First of all, thank you for being able to share the excellent work.
The performance results were poor, when I tested the data from my experimental platform using VGCIP, the first picture is the Horizon radar test result and the second picture is the VLP-16 result. The vehicle is moving forward at near constant speed. Attitude estimation often jumped in continuous motion and the elevation error was quite large. This is not as good as the results of some of the current purely geometric class methods. Can you provide some advice on how to use this in practice.
In addition, the size of the voxel downsampling will seriously affect the time consumption of the algorithm. Is there any way to improve that?
First of all, thank you for being able to share the excellent work. The performance results were poor, when I tested the data from my experimental platform using VGCIP, the first picture is the Horizon radar test result and the second picture is the VLP-16 result. The vehicle is moving forward at near constant speed. Attitude estimation often jumped in continuous motion and the elevation error was quite large. This is not as good as the results of some of the current purely geometric class methods. Can you provide some advice on how to use this in practice.
In addition, the size of the voxel downsampling will seriously affect the time consumption of the algorithm. Is there any way to improve that?