1. ORK
網址:https://wg-perception.github.io/object_recognition_core/
ORK (Object Recognition Kitchen) 是 ROS 集成的物體識別庫,當前 Kinetic 版本的 ROS 只集成了部分功能包的二進制安裝文件,所以需通過源碼編譯安裝。
- 安裝依賴庫
sudo apt-get install meshlab sudo apt-get install libosmesa6-dev sudo apt-get install python-pyside.qtcore sudo apt-get install python-pyside.qtgui
- 創建工作空間,下載功能包源碼,編譯
mkdir ork_ws && cd ork_ws wstool init src https://raw.github.com/wg-perception/object_recognition_core/master/doc/source/ork.rosinstall.kinetic.plus cd src && wstool update -j8 cd .. && rosdep install --from-paths src -i -y catkin_make
- 設置環境變量
echo "export ~/ork_ws/devel/setup.bash" >> ~/.bashrc source ~/.bashrc
2. CouchDB 建立模型庫
ORK 中的 LINEMOD 算法基於模板匹配,需要建立已知物體的數據模型,根據采集的信息逐一匹配,找到與數據庫中匹配的物體。首先,用 CouchDB 工具創建數據庫:
- 安裝
sudo apt-get install couchdb
- 測試是否安裝成功,如圖所示,說明安裝成功
curl -X GET http://localhost:5984
- 在數據庫中創建一條可樂罐模型的數據
rosrun object_recognition_core object_add.py -n "coke " -d "A universal can of coke" --commit
- 瀏覽器中查看 http://localhost:5984/_utils/database.html?object_recognition/_design/objects/_view/by_object_name,復制 id(一串類似亂碼的數字),下面要用
- 加載可樂罐的 3D 模型,在 ork_tutorials 中包含了一個可樂罐模型 coke.stl,直接下載使用。下載 ork_tutorials
git clone https://github.com/wg-perception/ork_tutorials
- 將 coke.stl 模型加載到數據中
rosrun object_recognition_core mesh_add.py bb01ae7a23033bdef1a1c3b76000092c ~/ork_ws/src/ork_tutorials/data/coke.stl --commit
- 再次在瀏覽器中打開上面的網址,如下圖所示
- 安裝 couchapp 工具,在瀏覽器中查看具體的模型
sudo pip install git+https://github.com/couchapp/couchapp.git rosrun object_recognition_core push.sh
- 在瀏覽器中查看
3. 模型訓練
rosrun object_recognition_core training -c `rospack find object_recognition_linemod`/conf/training.ork
訓練完如下圖所示
4. 物體識別
- 啟動 realsense d435i 相機
roslaunch realsense2_camera rs_camera.launch filters:=pointcloud
參考:https://github.com/IntelRealSense/realsense-ros
- 運行下述命令,進行物體檢測
rosrun object_recognition_core detection -c `rospack find object_recognition_linemod`/conf/detection.ros.ork
我們發現,detection.ros.ork 訂閱了下面幾個話題:
realsense 發布的對應的話題為:
/camera/depth/camera_info /camera/depth/image_rect_raw /camera/color/camera_info /camera/color/image_raw
因此需要用 topic_tools relay 做話題映射,如下所示:
rosrun topic_tools relay /camera/depth/camera_info /camera/depth_registered/camera_info rosrun topic_tools relay /camera/depth/image_rect_raw /camera/depth_registered/image_raw rosrun topic_tools relay /camera/color/camera_info /camera/rgb/icamera_info rosrun topic_tools relay /camera/color/image_raw /camera/rgb/image_rect_color
但每次都做映射太麻煩,干脆修改一下 detection.ros.ork 訂閱的話題名:
gedit `rospack find object_recognition_linemod`/conf/detection.ros.ork
修改如下:
- 做完話題映射或上述設置后,再次運行 detection 命令:
rosrun object_recognition_core detection -c `rospack find object_recognition_linemod`/conf/detection.ros.ork
此時會出現下圖所示信息:
- 打開 rviz 查看
- 添加 Pointcloud2,選擇話題 /camera/depth/color/points
- 添加 OrkObject,選擇話題 /recognized_object_array (若能夠成功識別,則會發布此話題)
- 查看 /recognized_object_array 話題的實時內容
rostopic echo /recognized_object_array
- 查看節點圖
rqt_graph
參考:
[1]. 官網:https://wg-perception.github.io/object_recognition_core/install.html#install
[2]. 《ROS機器人開發實踐》胡春旭
[3]. https://blog.csdn.net/weixin_40799950/article/details/81911877
[4]. https://blog.techbridge.cc/2016/05/14/ros-object-recognition-kitchen/