fusion processing unit of Neuvition LiDAR-based Train Detection System receives the point cloud and video stream data of the sensor in real-time, and performs temporal and spatial registration and correlation of the data information detected by the sensor, and then accurately recognizes the train body and divides the target detection area. The data fusion processing module can accurately obtain the contour of the train compartment and the train speed to reconstruct the train 3D model. The generated train 3D model is used to calculate the posture, volume ratio, and other information of the train compartment. Once the compared data exceeds the threshold set by the system, the system will issue an alarm and provide cargo compartment number information, which is convenient for railway inspectors to quickly find and deal with the fault location in time. The system can provide storage space of different capacities according to the needs of customers to save the detection information, which is convenient for inspection and analysis by maintenance personnel.

i. Target detection module: Using deep neural network algorithm to extract train contours in real time, the process includes feature extraction, border regression, and classifier classification;

ii. Rail segmentation module: Using deep neural network algorithm, process intelligent real-time identification of rails;

iii. Speculation and compensation module: Using Kalman filter algorithm to track and compensate the real-time position of the target object;

iv. Data fusion processing module: target recognition results will be transmitted to the data fusion processing module;