Vision Based Navigation for Autonomous Vehicles

 

Objectives:

 

The work starts by estimating position, orientation, and intrinsic parameters (calibration) of each camera mounted on a car. The cameras may have non-overlapping fields of view (FOVs). Calibration comprises both the intrinsic and extrinsic parameters. The non-overlapping FOVs are overcome by placing calibration patterns at known poses or by the motion of the car.

 

Given the calibration of the cameras, objects imaged by multiple cameras can be reconstructed in 3D. The 3D poses of objects (obstacles) are useful to estimate obstacle avoidance trajectories. Corresponded points of moving objects allow also to estimate time-to-contact information. An informal discussion on the use of sensors for autonomous vehicles navigation can be found in:

https://www.quora.com/Why-do-self-driving-cars-use-LIDAR-instead-of-depth-cameras-like-Kinect

 

Steering and speed control need to be adapted to the perceived environment. Trajectories planning and following are an essential component to develop upon the sensing devices.

 

Work-proposal detailed steps:

- Review the state-of-art on the calibration of multiple cameras on a car

- Setup of a simulation scenario based on applied games

- Calibration of the cameras

- Steering for obstacle avoidance

 

This work is conducted within the framework of the project VIENA, driving a FIAT electric car, representing the IST. Some information regarding the VIENA project can be found in

https://istpowerlab.wixsite.com/e2tes/viena

 

 

References:

 

[Silva12] M. Silva, R. Ferreira and J. Gaspar. Camera Calibration using a Color-Depth Camera: Points and Lines Based DLT including Radial Distortion. In WS in Color-Depth Camera Fusion in Robotics, held with IROS 2012.

 

[Hartley00] R. I. Hartley and I. Zisserman. Multiple view geometry in computer vision. pages 150–152, 2000.

 

 

Localization:

ISR/IST

 

 

-- More information in:

 

http://www.isr.tecnico.ulisboa.pt/~jag