Occupancy Set Key(OSK)

Occupancy set key is a LiDAR based place recognition algorithm had been accepted by IEEE TIM.

demo video: https://www.youtube.com/watch?v=4yEeCB2dhWg

Feature

1. Introducing set based location descriptor, and using locality sensitive hashing trick, we convert the 3D environment around the landmark into several unique hash values.
2. A simple but effective landmark selection strategy and by using RANSAC, algorithm can provide approximate 3DoF pose

How to use

Run on KITTI

We use point cloud download from Semantic KITTI

Download Data: LiDAR, calib&time, pose

1. In script/gt_kitti.py, set four path:lidar_folder_path(lidar .bin file's folder), timestamp_path, calib_cam_to_velo_path, cam0_to_world_path according to the dataset's path on your computer. Then create a folder and assign its path to output_folder_path.

   After these change run this python script, you will get some output file under output_folder_path including a xxx/out.txt, please copy the complete path of xxx/out.txt .

2. In launch/run_osk.launch, set the lidar_info_path using xxx/out.txt generated before, and set the result_save_path using an arbritrary xxx.txt file path:

   <param name="lidar_info_path" type="string" value="xxx/out.txt" />
   <param name="result_save_path" type="string" value="xxx.txt" />
   

3. in shell/ folder run ./run.sh. The result will be saved in result_save_path

Run On KITTI360

Official website of KITTI360. Please download Raw Velodyne Scans, Calibrations and Vechicle Poses

1. The only difference with Run on KITTI is step 1, Please use script/gt_kitti360.py and set the four path according to your dataset's path.

Evaluation

You can evaluate the performance(such as Precision-Recall) using the python script:

In script/evaluation.py, set the file_gt_sens_poses using xxx/out.txt and set file_outcome_osk using the result_save_path then run this python script.