5.1. Mono Camera Tools and Capture Node

5.1.1. Mono Camera Tools

  1. Camera Calibration: A Python tool for camera calibration is provided: $SDK_DIR/tools/mono_camera/camera_calibration.py. This tool calibrates the camera using a set of checkerboard images captured with the target USB camera and generates a camera_info.yaml file. Example calibration images are provided. While setting up the Docker environment, the calibration images are downloaded under $HOME/j7ros_home/data/calib_images. Below is the usage of the camera calibration tool.

    usage: camera_calibration.py [-h] [-p PATH] [-e EXT] [-r PTS_ROWS] [-c PTS_COLUMNS] [-s SQUARE_SIZE] [-v VISUALIZE] [-f SCALING_FACTOR] [-o OUT_FNAME] [-n CAMERA_NAME]
    
    optional arguments:
    -h, --help            show this help message and exit
    -p PATH, --path PATH  Path to the directory storing images for calibration
    -e EXT, --ext EXT     Extension of the files
    -r PTS_ROWS, --pts_rows PTS_ROWS
                            Number of inner corners per row
    -c PTS_COLUMNS, --pts_columns PTS_COLUMNS
                            Number of inner corners per column
    -s SQUARE_SIZE, --square_size SQUARE_SIZE
                            Size of a square (the edge, by mm) of the checker board
    -v VISUALIZE, --visualize VISUALIZE
                            Visualize the checker board or not
    -f SCALING_FACTOR, --scaling_factor SCALING_FACTOR
                            scaling factor after reading images
    -o OUT_FNAME, --out_fname OUT_FNAME
                            Path to the output file to store result (in YAML format)
    -n CAMERA_NAME, --camera_name CAMERA_NAME
                            Camera name
    

    For examples of how to use the tool, please refer to $SDK_DIR/tools/mono_camera/calib_C920.sh.

  2. Rectification Map Generation: A Python tool is provided for generating rectification map: $SDK_DIR/tools/mono_camera/generate_rect_map_mono.py. This tool takes the camera_info YAML file as input and generates a binary file for undistortion/rectification look-up-table (LUT). The generated LUT is required for offloading the undistortion/rectification processes to the VPAC/LDC hardware accelerator. Below is the usage of the tool.

    usage: generate_rect_map_mono.py [-h] [--input INPUT] [--name NAME]
    
    optional arguments:
    -h, --help            show this help message and exit
    --input INPUT, -i INPUT
                            YAML file name for camera_info. The file should be stored under ../config folder
    --name NAME, -n NAME  Camera name
    

    For examples of how to use the tool, please see $SDK_DIR/tools/mono_camera/calib_C920.sh.

5.1.2. USB Mono Camera ROS Node

This is a ROS node for USB mono cameras (Webcam) based on the OpenCV VideoCapture API. The node publishes raw images and camera information.

Note

This ROS node has only been tested with Logitech C920 and C270 webcams in ‘YUYV’ (YUYV 4:2:2) mode. ‘MJPG’ (Motion-JPEG) mode is currently not enabled or tested. If you want to use the camera in ‘MJPG’ mode, it is recommended to use the gscam / gscam2 ROS package instead.

5.1.2.1. Usage

  1. Build the USB mono camera ROS node

    cd $ROS_WS
    # ROS 1
    catkin_make --source /opt/robotics_sdk/ros1 --force-cmake
    source devel/setup.bash
    # ROS 2
    colcon build --base-paths /opt/robotics_sdk/ros2 --cmake-force-configure
    source install/setup.bash
    
  2. Launch the mono camera node

    # ROS 1
    roslaunch mono_capture mono_capture.launch
    # ROS 2
    ros2 launch mono_capture mono_capture_launch.py
    

5.1.2.2. Launch File Parameters

Parameter

Description

Value

model_str

Camera model string

string

cam_id

Camera ID number. Set cam_id = X if the device shows up as /dev/video-usb-camX on the target with TI Edge AI

string

camera_mode

Camera mode

‘FHD’ (1920x1080)

‘HD’ (1280x720)

‘VGA’ (640x480)

frame_rate

Frame rate at which raw images are published

int

encoding

Image encoding

‘yuv422’ (default) or ‘bgr8’

topic_ns

Topic namespace

‘camera’ (default)

Note

When encoding is set to ‘yuv422’, the pixel format YUV422::YUYV from the webcam is converted to YUV422::UYVY format to ensure compatibility with the LDC hardware accelerator.