The invention discloses a
magneto-optic
image enhancement method based on an
optical flow algorithm. Low-frequency AC excitation is used for excitation of a
magnet yoke, an image set is obtained under different
signal intensities through a periodic variable
magnetic field, and a low-frequency
signal enables magnetic
force lines to pass through a deeper region of a detected object, thereby enabling a
magneto-optic image to reflect the defect information of the deeper region, and reducing the
impact from the
skin effect. Meanwhile, the multi-point sampling and photographing (collection) can be carried out in one period, thereby increasing the
data information of the
magneto-optic image. However, the magneto-optic image obtained through the above method has the interference
noise caused by
magnetic domain movement under the action of a low-frequency AC
signal, and is liable to be mixed with the defect information, thereby reducing the
image detection information. Therefore, the method employs a
pyramid optical flow algorithm for the iteration
processing of the magneto-optic image, achieves the accumulation of feature points generating
optical flow motion, and obtains an optical flow enhanced image. The enhanced image is more apparent in
image gradient changes, greatly reduces the interference, and is more visual in defect information.