Вы находитесь на странице: 1из 4

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882

Volume 4, Issue 3, March 2015

AUTOMATIC CONTEXTUAL CROPPING AND FEATURE


EXTRACTION FOR PLANT LEAF RECOGNITION
VIJAYALAKSHMI B 1
1
Assistant Professor (Sr.Gr), Department of MCA,
K. L. N. College of Engineering,
Sivagangai District, TamilNadu, India

ABSTRACT
Plants are the mainstay of all life on Earth and an
important resource for human welfare. Plant
identification is really significant in agriculture for the
management of plant or plant cases. This paper
presents a simple and computationally efficient
method to plant identification using digital image
processing. The proposed approach consists of three
phases: find out the four points, crop the image by
using four points, and calculate some basic geometric
features.
Keywords: Plant Identification, cropping, feature
extraction, Euclidean distance.
1. INTRODUCTION
Plants are necessary to the balance of natural
surroundings and in peoples lives. They are the ultimate
source of food and metabolic energy for closely all
animals, which cannot make their own food. Thus the
study of plants is vital because they are an essential part
of life on Earth. A digital plant identification system can
be used for fast characterization of plant species without
needing the knowledge of botanists, thus atomizing their
work.
In order to extract any specific information, image
preprocessing steps are carried out before the actual
analysis of the image data. Preprocessing refers to the
initial processing of input leaf image to remove the noise
and accurate the distorted or degraded data.
Preprocessing techniques like grayscale conversion,
Smoothing, resize, filtering and cropping. Many times
digital images shot for Web use have a border of useless
space around the object(s) of interest. Rather than crop
to precisely the film or chip's border, crop contextually
down to the minimum dimensions that still impart the
substance or context of your icon.
Cropping simply mentions to removing unwanted
components of an icon. It can as well be utilized to
produce an image of a specific size or dimension. This

V. MOHAN 2
2
Professor and Head of Department,
Department of Mathematics,
Thiagarajar College of Engineering,
Madurai, TamilNadu, India

report describes our approach for automatic cropping


and extracting basic geometric features. The
arrangement of the report is as follows: section2 outlines
the need for automatic image cropping, section 3,
describe the proposed method with algorithm steps,
section 4, gives Experimental Result and discussions,
and section 5 brings up the overall conclusion and scope
for future inquiry.

2. NEED FOR IMAGING CONTEXTUAL


CROPPING
Cropping simply refers to eliminating unwanted portions
of an image. It can as well be utilized to produce an
image of a specific size or dimension
Thither are many reasons to crop an image; for
instance,
Fitting an image to fill a form,
Getting rid of a part of the background to
highlight the issue, and so on
In leaf identification system, the user has used mobile
phone, digital camera or notepad to acquire the leaf.
When they put-on the leaf, distance from acquiring the
instruments to the leaf are problematic. [1] This distance
may be small or far. Suppose the distance is far from the
object, there is useless space around the object(s) of
interest. It will increase the computation time also.
To avoid such problem many tools are available to
crop contextually down to the minimum dimensions that
still impart the substance or context of your icon.
Even though many tools are available, we will
conform to any one of the way in manually.
i)
Mention the parameter (x1, y1) as starting
passion of rectangle and (x2, y2) as an end
Position of the rectangle.
ii)
Click on the upper left hand corner of the area
you wish to keep. While controlling the mouse
button, drag toward the bottom right of the
picture.
Then, measuring physiological Width and the diameter
is the basic Geometric features in feature extraction of

www.ijsret.org

223

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 3, March 2015

leaf shape description.. A human must click the two


terminals of the main vein of the leaf via mouse click
[2] [3] [4]. This problem also happens with methods,
extracting features in the plant leaf recognition system
Previous works have some disadvantages. To overcome
manual interaction in the leaf identification system,
hence this paper proposes the automatic contextual
Cropping.

