Skip to content
Quadrotor planning for 3D Reconstruction
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
VirtualScan
cmake
external
include
plane_sweep_stereo
quad_flight
quad_planner
src
stereo
video
viewpoint_planner
.gitignore
CMakeLists.txt
README.md
RealResults.png
Screenshot.png
cmake.sh
cmake_jetson.sh
get_octomap.sh
make_jetson.sh

README.md

Quad3DR

Quad3DR enables planning of Quadrotor trajectories for dense 3D reconstruction of buildings and other structures.

Quad3DR computes a volumetric map of the environment from an initial flight (using a modified version of Octomap). The volumetric map is used to ensure that the planned trajectories are only in free space and not in occupied or unknown space. Next a viewpoint graph is built by performing a raycast on the GPU to measure visible voxels and a viewpoint score is computed. The viewpoint score is computed based on distance and incidence angle of viewpoint and voxel (the angle is computed by using a rendered Poisson-reconstructed mesh of the initial flight). Connections in the viewpoint graph are found using RRT*. The resulting viewpoint score is submodular and we compute a viewpoint path by using an adaptation of the recursive method in (Chekuri, Chandra, and Martin Pal. "A recursive greedy algorithm for walks in directed graphs." Foundations of Computer Science, 2005. FOCS 2005. 46th Annual IEEE Symposium on. IEEE, 2005.).

The planned viewpoint paths were evaluated on synthetic scenes using Unreal Engine and on real scenes using a DJI Matrice 100.

Screenshot

Screenshot

Some results from real a scene

Result

You can’t perform that action at this time.