Shape Evaluation by Digital Camera for Grape Leaf

Document Sample
Shape Evaluation by Digital Camera for Grape Leaf Powered By Docstoc
					      Shape Evaluation by Digital Camera for Grape Leaf
                 Hiroya Kondou , Hatuyoshi Kitamura and Yutaka Nishikawa
                        Iga Agriculture Laboratory, Agricultural Research Division,
                  Science and Technology Promotion Center, Mie Prefectural Government
                                          Yoshitaka Motonaga
                                 Faculty of Agriculture, Niigata University,
                                           Atsushi Hashimoto
         Kenichi Nakanishi and Takaharu Kameoka,Faculty of Bioresources, Mie University

Abstract: In the production field of the agriculture, excellent farmers precisely catch the change of the
crops in the growing process and they manage the cultivation in proportion to the change in order to
cultivate the agricultural products of high quality. Since sensing the delicate change of crops is acquired
through the observation by the visual sense in their long cultivation experience, it is difficult for them to
transmit the understood technique to future generations as a general cultivation one.
     In this study, using the grapes as the experimental materials, it was examined a new technique for
getting the information on the tree growth using a digital camera instead of the human eyes. We studied the
nutrient conditions by the image analysis of the grape leaf. Because the shape and the color belong to the
sensuous information, it is difficult to analyze these information and to determinate a standard value. In this
experiment, the outline data of the leaf was analyzed in the polar coordinate so that the character of the
shape was quantitatively classified. In conclusion, it was shown that the possibility of quantifying the
important factors in the cultivation, which farmers usually acquired visually from grape tree.

Keywords: grape; cultivation; digital camera

      Grapes have the highest production among fruits in the world, and most of those are produced as an
ingredient of a wine. In Japan, most of them are produced as table varieties, and the big berries exceeding
12 g are popular among consumers as the high-grade fruits. The cultivation of the varieties with big berries
seeded is one of the most advanced techniques for the tree fruit production in Japan. The representative
variety of large-berries is 'Kyohou', which has a black skin, and 'Aki Queen' (Yamane et al., 1992), which
resulted from the selfing of 'Kyohou', has been recently interested as a high-grade grape with red skin.
      We made the fruit color chart of the 'Aki Queen' with the technique of the digital image (Kondou et al.,
1998), (Motonaga et al., 1998). In this color chart, each coloring stage is expressed by the corresponding
color sample which displays a standard berry shape and the texture information of the skin. In this
experiment of making the color chart for fruits of 'Aki Queen', the acquisition of a berry image was
carried out by digital camera in the laboratory because the influence of the surrounding light condition
should be removed.
      Digital camera has become very popular among us because of its low price and high performance.
This situation is causing the expectation of the researchers in the agricultural field that the skin color
assessment in the natural outdoor scenes could be made by digital camera. Generally speaking, however,
digital camera has still many problems to be used as a color measuring equipment in the natural outdoor
scenes where the object color is governed by the surrounding sun light condition. Therefore, the calibration
method to utilize digital camera as color assessment equipment in the natural outdoor scenes was developed.
Then we applied this method to the color evaluation of grape fruits using a digital camera in order to
quantify the coloring change with the growing of grape tree, which has been judged by visual sense of
cultivator (Kondou et al., 2000), (Hashimoto et al., 2001).
      Control of tree vigor is very important to keep a stable cultivation for the both. Therefore it is
necessary to take an accurate grasp of the tree growth change and to make a suitable cultivation
management for the change. Cultivator has been learning to sense a delicate change of the plants through
his ripe experience, which is mainly based on visual observation. So far the information visually observed
by the experienced cultivator could not have been clarified because it is difficult to quantitatively evaluate
the information such as the color and shape.
      The purpose of this work is to propose how to
utilize a popular digital camera for shape evaluation and
to quantify the shape of grape leaf blade, because the
morphological information such as width of petiolar                      width
sinus and length of sinus reflects the grape tree vigor.

Experiments and materials                                                               length

