VisLab Embedded Lane Detector

The aim of the VELD system is to detect lane markings for automotive applications by means of one single camera and a DSP board. It is a system derived from VIDA (VisLab Intelligent Driver Assistant). The main difference is that the standard version does not use CAN data for ego-motion compensation and the output is composed by raw data (i.e. position of lane markings) on ethernet to be used in further processings.

Add-ons

• Stop line detection
• CAN data input for ego-motion compensation
• Output on CAN

Possible applications

• Sensor for a Lane Departure Warning system
• Sensor for an autonomous vehicle
• Line detector for industrial UGV

System overview

The system can detect up to four simultaneous bright lane markings (i.e. it can handle three lanes); the lane markings can be both solid and dashed. Vehicle pitch variations in normal driving conditions are automatically compensated (no extra inertial sensors are needed).

Being a VIDA sub-product, the VELD output is provided on an ethernet port as a stream of packets containing, for, each line, their position with respect to the vehicle at fixed distances, which can be further interpolated to build a geometric model of the road (e.g. a polynomial function). The algorithm output superimposed on the acquired image can be accessed by connecting a NTSC display to the video out plug.

Resources

Datasheets & manuals

• VELD Specifications
• eGOLD Client v20100916
• eGOLD Client v20090720
• eGOLD Client v20090710

Tools

• Online Calibration tool (available form your private section after login). This tool allows customers to upload an image acquired by their VELD camera to the VisLab website, and -thanks to an internal calibration engine- retrieve the correct camera parameters to be included in their VELD configuration.