TUeLaneTracker

This is a software application that detects and tracks lane boundaries. The underlying algorithm is a probabilistic classifier which was originally develope at Eindhoven University of Technology (TU/e), under the strategic area of Smart Mobility. The algorithm exploits the concept of hierarchical classification from deep learning, however, unlike deep learning, classification at each level is engineered instead of being trained through images. The software application is completely object oriented and follows various software design principles recommended by the safety standard ISO26262.

This application provides a loose coupling between the software control flow and the algorithm, making it possible to generate various target specific implementations for the algorithm. Besides this generic implementation, an accelrated version for the NXP-BlueBox (s32v) is also available. The s32v specific implementation makes use of the APEX accelrators to speed-up the vision processing. The APEX architecture blends scalar and vector processing capabilities within the two fully programable cores, achieving an effective 5-10x speed-up for the algorithm.

Note: A ROS package for this lane tracker is available under tue_lane_tracker repository.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes. See deployment for notes on how to deploy the project on a live system.

Prerequisites

What libraries do you need, in order to build and install the TUeLaneTracker

* OpenCv-3.1.0
* Boost-1.62.0

How to install them:

OpenCV Installation Guide

Boost Library Sources

Build, Install and Run the Application

Follow the following steps to run the software application

clone this repository:

git clone https://github.com/RameezI/TUeLaneTracker.git

Go to the parent directory of the cloned repository:

cd TUeLaneTracker

Create a build subdirectory and invoke cmake:

mkdir build
cd build
cmake ..

Build and Install the executeable:

make install

Run the executeable:

cd ../install/bin
./TUeLaneTracker

Modifying the Frame Source for the Application

The TUeLaneTracker has three modes for recieving the input frames :

imgstore [Images in a directory]
stream [NetworkStream, VideoFile, V4LCapture]
gmsl [High-Speed LVDS Camera Interface] [In-Progress]

For usage instructions use the --help switch

./TUeLanetracker --help

Callibrarting TUeLaneTracker

The algorithm is parameterised with a minimal set of parameters. These parameters includes, among others, intrinsic and extrinsic camera specifications. At the boot-time the algorithm takes into account these parameters, whcih are set via the include/Config.h file.

This provides a way to reconfigure the algorithm but only at the compile time. However, in case of the ROS (Robot Operating System) package for the lane tracker it is possible to dynamically reconfigure and reboot the algorithm, with new configuration, at runtime. The ros package for this lane tracker is available under the tue_lane_tracker repository.

Youtube Videos

These videos showcase the functional performance of the algorithm:

Eindhoven

Built With

OpenCV3.1 - Copmputer vision library
CMake - Dependency management and makefiles generation
Eigen 3.3.3 - Linear Algebra library
Boost 1.62.0 - Provides Program Options

Contributing

Please read [CONTRIBUTING.md](TO BE DEFINED) for details on our code of conduct, and the process for submitting pull requests to us.

Authors

Rameez Ismail See also the list of contributors who participated in this project.

License

This project is licensed under the MIT License - see the LICENSE.md file for details

Acknowledgments

This research and design project was supported by NXP Semiconductors and Eindoven University of Technology.

DodTech / TUeLaneTracker