3. PROPOSED METHODOLOGY
3.1Computation modelfor cropping leaf

Figure 1: Sequential Steps to Crop the image


3.2Algorithm steps:
1. Load the image data
2. Change color to gray image.
3. Determine the threshold of the image
(useostu threshold here)
4. Predict the first role (ex: named as findrecpoints)
which is employed to find out the four points
from the image to clip.
Pass the loaded image and threshold value
as parameters to the subroutine.
This function should return the 4 values.
(ex: [x1 y1 x2 y2] = findrecpoints(image,
threshold)
5. Call the second function which is used to crop the
image.
Pass the loaded image and four values as
parameter (which is getting even by first
function).
This subroutine should return the trimmed
picture
6. You can also save or display the cropped image.

www.ijsret.org

Steps to the first function:


1. Take the image and threshold value.
2. Initialize leftx= 0 (zero) and lefty= 0
(null), rightx=0 and righty = 0.
3. Start raster scanning (column wise) of the image
matrix from the protruding location.
i. For i=2: n-1
ii. For j = 2: m-1
4. Determine for each pixel value of an image
having greater than threshold value or lessthen
the threshold value. (IE: if image (j, I) <
threshold)
5. If it is larger than threshold go to step 13
6. If it is to a lesser extent than the threshold value,
Check the variable left
also equal to zero.(ie: if leftx = = 0)
7. if step 6 is true, then check ith position
value greater than 10.(ie: if ( i> 10)
8. if step 7 is true, Assign leftx= i -8 and lefty= j .
9. if step 7 is false, assign leftx = 2 and lefty = j.
10. End
11. End
12. Then, Assign rightx =I and righty = j.
13. End to step 4.
14. Repeat step 3 to 13 until reaches the final status
of the image (bottom right).
15. Initialize topx= 0 (zero) and topy= 0 (null),
bottomx= 0, bottomy =0.
16. Start raster scanning (row wise) of the image
matrix from the protruding location.
i. For i=2: m-1
ii. For j = 2: n-1
17. Determine for each pixel value of an image
having greater than threshold value or lessthan
the threshold value.( ie: if image(i,j ) < threshold
)
18. If it is larger than threshold go to step 26.
19. if it is to a lesser extent than the threshold
value,Check the variable also equal to zero.(ie:
if topx = = 0)
20. if step 19 is true, then check ithposition
value greater than 10.(ie: if ( i> 10)
21. if step 7 is true, Assign topx= i -8 and topy= j .
22. if step 7 is false, Assign topx = 2 and topy = j.
23. End
24. End
25. Then, Assign bottomx =I and bottomy = j.
26. End to step 17.
27. Repeat step 16 to 26 until reaching the last
position of the image (bottom right).
28. Return x1= topx, y1=bottomx, x2= leftx,
y2= rightx.
29. Stop.

224

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 3, March 2015

Table 1
Sample Original image and Cropped image

Steps to Second Function:


1. Get the four values from the first function and
image
2. Initialize a=1 and b=1;
3. Start looping from top x to bottom x and left x to
right x.
4. For i = topx : bottom x
5. For j = leftx to right x
6. Croppedimage(a,b) = image(i , j)
7. Then increase b by one.(ie: b=b+1)
8. End to step 5
9. Then increase a by one (ie: a=a+1)
10. Assign b=1.
11. End to step 4.
12. Return the trimmed picture.
4. EXPERIMENTAL

RESULT

Name
of
image

The experiment the proposed method, the dataset


named Flavia, which can be downloaded from [5] has
been used. This dataset contains 32 varieties of plants,
leaves.. The entire algorithm was carried out and tested
using MATLAB R2009b.

Elapsed
Time
0.891000

Size : 461 x
1241

Size:1200x
1600
4.jpg

2.125000

Size:1200x
1600

Size:1089x1202

5.jpg

In that respect are different publicly available


leaf image datasets such as Flavia dataset, Leafsnap
dataset, Image CLEF dataset, One-hundred plant species
leaves data set Data Set (100 leaves, plant species), ICLleaves dataset. The execution of this experiment is
evaluated using Flavia Dataset.

0.797000

Size : 424 x 826


Size:1200x
1600
10.jpg

1.156000

Size:1200
1600

Size : 589 x 1370


1.73400

14.jpg

To estimate the overall accuracy of our system we


have employed the following recipe.
Accuracy =

Examine the
output image

3.jpg

AND

DISCUSSIONS

Input image

Size:1200
1600

Size : 833 x 1461

( 1)
15.jpg

Some of crops images are given table 1. We have


listed the proposed techniques and their result as given
in Table 2. Out of 32 leaves, 30 leaves are cropped well
and 1 leaf cropped are partially correct. The proposed
method performance accuracy is 96.87 %. Physiological
Width and diameter also measured the distance from
(left x, left y) to (right x, right y) and (top x, top y) to
(bottom x, bottom y).Based on the table2, the proposed
method that is certainly encouraging for plant
identification.

