plane_localization

This is a ROS package for indoor global localization (relocation) based on plane octree and plane features. The test results show that the global localization (relocation) algorithm can achieve about 3.5 seconds of octree construction and global plane feature extraction (only need to run once) in a map composed of 372 keyframes, complete the reading of plane features and octree in about 0.02 seconds, and complete global localization (relocation) in about 1.0 seconds. It is a robust and fast global localization (relocation) algorithm for indoor or other structured scenes with more plane features.

Usage

1. Requirement

1. ROS
2. PCL

2. Build

cd ${YOUR_PROJECT_DIR}
catkin_make

3. Run

### 1.Prepare keyframe point clouds and keyframe poses and place them in the data directory
### 2.Modify parameters in the config directory
### 3.run the program
source ./devel/setup.bash
roslaunch plane_localization start.launch
### 4.Obtain the results
---------------------------
transformation_matrix_vector size: 49892
cluster_transformation_matrix_vector size: 19614
cluster_transformation_time: 0.034689
---------------------------
fine_verify_octree_time: 0.084577
best_transformation_score: 1865 best_transformation_matrix: 
   -0.417447     0.908311    0.0266064   -0.0180196
   -0.908634    -0.417593 -9.00216e-05     0.061221
   0.0110289   -0.0242131     0.999646    -0.130588
           0            0            0            1
localization_time: 0.772197

4. Example

The above figure achieves correct global localization (relocation) in an indoor scene of approximately 2500 square meters