2011 Third International Conference on Measuring Technology and Mechatronics Automation

Smart Scarecrow

Chomtip Pornpanomchai, Malinee Homnan, Navarat Pramuksan and Walika Rakyindee

Faculty of Information and Communication Technology, Mahidol University

Rama 6 Road, Rajchathawee, Bangkok 10400, THAILAND

itcpp@mahidol.ac.th, {u4988042, u4988135 and u4988205} @student.mahidol.ac.th

Abstract - Thailand is an agricultural country, where is different for the image analysis and pattern matching
located in Southeast Asia. We can produce various to find the objects. The project is still in the prototype
kinds of food in not only a good quality but also a huge mode. We require more research and development in
quantity. One problem of both quality and quantity order to reduce the system limitation and enhance the
control of our food products are the food harmful pests
efficiency of the system to be able to perform for a
such as bird, ant, weevil, aphid, grasshopper etc.
Therefore, this project intends to develop the computer real-world application. The topic of further
system that can be chased birds from a farm. The enhancement will be discussed at the end of the
smart scarecrow is developed by using an image document. [1][2][3]
processing technique. Overall works are software This project is about developing an image
development. The system is designed to detect pest birds processing system that can detect the pest birds from
from a real time video frame after it detects the birds the video scene, which is running on specific
then it generates a loudly sound to chase them. The constraints. The system is intended to replace the
system consists of four major components: 1) image normal scarecrow and sensor equipment, which has
acquisition 2) image preprocessing, 3) bird recognition
some weak points as explained before. The system
and 4) generating sound. The experiment has been
conducted in order to access the following qualities: 1) will be created in the form of software which consists
usability, to prove that the system can detect and scare of input and output parts. The input part accepts a
pest birds and 2) efficiency, to show that the system can video stream from the video recorder. The output part
work with a high accuracy. is responsible for generating a very loud sound to
chase the pest birds.
Keywords-Siphonic Smart Scarecrow, Image
Processing, Sound Generating
II. LITERATURE REVIEWS
I. INTRODUCTION Due to the world food crisis, we need to produce
The idea of using the video recorder to detect and more and more food that is enough to feed all people
scare pest birds for protecting our food has been around the world. One way to increase our food is to
proposed. The current detection and chasing keep our food save from all pests. Many scientists try
approaches are based too much on the sensor to apply many methods for building the pest control
equipment. The use of sensor equipment to detect the system. The applicable techniques to detect the pests
objects has been widely spread into different kinds of are described as follow:
industries. Even though the equipment itself provides
a benefit to the user, there are also some 2.1 Apply Signal Processing Methods
disadvantages. The sensor equipment is only used for A signal processing technique is a good method to
real time - the user cannot insert other kinds of input detect pest inside a food product that is no need to see
such as video file to the sensor equipment in order to any pest bodies. But one disadvantage of this
detect and record the object. technique is to detect the dead pests that contaminate
An image processing is the technique which is in a food product because no any signals transmit
based on the software component that does not from the dead pests. [4][5][6][7][8][9][10]
require special hardware. With a typical video
recording device and a normal computer, we can 2.2 Apply Image Processing Methods
create the detection device. By using the basic pattern An image processing method applied with a
matching theory, we can find the number of objects standard camera or video recorder, a personal
in the video scene from the known size of the object. computer and some image processing techniques to
Few image processing key methodologies have detect the pests. All these hardware devices and
been applied to this project. The image differentiation software are reliable and affordable for the farmer.
is used in the object detection process, level of But one weak point of this technique is it can detect

