Tested the performance in some more complex scenarios

[HowieHepburn]

First and foremost, I would like to extend my sincere gratitude and appreciation for your hard work and dedication in developing DipG-seg. This open-source ground segmentation method is a valuable contribution to the community, and I have been excited to test it out.

During my testing of DipG-seg for ground segmentation in an autonomous driving scenario, I encountered several issues that I hope you can help me address:

1)Vehicle Front and Other Vehicles Segmented as Ground
I noticed that the front part of the vehicle and parts of other vehicles are being incorrectly segmented as ground points. This misclassification is affecting the accuracy of the segmentation.

UrbanNav-dataset, with LiDAR HDL-32E, and I have defined NUSCENES in prijection_param.h

2)Distant Trees and Lawns Segmented as Ground:
Trees and grass areas at a certain distance are also being segmented as ground. This is causing a significant amount of non-ground points to be included in the ground segmentation.

3)Obstructed Areas Identified as Ground:
Sections that are obstructed by the LiDAR installation are being identified as non-ground points, which is not expected.

Could these issues be arising due to parameter configuration? If so, could you provide some guidance or suggestions on how to adjust the parameters to improve the segmentation accuracy? Any insights or tips on fine-tuning the parameters for better performance would be greatly appreciated.

Thank you once again for your amazing work and for providing this tool to the community. I look forward to your response and any advice you can offer.