5. Data Output

5.1 Coordinate Frame Definition

The Odin1 Lite Cartesian coordinate definition is shown below. I is the IMU (i.e., Body; the two are at the same physical location and differ only in name) coordinate frame, L is the point-cloud coordinate frame, and C is the camera coordinate frame.

5.2 Extrinsic Parameter Description

Extrinsic parameter notation

We use $\mathbf{T}^{\text{A}}_{\text{B}}$ to denote coordinate transformation, meaning:

\[\mathbf{P}_{\text{A}} = \mathbf{T}^{\text{A}}_{\text{B}} \cdot \mathbf{P}_{\text{B}}\]

where $P_A$ and $P_B$ are the coordinates of the object in frame A and frame B, respectively.

Camera - LiDAR

The extrinsic parameter $\mathbf{T}^{\text{camera}}_{\text{lidar}}$ varies among devices. After the driver starts, obtain it from config/calib.yaml (Tcl_0); see the driver for details.

Body - LiDAR

Note: The Body here is the former IMU, differing only in naming (expression); it corresponds to the same physical location.

The extrinsic parameter from Body (formerly IMU) to LiDAR, $\mathbf{T}^{\text{body}}_{\text{lidar}}$ (body_T_lidar in the calibration file), is:

\[\mathbf{T}^{\text{body}}_{\text{lidar}} = \begin{bmatrix} 0.9999543679 & -0.0086157257 & 0.0041269153 & -0.0305398721 \\ 0.0086800624 & 0.9998369581 & -0.0158339418 & 0.0234679052 \\ -0.0039898215 & 0.0158690412 & 0.9998661185 & 0.0333162287 \\ 0 & 0 & 0 & 1 \end{bmatrix}\]

Others

  • If you need the Camera-to-Body extrinsic parameter, calculate it yourself from the above.

5.3 Intrinsic Parameter Description

Intrinsics mainly refer to the camera intrinsics; LiDAR intrinsics are not discussed.

5.3.1 Odin1 Lite Camera Intrinsics

  • Methods to obtain intrinsics:
    • Method 1 (runtime): After the driver starts, it pulls the calibration file from the device and saves it to ~/catkin_ws/src/ros_driver/config/camera_calib_<SN>.yaml; this varies per device.
    • Method 2 (ROS service): Run ros2 service call /manifold/device0/get_calibration odin_ros_driver_rev1/srv/GetCalibration "{}" to get the YAML string directly. The actual service prefix is subject to the startup log or ros2 service list.
    • Method 3 (HTTP Dashboard): Open a browser and visit http://192.168.1.251:8080, then download from Dashboard → Calibration File Management.
# O1L-P040100047
cam_num: 1
img_topic_0: /camera/rgb
Tcl_0: [
  0.01814, -0.99980, -0.00813, 0.00579,
  -0.00071, 0.00812, -0.99997, -0.03109,
  0.99984, 0.01815, -0.00057, -0.01629,
  0.00000, 0.00000, 0.00000, 1.00000
]

cam_0_640_544:
        cam_model: FishPoly
        image_width: 640
        image_height: 544
        k2: 3.9309765393149745e-02
        k3: 1.5817608030575755e-01
        k4: 5.8836988291029024e-01
        k5: -8.2827749950756613e-01
        k6: 3.8543134263557260e-01
        k7: -6.3564147672175983e-02
        p1: 0.
        p2: 0.
        A11: 2.3172737533779440e+02
        A12: -4.1424461677121709e-02
        A22: 2.3167070553813502e+02
        u0: 3.1331230979976016e+02
        v0: 2.7959256780073559e+02
        isFast: 0
        numDiff: 3000
        maxIncidentAngle: 120

cam_0_1280_1088:
        cam_model: FishPoly
        image_width: 1280
        image_height: 1088
        k2: 3.9309765393149745e-02
        k3: 1.5817608030575755e-01
        k4: 5.8836988291029024e-01
        k5: -8.2827749950756613e-01
        k6: 3.8543134263557260e-01
        k7: -6.3564147672175983e-02
        p1: 0.
        p2: 0.
        A11: 4.6345475067558880e+02
        A12: -8.2848923354243417e-02
        A22: 4.6334141107627005e+02
        u0: 6.2662461959952032e+02
        v0: 5.5918513560147119e+02
        isFast: 0
        numDiff: 3000
        maxIncidentAngle: 120


    

5.3.2 Camera Model Description

Odin1 Lite RGB Camera Model and Intrinsics

This document describes the projection model and intrinsic parameters of the Odin1 Lite RGB camera for engineering reference.

1 Camera Projection Model

Odin1 Lite RGB uses a polynomial fisheye projection model to describe wide-field imaging. After undistortion, this model can be simplified to a standard pinhole model.

