Object-detection-in-Point-Cloud-road-boundary

Object detection in Point Cloud is popular in HD Map and sensor-based autonomous driving. There basically four types of object you can obtain in daily scenario: road surface - contains painted lane marking and pavement area, support facility - contains road boundary (guardrail and curb), road sign, light pole, etc., uncorrelated object - for example, sidewalk, building, etc., and moving object - such like pedestrian, vehicle, bicycle, etc.

In this project, please search references, design and prototype your road boundary (guardrail) detection algorithm.

Run Instructions

1. Open a Terminal at the project directory, run:

$ python3 CoordTransform.py

This python script will read original point cloud data, do coordinate transformation (LLA -> ECEF -> ENU -> Camera), and generate a point_cloud_camera_coord.csv file, which has 4 columns (X-coord, Y-coord, Z-coord, Intensity) for each data item.

2. Build the project "cloud_viewer" with the CMakeLists.txt

3. Run the project solution

Notes:

1. Make sure the file final_project_point_cloud.fuse is in the directory ./final_project_data/

2. Make sure the point_cloud_camera_coord.csv is under the correct directory

3. If you already have point_cloud_camera_coord.pcd file, you can commemt the line 335 to line 339

Required Environment

Python 3.6
C++
PCL 1.8.1 for C++
opencv for C++

Notes

1. It is recommended that you use cmake to build the project

2. We only test the project under windows system

Project Files

  CoordTransform.py
  point_cloud_camera_coord.csv
  cloud_viewer.cpp
  CMakeLists.txt

Reference

1. http://www.jeffdelmerico.com/wp-content/uploads/2014/03/pcl_tutorial.pdf
2. http://pointclouds.org/