Skip to main content

Realsense2 Camera

The robot is equipped with an Intel RealSense D435 camera, which can obtain RGB and depth images. The realsense_ros package is a ROS functionality package provided by Intel for calling the D435 camera.。

Starting the Camera​

  • Package Location

    /home/lemon/robot_ros_application/catkin_ws/src/realsense/realsense2_camera

  • Launch File Path

    /home/lemon/robot_ros_application/catkin_ws/src/realsense/realsense2_camera/launch/RobanD435camera.launch

  • Starting the Camera with the Launch File

    • Due to the mechanism of ROS, nodes with the same name can affect communication, so only one node with the same name can exist. Since the robot automatically starts all functional nodes, including the depth camera node, upon normal startup, you should ensure that the depth camera node started at boot is closed when running the above launch file.

      # To close the depth camera node
      rosnode kill /camera/realsense2_camera_manager

    • Starting the Depth Camera Node

      source ~/robot_ros_application/catkin_ws/devel/setup.bash
      roslaunch realsense2_camera RobanD435camera.launch

      • If the following error occurs, it indicates that the depth camera node was started repeatedly, causing an abnormal startup.

Node Output Topics​

The main topics published by the node are:

  • Camera intrinsic parameters

    /camera/color/camera_info

  • RGB image information

    /camera/color/image_raw

  • Unaligned depth information

    /camera/depth/image_rect_raw

  • Depth information aligned with RGB images

    /camera/aligned_depth_to_color/image_raw

    This topic is only output if the align_depth parameter is set to true in the launch file.

Subscribing to RGB Images​

  • Python Example

    #! /usr/bin/env python
    # coding=UTF-8

    import rospy                                        # ROS Python library
    from cv_bridge import CvBridge, CvBridgeError       # Library for converting msg data type images to OpenCV image format
    from sensor_msgs.msg import Image                   # Message type for the topic
    import cv2                                          # OpenCV-Python library
    import numpy as np

    class Vision():
        def __init__(self):
            self.topic_name = '/camera/color/image_raw'      # Topic name
            self.__cv_bridge = CvBridge()
            rospy.Subscriber(self.topic_name, Image, self.__image_callback, queue_size=1)   # Define the subscriber, including topic name, message type, callback function, and queue size

        def __image_callback(self, msg):                # Callback function for the message, the received message is in msg
            try:
                image_origin = self.__cv_bridge.imgmsg_to_cv2(msg, msg.encoding)   # Convert to numpy matrix format through cv_bridge
            except CvBridgeError as err:
                rospy.logerr(err)
                return
            print("image shape is {}".format(image_origin.shape)) # Print the shape of the image
            cv2.imshow("image",image_origin)            # Display the image
            cv2.waitKey(1)

    def rosShutdownHook():                              # This will be called when the node exits
        rospy.loginfo("VisionTest node is killed.")

    if __name__ == "__main__":
        rospy.init_node("VisionTest", anonymous=True)   # Initialize the ROS node
        rospy.on_shutdown(rosShutdownHook)              # Set the callback when the node exits
        chinV = Vision()
        rospy.spin()

    The above example subscribes to the /camera/color/image_raw topic to obtain RGB image messages, converts ROS messages to OpenCV image format (i.e., numpy arrays) through cv_bridge, and displays them.

Subscribing to Depth Information​

  • Python Example

    #! /usr/bin/env python
    # coding=UTF-8

    import rospy                                        # ROS Python library
    from cv_bridge import CvBridge, CvBridgeError       # Library for converting msg data type images to OpenCV image format
    from sensor_msgs.msg import Image                   # Message type for the topic
    import cv2                                          # OpenCV-Python library
    import numpy as np

    class Vision():
        def __init__(self):
            self.topic_name = '/camera/aligned_depth_to_color/image_raw'      # Topic name
            self.__cv_bridge = CvBridge()
            rospy.Subscriber(self.topic_name, Image, self.__image_callback, queue_size=1)   # Define the subscriber, including topic name, message type, callback function, and queue size

        def __image_callback(self, msg):                # Callback function for the message, the received message is in msg
            try:
                image_origin = self.__cv_bridge.imgmsg_to_cv2(msg, msg.encoding)   # Convert to numpy matrix format through cv_bridge
            except CvBridgeError as err:
                rospy.logerr(err)
                return
            print("depth shape is {}".format(image_origin.shape))

    def rosShutdownHook():                              # This will be called when the node exits
        rospy.loginfo("VisionTest node is killed.")

    if __name__ == "__main__":
        rospy.init_node("VisionTest", anonymous=True)   # Initialize the ROS node
        rospy.on_shutdown(rosShutdownHook)              # Set the callback when the node exits
        chinV = Vision()
        rospy.spin()

    Through the above example, you can obtain the depth image matrix aligned with the RGB image from the /camera/aligned_depth_to_color/image_raw topic, with a size of (height x width) = (480 x 640). After obtaining this matrix, you can extract the distance of each pixel point in the image individually, with the unit being mm (note that if the distance of some areas is 0, it indicates that the distance of the pixel point is invalid, and the minimum depth distance is approximately 0.28m).