Vision — ROS 2 nodes¶

Ready-to-run nodes that wrap the vision algorithms and cameras as ROS 2 executables (ros2 run nectar <node>.py). Each drives an ImageHandler internally and publishes to nectar_interfaces messages.

See also: Cameras · Algorithms · Vision overview.

Architecture¶

classDiagram
    class ArucoNode {
        -aruco Aruco
        -img_handler ImageHandler
        -pose_estime_pub Publisher
        +process_image(img)
        +cleanup()
    }

    class LineDetectionNode {
        -line_detectors Dict~str,LineDetector~
        -estimation_methods Dict~str,ILineEstimationMethod~
        -image_handler ImageHandler
        +process_image(img)
        +initialize_color_detector(color, color_space)
        +run()
        +cleanup()
    }

    class ColorCalibrationNode {
        -color_detector ColorDetector
        -img_handler ImageHandler
        -_process(img)
    }

    class CameraPublisherNode {
        -camera AbstractCam
        -image_handler ImageHandler
        -publisher Publisher
        +cleanup()
    }

    ArucoNode o-- Aruco
    ArucoNode o-- ImageHandler
    LineDetectionNode o-- LineDetector
    LineDetectionNode o-- ImageHandler
    ColorCalibrationNode o-- ColorDetector
    ColorCalibrationNode o-- ImageHandler
    CameraPublisherNode o-- ImageHandler

ArUco detection node¶

ros2 run nectar aruco_node.py --ros-args \
    -p image_source:=webcam -p marker_dict:=5 -p tag_size:=0.05

Publishes /aruco/pose_estimate (nectar_interfaces/ArucoTransforms).

Line detection node¶

ros2 run nectar line_detection_node.py --ros-args \
    -p line_colors:="blue,red" -p method:=HoughLinesP \
    -p spaces:="hsv,lab" -p image_source:=webcam -p show_visualization:=true

Methods: HoughLinesP, RotatedRect, FitEllipse, RansacLine, AdaptiveHoughLinesP. Per color it publishes /line_state/{color} (nectar_interfaces/LineInfo) and /line_detect/{color} (std_msgs/Bool).

Color calibration node¶

ros2 run nectar color_calibration_node.py --ros-args \
    -p image_source:=webcam -p color_space:=hsv -p flood_tolerance:=15

Interactive HSV/LAB calibration in one window: left-click a colored region to auto-compute thresholds via flood fill, then fine-tune with the six channel trackbars. The stacked view shows original | mask | result.

Saved colors are written to the shared algorithms/color/color_calibration.json and load via ColorDetector(mode="preset", color=<name>) and the line detection node.

Note: this node requires an OpenCV GUI (mouse + trackbars).

Camera publisher node¶

Publishes any CameraFactory source as a ROS 2 image topic. Select the driver with camera_source (webcam, realsense, oakd, c920, imx219, ros, ...) and set the source-specific parameters.

ros2 run nectar camera_publisher_node.py --ros-args \
    -p camera_source:=webcam -p device_index:=0 \
    -p width:=1280 -p height:=720 -p fps:=30 \
    -p use_compression:=true -p jpeg_quality:=80 -p threaded:=true

Publishes image_raw/compressed (sensor_msgs/CompressedImage) or image_raw (sensor_msgs/Image, when compression is disabled).