978-0-7695-4296-6/11 $26.00 2011 IEEE 294

DOI 10.1109/ICMTMA.2011.644
the pest only on an external surface of the food component, 2) Image Preprocessing component, 3)
product. It cannot detect the pests inside the food Bird Detection component and 4) Sound Generating
product because it cannot get the pest image that component. Each component has the following
move inside the food product. [11][12][13] details
2.3 Apply Auditory Methods
This technique uses loudly sound to scare pests,
for example gas cannons, pyrotechniques and bio-
acoustics. This technique may be worked in short-
term period because of habituation. [14]
2.4 Apply Visual Methods
There are many visual models, which scare many
pests, such as lasers dogs, human disturbance,
scarecrows, raptor models, corpses, balloons, kites,
falconry, lights, mirrors/reflectors etc. These methods
need to study the pests behavior for example small Figure 2 Electronic scarecrow system structure chart
birds are afraid of a falcon model. [14][15][16]
3.2.1 Image Acquisition
2.5 Apply Chemical Methods The system receives data in form of a video
The chemical technique works very well in stream in real time environment and converts a video
laboratory but less effective in the field due to stream into sequences of video frames. The image
practical problems, for example washes off with rain. acquisition process consists of two sub-processes,
These techniques not only harm pests but also give which are 1) A video streaming sub-process and 2)
some side effects to the environment. [14] frame capturing sub-process. A video streaming sub-
process captures a video file from a computer
III. METHODOLOGY webcam. The frame capturing sub-process converts a
This section presents research methodology video frames from the video stream into a sequence
which includes the conceptual diagram and the of frames.
structure chart.
3.2.2 Image Preprocessing
3.1 Conceptual Diagram Image preprocessing consists of five sub-
Based on conceptual diagram in Figure 1, the processes which are: 1) Grayscale conversion 2)
objective of this system is to develop computer Binarization, 3) Object finding, 4) image cropping
software that can detect and scare birds based on a and 5) image resizing. Each sub-process has the
real time video frame. The computer system takes a following details:
bird image by using computer Webcam after that a) Grayscale conversion the equation for
uses image processing technique to detect the birds in transforming RGB color into grayscale color is
a video frame. Finally, the system generates a loud shown in equation 1 and the original and result image
noise to scare the birds in a field. of this transformation is shown in Figure 3(a) and (b).

Grayscale = 0.299*R + 0.587*G + 0.114*B (1)

Where
R, G, B = the pixel in three components (red,
green and blue colors).

b) Binarization This sub-process defines the

threshold value first, after that uses a threshold value
to convert a gray scale image into a binary image.
Figure 1. Conceptual diagram
The method to convert grayscale image to binary
3.2 Structure Chart image is to compare every pixels color value with
From the overall processes of the conceptual threshold value. If the color value of a pixel is less
diagram in the Figure 1, we can map to the structure than the threshold value, then converts that pixel into
chart for the electronic scarecrow as shown in Figure 0 or white color, otherwise converts that pixel to 1 or
2. The structure chart of the system consists of four black color. Figure 4 (a) and (b) are shown the
components, which are 1) Image Acquisition

295
grayscale and black and white bird picture, means that unknown object and the bird template are
respectively. possible to be the same object. After the system
recognizes the birds in the video frame, then the
system will generate a loudly sound to chase the
birds.

Figure 3 The bird image (a) RGB color (B) grayscale

Figure 6 Extract a number of white pixels in four corners of bird

cropping image.

3.2.4 Sound Generating

The system uses six types of sound to chase the
birds from a video frame. The six generating sound
are 1) a gun shooting sound, 2) a cannon shooting
sound, 3) a machine gun shooting sound, 4) a
Figure 4 The bird picture in (a) grayscale image (b) black and
white image
firecracker sound, 5) a dog barking sound and 6) girls
c) Object finding this sub-process finds all scream sound. We try to generate the sound randomly
objects in the image, which includes birds, trash bin, to chase the birds in order to avoid their habituation.
trees etc. IV. TESTING AND EVALUATION
d) Image cropping this sub-process crops only
This part presents the experimentation on the
the bird area for recognition in the next step. The
electronic scarecrow system, which is developed and
cropping bird image size has the height around 80-
based on the concepts and design mentioned in the
160 pixels and the width around 150-300 pixels. The
previous part. In this system, the experimental are
cropping bird image result is shown in Figure 5 (a)
focused on the usability and the effectiveness of the
and each cropping bird is shown in Figure 5(b)5(e).
system.
4.1 Usability Proof
In this section, we will analyze the usability of
our system. First, we used our system to captures
video frames from web camera with many birds in a
park. After that, we use our program to recognize the
birds in the video frame. Finally, we observe the
system to generate a loudly sound to chase the birds
away. We tested the system in many locations and
Figure 5 Image cropping sub-process result
many times to make sure that the system work
e) Image resizing - normally, the bird input properly.
images have a variety size that can be affected to the 4.2 Effectiveness Test
recognition result. We adjust an iris image with the
width X height equal to 50 X 50 pixels. This section, the effectiveness test will be done
same as section 4.1, but we focus on the correctness
3.2.3Bird Recognition of the final result. We tested the system from
The bird recognition process consists of two sub- different sample of video files which were recorded
processes, which are 1) feature extraction sub-process from different background and difference
and 2) template matching sub-process. Each sub- environment to prove that our system can be used to
process has the following details. perform the bird detection and chase them away. The
a) Feature extraction this sub-process separates experimental results are shown in Table 1.
the four corners of a picture into the same size as Based on Table 1, there are six columns, which
shown in Figure 6. After that the system stores a consist of: 1) video number, 2) number of birds in
number of white pixels in each corner in a database. each video, 3) number of chase birds, 4) percentage
b) Template matching this sub-process finds of chase birds, 5) percentage of chase birds calculated
the sum of minimum value comparing between from number videos and 6) type of generating sound.
unknown object image with the bird templates in the The experiment conducted on 21 videos and the
database. If the minimum value is nearly zero, it system can chase the birds in 19 from 21 videos or

