elevation and flatness conditions

[I3aer]

Hi!

flatness and elevation thresholds are updated iteratively. Therefore, they are set to zeros:

elevation_thr = {0, 0, 0, 0}; // threshold of elevation for each ring using in GLE. Those values are updated adaptively.
flatness_thr = {0, 0, 0, 0}; // threshold of flatness for each ring using in GLE. Those values are updated adaptively.

In the estimateGround member function, you compare the ground_elevation and ground_flatness variables with those thresholds using the concentric_idx:
if (concentric_idx < params_.num_rings_of_interest)
{
is_not_elevated = ground_elevation < params_.elevation_thr[concentric_idx];
is_flat = ground_flatness < params_.flatness_thr[concentric_idx];
}

I have two observations:

1. Since the z-axis is upward, the ground_elevation for ground planes should be always negative. Therefore, is_not_elevated is always True. On the other hand, if the least significant eigenvalue of a plane ground_flatness is positive, the is_flat is always False.
2. concentric_idx is increaded under the ring and sector loops. Then, the threshold arrays should have more than 4 values.

[LimHyungTae]

At first, yes, but aren't they updated at the next step?

[I3aer]

I did not check if they are updated or not as I am interested in a single scan pcd. Another question, why I get non orthogonal normals even though I set uprightness threshold to 0.99?

patch_normals
array([[-0.09144519, 0.05306055, 0.99439555],
[ 0.69917166, -0.604823 , 0.3812459 ],
[-0.01123331, 0.04285617, 0.9990182 ],
[-0.09687269, 0.05675028, 0.99367774],
[-0.10806715, 0.055059 , 0.9926178 ],
[-0.09962063, 0.06085207, 0.99316317],
[-0.06921859, 0.05507125, 0.9960804 ],
[-0.06099597, 0.05220991, 0.99677163],
[-0.04113085, 0.03393086, 0.99857754],
[-0.09267069, 0.04970804, 0.9944553 ],
[-0.09723388, 0.04760678, 0.9941223 ],
[-0.07620493, 0.06242752, 0.995136 ],
[-0.03467678, 0.0526633 , 0.99801 ],
[-0.04037486, 0.04701625, 0.9980778 ],
[-0.06198691, 0.06860244, 0.9957165 ],
[-0.08162168, 0.01962108, 0.99647033],
[-0.07868537, 0.0243757 , 0.99660146],
[-0.06772294, 0.02173795, 0.9974672 ],
[-0.00943837, 0.0398108 , 0.9991627 ],
[-0.01602176, 0.03286225, 0.9993314 ],
[-0.03886676, 0.06128451, 0.9973634 ],
[-0.07812789, -0.00297109, 0.99693906],
[-0.0880304 , 0.02805167, 0.99572283],
[-0.06525222, 0.04650887, 0.9967845 ],
[-0.6830666 , -0.7300232 , 0.02205181]], dtype=float32)

[LimHyungTae]

I guess those might be filtered out later after the prediction of all normals. Could you also share the exact part that you want to discuss?