IWR6843ISK: height of sensorPosition and boundayBox in Long Range People Detect demo

Thanks for reply,

To be sure, let me ask one more time.

The objects are within the defined boundary box. but when z-value of "sensorPostion" command is equal or larger than z-value of "boundaryBox" command, it isn't detected. Is that correct? 

I have an additional question.

I'm using 100m cfg of Long Range People Detect demo. 

When detecting outside, an object shaken by the wind is also allocated as an tracked object.

Could you please advise me how to filter these objects(non-people)?

Thanks.

Thank you for reply,

I read example file you recommended, but It isn't include how to mitigate False Detection. The example file says "by intelligently grouping and tracking the detected points, the software can successfully track the human without tracking the trees and shrubs". Is that means Gtrack software is already can track the human without tracking trees and shrubs?

In my test case, The object tracking the human remained on the points of the trees or sometimes objects are allocated due to points of the trees.

Below case is that object tracking the human remained on the points of the trees.

Initially there are only points of the trees.

Human pass by and the tracker starts tracking human.

Thanks for your reply,

Sudharshan KN said:

The tracker might retain the points for few frames before it can confirm to remove the objects from the scene. This could be the reason for having latencies.

Does that mean, the parameter of the 'stateParam' command must be set to 0 or 1 to immediately remove the object from the scene?

JINHYEONG LEE said:

I read example file you recommended, but It isn't include how to mitigate False Detection. The example file says "by intelligently grouping and tracking the detected points, the software can successfully track the human without tracking the trees and shrubs". Is that means Gtrack software is already can track the human without tracking trees and shrubs?

and please answer this question as well.

Thanks.

Thank you for reply,

Jackson Thomas said:

Also, in the top of the post you modify the boundary box, but in the GUI screenshots it looks like there are tracks created well outside of these bounds. Are you still using the boundary box when running this? This should help to filter some of these tracked objects.

There was an error in that test. The test was run with the Long Range People Detect 100m 2D example, but I adjusted the height values of the bounding box and sensor. I'll try run with 3D example to filter some of objects.

Jackson Thomas said:

There is no feature in gtrack right now to classify whether there is a human or tree/shrub. You may be able to eliminate some of these tracks by modifying the allocationParams or gatingParams. Please see the below document for more info on these.

Is it helpful to adjust gatingParams in cases(as mentioned above) where human-tracking objects remain at points in the tree, or sometimes objects are assigned due to points in the tree?