Neousys Nuvis-5306RT GPU Machine Vision Inspection Controller Application on Railway Vehicle Inspection System

Railway and railway vehicle inspection have always relied on human onsite visual inspection where the environment may be demanding and rigorous which is dangerous and inefficient.
Neousys' advanced GPU-aided machine vision controller is designed specifically for visual inspection tasks. Neousys Nuvis-5306RT is an ultra-compact embedded GPU-aided machine vision computer that requires minimum maintenance. The system was incorporated into the intelligent rail vehicle maintenance system created by renowned China Railway Technology Research Institute. The intelligent rail vehicle maintenance system can shuttle freely between tracks and utilizes cameras to scan/ inspect the underside of the carriage and bogie which expedites the inspection/ maintenance work, improves efficiency and offer work safety via remote visual inspection capability.
The renowned China Railway Technology Research Institute developed an intelligent rail vehicle inspection system with integrated vision control console and an image acquisition robot that uses Neousys' Nuvis-5306RT GPU-aided machine vision controller. The image acquisition robot components include an autonomous driving platform, industrial digital line-scan cameras, area array camera, position sensor, obstacle sensor, monitor lens, lift level, multi-angle robotic arm, image acquisition computer and travel device.
Nuvis-5306RT GPU-aided machine vision platform and intelligent rail vehicle inspection system:


Imagery Acquisition Robot for Visual Inspection
The imagery acquisition robot moves into position at the front end of the track and the digital line-scan camera begins to capture the underside imagery of the carriage, moving from front to rear. Once the image acquisition operation is at the rear of the carriage, the image acquisition robot will decelerate and stop at the fixed point, extend the arm back to the standby position.
The image capturing robot starts to move using the laser to accurately guide itself to the predetermined position while extending the image capturing device mounted on the robot arm to between the bottom layer of the carriage and the bogie for scanning.
The area array camera captures component imageries above the bogie; when completed, the robot moves to the other end of the bogie, captures underside images of the bogie. All scans and images captured are fed live onto a computer monitor. When all cars have been inspected, the robot will automatically return to the initial position and shut down.
Under normal conditions, the robot arm will move according to the predetermined scan path. If an obstacle is detected in the scan path, the robot arm will automatically register and return the obstacle information and re-plan the path (avoid and go-around the obstacle). The obstacle sensor is also installed on the front and back of the driving platform of the image capturing robot to avoid a collision.


High-Resolution Image Acquisition and Fault Identification
The robotic arm of the image acquisition robot can rotate at multiple angles to capture high-resolution images of key components such as the top, sides and bottom of the bogie. Its intelligent platform is equipped with line-scan cameras to scan the underside of the carriage and area array cameras to scan the underside of the bogie.
The track intelligent maintenance system has advanced image recognition and machine deep learning technology. It can automatically identify damaged components/ equipments from captured images and automatically generate maintenance work orders.
Rail Compatibility
The image acquisition robot can be driven by two types of rails, vertical lifting or plane laying rails. It is compatible with a variety of grooves and can be customized according to onsite conditions.

Neousys Technology Nuvis-5306RT
The Neousys Technology Nuvis-5306RT is the perfect intelligent visual inspection solution. It features a 75W NIVIDIA® GPU with Intel® 6th generation Skylake processor with dedicated machine vision I/Os for real-time control and GPU-aided computing tasks.