Point Cloud Optimization Producing Hazy Results on Custom Dataset

[sahith-chada]

Hi Team,

First of all, thank you for your incredible work on this project!

I am attempting to use your optimization framework on my own dataset, but I’ve encountered an issue where the optimized point cloud appears hazier compared to the initial input.

Details about my setup:

Sensors:

Velodyne LiDAR- 128 channel OXTS Global Positioning System

Parameters Used:

max_iter = 30  
downsample_size = 0.05  
voxel_size = 4.0  
eigen_ratio = 0.1  
reject_ratio = 0.05  
WIN_SIZE = 15  
GAP = 5  
layer_limit = 2  

Attached Resources:

Input point clouds and poses and Mapped output (before and after optimization)

Screenshots of the maps (before and after optimization): Before optimization:

After optimization:

Problem Description:

After running the optimization, the resulting point cloud appears significantly hazier and less defined than the original. I expected the optimization to improve clarity and accuracy, but it seems to have the opposite effect.

Could you please help me identify what might be causing this issue? Are there specific parameter adjustments or dataset preprocessing steps that could address this?

Thank you in advance for your guidance!