2 Polynomial Fisheye Model

2.1 Projection Formula

Given a 3D point in the camera coordinate frame $\mathbf{P_c} = (X, Y, Z)^T$, compute the incident radius:

\[r = \sqrt{X^2 + Y^2}\]

Compute the incident angle:

\[\theta = \arccos\left(\frac{Z}{\sqrt{X^2 + Y^2 + Z^2}}\right)\]

Polynomial radial mapping:

\[\theta_d = \theta + k_2\theta^2 + k_3\theta^3 + k_4\theta^4 + k_5\theta^5 + k_6\theta^6 + k_7\theta^7\]

Project to the distortion plane:

\[x_d = \frac{\theta_d}{r} \cdot X, \quad y_d = \frac{\theta_d}{r} \cdot Y\]
2.2 Affine Transformation
\[\begin{bmatrix} u \\ v \end{bmatrix} = \begin{bmatrix} A_{11} & A_{12} \\ 0 & A_{22} \end{bmatrix} \begin{bmatrix} x_d \\ y_d \end{bmatrix} + \begin{bmatrix} u_0 \\ v_0 \end{bmatrix}\]
Parameter Description
k2–k7 Polynomial distortion coefficients
A11 X-axis focal length (horizontal scale)
A12 Non-orthogonality compensation (skew)
A22 Y-axis focal length (vertical scale)
u0, v0 Principal-point coordinates

2.3 Pinhole Model After Undistortion

\[u = f_x \frac{X}{Z} + c_x, \quad v = f_y \frac{Y}{Z} + c_y\]

2.4 Coordinate Frame Convention

Odin1 Lite RGB follows the OpenCV camera coordinate frame: X → right, Y → down, Z → front (right-handed coordinate frame).

2.5 Engineering Notes

  • Undistortion does not change extrinsics.
  • For high-precision reprojection, use the polynomial fisheye model.
  • For detection, SLAM, fusion and other algorithm frameworks, use the undistorted pinhole model.
  • Numerical instability may occur when $\theta \to 0$ ($r \to 0$ causes division by zero).

5.4 Data Description

The raw point cloud (cloud/raw, resolution 256×192) PointCloud2 fields are as follows:

Field Data Type Description
x FLOAT32 X axis, unit: meter
y FLOAT32 Y axis, unit: meter
z FLOAT32 Z axis, unit: meter
intensity UINT8 Reflectivity, range 0–255
confidence UINT8 Point confidence, range 0–255; higher is more reliable. Suggested threshold: 4 for indoor scenes; adjustable via confidence_threshold
offset_time FLOAT32 Time offset relative to the frame baseline timestamp, unit: second

Note: When a point’s confidence is below confidence_threshold, the driver will set the point’s x/y/z to 0. point_step = 18 bytes.

5.5 ROS Topics and Services Overview

Package name: odin_ros_driver_rev1, main node: odin_ros_driver_node, post-processing node: post_process_node.

All topics and services carry a per-device prefix {prefix} = /{topic_prefix}/device{N}/ (default /manifold/device0/). Only the suffixes are listed in the tables below.

ℹ️ The device model no longer appears in topic/service paths; the model and SN are published in the JSON of /{topic_prefix}/driver/device_online.

ℹ️ The imu that appears in topic/field names and the Body referred to in this document are the same concept (same physical location, differing only in name).

5.5.1 Main Driver Node (odin_ros_driver_node)

Topic suffix Type Description
cloud/raw sensor_msgs/PointCloud2 Raw point cloud (256×192)
cloud/slam sensor_msgs/PointCloud2 SLAM point cloud
camera0/compressed sensor_msgs/CompressedImage Left RGB camera JPEG compressed image
camera1/compressed sensor_msgs/CompressedImage Right RGB camera JPEG compressed image (valid only on binocular devices)
gray sensor_msgs/Image Grayscale image (mono8)
imu sensor_msgs/Imu IMU (acceleration + angular velocity)
odometry_lf nav_msgs/Odometry Low-frequency odometry (includes pose covariance)
odometry_hf nav_msgs/Odometry High-frequency odometry (includes pose covariance)

In addition, the main driver publishes the following topics under the driver-level namespace /{topic_prefix}/driver/ (no device prefix, shared by all devices):

Topic Type Description
/{topic_prefix}/driver/device_online std_msgs/String Device online event (latched, JSON contains SN/model/prefix/IP)
/{topic_prefix}/driver/device_offline std_msgs/String Device offline event (latched, JSON contains SN/model)
/{topic_prefix}/driver/resolution_change std_msgs/String Resolution change event (latched)

