hku-mars / m-detector Goto Github PK

Great work, @HuajieWu99, @XW-HKU, @Ecstasy-EC

• Can this be used to remove dynamic environmental obscurants like dust/smoke/fog/rain etc? Standard mechanical LIDAR are still prone to false target detections due to these obscurants.

• Have you faced similar problems?

• Is M-detector an independent preprocessing module to filter out such obscurants and output static/moving pointcloud separately?

hello, in fast_lio embeded, revise_kitti.cpp,
why the calibration of vertical angle of original pcl is 0.15deg?
and why the vertical_angle is atan2(pi.z, range) ? Maybe asin?
Looking forward to your reply, thx!

您好，我只用velodyne雷达跑配套fast_lio和m-detector的时候资源占用率比较高，显得有点卡，请问后面会优化这部分吗？如果调试参数，需要考虑修改哪些参数？

Nice work! Thanks for your sharing.

I am trying to run the M-detector with the suggested odometry (Fast-Lio) and the provided nuScenes dataset. However, I found the provided nuScenes dataset didn't contain the IMU data. So it can not be used by Fast-Lio. Did you use another odometry (like A-LOAM) for the nuScenes dataset?

Thanks!

There is no imu data in kitti rosbags you provided, and fast-lio is not running properly. Please re-upload the correct kitti rosbag thx.

Hi, @HuajieWu99 @XW-HKU @Ecstasy-EC. Kudos for the great work.

Supplementary Information document states that:

For input, M-detector accepts either an individual point or a frame of
points, with the sensor ego-motion compensated in advance. The result of ego-
motion compensation can be provided by a LiDAR-based odometry or a full
Simultaneous localization and mapping (SLAM) system.

• If that is the case, won't the ego-motion estimate from LiDAR-based odometry/SLAM be untrustworthy due to dynamic objects in the raw pointcloud data?
• What's the impact of additional uncertainity in position estimate from LiDAR-based odometry/SLAM due to dynamic events on M-detector's performance?
• Can we use M-detector without prior ego motion information(e.g. from FAST_LIO)? Or, Use it as a preprocessing module which removes dynamic events from the pointcloud given to LiDAR-based odometry/SLAM?
• Is robust localization also a use case? Or, mapping only?

Great research work, I have a question to ask you? It can be used for a 32-wire laser radar?or How should I configure the parameters for a 32-wire laser radar, such as Hesai lidar pandar?

Hey, thanks for your amazing works, i think it will me a lot in my program.
In the readme file, you mentioned there is a embeded version of m-detector in the releases, but I can't find it in the releases. I would appreciate it if you would check it out once.
Best wishes.

Very nice work! Would you provide a parameter table which shows the relationship between param in paper and param in your code? It's hard to keep track with them by reading the code. Thank you so much!

depth_map_list的大小由参数max_depth_map_num，在Case1中当前的点投影到最近的HASH_PRIM个深度图，当max_depth_map_num修改大于HASH_PRIM/19时，我设置的25，会出现越界的问题如下(KITTI1.yaml数据结果)：

你好，我看到cmake中注释掉了这个，然后编译会找不到python，是不是应该取消注释呢

# find_package(PythonLibs REQUIRED)
# find_path(MATPLOTLIB_CPP_INCLUDE_DIRS "matplotlibcpp.h")

Hi, dear author,

 Nice work! Thank you for your contributions to the open source community.
 
  Could you provide a Baidu Cloud link for mainland users? The data is too large for mainland users to download.

你好，很优秀的项目。
我用你们的filter接口跑自己的激光点云数据程序蹦了，定位到DynObjFilter.cpp619行出现非法访问指针错误
，可能是max_pointers_num相关参数设置不对？

I try to run M-Detector and fast-lio on the kitii and waymo dataset, but the dataset doesn't contain IMU data. How to deal with this problem?

