search

koide3 / hdl_people_tracking

Real-time people tracking using a 3D LIDAR
BSD 2-Clause "Simplified" License
288 stars 94 forks source link

Hi,what‘s the number of the the lidar lines you are using? VLP64? #18

Open Zhang-Qinghui opened 5 years ago

koide3 commented 5 years ago

Hi @Zhang-Qinghui , We mainly use 32 and 16-line velodyne LIDARs (HDL32e and VLP16).

Zhang-Qinghui commented 5 years ago

yes,it is amazing. But why the piontcloud is so dense？

koide3 commented 5 years ago

If you are referring to the video on youtube, it's a standard 32line LIDAR, and we didn't do anything special.

Zhang-Qinghui commented 5 years ago

Thanks for your reply. I got it. I have try to run your code with my data bag from VLP16.But the result is not so good. What's your human body detection precision ?

koide3 commented 5 years ago

Some parameters are tuned for 32line, and you may need to tweak them for your sensor. Please see the launch file to find the params to be tuned.

Zhang-Qinghui commented 5 years ago

Thanks for your reply.I did not find the parameter in 'hdl_people_tracking_static.launch' or 'hdl_people_tracking.launch' to adjust the lines of velodyne. In addition, I can detect some human body without adjust params, but I can not detect it when the human body is fragmentary or the human body is sparse. I want to know the detection precision and detection distance of human body.Thank you in advance.

koide3 commented 4 years ago

Probably, the false negatives are caused by the classifier which was trained with 32line data. Try to set "enable_classification" to false to see if it works.

We have not evaluated the precision-vs-distance curve. I would be glad if you could do some tests to check it out.