Matlab/OpenCV stereo vision distance measurement is not accurate -
after stereo calibration, when run matlab example stereo depth estimation (sde), distances wrong: @ 2 meters, reports distance less 1m. , 3d scene reconstruction looks cone-shaped instead of real scene. disparity map noisy (non-smooth), resembles scene.
if 'feed' sde script example file instead of webcam input, runs okay, looks great; when feed 2 webcams ( 'logitech hd pro webcam c920' ) that's when above bad results, beginning rough disparity map.
i've tried many different calibration attempts few images 60, matlab's checkerboard pattern @ different angles (never > 45) , distance cameras 8 20'. camera lenses 3.8175" apart always, , mounted top edge of laptop. followed matlab's recommended workflow.
what doing wrong in calibration?
matlab r2015a. laptop windows 7 64-bit checkerboard pattern 37" x 27"
............just discovered problem: creating disparity map this: disparitymap = disparity(frameleftgray, framerightgray); however, camera #1 on right, , matlab says default disparity range [0 64] , cam #1 right should [-128 0], changes disparity map uniform blue.
i got working. (1) left/right of calibration , images , detection data structures must match. (2) use mm checkboard square size. inches causes malfunction, because else in mm.
Comments
Post a Comment