search

VincentCheungM / Run_based_segmentation

An ongoing implementation ros node on `fast segmentation of 3d point clouds: a paradigm`...
MIT License
389 stars 121 forks source link

Unable to view scan line run clustering? #6

Open chowkamlee81 opened 5 years ago

chowkamlee81 commented 5 years ago

Hai,

I got simulated ros codes ground plane fit and scanline codes. All the topics are published but i don't know how to view the results... I already installed visualization ros toolkit. Kindly help us....

VincentCheungM commented 5 years ago

Hi, for visualization you should launch Rviz and subscribe the corresponding topics. The following may help you work it out, in case you are not familiar with ROS. Before running Rviz, you should have the ROS master process run first, which can be done by typing

roscore

After that, you can launch the Rviz for visualization by typing

rosrun rviz rviz

in another terminal. After that, you can see an Add button in the view, and just press it to choose the topics you are interested topics through somewhat By topics.

chowkamlee81 commented 5 years ago

Thanks for your reply. I got the results of ground plane and non-ground plane in RVIZ.

But to view the results of segmentation i subscribed "/all_points" topic in RVIZ. Iam unable to observe the results as it is just the same input pointcloud.

Am i missing anything. KIndly let us know.

VincentCheungM commented 5 years ago

The clustered point cloud object is publish as topic /slr in default. see

kxhit commented 5 years ago

@VincentCheungM Thanks for opening source code! I run the codes with the example.bag but can't get the same performance as the video shows. Is there any parameters that needs to be changed? Here is the result in rviz: slr_example1 slr_example2

Waiting for your reply! Thanks!

VincentCheungM commented 5 years ago

@kxhit @chowkamlee81 Hi, sorry for replying you so late. Each point was given a cluster ID as label value, and RVIZ can render that property by selecting label (maybe it's intensity in default). And as there are too many clusters like 2000 or more (it can be seen as 'Max intensity' on RViz pannel), and it's hard to tell a difference between cluster ID 1 and cluster ID 2 by your eyes, after those 2000 or more clusters are scaled. So they are simply looks the same, and in red.. While talking about the img in readme, it was a screen shoot captured by pcl_viewer. I just simply viewing one of the segmented pcd file, and render it by its channel.
And as this repo it's just some toy codes, you could add some bounding boxes for each cluster for helping you guys have better visualization experience. I've applied and simplified a version of object builder in apollo, it looks good. As there are some license problems, I can't provide it here. Maybe you could refer to apollo, I guess you have done, or this lidar perception repo from my friend @Durant35.