Please find the full documentation in the documentation folder. This was made for the course HW/SW co-design with a LEGO car (IN2106, IN0012, IN4154) at TUM, in the Winter semester 2016.
This program calibrate a stereocamera (intrinsic and extrinsic), calculate a depth map and provides a top view of the depth map so the automated car can decide where to go. Can only be used with the plugged stereo camera. You can see the result video here.
Install following libraries:
- Boost C++
- OpenCV 2.7
Boost is only for the calibration process. Once the calibration is done, the intrinsics and extrinsics files can be used without using this library.
- Go into at the root of the folder in a terminal, and make sure the file build.sh as the proper access with
sudo chmod 777 build.sh - Then send the command
./build.sh - Go into the folder build with
cd build - If you wanna do the calibration process send
./sv 1but make sure you have the proper pictures taken with the stereocamera in the calibration folder. A set of at least 7 pairs is necessary. - If you wanna just use the program, send
./sv - Enjoy the depth map and the top view!
- In the main.cpp verify the two camera index.
- If you use the calibration, verify the chessboard size in the cameracalibrator.hpp file (default is 9.6) and the square size of the chessboard in the cameracalibrator.cpp (default is 2.5 cm)
- To change the default matching method, go in depthsubstraction.cpp, by default it's the BM method.
- If you wanna use the SGBM algorithm, you have to change de disparity parameters, because the default parameters are changed to work with the BM algorithm only.
This code was not only developped for a university course but also for a master student thesis, Roger Rösch. You can find part of my code he refactored for OpenCV 3.1.0 here. You can also find his full master thesis about the Development and Construction of an Autonomous Car with Low Cost Components here.