Professional Documents
Culture Documents
Abstract: More selective methods for applying If a plant or row of plants can be located in a
agricultural herbicides on fields can result in field using vision then it is possible for a tractor
substantial cost savings. Three image or other vehicle to automatically steer itself
processing methods were tested for their along the row and free the driver to do other
ability to identify four different images of tasks. The ability to detect weeds from within
plant species. First two images were different crops could lead to the possibility of localized
and the other two were similar. The images spraying of herbicides thus reducing chemical
are preprocessed by segmentation and spatial waste, crop damage and environmental
filtering using the Color Chromaticity Chart. pollution.
The test results provide evidence that texture
based methods can provide a useful metric for The general objective of this research is to
distinguishingbetween some species of plants. investigate the potential of using image
processing techniques on plants spatial
I. INTRODUCTION characteristics as an alternative or supplemental
plant identification method. It also evaluates
0 1995 IEEE
IEEE CAT, NO. 95CH3581-6/0-7803-2741-1/95/$3.00
404 EEE WESCANEX '95 PROCEEDINGS
the color cromaticity chart. The plant color is black and white image in our case was generated
assumed green. The equation for separating the from color RGB image.
green color information from the image was
developed as follows. Using a Fast Fourier Transform Annular
Band Approach &e input image is converted
Threshold limit = 0.83R + Intercept (1) from spatial domain to frequency domain. This
frequency domain image is analyzed for plant
The characteristics to differentiate one color identification. The advantage of using Annular
from another are brightness, saturation, hue. Band Approach is rotational invariance. This
Hue and saturation together are called was important as the plant are not in a fix
chromaticity. The color chromaticity diagram position [3]. After transforming the image in
shows color composition as a function of R, G frequency domain, the power spectrum inside
and B plane [3].As green color is predominant the annular band of radius R1 to R2 is
in the upper part of the chromaticity chart the calculated. The center of the image is the center
equation of the line will separate green plant point of the annular band. This user defined
from the rest of the image by varying Y band keeps on sliding towards the edge of the
intercept. The original images are as shown in image. A plot of power capsulated in the
Fig. 4, Fig. 5, and Fig. 6. The segmented image annular band verses radius is plotted. The
is shown in Fig. 7. relative frequency is observed. This relative
frequency forms the basis of identification plants
The use of spatial masks for image processing [ 5 ] . A graph power spectrum verses radius was
usually is called spatial filtering and the masks plotted. One way to analyze spatial variations is
themselves are called spatial filters. Smoothing the decomposition of an image functions into a
filters are used for blurring and for noise set orthogonal functions, one such set being
reduction. Blurring is used for removal of small fourier (sinusoidal) functions. The fourier
details from an image prior to object extraction. transform may be used to transform the intensity
The extra information present in the image like image into the domain of spatial frequency Fig.
ground, stones has to be removed as it may lead 9.
us to wrong results fig. 8. One of the principal
difficulties of the smoothening method discussed Let f(x) be a continuous function of a real
is that it blurs edges and other sharp details. The variable x. The fourier transform of f(x),
method used is particularly effective when the denoted by f[f(x)], is defined by
noise pattern is due to small pebbles, soil
patches with similar color with plant that
consists of strong spike like component. The
f [(f(x))]= F(u)= Jz f(x)'-'2*ux1dx (2)
where j=d- 1
characteristics to be preserved is texture. The
window size and threshold level is selected by
If a texture is at all spatially periodic or
the user. If more information in terms of gray directional, its power spectrum will tend to have
levels is present in the window then the window peaks for corresponding spatial frequencies.
is replaced by the texture information from the These peaks can form the basis of features of a
black and white input image. This method filters pattern recognition discriminator. In this
noise or scattered patches created due to stones method the fourier space is partitioned in to
on the ground. Software for segmentation, radial bins. These bins together with fourier
spatial filter, and thresholding was developed power spectrum are used to define features. if F
using C language. is the fourier transform, the fourier power
spectrum is given by IF12. Radial features are
given by
III. Frequency Domain Method
Vrlr2= fJ IF(u,v)12}dudv (3)
The fourier transform is a single-band
processes. The 3-band color images needed to be
Where the limits of integration are defined by r2
adapted by either selecting only one of the
IF(u,v)l*dudv,r2 5u2, v2< rz, 0 l u .
original
- red, green or blue (R, G, or B) bands, or
generating a forth "band". Fourth band which is
SHULIN, D.er al.: IMAGE PROCESSING METHODS FOR IDENTIFYING SPECIES OF PLANTS 40.5
Where [r1,r2]is one of the radial bin and V is The model changes only when surfaces are
the vector defined by different values of r l and considered.
r2. Radial features are co-related with texture
coarseness. A smooth texture will have high S(e) = he"-D' (4)
values of Vrlrz for small radii, where-as a Where S(e) = Surface area. D= 2.00 implies a
coarse, grainy texture will tend to have smooth planer surface and Dz2.99 very rough
relatively higher values for larger radii. 3D surface.
CO-relation
Fig. 6 Plant3
l:r-.-
Fig. 7 Segmented plant 1 Fig 8 Filtered Plantl Fig. 9 FFT Plantl
0" 150
100 Series1
50
0
0 io 20 30 40 50 M) 70 Bo 90 100 110 120
RadlUS
Radius