1.82900

Size:1200
1600

Size:895x
1386

17.jpg

www.ijsret.org

1.85900

Size 1200
1600

Size:895X141
3

225

International Journal of Scientific Research Engineering & Technology (IJSRET), ISSN 2278 0882
Volume 4, Issue 3, March 2015

5. CONCLUSION

REFERENCES

In this report, we have introduced a novel glide path to


automatic cropping the leaf image for plant identification
system. We reason that automatic cropping is a viable
option for appropriate cropping. The proposed method
performance accuracy is 96.87 %.
However, they can be computed basic geometric
features in parallel and as the computational
performance of computers increase, the time necessary
for their calculation perhaps wont be a trouble in the
nearby future. Although the operation of the system is
proficient enough, we consider that the performance still
can be bettered.
Table2: Comparison of image size
S.No

Name of
the Image

Size of Input Image

Size of Output Image

1.

1.jpg

1200 x 1600

1043 x 1450

2.

2.jpg

1200 x 1600

1041 x 1424

3.

3.jopg

1200 x 1600

461 x 1241

4.

4.jpg

1200 x 1600

1089 x 1202

5.

5.jpg

1200 x 1600

424 x 826

6.

6.jpg

1200 x 1600

1040 x 1233

7.

7.jpg

1200 x 1600

1124 x 1501

8.

8.jpg

1200 x 1600

1137 x 1426

9.

9.jpg

1200 X 1600

972 x 1442

10.

10.jpg

1200 x 1600

589 x 1370

11.

11.jpg

1200 x 1600

1147 x 1448

12.

12.jpg

1200 x 1600

1024 x 1475

13.

14.jpg

1200 x 1600

833 x 1461

14.

15.jpg

1200 x 1600

895 x 1386

15.

16.jpg

1200 x 1600

1087 x 1548

16.

17.jpg

1200 x 1600

895 x 1413

17.

18.jpg

1200 x 1600

971 x 1417

18.

19.jpg

Partially cropped

19.

20.jpg

1200 x 1600

1051 x 1404

20.

21.jpg

1200 x 1600

1027 x 1482

21.

22.jpg

1200 x 1600

939 x 1402

22.

23.jpg

1200 x 1600

1062 x 1479

23.

24.jpg

1200 x 1600

1042 x 1466

24.

25.jpg

1200 x 1600

1028 x 1483

25.

26.jpg

1200 x 1600

1093 x 1532

26.

27.jpg

1200 x 1600

1131 x 1556

27.

28.jpg

1200 x 1600

1109 x 1239

28.

29.jpg

1200 x 1600

1125 x 1485

29.

30.jpg

1200 x 1600

1051 x 1467

30.

31.jpg

1200 x 1600

1093 x 1105

31.

32.jpg

1200 x 1600

1104 x 1306

32.

33.jpg

1200 x1600

1089 x 1535

[1] SandeepKumar. E,Leaf Color, Area and Edge


Features based Approach for Identification of Indian
Medicinal Plants, Indian Journal of Computer Science
and Engineering (ISSN: 0976-5166), Vol.3 No.3 Jun-Jul
2012
[2] ShayanHati, Sajeevan. G, Plant Recognition from
Leaf Image through Artificial Neural Network ,
International journal of Computer Applications (09758887), volume 62-No 17, January 2013.
[3] S. Wu, F. Bao, E. Xu, Y.X. Wang, Y.F. Chang,
and Q.-L. Xiang, A leaf recognition algorithm for plant
classification using the probabilistic neural network,
Signal Processing and Information Technology, 2007
IEEE International Symposium on, Dec. 2007, 11 16.
[4] Nikesh. p, Nidheesh. p, sugar. M, Leaf
identification using Geometric and Biometric Features,
ASMs International E-journal of Ongoing Research in
Management And IT (e-ISSN-2320-0065), INCON13IT-046, 2013.
[5] http://flavia.sourceforge.net/.

www.ijsret.org

226

Вам также может понравиться