This is to reduce failure cases, caused by noisy point clouds.
In particular, we experienced that sometimes the disparity map of the selected blob contains some noisy sub-blobs (usually belonging to the scene background). The filter currently implemented is not enough for detecting this type of failure, since it requires an a-priori tuning of the parameters.
We need to find a new filtering approach to be overlapped on the old filter.
pattacini
commented
7 years ago
This is to reduce failure cases, caused by noisy point clouds.
In particular, we experienced that sometimes the disparity map of the selected blob contains some noisy sub-blobs (usually belonging to the scene background). The filter currently implemented is not enough for detecting this type of failure, since it requires an a-priori tuning of the parameters.
We need to find a new filtering approach to be overlapped on the old filter.