Dynamic Vehicular Pose Estimation with Heuristic L-Shape Fitting and Grid-based Particle Filter
The example vedio for the vehicle pose estimation can be found on the Youtube.
This reposetary contains the implement detais abount the paper" Dynamic Vehicular Pose Estimation with Heuristic L-Shape Fitting and Grid-based Particle Filter".
The following image denotes the illustration of the process for the geometric shape classifier. (a) shows the processing of endpoints extraction and width calculation, (b) shows the step of cluster classifying according to extracted contours, the solid blue points in each segment represent contours. (c) shows a real-world traffic scene on the road, there are various clusters on it, the cluster in black rectangle represents a symmetrical-like cluster.
platform: ubuntu14.04 or higher ubuntu version., CMakeList 2.8.3 or higher.
third party:
Qt5.6: you should change CMakeList.txt content Qt5_DIR to Qt install dir.
OPENCV3： you should change CMakeList.txt content OpenCV_DIR to Opencv install dir.
QGLVIEWER is a powerful and high-performance point cloud image visualization.
sudo apt-get install libqglviewer-dev libboost-all-dev libeigen3-dev freeglut3-dev libgl1-mesa-dev libgtest-dev2
when all dependencies are met, type this in root code dir， you will get a bin exec file to run.
mkdir build && cd build && cmake ..
we used kitti dataset, example data can be done here. link， Register and download the data for this section 2011_10_03_drive_0047 (3.3 GB).
change data to this format
dataset/velodyne
dataset/image_2
dataset should change param.h kitti_velo_dir, this var will be use in mainwindow.cpp => onOpenFolderToRead