Find the 3-D position of objects using two cameras. The first part, following Machicoane et al. [1], enables the mapping of camera pixel locations to optical paths in 3-D space, given a set of calibration images that can be obtained by moving a planar calibration target along an axis. The second part enables the "matching" of objects between two 2-D views, yielding their 3-D position, by finding the near-intersection of the optical paths corresponding to the detected object in each view.
This implementation approximates the optical paths as straight lines, so it cannot be used for systems in which there are changes in the index of refraction within the measurement volume. (Changes in the index of refraction between the measurement volume and the cameras, however, are not an issue.)
from stereo.camera import camera
from stereo.stereo_system import StereoSystem, Matcher
# load camera calibration information that can be created using the calibration GUI
calib_A = camera.calibration_from_dict('camera_A.pkl')
calib_B = camera.calibration_from_dict('camera_B.pkl')
# create a "stereo system" comprised of the two cameras
ss = StereoSystem(calib_A,calib_B)
# load DataFrames giving the location and size of 2-D objects you've detected in each view
df_A, df_B = your_function_to_load_your_2D_data()
# match detected objects
m = Matcher(df_A,df_B,ss)
df_3d = Matcher.match()The calibration procedure described in [1] involves taking images of a calibration plate as it is traversed through the measurement domain. Here, we define the plate to be parallel to the x-z plane, so the orthogonal direction in which it is traversed is the y direction. The points marked on the calibration plate should lie on rows corresponding to the x direction and columns corresponding to the z direction.
After obtaining images of a calibration target obtained following [1], save them as .tif files and place them in directories BASE_FOLDER\CAMERA_NAME_1\ and BASE_FOLDER\CAMERA_NAME_2\, each named y_Ymm.tif, where Y is the y-position of the plate in that image, in mm.
Then, call the calibration GUI in stereo/camera/calibration_gui.py once for each camera. The GUI facilitates the construction of the calibration files for each camera that are used in initializing a StereoSystem.
Download the code to your computer, then run pip install -e . .