5.5.2 Post-processing and Functional Nodes

Topic suffix Type Description
camera0/raw sensor_msgs/Image Left RGB camera decoded image
camera1/raw sensor_msgs/Image Right RGB camera decoded image (valid only on binocular devices)
camera0/undistort sensor_msgs/Image Left RGB camera undistorted image
camera1/undistort sensor_msgs/Image Right RGB camera undistorted image (valid only on binocular devices)
camera0/rect sensor_msgs/Image Rectified left image
camera1/rect sensor_msgs/Image Rectified right image (valid only on binocular devices)
camera0/camera_info sensor_msgs/CameraInfo Rectified left camera intrinsics
camera1/camera_info sensor_msgs/CameraInfo Rectified right camera intrinsics (valid only on binocular devices)
cloud/render sensor_msgs/PointCloud2 Colored point cloud (point cloud + image fusion, LiDAR coordinate frame)
reprojection/overlay sensor_msgs/Image Point cloud reprojected onto camera image overlay result
pointcloud2depth/depth sensor_msgs/Image (32FC1) Dense depth map, unit meter; 0 means invalid pixel
pointcloud2depth/depth_color sensor_msgs/Image (bgr8) JET colormap depth map
pointcloud2depth/depth_overlay sensor_msgs/Image (bgr8) Depth map and camera image overlay result

ℹ️ Odin1 Lite is a single-RGB device. The camera1/*, stereo/* and other binocular / disparity-related topics retained in the driver code will not publish data on Odin1 Lite.

5.5.3 Point Cloud Compressed Transmission

The base topics cloud/raw and cloud/slam are always standard sensor_msgs/PointCloud2 and can be subscribed directly in RViz. After installing the compression plugin and restarting the driver, additional /zlib, /zstd, /draco compressed sub-topics are provided:

# ROS2 Humble; replace humble according to your ROS distribution
sudo apt install ros-humble-point-cloud-transport-plugins

RViz cannot display compressed point clouds directly; to visualize them, use point_cloud_transport republish to decompress to a normal PointCloud2. No plugin installation is needed if only uncompressed point clouds are used.

5.5.4 Services

Service suffix Type Description
get_calibration odin_ros_driver_rev1/srv/GetCalibration Get camera calibration YAML content
set_sensor_mode odin_ros_driver_rev1/srv/SetSensorMode Switch Sensor Mode (0=HDR / 1=High Peak / 2=Low Peak / 3=Adaptive)
reset_initial_pose odin_ros_driver_rev1/srv/ResetInitialPose Reset SLAM initial pose, map and factor graph
get_adaptive_param / set_adaptive_param GetAdaptiveParam / SetAdaptiveParam Query / set adaptive-mode parameters; effective only in Adaptive mode
get_sync_mode / set_sync_mode GetSyncMode / SetSyncMode Query / set default time-sync mode; device restart required after setting
get_ptp_sync_status odin_ros_driver_rev1/srv/GetPtpSyncStatus Query PTP sync status, offset, path delay and frequency adjustment
restore_default_config odin_ros_driver_rev1/srv/RestoreDefaultConfig Restore factory configuration; device restart required after execution
device_mode odin_ros_driver_rev1/srv/DeviceMode Set device mode (0=Standby / 1=Normal / 2=Upgrade / 3=SLAM / 4=Relocation / 5=ReMap)
save_map odin_ros_driver_rev1/srv/SaveMap Save loop-closure map to host file
load_map odin_ros_driver_rev1/srv/LoadMap Load map from host file for relocalization or ReMap
exploy_map odin_ros_driver_rev1/srv/ExployMap Export current loop-closure map to host file (specific format determined by plugin)
set_diag_monitor odin_ros_driver_rev1/srv/SetDiagMonitor Enable / disable 1 Hz frame-rate and latency diagnostics, and publish /diagnostics
set_channel_config / get_channel_config SetChannelConfig / GetChannelConfig Batch set / query main driver layer-A data channel switches
set_post_process_config / get_post_process_config SetPostProcessConfig / GetPostProcessConfig Batch set / query post-processing layer-B switches; service prefix is /{topic_prefix}/postprocess/

5.6 ROS Driver Acquisition

Odin1 Lite ROS driver: https://git.manifoldtech.cn/manifoldtech-sdk/odin_ros_driver2

Odin1 Lite SDK: https://git.manifoldtech.cn/manifoldtech-sdk/odin-sdk2

For detailed deployment and running steps, see 7. Odin1 Lite Quick Start (Ubuntu).


← Previous Chapter:Power and Connection | Next Chapter:Mapping and Relocalization API Workflow →


中文 | EN
This site uses Just the Docs, a documentation theme for Jekyll.