A fast algorithm for comupting the normal of the depth img. And warp it with pybind11. Modified from the code in line 702 of PCL linemod package and patch_linemod.
- opencv3
- numpy
- pybind11:
pip3 install "pybind11[global]" # Or install from source code: git clone https://github.com/pybind/pybind11.git cd pybind11 && mkdir build && cd build && cmake .. sudo make install pip3 install pybind11 --user
In bash:
cd normalSpeed
python3 setup.py install --useror
cd normalSpeed
python3 setup.py build_ext --inplace
cd .. && cp normalSpeed/normalSpeed.* ./In python:
"""
The coordinate of depth and normal is in cv coordinate:
- x is horizontal
- y is down (to align to the actual pixel coordinates used in digital images)
- right-handed: positive z look-at direction
"""
fx = 1736.0597
fy = 1740.7188
k_size = 5 # pixel, kernal size for computing the normal
distance_threshold = 2000 # mm,
# Ignore pixels beyond this distance.
difference_threshold = 20 # mm, When computing normals,
# ignore contributions of pixels
# whose depth difference with the
# central pixel is above this threshold.
point_into_surface = False # Regular normal point out from the mesh
# surface
# depth - np.array, h x w depth image, in mm
normals_map = normalSpeed.depth_normal(
depth, fx, fy, k_size, distance_threshold, difference_threshold, point_into_surface
)python3 example.pyYou will get the normal map of a depth map as follows:
- Input depth map:
- Output normal map:
