Visual Perception of the road onboard the AtlasCar

This is my Mechanical Engineering Masters Thesis Repo in the field of Artificial Vision applied to Autonomous Driving.

Where and when was it made?

Department of Mechanical Engineering (DEM), University of Aveiro (UA)

LAR: Laboratory of Automation and Robotics

2019

Advisor

Profesor Vitor Santos

DEM, UA

Aveiro, Portugal

Prerequisites

What things that are needed:

• ROS-melodic - The ROS version used.
• FlyCapture 2.x
• Camera Pointgrey_driver - The driver used for the camera.

Hardware:

• Camera PointGrey FL3-GE-28S4-C - The camera used

Project Guide

After all the prerequisites are installed, the first step is to calibrate the cameras in order to get the intrinsics parameters. This was based in this Tutorial.

To run the camera_calibrator package to stereo calibration:

rosrun camera_calibration cameracalibrator.py --approximate 0.1 --size 8x6 --square 0.105 right:=/top_right_camera/image_raw left:=/top_left_camera/image_raw right_camera:=/top_right_camera left_camera:=/top_left_camera

To get the undistorted image (this step is already included on the my camera_driver launch file):

$ ROS_NAMESPACE=my_camera rosrun image_proc image_proc

In order to get faster results, I rescale the initial image inside of my architecture. At this moment, to change the resolution it is necessary to follow the steps below:

1. Accordingly,comment/uncomment the perspectiveSrc and the pespectiveDst variables in advanced_lane_detection/src/main.cpp because they respresent the perspective transformation (IPM);
2. In advanced_lane_detection/src/laneDetection.cpp (lines 383/384) uncomment/comment the lines to show the right result;
3. In Lane_Detector/params/Detector2.yaml change the value's variable ipmRight to number of columns of the image;
4. In Lane_Detector/cfg/Detector.cfg (lines 47-70) uncomment or comment the lines based on the image resolution;
5. On the launch files: -algorithm1.launch; -data_treatment.launch; -advaced_lane_detection.launch; Change the paramaters related to the image resolution.

Usage

To launch cameras:

roslaunch atlas2_bringup drivers.launch

To launch the architecture:

roslaunch data_treatment road_visual_perception.launch 

Algorithms that were used

This project uses algorithms from different authors:

• Nsteel/LaneDetection
• HsucheChiang/Advanced_Lane_Detection

Author

• Tiago Almeida - My Thesis Blog