A ros demo for people detection and tracking on DJI M100 drone
mkdir -p ~/catkin_ws/src
next, copy all packages to /catkin_ws/src
and
catkin_make
-
Set your own image topic
find ros_people_detect.launch, ros_kcf_node.launch, dji_sdk_client.launch, and replace
/image_pub/image
with your own image topic. -
Run dji_sdk_client and dji_sdk
roslaunch dji_sdk_demo dji_sdk_client.launch
roslaunch dji_sdk sdk_manifold.launch
-
Run people_detect node
roslaunch people_detect ros_people_detect.launch
-
Run ros_kcf node
roslaunch ros_kcf ros_kcf_node.launch
-
Remote control
Now, you can use your remote control with custom functions to control the stop and start of tracking and detection. The control interface is defined in client.cpp.
void StartMission1Callback(DJIDrone *drone) { drone->request_sdk_permission_control(); sleep(1); ros::Rate loop_rate(50); while(ros::ok()) { ros::spinOnce(); drone->attitude_control(0x4B,forwardV,leftrV,heightV,yawV); cout<<fixed<<setprecision(2)<<" "<<forwardV<<" "<<leftrV<<" "<<heightV<<" "<<yawV<<endl; loop_rate.sleep(); } }
This callback function is used to start the task, that is to start the autonomous detection and tracking.
void StartMission2Callback(DJIDrone *drone) { std_srvs::Empty srv; start_tracking.call(srv); }
This callback function is used to start tracking, that is, people_detect node give a target to kcf_track node. In fact, this goal is the closest person to the center of the image, and if there is no people is detected, the tracking program will not run, but when any people is detected again, the tracking program will automatic run.
void StartMission3Callback(DJIDrone *drone) { std_srvs::Empty srv; if(stop_detecting.call(srv)) { sleep(1); detect_stop.call(srv); } }
This callback function is used to start and stop people detecting.
void StartMission4Callback(DJIDrone *drone) { std_srvs::Empty srv; if(stop_tracking.call(srv)) { sleep(1); track_stop.call(srv); } }
This callback function is used to stop kcf tracker.