The bounding box is unstable

[BigRayLee]

when I run the code with my rosbag, the orientation of bounding box is changing sometimes? Have you encountered a similar problem?

[Durant35]

Is under static scenario? Similar problem, but I think the source is on segmentation.

[BigRayLee]

I think it is under static scenario, firstly I think the smallest area can represent the true orientation of the obstacle, but eventually I find that it may not be the same.

[BigRayLee]

I used the method proposed in paper（ An Orientation Corrected Bounding Box Fit Based on the Convex Hull under Real Time Constraints） can achieve the correction of the bounding box direction.

[Durant35]

Just make some improvement upon this repo's implementation? Hope to learning from you, great!

[BigRayLee]

I used the method in the paper to get the direction of the bounding box and correct the direction of the minimum area bounding box, but now the algorithm is not fast enough.😂

[Durant35]

Never mind, good job! You can optimize it.

[BigRayLee]

Thank you!

[BigRayLee]

I want to see the boundingbox in rviz, what kind of message shoube used to publish the boundingbox to rviz? Thank you!

I think you can use this one: jsk_recognition_msgs/BoundingBoxArray

[BigRayLee]

I have tried to use the message type you mentioned, but that message does not have the orientation information, I mean the boundingbox of that message is always parallel to the x y axis.

Oh, I know what you mean. We just create a message for visualize the orientation of bounding box. You can try this one: vision_msgs/BoundingBox3D,it includes the 3D position and orientation of the bounding box center.