Is there suggestion to mitigate unstable of heading of obstacle bounding box?

[Charlie715]

perception module as the heading of bounding boxes from lidar detector sometimes jitter (especially trucks), and the only refinement of heading in the recognition module and fusion module is simple moving average, is there any suggestion for the improving the precision and stability of heading of bounding boxes? Thanks.

[daohu527]

Can we replace the question with:

question

known conditions

• measurement heading itself is inaccurate
• objects's type, size, speed

we want

• stable heading

solution

Because the measurement is inaccurate, we need to filter outliers and smooth errors.

Use another method like compare the before and after frames to determine whether the heading has changed. And the initial orientation does not need to be so precise.