search

stereolabs / zed-ros-wrapper

ROS wrapper for the ZED SDK
https://www.stereolabs.com/docs/ros/
MIT License
447 stars 391 forks source link

[Question] How to adjust the density of the point cloud? #558

Closed willzoe closed 4 years ago

willzoe commented 4 years ago

I want to use point cloud as input for local costmap.

Is it possible to downsample a point cloud by setting certain parameters?

Myzhar commented 4 years ago

Hi @willzoe you can downsample the pointcloud (without lowering the camera resolution) using the depth/depth_downsample_factor parameter in common.yaml: https://github.com/stereolabs/zed-ros-wrapper/blob/master/zed_wrapper/params/common.yaml#L44

willzoe commented 4 years ago

Thank you very much, I will try it.

Resample factor for depth data matrices [0.01,1.0] The SDK works with native data sizes, but publishes rescaled matrices (depth map, point cloud, ...)

  1. Does this mean that after modifying this parameter, not only the point cloud is downsampled, but the depth map is also downsampled?

I found that rtabmap_ros has a nodelet point_cloud_xyz, which can generate a point cloud from depth or disparity images and downsample it. There are two types of input, one is depth image, and the other is disparity image.

  1. For Zed cameras, which image is better(more accurate results and faster calculations) for input?
Myzhar commented 4 years ago

  1. Yes, using the depth/depth_downsample_factor you affect all the data structures involved in the depth estimation (depth map, disparity map, point cloud, Confidence map, ...)

  2. You can use both, but depth image does not require conversion from "pixel to depth". However I suggest to not use it since the ZED SDK provides the pointcloud and you add heavy calculations for something that you already have.

willzoe commented 4 years ago

  1. I use the 2.8.5 version of the sdk and its associated zed_wrapper. The parameter depth / depth_downsample_factor was not found in the file common.yaml.

  2. If I don’t subscribe to topics related to depth and point cloud, and only subscribe to the left and right images(/zed/zed_node/left/image_rect_color /zed/zed_node/left/image_rect_color) and their info(/zed/zed_node/left/camera_info /zed/zed_node/right/camera_info), will the ZED SDK continue to calculate and publish messages such as depth or point cloud that require a lot of computing resources?

Myzhar commented 4 years ago
  1. You are using a very old version. This parameter is available starting from v3.1
  2. To totally disable the depth module you must set depth_stabilization to 0 and quality to 0.
willzoe commented 4 years ago

  1. It's really old. I previously encountered some problems when deploying the turtlebot2 related package on melodic, so I changed to kinetic. So I had to use Ubuntu16 and jetpack3.3.

  2. Sorry, I did not make it clear just now. In addition to the four topics mentioned above, because nodelet point_cloud_xyz is used, I think I need to use /zed/zed_node/depth/camera_info and /zed/zed_node/depth/camera_info. In order to reduce the calculation, should I close the topics related to confidence, disparity, odom, point_cloud and pose?

Myzhar commented 4 years ago

If a topic is not subscribed the wrapper does not perform computations related to it, to reduce the required computational power simply do not subscribe to topics that you do not use. If you need /zed/zed_node/depth/camera_info you must subscribe to depth otherwise it will not be published... however the depth camera info is equal to the left camera camera info

willzoe commented 4 years ago

Sorry, I wrote it wrong ... What I need is /zed/zed_node/depth/camera_info and /zed/zed_node/depth/depth_registered.

Thanks for your answers~