4. Conclusions
With the adaptive segmentation procedure developed in this
research, we can effectively cope with the problem of grey
level gradient in the X-ray images due to shape or uneven
thickness of fruit for most of tested cases. However, the infestation
sites could not be segmented using the traditional
standard thresholding algorithm. Consequently, variations in
fruit characteristics and extent of infestation may be dealt
with by adjusting the iteration number of morphological filtering
or the sub-image size for adaptive thresholding. The
algorithm is fast in computation time and was implemented
in the X-ray scanner for real-time quarantine inspection at
a scanning rate of 1.2 m/min. Suspected sites of infestation
inside fruit can be accurately marked on the acquired X-ray
image to aid the quarantine officer during inspection. For
different kinds of fruits, initial tests and choices of operation
parameters need to be determined, including number of
iterations for morphological filtering, operational sub-image
size, and interpolation grid size. Once the parameters are
configured, they are saved and used in subsequent quarantine
inspection. The effect of operational parameters was also
examined by comparing the computation time and threshold
value map using different combinations of parameters.
Experimental results revealed that the effect of sub-image
size and interpolation grid size has little effect on the computational
time when the interpolation grid size is greater
than 8×8 pixels. The adaptive thresholding algorithm is
stable judging from the insignificant difference of thresh old value maps created using various sub-image size and
interpolation grid size. The algorithm resolves the frequent
problem of segmenting object from X-ray image using global
thresholding approach.Without substantial modification, the
algorithm can be applied to X-ray inspection of various
agricultural products having uneven thickness and variable
densities