《机器人实践开发③：Foxglove可视化机器人的眼睛-视频》

导语：

视频是机器人感知系统最核心的模块之一。本篇将讲解 Foxglove 如何处理摄像头视频流，包括编码格式、消息封装、bridge 转发、前端渲染机制，并对常见问题（如卡顿、颜色不对）给出解决方案。

---

本篇依赖前两篇的内容，请读者前往阅读 👀

《机器人实践开发①：Foxglove 开发环境完整搭建指南（含常见坑位）》
《机器人实践开发②：Foxglove 嵌入式移植 + CMake 集成》

第一种方式比较简单，通过 ros2 的方式推送图片流

注意：

这种方式比较耗流量，在网络不好的情况下会导致异常显示

需要结合 opencv2 处理图片

需要在机器人端运行桥接指令 ros2 launch foxglove_bridge foxglove_bridge_launch.xml port:=8765

#include <rclcpp/rclcpp.hpp>
#include <sensor_msgs/msg/image.hpp>
#include <cv_bridge/cv_bridge.h>
#include <opencv2/opencv.hpp>

class CameraPublisher : public rclcpp::Node {
public:
    CameraPublisher()
            : Node("camera_publisher")
    {
        publisher_ = this->create_publisher<sensor_msgs::msg::Image>("/camera/image_raw", 10);

        cap_.open(0);  // 默认摄像头
        if (!cap_.isOpened()) {
            RCLCPP_ERROR(this->get_logger(), "Cannot open camera");
            rclcpp::shutdown();
        }

        timer_ = this->create_wall_timer(
                std::chrono::milliseconds(33),  // ~30 FPS
                std::bind(&CameraPublisher::publish_frame, this)
        );
    }

private:
    void publish_frame() {
        cv::Mat frame;
        cap_ >> frame;
        if (frame.empty()) {
            RCLCPP_WARN(this->get_logger(), "Empty frame captured");
            return;
        }

        auto msg = cv_bridge::CvImage(std_msgs::msg::Header(), "bgr8", frame).toImageMsg();
        publisher_->publish(*msg);
        RCLCPP_INFO(this->get_logger(), "Published a frame");
    }

    rclcpp::Publisher<sensor_msgs::msg::Image>::SharedPtr publisher_;
    rclcpp::TimerBase::SharedPtr timer_;
    cv::VideoCapture cap_;
};

int main(int argc, char * argv[])
{
    rclcpp::init(argc, argv);
    auto node = std::make_shared<CameraPublisher>();
    rclcpp::spin(node);
    rclcpp::shutdown();
    return 0;

编写 CmakeList.txt文件

cmake_minimum_required(VERSION 3.8)
project(camera_publisher)

find_package(ament_cmake REQUIRED)
find_package(rclcpp REQUIRED)
find_package(std_msgs REQUIRED)
find_package(example_interfaces REQUIRED)
find_package(visualization_msgs REQUIRED)
find_package(cv_bridge REQUIRED)
find_package(OpenCV REQUIRED)



add_executable(camera_publisher camera_publisher.cpp)
ament_target_dependencies(camera_publisher rclcpp std_msgs visualization_msgs cv_bridge OpenCV)

install(TARGETS
        camera_publisher
        DESTINATION ${CMAKE_CURRENT_SOURCE_DIR}/lib/${PROJECT_NAME})
ament_package()

---

第二种方式，使用 foxglove sdk 的方式集成，使用 H264/H265视频流的方式，推荐！！

有效的节省带宽，运动特性更好

视频指令可控，帧率/码流/分辨率

功耗低，运行比第一种方式节省 cpu 资源

定义视频类

std::unique_ptr<foxglove::schemas::CompressedVideoChannel>                          visualization_compressed_main_video_channel;

封装视频发送接口

核心接口是visualization_compressed_main_video_channel>log(compressed_video)

注意填充数据 一定要将timestamp 填充对，否则没有看不见视频

int FoxgloveVisualizationService::startVideoChannelService(int index,uint8_t *video_buffer, uint32_t audio_size) {
    auto time_since_epoch = std::chrono::system_clock::now().time_since_epoch();
    auto seconds = std::chrono::duration_cast<std::chrono::seconds>(time_since_epoch);
    auto nanoseconds =
          std::chrono::duration_cast<std::chrono::nanoseconds>(time_since_epoch - seconds);
    foxglove::schemas::Timestamp timestamp;
    timestamp.sec = static_cast<uint32_t>(seconds.count());
    timestamp.nsec = static_cast<uint32_t>(nanoseconds.count());


    std::vector<std::byte> data(audio_size * sizeof(uint8_t));
    std::memcpy(data.data(), video_buffer, audio_size * sizeof(uint8_t));

    foxglove::schemas::CompressedVideo compressed_video;
    compressed_video.format = "h264";
    compressed_video.timestamp = timestamp;
    compressed_video.frame_id = visualization_frame_id;
    compressed_video.data = data;
    index==0?visualization_compressed_main_video_channel->log(compressed_video):visualization_sub_compressed_video_channel->log(compressed_video);
}

使用工具可视化视频