Acquisition of Leaf Image
      We have made examinations in the Iga Agriculture
Laboratory, the adult grape tree of 'Kyohou' was used for
the examination. Four trees of grape 'Kyohou' in the rain
cut house were provided. Leaves were sampled five
times, namely before flowering (May 18, 2000), after the      Fig. 1 Measurement part of grape leaf
physiological fruit drop (June 22), at the early coloring
stage (July 26), at the harvest time (September 7) and at
the defoliation time (October 22). Six to nine new shoots
growing from terminal bud and proximal bud of bearing
base branch were sampled from each tree.
      After all leaves were cut off from the new shoots,
petiole was removed from each leaf blade. The images of
144dpi were acquired with a digital image scanner
(GT-5500, Seiko Epson). The representative length and
width of each leaf were measured at the same time, as
shown in Fig. 1.
      On May 22nd, 2001, the leaf blades were sampled
from the 5 new branches cut off each of the 4 trees
‘         .
 Kyohou’The images of the leaf blades were acquired
with an image scanner(resolution : 144 dpi ) and a digital
camera( picture element number : 1440 × 960 shutter
speed : 1/22 Restriction : 6 ) in the laboratory. Fig. 2
showed the schematic diagram of image acquisition             Fig. 2 Setup of image acquisition by
system with a fluorescent light (TRUE-LITE                           digital camera
DURO-TEST, Co., Ltd.) and a digital camera (FinePix S1
Pro, Fuji Photo Film Co., Ltd.). The light sources for the
illumination, which had 5500 K color temperature, were located both sides of the sample. The light was
scattered diffusely from the diffuse reflectors placed over
the light sources in order to prevent the specula reflection
at the sample surface. Furthermore, the sample was
enclosed with the diffuse box, which was made from white
paper. Length and width of each leaf blade were measured
before image capture.                                                               φ
Method of Processing Leaf Shape Image                                                  r
    Leaf shape was analyzed by using the shape analytical
tool mounted on the Agroinformatic Management System                                 O
(Motonaga et al., 1998, 2001). This tool is based on the
r-- coordinate system in Fig. 3. The origin was set on
the center of gravity of the target shape. A vector from the
origin to a point on the curve was auxiliary drawn as an
arrow. Let ,  and f be radius, angle of the vector and
                                                                       Fig. 3 r-- coordinate system
angle between the vector and tangent line at the point
along the curve. Then the curve can be projected into the
coordinate system with parameters,  and . This matter
signifies that - data have only the shape information
without the size one. In addition, it is becomes possible that the calculation of the average shape by finding

the average value of  and  which are different sizes.
     The relationships between the x-y coordinate system and the r-θ-φ coordinate system are shown as
following equations.
      x  r  cos θ                                                                   (1)
      y  r  sin θ                                                                   (2)
               tan( φ            θ )                                                                       (3)
By differentiation of Eqs.(1) and (2), and substitute these into Eq.(3)
(      tanφ  r ){1  tanθ  tan( θ)}  0
                                φ                                                                                 (4)
Eq.(4) has some analytical solutions. Representative solutions were shown as follows.
     If =/2, the shape is a circle whose center is on the center of gravity. If  is constant, the shape is a
logarithm spiral. If= and , the shape is a straight line. If and , the shape is a circle
of which the center is on the circumference. If  and 2, the shape is a hyperbola. If and
1/2 and 0, the shape is a parabola.
     Since this tool also has a sophisticated human interface displayed in Fig. 4, a pretreatment, removing

                                    Fig. 4 Interface of shape analytic tool
background, threshold and outline extraction, is easy to proceed with the menu button. The shape data of
each sample on - coordinate system can be calculated
automatically. Moreover it has the function of consecutive       200
processing of many images loaded at one time.
                                                                                                            Intermediate Type
                                                                 Number of Images

Results and discussion
Morphological Analysis of Leaf
     The leaves were classified into three categories based                          50
on the L/W value, the ratio of the leaf length (L) to the                                 Wide Type
                                                                                                                   Long Type
width (W). Fig. 5 shows the histogram of the values for
987 leaf blade images. Since the value of the wide type                               0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
leaf was lower than 0.65, the long type one higher than                                               L / W [-]
0.95, the rest were treated as the intermediate type one.
                                                                 Fig. 5 The histogram of the L/W value for
                                                                        987 leaf blade images.
                                                                        L/W value means the ratio of the leaf
                                                                        length to the leaf width.
                                                                                                                               3.5                                                     Wide type


                                              -4     -3        -2        -1       0   1       2   3   4                              -4   -3   -2   -1    0    1   2   3   4
                                                                             θ[rad]                                                                  θ[rad]

                                                                                                                                                                                     Intermediate type
                                         3.5                                                                                   3.5
                                         3.0                                                                                   3.0
                          φ[rad]                                                                                               2.5

                                               -4        -3        -2    -1       0       1   2   3   4                              -4   -3   -2   -1    0    1   2   3   4
                                                                             θ[rad]                                                                  θ[rad]


                                                                                                                               3.0                                                        Long type

                                   0.5                                                                                         0.5
                                   0.0                                                                                         0.0

                                         -4         -3        -2        -1    0       1       2   3   4                              -4   -3   -2   -1    0    1   2   3   4
                                                                         θ[rad]                                                                      θ[rad]

            Coordinate                                                                                    Averaging                                       Inverse transformation
                                                Fig. 6 Processing of shape analysis on grape leaf

     The leaf images were processed using the shape analytical tool as follows (Fig. 6). Transformation
from orthogonal coordinate system to the r-- one was done after outline image of leaf was smoothed with
10% of total outline image pixels. The f values around the leaf top, leaf sinus and petiolar sinus have
significant fluctuations for all the three types. The fluctuation for the long type was the most significant and
that for the wide-type was negligible. Ten - data of 10 leaves for each type were equalized on the -
coordinate system. Then the inverse transformation was performed to obtain the averaged shape for each
type. Each averaged image can be obviously distinguishable each other to classify into three types: the long
type, the intermediate type and wide type. The averaged image data well expressed the shape characteristics
of each type of leaf, as shown in Fig. 6.

