tangshp / Lidar-Obstacle-Detection

Lidar-Obstacle-Detection

This project processes a stream of pointclouds. The following steps are done to track the objects moving in the pointcloud:

1. Load current PointCloud from the stream
2. Segment into ground plane and residual data (ransac)
3. Cluster Objects in the residual data (euclidean clustering + kdtree)
4. Define bounding boxes around each cluster

Dependencies

A Dockerfile is provided under .devcontainer/Dockerfile. In Visual Studio Code this directory can be used to automatically setup a developement container. Dependecies:

• CMake 3.4
• PCL 1.8
• OpenMP

Profiling

This project implements profiling capabilities. For that it defines the two macros PROFILE_FUNCTION() and PROFILE_SCOPE("name").

• PROFILE_FUNTION(): Records the time required by a function and saved the data to ´profile/profile.json´ under the function name.
• PROFILE_SCOPE("name"): This macro allows to record the time taken by random peaces of code. This can be done by encapsulating that code into its own scope. The data is saved to ´profile/profile.json´ under the passed name.

The output data is parsed in such a way, that chrome tracing tool can interpret it. To visualize the resulting profile, open the chrome tracing tool in the chrome browser under the following link. Then load the profile.json file into the tool.