您好 @HuajieWu99 ，
请问为何fast-lio release版本编译通过后依旧无法找到node？已经source 过工作空间了，这个issue#183下的方法也无法解决，以下是运行时信息：

roslaunch fast_lio mapping_kitti.launch 
... logging to /home/robot-nuc12/.ros/log/c1561a6c-aeb8-11ee-a946-291d976a18a1/roslaunch-robot-nuc12-3171814.log
Checking log directory for disk usage. This may take a while.
Press Ctrl-C to interrupt
Done checking log file disk usage. Usage is <1GB.

started roslaunch server http://robot-nuc12:35365/

SUMMARY
========

PARAMETERS
 * /common/imu_topic: /quadrotor_simula...
 * /common/lid_out_topic: /cloud_registered...
 * /common/lid_topic: /velodyne_points_...
 * /common/odom_out_topic: /aft_mapped_to_init
 * /common/odom_topic: //odometry_gt_
 * /common/points_topic: /velodyne_points
 * /common/time_sync_en: True
 * /cube_side_length: 50.0
 * /filter_size_map: 0.5
 * /filter_size_surf: 0.5
 * /mapping/acc_cov: 0.5
 * /mapping/b_acc_cov: 0.0001
 * /mapping/b_gyr_cov: 0.0001
 * /mapping/det_range: 100.0
 * /mapping/extrinsic_R: [1, 0, 0, 0, 1, 0...
 * /mapping/extrinsic_T: [0.0, 0.0, -0.0]
 * /mapping/fov_degree: 180
 * /mapping/gyr_cov: 0.2
 * /mapping/imu_en: False
 * /mapping/init_vel_x: 10.0
 * /mapping/init_vel_y: 0.0
 * /mapping/init_vel_z: 0.0
 * /max_iteration: 4
 * /pcd_save/interval: -1
 * /pcd_save/pcd_save_en: False
 * /point_filter_num: 1
 * /preprocess/blind: 0
 * /preprocess/feature_extract_en: False
 * /preprocess/lidar_type: 4
 * /preprocess/scan_line: 6
 * /publish/dense_publish_en: True
 * /publish/path_en: False
 * /publish/scan_bodyframe_pub_en: False
 * /publish/scan_publish_en: True
 * /rosdistro: noetic
 * /rosversion: 1.16.0
 * /voxel/layer: 2
 * /voxel/match_eigen_value: 0.005
 * /voxel/match_s: 0.9
 * /voxel/min_eigen_value: 0.005
 * /voxel/pub_plane_en: False
 * /voxel/sigma_num: 81
 * /voxel/voxel_map_en: False
 * /voxel/voxel_size: 3

NODES
  /
    laserMapping (fast_lio/fastlio_mapping)
    revise_kitti (fast_lio/revise_kitti)

ROS_MASTER_URI=http://localhost:11311

ERROR: cannot launch node of type [fast_lio/fastlio_mapping]: Cannot locate node of type [fastlio_mapping] in package [fast_lio]. Make sure file exists in package path and permission is set to executable (chmod +x)
ERROR: cannot launch node of type [fast_lio/revise_kitti]: Cannot locate node of type [revise_kitti] in package [fast_lio]. Make sure file exists in package path and permission is set to executable (chmod +x)
No processes to monitor
shutting down processing monitor...
... shutting down processing monitor complete

有没有相关的参数配置？

Hello,

Can you let me know when the code will be released? I am facing an issue in my current setup regarding moving objects.

Thank you

[dynfilter-1] process has died [pid 74452, exit code -11, cmd /home//m-decetor/devel/lib/m_detector/dynfilter __name:=dynfilter __log:=/home/*/.ros/log/0754e268-dd27-11ee-a00a-d191908e71e6/dynfilter-1.log].
fast-lio work is fine,but not m_detector,how can I fix this problem?

参照Supplementary中的讲解，该如何对应config/.yaml中的参数？有没有相关的说明文档

The kitti bag I used was downloaded according to your link, and the parameters have not been adjusted. They were pulled from the original code. You can refer to the picture for the effect. You can see that there are a lot of ground points that are judged to be dynamic points. Please tell me how should I solve this problem

按照5.1描述后，并下载作者提供的kitti包，发现未高亮动态点云，但是有检测到动态点的个数。

作者你好，运行并看了你的框架，发现效果确实很不错，但是好像没有提供如何计算真实标签的代码吧？在你补充中有说明，你是利用开源数据集提供标注框的速度来标记事件点作为真实标签，可以公布一下这部分代码吗？