开发流水线应用

​

流水线应用工作流

1. 定义应用的目标。
2. 检查 QIR SDK 提供的可用 API。
3. 决定流水线。

​

定义目标

​

检查 API

1. 检查 QIR SDK 提供的 API，参见 QIR SDK API 列表。首先发现 qrb_ros_camera 能够从设备内置摄像头捕获数据并将其发布到 /cam0_stream1 topic。但是，此 topic 默认使用 NV12 编码格式，而应用需要 RGB888 格式。
2. 分析其他可能有用的 API。由于 qrb_ros_camera 不输出所需格式的图像，进一步检查发现以下匹配的 ROS 节点：qrb_ros_colorspace_convert 能够将图像编码从 NV12 转换为 RGB888。
3. 找出实现的关键点。qrb_ros_camera 发布到 /cam0_stream1 topic，但 qrb_ros_colorspace_convert 的输入是 /image_raw。因此，示例应用需要通过额外处理来处理这个冲突。

​

决定流水线

1. 从 qrb_ros_camera 捕获图像并将其发布到 /cam0_stream1 topic。
2. 使用 qrb_ros_colorspace_convert 将图像从 NV12 转换为 RGB888 格式，并将 RGB888 格式的图像发布到 /image topic。

​

示例实现

​

设置交叉编译环境

cd <decompressed_workspace>/target/qcs9075-iq-9075-evk/qirpsdk_artifacts/qcs9075-iq-9075-evk
tar -zxvf qirp-sdk_<qirp_version>.tar.gz
cd qirp-sdk
source setup.sh

​

导入示例项目

cd <qir_sdk_path>/qirp-samples/ai_vision/sample_template

​

自定义 launch 文件

1. 创建实现以下步骤的 launch 文件。此示例使用两个节点 qrb_ros_camera 和 qrb_ros_colorspace_convert。它将 qrb_ros_camera 的 /cam0_stream1 topic 重映射到 /image_raw topic，然后使用 qrb_ros_colorspace_convert 将 RGB888 格式的图像发布到 /image topic。
   注意有关如何编写 launch 文件的信息，请参见 ROS 2 documentation — Creating a launch file。
   mkdir launch
   vim launch/rgb8_image_publisher_launch.py
   # Copy the following code to the file
   
   rgb8_image_publisher_launch.py

   # Copyright (c) 2025 Qualcomm Innovation Center, Inc. All rights reserved.
   # SPDX-License-Identifier: BSD-3-Clause-Clear
   import os
   import launch
   from launch import LaunchDescription
   from launch.actions import DeclareLaunchArgument
   from launch.substitutions import LaunchConfiguration
   from ament_index_python.packages import get_package_share_directory
   from launch_ros.actions import ComposableNodeContainer
   from launch_ros.descriptions import ComposableNode
   def generate_launch_description():
       # Get the path to the camera info configuration file
       camera_info_config_file_path = os.path.join(
           get_package_share_directory('qrb_ros_camera'),
           'config', 'camera_info_imx577.yaml'
       )
       # Define the composable node container to hold the camera node
       camera_container = ComposableNodeContainer(
           name="container",
           namespace='',
           package="rclcpp_components",
           executable="component_container",
           output='screen',
           composable_node_descriptions=[
               ComposableNode(
                   package='qrb_ros_camera',
                   plugin='qrb_ros::camera::CameraNode',
                   name='camera_node',
                   parameters=[{
                       'camera_id': 0,
                       'stream_size': 1,
                       'stream_name': ["stream1"],
                       'stream1':{
                           'height':480,
                           'width':640,
                           'fps':30,
                       },
                       'camera_info_path': camera_info_config_file_path,
                   }],
                   remappings=[
                       ('/cam0_stream1', '/image_raw'),
                       ('/cam0_stream1_camera_info', '/qrb_ros_camera_info')
                   ],
               )
           ]
       )
       # Declare launch arguments for conversion type and latency FPS test.
       # These arguments can be set when launching the file to customize behavior.
       # Default conversion type is set to 'nv12_to_rgb8' and latency FPS test
       # is enabled by default.
       conversion_type_arg = DeclareLaunchArgument(
           'conversion_type',
           default_value='nv12_to_rgb8',
           description='The type of conversion'
       )
       latency_fps_test_arg = DeclareLaunchArgument(
           'latency_fps_test',
           default_value='true',
           description='Enable or disable latency FPS test'
       )
       # Define the composable node container for the colorspace conversion node
       colorspace_convert_container = ComposableNodeContainer(
           name='component_colorconvert_container',
           namespace='',
           package='rclcpp_components',
           executable='component_container',
           composable_node_descriptions=[
               ComposableNode(
                   package='qrb_ros_colorspace_convert',
                   plugin='qrb_ros::colorspace_convert::ColorspaceConvertNode',
                   parameters=[{
                       'conversion_type': LaunchConfiguration('conversion_type'),
                       'latency_fps_test': LaunchConfiguration('latency_fps_test'),
                   }],
                   extra_arguments=[{'use_intra_process_comms': True, 'log_level': 'INFO'}],
               ),
           ],
           output='screen',
       )
       return LaunchDescription([
           camera_container,
           conversion_type_arg,
           latency_fps_test_arg,
           colorspace_convert_container,
       ])
   
2. 更新 setup.py 文件。
   vim setup.py
   # Copy the following code to the file
   
   setup.py

       from setuptools import find_packages, setup
       package_name = 'sample_template'
       setup(
           name=package_name,
           version='0.0.0',
           packages=find_packages(exclude=['test']),
           data_files=[
               ('share/ament_index/resource_index/packages',
                   ['resource/' + package_name]),
               ('share/' + package_name, ['package.xml']),
               ('share/' + package_name, ['launch/' + "rgb8_image_publisher_launch.py"])
           ],
           install_requires=['setuptools'],
           zip_safe=True,
           maintainer='your_name',
           maintainer_email='your.email@example.com',
           description='Package description',
           license='License declaration',
           tests_require=['pytest'],
           entry_points={
               'console_scripts': [
               ],
           },
       )
   

​

构建和部署

cd <qir_sdk_path>/qirp-samples/ai_vision/sample_template
colcon build

cd <qir_sdk_path>/qirp-samples/ai_vision/sample_template/install/sample_template
tar -czvf sample_template.tar.gz lib share
scp sample_template.tar.gz root@[ip-addr]:/opt/

# Install sample package
tar --no-overwrite-dir --no-same-owner -zxf /opt/sample_template.tar.gz -C /opt/

# Configure according to qrb_ros_colorspace_convert
# Set up Weston environment
export XDG_RUNTIME_DIR=/dev/socket/weston/
mkdir -p $XDG_RUNTIME_DIR
export WAYLAND_DISPLAY=wayland-1
# Set up shared memory
export FASTRTPS_DEFAULT_PROFILES_FILE=/usr/ros/jazzy/share/qrb_ros_colorspace_convert/config/large_message_profile.xml
# Launch the application
export ROS_DOMAIN_ID=123
source /usr/share/qirp-setup.sh
ros2 launch sample_template rgb8_image_publisher_launch.py

export ROS_DOMAIN_ID=123
source /usr/share/qirp-setup.sh
ros2 topic echo /image_raw | grep "encoding"
ros2 topic echo /image | grep "encoding"

​

后续步骤

• 使用其他 QIR API，例如传感器数据流、诊断或控制接口。
• 组合多个模块以创建混合应用（例如，自主导航 + 健康监控）。
• 学习如何使用 QIR SDK 开发 AI 流水线应用。参见 QIR SDK 示例应用。
• 查看 QIR SDK API 列表 以获取可用 API 的完整列表。