how to improve reconstruction precision

[dbdxnuliba]

When asking for help or submitting a bug report, please provide the following important information about your system:

• Operating system version (for example, Ubuntu 18.04 LTS)
• ROS version (for example, ROS Melodic)
• CUDA version (for example, CUDA 11.4)
• Nvidia driver version (for example, Driver 470.141)
• _Graphics card type (for example, Nvidia GeForceGTX 1650)

this is the outpu ply file by

I use franka +L515 +yak_ros with two pc , one of them run franka_ros and L515 and another pc run yak_ros ,subscribe /camera/depth/image_rect_raw,

and this is the real environment:

and this is the launch file <?xml version="1.0"?>

[457.046875, 0.0, 330.37109375, 0.0, 456.90625, 249.544921875, 0.0, 0.0, 1.0]

we can see many large holes and in the surface after resconstruction, and plane becomes curved can you give me some advice to improve the reconstruction precision

and quesiont2: there are so many faces in the ply file ,how can we combine them into a few surfaces

[Levi-Armstrong]

Increase the tsdf resolution from 0.002 to something larger that the sensor depth noise. If this is smaller that the expected noise in the depth then the noise decays seen areas resulting in holes in the data. Also this is directly propositional to the mesh resolution since it use marching cubes using the tsdf to generate the mesh.