lidar_euclidean_cluster_detect output topics are not transformed properly.

/label ~bug

Required information:

• Operating system and version: Ubuntu 18.04
• Autoware installation type: Source
• Autoware version or commit hash OAP7
• ROS distribution and version: Ros melodic
• ROS installation type: APT
• Package or library, if applicable: lidar_euclidean_cluster_detect

Description of the bug

lidar_euclidean_cluster_detect output topics are not fully transformed.

CloudCluster in /detection/lidar_detector/cloud_clusters

Transformed properly :

• min_point
• max_point
• avg_point
• centroid_point

Not transformed :

• convex_hull
• bounding_box

Because this message is not TFed properly the messages published in /detection/lidar_detector/objects are also wrong :

DetectedObject in /detection/lidar_detector/objects

Transformed properly :

• None

Not transformed :

• pose
• convex_hull

Steps to reproduce the bug

• Configure lidar_euclidean_cluster_detect output frame different from the input frame.
• Publish pointclouds to the /points_raw topic.
• Open rqt and notice that some data is not tfed.

Screenshots

The points are still in Lidar frame and not map frame, even if the header says so ! It really looks like a bug.

Additional information

I have corrected the bug. I can make a PR. One question though, what is the best way to TF a Point32 ? Because the TF listener has no transformPoint32 function.

Thanks !