TextMap

TextMap builds 3D text landmarks for indoor navigation.

This package is the mapping-side component in the text navigation pipeline:

• Consumes text detections from NavOCR, and pose from an external SLAM
• Lifts detected text into 3D landmarks in the map frame
• Merges repeated observations into stable landmark entries

This README only summarizes the package role and ROS interface.

For dependencies, build instructions, and end-to-end launch examples, see the TextMap_Examples repository.

Interface

Interface with SLAM / Localization

TextMap does not perform SLAM itself. Currently, it supports rtabmap and slam_toolbox as slam_backend options, and custom SLAM backends can also be integrated.

Subscribed Topics

Topic	Type	Description
/navocr/detections	vision_msgs/Detection2DArray	OCR detections consumed as landmark observations
/camera/depth/image_rect_raw	sensor_msgs/Image	Depth image used to project detections into 3D
/camera/infra1/camera_info	sensor_msgs/CameraInfo	Camera intrinsics for depth projection
/odom	nav_msgs/Odometry	Robot odometry used for motion filtering and pose buffering
/mapGraph	rtabmap_msgs/msg/MapGraph	RTAB-Map pose graph used for landmark re-anchoring
/info	rtabmap_msgs/msg/Info	RTAB-Map graph update timing / synchronization info

Published Topics

Topic	Type	Description
/textmap/markers	visualization_msgs/msg/MarkerArray	Landmark markers for RViz visualization

Services

Service	Type	Description
/textmap/save_landmarks	std_srvs/srv/Trigger	Saves the current landmark set to YAML

Key Parameters

Parameter	Description
slam_backend	Selects the localization / SLAM backend used by TextMap, such as rtabmap or slam_toolbox
landmark_save_path	Output path for the exported landmarks.yaml file

Output

The text landmark mapping process can be monitored in RViz through the published landmark markers.

TextMap produces landmarks.yaml, which is consumed by text_nav_bridge for text-command navigation.

Example output:

landmarks:
  - id: 1
    text: "restroom"
    position: {x: 3.21, y: -1.05, z: 0.82}
    confidence: 0.92
    observation_count: 5

License

Apache License 2.0