• State Resolved
  • Locked Locked
  • Replies 2 replies
  • Subscribers 30 subscribers
  • Views 554 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR1642BOOST: I'd like to understand the relation between Tracking and clustering of lab0011-pplcount .

user5205609
Intellectual 960 points
Part Number: IWR1642BOOST

Hi,

I use IWR1642BOOST(ES1.0),with Tool Box 2.3.0 and SDK 1.2.0.5.
I get source codes of lab0011-pplcount (dss and mss) by CCS 8.xx.

I'd like to understand the relation between tracking and clustering ,
and the timing of processing these.

I check The pplcount_customization_guide.pdf and source codes.
But these explanation of processing order are difficult to understand.

[1].Consideration from project source code
 (1).task_mbox.c
  It set datas input from dss via mailbox to works for task_app.

(2).task_app.c
　It executes tracking with gtrack_moduleStep ().

 Since the clustering API is included in dss.
  It looks like clustering on dss side before tracking on mss side.


[2].Consideration from pplcount_customization_guide.pdf.
  I referred to the following.
   Page 15 :"Allocation Parameters"
    "Points that don’t get associated are subjects for the allocation decision.
     Each candidate point is clustered into an allocation set. "
  Page 22 : "Tuning Example – Increase Range to 12 m"
   "By lowering the Det2Active threshold, the tracker will classify
   a detected cluster as human after less frames have transpried,
   meaning an approaching person will be detected farther away"

 Is clustering performed after tracking?
 Then, where is the processing timing on the source code?
 Dss(call clustering api) will be executed before mss(call gtrack for tracking)?

Regards,
user5205609

  • 0
    TI__Mastermind 28055 points

    Hi Hanako,

    For a more detailed explanation of the tracker, please see our TI Design. 

    The tracker has 4 main steps:

    1. Update position of tracked people based on information in Kalman filter
    2. Associate points in the point cloud with the tracked people
    3. Attempt to cluster remaining points - clusters that meet allocation thresholds (set by allocation parameters) will be added to list of tracked people
    4. update state of tracked people in Kalman filter

    You can see that clustering and tracking both happen in the MSS. Clustering happens in two stages. The first stage is the second step above, where points are grouped with already tracked objects.  The second stage is step 3 above where points are clustered in an attempt to find new objects to track.

    The DSS is responsible for generating the point cloud, which it then sends to the MSS, which is responsible for tracking and clustering, as well as sending data to the PC.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Intellectual 960 points

    Hi,

    Thank you very much.


    I read the design guide.
    It seems that there are many things to study, so I will try it carefully.


    Regards,
    user5205609