• State TI Thinks Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 2 answers
  • Subscribers 53 subscribers
  • Views 250 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

CCS/IWR6843ISK: Mapping logic between People count with Point cloud index in IWR6843ISK

Vivek Marimuthu
Prodigy 150 points
Part Number: IWR6843ISK

Tool/software: Code Composer Studio

Hi,

Currently, I am working on 68xx ISK with the industrial tool box 4.1.0. My requirement is to get the point cloud data of a person and I have to map to concern person accordingly.

I have flashed the pre build binary image from the lab and run oob_parser to collect the data.

Using the below parser code, I am dumping the people count and cloud point data into txt file.

And Every time I got Point Cloud[5] array, even if People count  0,1 or 2. Please see the samples below.

PointCloudDump.zip

 And in sometimes I am able to get proper values in both the indexes of PointCloud array. But not sure which index I have to take it.

Could you please help on the following?

1. How to map people index with point cloud index. People count {Y} = PointCloud[ x ]. How to relate X and Y.

2. How to validate PointCloud[Index] having valid data or junk data.

3. How to retrieve height of the target from the PointCloud [index].

4. Why I am getting PointCloud [5] array

Looking forward your response.

Regards,

Vivek  

 

  • 0
    TI__Mastermind 28055 points

    Hi Vivek,

    With the 3D People Counting lab, there is a static chain that runs in the background. This will always return detected points, but these are hidden by default in the visualizer. -> This is why you always get detected points.

    1. Each target has a Target ID (TID). This is transferred as the first value, so it is in targets[0].  Point Index TLV associates the points and the targets. It is a list of TIDs, point index[5] = 3 means that pointCloud[5] is associated with target ID 3. Target List and target index are sent one frame after the point cloud is sent, i.e. point cloud is frame N and target info is frame N-1, so you need to save the point cloud from the previous frame to properly associate them.
    2. You can run in debug mode, and check the point cloud index data to see that it matches the data you received from the device. I don't expect that you should have any junk data. Point Index values will be 254, 255, 256 is the point is not associated with a target.
    3. Take all the points associated with a track, find the highest, then average over 10 frames.
    4. This is the X,Y,Z, Velocity. SNR value of each point.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 150 points

    Hi Justin,

    Thanks for the detailed explanation. I am working on the parsing part based on your explanation.

    We will try out with multiple sample data and inform you

    Regards,

    Vivek

  • 0 in reply to Vivek Marimuthu
    Prodigy 150 points

    Hi Justin,

    As you suggested, I am following below steps for calculating height of the frame.

    • Get Point cloud data (x, y, z and velocity,snr) of each point
    • Point cloud of Frame N and Target info of N - 1 is mapped and got properly associated point cloud
    • To get the height point I am doing Z - Co ordinate sorting of Point cloud data of Nth frame.
    • Then with the Highest and lowest point I am calculating distance.

    I am having couple of queries here.

    1. As per my assumption  X axis is for Length, Y for Breadth and Z for height of a person. Thatsy I am doing sorting of point cloud data with Z axis data. Please correct me if I am wrong.

    2. Even I mapped the target index with point cloud properly, when I calculate the height of a person in a standing mode. The distance is not quite similar. It is varying a lot. Could you please guide what filter I am missing?

    Given below distance values captured for different frames

    Looking forward your response.

    Regards,

    Vivek

  • 0 in reply to Vivek Marimuthu
    TI__Mastermind 28055 points

    Hi Vivek,

    Do you have stored data from the frames where you get a very large value? You shouldn't be seeing any values that large.  I recommend saving the point cloud for a few frames, so you can see what data point is causing your value to be so large.

    Regards,

    Justin

  • 0 in reply to Justin Curewitz
    Prodigy 150 points

    Hi Justin,

    Thanks for the prompt response. Yes I have dumped the Point cloud data and attached here for your reference.

    But I just wanted to know, why sudden change in the point cloud data for particular frame?

    textfile_7.zip

  • 0 in reply to Vivek Marimuthu
    TI__Mastermind 28055 points

    Hi Vivek,

    I don't expect to see a major change in point cloud data. With the restrictions on point location in GTrack, any point with a large X, y, or Z should be ignored. Which frames in your file had the large values?

    Regards,

    Justin