Feature Extraction of Leaf Shape
     A relationship between L/W value and fluctuation
in the vicinity of =0 was found when the image of the                                                                                                                          
                                                                                                                                                    3.0                                         sinΘ+π/2
leaf blade was compared with - data. In short, the
fluctuation is the most significant and that for the wide                                                                                           2.5
type     is    negligible.   Then,         is    defined

as   |   sin    / 2  |d
                                                                                             (5)                                                   1.5
and  is also defined in Fig. 7 graphically. When the                                                                                               1.0
relation between L/W value and  was examined for                                                                                                   0.5
each 5 samples of the wide type leaf blade image (L/W
value:0.62 ), intermediate type (L/W value:0.87 ), long                                                                                                        -3          -2   -1    0     1    2     3
type (L/W value:0.98 ), as shown in Fig. 8, the positive                                                                                                                         Θ[rad]
correlation was confirmed. Therefore the methodology
provided here is applicable to extract morphological
characteristics of leaf and it could analyze the leaf shape                                                                                                   Fig. 7 Calculation of 

Image Acquisition by Digital Camera

      Leaf blade images acquired in both of digital
camera and image scanner were classified into                  1.0
wide type, intermediate type, and long type in the
visual observation, and 16 images, 8 images, and               0.9
7 images were offered as sample for each type

respectively. Then, images were converted into the             0.8
- coordinate system and  was calculated after
the smoothing treatment was carried out using the              0.7                       Wide type
above-mentioned shape analytical tool at the 10%                                         Intermediate type
of total outline image pixels. Although  value                0.6                       Long type
calculated from the image obtained by image
scanner is smaller than the value by the digital                 100    200      300    400      500      600
camera image for the same leaf blade, there was                                      β
the fixed relation as shown in Fig. 9. As the result,
it was shown that the feature of the leaf blade               Fig. 8 The relationship between 3 types of
shape could be extracted based on the -                               L/W and 
coordinate system even if the digital camera is
used for the image acquisition of the leaf blade of the grape.

Conclusions                                                                   500

      With using the grapes as the experimental
materials, it was examined a new technique for
                                                         β by Image Scanner

getting the information on the tree growth using                              400
a digital camera instead of the human eyes. We
analyzed the shape of the grape leaves, which is
a very important factor in order to judge the tree
vigor of the cultivating grapes, applying the                                 300                                   **
      The r--system for their image data. As a
result, we can obviously remark the differences
of the leaf shapes and easily classify the shapes                               200   300   400   500   600   700        800   900
into three types; the long, intermediate and width                                           β by Digital Camera
leaves. This meant that it could be possible to
extract the shape characteristics of the leaves and                           Fig. 9 The correlation between  by image
to process the data quantitatively.                                                 Scanner and  by digital camera

1 Hashimoto, A. et al.: Evaluation of Tree Vigor by Digital Camera Based on Fruit Color and Leaf Shape,
Proceedings of the 1st World Congress of Computers in Agriculture and Natural Resources, pp.70-77
                                                     Aki         by
2 Kondou, H .et al.: Color Chart for Fruits of Grape ‘ Queen’ Digital Image Processing, Proceedings
of Agricultural Information Technology in Asia and Oceania 1998, pp.197-202 (1998)
3 Kondou, H .et al.: Color Evaluation by Digital Camera for Fruits in Natural Outdoor Scenes, Proceedings
of Agricultural Information Technology in Asia and Oceania 2000, pp.432-442 (2000)
4 Motonaga, Y. et al.: Determination of the Standard Shape and Color of Agricultural Products, Proceedings
of QCAV'98, pp.29-34, (1998)
5 Motonaga, Y. et al.: Agroinformatic Management System with Quality Analysis Function, Proceedings of
the 1st World Congress of Computers in Agriculture and Natural Resources, pp.580-587 (2001)
                                             Aki       ,
6 Yamane, H. et al.: New Grape Cultivator ‘ Queen’Bull. Fruit Tree Res. Stn. 22, pp.1-11 (1992).


Shared By:
Description: Digital cameras, is a use of electronic sensors to the optical image into electronic data camera. Use is divided into: SLR camera, card camera, telephoto cameras and home cameras.