Skip to content

Object Detection

Summary

This node detects buoys and boats on the water and estimates their bearing (angle) and position.
Detection is powered by an Ultralytics YOLO26 model. The pipeline is built with Deepstream and Gstreamer.

The node runs on ROS 2 and publishes:

  • Bearing to target (angle) — relative to the camera/boat frame.

  • Triangulated Position — Absolute position using camera GPS and heading data.

  • Per-detection metadata — confidence score, relative x/y of the bounding box, etc.

What it publishes

  • autoboat_msgs/msg/ObjectDetectionResultsList
    • Topic: /object_detection_results_list
    • detection_results[]: { detector_confidence, tracker_confidence, x_position, y_position, width, height, object_id, class_id, angle_to_object }
    • timestamp
    • model_name
    • yolo_version
    • threshold
  • autoboat_msg/msg/TriangulationResultsList
    • Topic: /triangulation_results_list
    • triangulation_results[]: { object_id, class_id, label, latitude, longitude }
    • iou_threshold

What it subscribes to

  • /cv_parameters
    • buffer_window_size. The number of object detections to save for triangulation.
    • iou_threshold. If two objects are triangulated to be within this distance in meters, they are assumed to be the same physical object and the older one is dropped.
    • update_rate. How often to perform triangulation and publish triangulation results in seconds.
    • model_name. The name of the model to use for inferencing without the file extension (Ex: yolo26s.pt -> yolo26s)
    • threshold. The detection threshold for the primary detector
  • /heading
  • /position

Running the node

Follow the DeepStream installation instructions

ros2 run object_detection object_detection

Optional environment variables

Set these before running the ROS2 node

  • export YOLO_VER=11 or export YOLO_VER=26: The default yolo version is 26. Set this to use Yolo11.
  • export INFERENCE=false: Disables inference and triangulation.
  • export CAMERA=false: Disables the camera.

Changin CV parameters while running

Open a new terminal and navigate to to the object_detection folder. 

cd /home/ws/src/object_detection/object_detection
Run the publish_cv_params.py script. 
python publish_cv_params.py [-m MODEL] [-t THRESHOLD] [-b BUFFER_SIZE] [-u UPDATE_RATE] [-i IOU_THRESHOLD]
Choose any number of these options.

  • -m. The model name to switch to. Do not include the file extension (Ex. yolo26s.pt -> yolo26s). The model can be either Yolo11 or Yolo26, as long as it has been previously build into an engine file.
  • -t. The threshold for the detector. Enter a number in the range (0, 1)
  • -b. The number of detections to save for triangulation. The default is 300 frames (approximately 10 seconds).
  • -u. How often to triangulate objects in seconds. Default is every 0.5 seconds.
  • -i. During triangulation, if two objects are within this distance in meters, they are considered the same object and the older one is disregarded. Default is 10 meters.