Blog

Multi-sensor vision system for corridor mapping

Dutch 3D-ONE developed a multi-sensor embedded vision system for the acquisition of all assets alongside the rail-track from a driving train (corridor mapping). An oblique video is generated from 3 aligned image sensors and tagged with a GPS time stamp. Video and GPS data are stored in real time to an onboard mass storage device. Different compression and encryption techniques are available and can be applied on request. The development has been conducted in close cooperation with Fugro Raildata who are deploying the system world-wide.

Positioning accuracy
A stringent requirement for this development was to allow operation of the system from any passenger train driving its regular schedule. This meant the system is specified to acquire video with high positioning accuracy while driving up to 160km/h. To produce valuable video data at this speed all 3 cameras are synchronized to capture their images with a camera-to-camera delay of <10µS while the GPS time stamp is synchronized to the cameras within 50µS. This allows for a very accurate stitching and reproduction of the exact location of each image during post-processing. To prevent for any interruption (data-loss) the image acquisition and GPS synchronization is implemented as a core process at FPGA level.

Changing light conditions
Driving the system for hours means continuous changing light conditions. These can be caused by passing a tunnel, a changing position of the sun or change due to clouds or rain. To cope with these changing conditions 3D-ONE implemented a range of exposure schemes. These schemes are user selectable to provide maximum flexibility during stitching in post processing.

Post processing
After completing the ride all data is downloaded from the embedded vision system for post processing together with the laser point cloud generated by the LiDAR system. The information will be used to create a digital terrain model. The combination of the point cloud and imagery data allows to extract features with exact XYZ location and to measure distances in the image in real time.

Video

About 3D-ONE
3D-ONE develops embedded systems for visual intelligence from multi-sensor imaging systems. Our solutions include data acquisition, signal processing and data analysis where we strive to reach the optimal ratio between size, weight and performance.
3D-ONE applies vision technologies in the visible, hyperspectral and IR regime. We are a strategic partner to technology leading OEM customers who are developing vision enabled products.

About Fugro Raildata
Fugro Raildata provides a safer, faster and more affordable way of delivering accurate up-to-date rail infrastructure information for asset management and engineering needs.  Our 3D virtual model of the railway corridor helps clients worldwide to design, build, upgrade and maintain rail infrastructure.