The objective of this research was to develop a method
for segmenting rice seedlings from background in a multispectral
image of paddy field. Multispectral images of paddy field were
taken by a high resolution multispectral camera. The DVI
(Difference vegetation index) image which could reduce the
image noises was suitable for seedlings recognition by analyzing
spectral characteristics of the objects in near infrared, red, and
green wavebands. The difference vegetation index was calculated
by subtracting a red image from a near infrared image and the
segmentation threshold was confirmed based on fuzzy C
clustering arithmetic. 212 images were acquired to do validated
test and 210 images got satisfied segmentation results. The
accuracy of image segmentation was above 99%.