ROS学习 二、Debian10虚拟机下部署ROS YOLOv3

前言

现在的机器人定位、导航等技术都会涉及到感知、规划、控制算法。本次将在ROS中配置yolov3目标检测算法，以此学习视觉感知算法在ROS中的搭建。

本文已经提前将ROS环境搭建好了。

darknet_ros 源码编译

首先创建一个ros工作空间，然后下载源码，

mkdir -p ~/yolo_ws/src
cd ~/yolo_ws/src
git clone https://github.com/leggedrobotics/darknet_ros.git

为了在加速编译，先提前把yolo的权重下载下来，

cd darknet_ros/darknet_ros/yolo_network_config/weights/
wget http://pjreddie.com/media/files/yolov2.weights
wget http://pjreddie.com/media/files/yolov2-tiny.weights
wget http://pjreddie.com/media/files/yolov3-tiny.weights
wget http://pjreddie.com/media/files/yolov3.weights

然后就可以开始编译了，

catkin_make -DCMAKE_BUILD_TYPE=Release
# 编译成功后，准备进行测试

运行ROS YOLOv3

首先配置darknet_ros的运行环境

source ~/yolo_ws/devel/setup.bash

然后选择你要使用的yolo模型，这里可以选择之前下载的yolov2、yolov3以及tiny模型，通过修改darknet_ros.launch文件来进行切换，

rosed darknet_ros darknet_ros.launch

该launch文件的注释如下：

  
  
  
# 定义yolo结构和权重的路径
  
  
  
# 定义ros通信的方式 与 yolo的配置
  
  
  
# 载入ros和yolo配置
  
  
  
# 开启节点，接收摄像头图像
  
  
    
    
    
  

  -->

由于我们想使用yolov3，因此将launch中的yolov2-tiny.yaml改为yolov3.yaml

接着测试一下usb摄像头是否能够打开，并找到摄像头发布的话题名，并修改darknet_ros中的话题订阅

# 测试摄像头
roslaunch usb_cam usb_cam-test.launch 
# 找到摄像头发布的话题
rostopic list
# 一般是类似 /usb_cam/image_raw 这种话题
# 然后编辑ros yolo需要接收图像时需要订阅的话题
rosed darknet_ros ros.yaml
# camera_reading中的topic改成找到的摄像头发布的话题

然后测试ros yolov3，

# 打开两个terminal
roslaunch darknet_ros darnet_ros.launch
roslaunch usb_cam usb_cam-test.launch
# 看到出现了一个名为yolo的新视图，则表明运行成功

此外，我还打开了一个新窗口查看运行中的topic，

# rostopic list -v
Published topics:
 * /rosout_agg [rosgraph_msgs/Log] 1 publisher
 * /rosout [rosgraph_msgs/Log] 3 publishers
 * /darknet_ros/found_object [darknet_ros_msgs/ObjectCount] 1 publisher
 * /darknet_ros/bounding_boxes [darknet_ros_msgs/BoundingBoxes] 1 publisher
 * /darknet_ros/detection_image [sensor_msgs/Image] 1 publisher
 * /darknet_ros/check_for_objects/result [darknet_ros_msgs/CheckForObjectsActionResult] 1 publisher
 * /darknet_ros/check_for_objects/feedback [darknet_ros_msgs/CheckForObjectsActionFeedback] 1 publisher
 * /darknet_ros/check_for_objects/status [actionlib_msgs/GoalStatusArray] 1 publisher
 * /image_view/parameter_descriptions [dynamic_reconfigure/ConfigDescription] 1 publisher
 * /image_view/parameter_updates [dynamic_reconfigure/Config] 1 publisher
 * /image_view/output [sensor_msgs/Image] 1 publisher
 * /usb_cam/image_raw/compressedDepth [sensor_msgs/CompressedImage] 1 publisher
 * /usb_cam/image_raw/compressedDepth/parameter_descriptions [dynamic_reconfigure/ConfigDescription] 1 publisher
 * /usb_cam/image_raw/compressedDepth/parameter_updates [dynamic_reconfigure/Config] 1 publisher
 * /usb_cam/image_raw/compressed [sensor_msgs/CompressedImage] 1 publisher
 * /usb_cam/image_raw/compressed/parameter_descriptions [dynamic_reconfigure/ConfigDescription] 1 publisher
 * /usb_cam/image_raw/compressed/parameter_updates [dynamic_reconfigure/Config] 1 publisher
 * /usb_cam/image_raw [sensor_msgs/Image] 1 publisher
 * /usb_cam/image_raw/theora [theora_image_transport/Packet] 1 publisher
 * /usb_cam/image_raw/theora/parameter_descriptions [dynamic_reconfigure/ConfigDescription] 1 publisher
 * /usb_cam/image_raw/theora/parameter_updates [dynamic_reconfigure/Config] 1 publisher
 * /usb_cam/camera_info [sensor_msgs/CameraInfo] 1 publisher

Subscribed topics:
 * /rosout [rosgraph_msgs/Log] 1 subscriber
 * /usb_cam/image_raw [sensor_msgs/Image] 2 subscribers
 * /darknet_ros/check_for_objects/goal [darknet_ros_msgs/CheckForObjectsActionGoal] 1 subscriber
 * /darknet_ros/check_for_objects/cancel [actionlib_msgs/GoalID] 1 subscriber

后记

过段时间我会对darknet_ros进行更细致的源码和配置文件解析~