Point clouds from SLAM scanner recordings.
Windows · Android · free
⬇ Download for Windows (.exe) · ⬇ Download for Android (.apk)
A2 mapper builds a dense point cloud (.las) and a trajectory (.txt) from
raw SLAM scanner recordings. It reads:
- ROS1 bag files (
.bag, uncompressed) with the topics/livox/lidar(livox_ros_driver2/CustomMsg) and/livox/imu(sensor_msgs/Imu) — recordings of SHARE S20 SLAM scanners (Livox Mid-360); .binrecordings of A2 scanners — see the OpenSLAM project.
- Colour for SHARE S20 scans. Choose the whole scan folder (the one with the
all_*.bagand theinfofolder) and the point cloud comes out coloured from the scanner's cameras. Choose a single file and you get a grey cloud. - Control points (RTK). The trajectory can be corrected by your control points, and the cloud comes out in their coordinates — see below.
- Closed loop. Start and finish are tied together when the recording is made as described below.
- Private. Everything runs on your own device. No account, no network, no cloud: your recordings never leave your computer or phone.
The odometry algorithm was designed and written by the author independently. It is not derived from, and does not reuse code of, well-known open-source SLAM implementations.
| Platform | File | Notes |
|---|---|---|
| Windows 10 / 11, x64 | v4/A2-mapper-v4.exe |
one file, 1.6 MB, nothing to install |
| Android 8.0+, arm64 | v4/A2-mapper-v4.apk |
the same engine on the phone (no control points) |
SHA-256 checksums: v4/SHA256SUMS.txt. The program is not
code-signed, so Windows SmartScreen may warn you on first start: choose
More info → Run anyway. On Android, allow installation from this source.
- Start A2-mapper-v4.exe.
- Choose the recording:
- File — a
.bagor.binrecording → grey point cloud; - Folder — a whole SHARE S20 scan folder → coloured point cloud;
- or just drop a file or a folder into the window.
- File — a
- The output folder is suggested automatically. Press Build.
- You get two files in the output folder:
cloud.lasandtrajectory.txt.
trajectory.txt is plain text, one pose per line:
x y z roll pitch yaw qx qy qz qw timestamp (metres, radians, quaternion in
x y z w order, timestamp in seconds). The theme follows Windows (light or
dark).
The cloud is best when the recording is closed:
- Stand still at the start for 10 seconds.
- Walk the whole route.
- Come back to exactly the same spot you started from.
- Stand still for 10 seconds again, then stop the recording.
A2 mapper finds the two stationary segments and ties the finish to the start, so the drift of the loop is closed.
Tick Correct by control points and the trajectory is pulled onto your control points. The cloud and the trajectory are then written in the coordinates of your points (any metric coordinate system: local, UTM, a national grid…). The offset between the scanner and the marked point is found automatically.
The program looks for the points file next to the recording.
File name: ctrl_points_edit.csv (or control_points.csv), in the same
folder as the all_*.bag.
Name,timestamp,x,y,z
PT1,1735725777.104061985,2170499.478,312016.336,240.424
PT2,1735725849.091454401,2170487.621,311994.746,240.505
PT3,1735725916.910032466,2170468.898,312007.000,240.282
PT6,1735725968.217876747,2170476.178,311990.227,242.312
PT7,1735726039.760526743,2170490.040,312018.295,242.373timestamp— the moment the point was marked during the scan, in seconds, on the scanner's clock. SHARE saves these moments incontrol_points.csvwith emptyx,y,z: just fill the coordinates in and save the file asctrl_points_edit.csv.x, y, z— coordinates of the mark in metres; they are used exactly as written. Rows with empty coordinates are skipped.- At least 3 points are required; 5–7, spread along the route, work best. Column names are case-insensitive, extra columns are ignored.
File name: scan_NNNNN.gcp.csv, next to scan_NNNNN.bin. The A2 scanner
writes it itself: the RTK position at every press of the control-point button.
Used columns: board_real_ns, lat_deg, lon_deg, h_ellps_m (optionally
hacc_m, vacc_m, carr_soln). Only RTK-fixed points (carr_soln = 2) are
used; the result is in local east-north-up metres around the first point.
A2 mapper is an independent project. SHARE, SHARE Studio and Livox are trademarks of their owners; they are named here only to say which recordings the program reads.
Free to use, for personal and commercial work. Reverse engineering, decompiling, disassembling and modification are not allowed. Full text: LICENSE.
- Telegram — https://t.me/a2blog
- YouTube — https://www.youtube.com/@a2-lab
- OpenSLAM — https://github.com/Werdna0107/OpenSLAM
#SLAM #LiDAR #PointCloud #Livox #Mid360 #SHAREslam #ROS #rosbag #LAS #Odometry #Mapping #3DScanning #Surveying #Geodesy #RTK #GNSS #GCP #ControlPoints #PointCloudProcessing #Windows #Android

