The invention discloses a method of detecting fruit surface defects based on
low pass filter. Binary
mask image B' is constructed based on B component of fruit colorful image. R component f(x, y) of removed background is obtained by the dot product of B' and R component. Central transformation of f(x, y) is performed to obtain f'(x, y) and discrete Fourier transformation of f'(x, y) is performed to obtain F(u, v).
Low frequency component G(u, v) is obtained by the multiplication of F(u, v) and
low pass filter H(u, v). g(x, y) is obtained by negative discrete Fourier transformation of G(u, v). i'(x, y), which is the
surface brightness image of f(x, y), is obtained by the multiplication of g(x, y) and the function of (-1)x + y. Homogenization brightness image f''(x, y) is obtained by the division of f(x, y) and i'(x, y). Fruit surface defects are detected by single threshold method of f''(x, y). Fruit surface defects are easily detected in the invention by single threshold method, wherein homogenization correction of fruit
surface brightness is carried out. The method of the invention overcomes the complexity of conventional defect inspection
algorithm and is free of considering the sizes and shapes of fruits to be detected, thus is better adjusted than standard ball brightness adjustment method. The images are acquired by a
single camera, thereby avoiding the problem of low
processing speed due to the
information fusion of images from a plurality of cameras.