296
the precision is 90.47 percent. In 21 videos contain National Computer Science and Engineering Conference
199 birds and the system can chase 147 birds or the (NCSEC 2009), pp. 293-298, 2009
precision is 73.86 percent. There is no sound number [2] Chomtip Pornpanomchai, Fuangchat Stheitsthienchai, and
Sorawat Rattanachuen Object Detection and Counting
in the line number 3 and 4 because the system cannot System, The IEEE International Congress on Image and
detect any birds in the video frames. Signal Processing, p.p. 61-65, 2008.
Table 1 The effectiveness testing results [3] Chomtip Pornpanomchai, Thitinut Liamsanguan, and
Vissakorn Vannakosit, Vehicle Detection and Counting
from a Video Frame, The IEEE International Conference on
Wavelet Analysis and Pattern Recognition, p.p.356-361,
2008.
[4] Dimitris Psychoudakis, William Moulder, Chi-Chin Chen,
Heping Zhu, and John L. Volakis, A Portable Low-Power
Harmonic Radar System and Conformal Tag for Insect
Tracking, The IEEE Antennas and Wireless Propagation
Letter. Vol. 7, p.p. 444-447, 2008.
[5] Haobin Dong, Chuanlei Wang, Huaping Wang, Xinhui Cai,
and C. Richard Liu, Two-Dimensional Resistivity Imaging
Survey for Detecting Termitaria in a Dam, The IEEE
International Conference on Networking, Sensing and
Control, p.p. 846-849, 2006.
[6] S. Neethirajan, C. Karunakaran, D.S. Jayas, and N.D.G.
White, Detection technique for stored-product insects in
grain, International Journal in Food Control, p.p. 157-162,
2007.
[7] Anshul Mittal and Aseem Singh, Microcontroller based pest
management system, The second International Conference
on System, p.p. 43-46, 2007.
[8] Jie Hu, Application of PCA Method on Pest Information
Detection of Electronic Nose, The IEEE International
Conference on Information Acquisition, p.p. 1465-1468,
2006.
[9] Sheng Ye and Jie Hu, Insect herbivory information
detection by Principal Component Analysis on Electronic
Nose System, The International Conference on Neural
Networks and Brain, p.p. 401-404, 2005.
[10] Kevin M. Coggins and Jose Principe, Detection and
V. CONCLUSION Classification of Insect Sounds in a Grain Silo using a Neural
Network, The IEEE World Congress on Computational
Based on the experimental results in the previous Intelligence and The IEEE international Joint Conference on
Neural Networks, p.p. 1760-1765, 1998.
section, we fulfill our research objective. We can
[11] C. Ridgway, E.R. Davies, J. Chambers, D.R. Mason and M.
conclude that the correctness of result depends on Bateman, Rapid Machine Vision Method for the Detection
three factors: of Insects and other Particulate Bio-contaminants of Bulk
1. The system may not differentiate some birds Grain in Transit The International Journal on Biosystems
Engineering, p.p. 21-30, 2002.
of which the color is similar to that of the tree in the
[12] Long Zhou, Ming Fang, Xue-Zhi Wang, Mou Yi and Zhan-
background image. Wu Yu, Application of gray relation analysis in images
2. The system cannot crop some bird that is edge detection of pests in stored grain, The International
move very close to other birds or some bird that Symposium on Intelligent Information Technology
move outside a video frame. Application Workshops, p.p. 405-407, 2008.
3. The brightness of surrounding environment is [13] Hongtao Zhang and Hanping Mao, Feature Selection for the
Stored-grain Insects Based on PSO and SVM, The second
one factor that can affect the result because the International Workshop on Knowlwdge Discovery and Data
brightness of environment affects the color value of Mining, p.p. 586-589, 2009.
both bird and background in a video frame. [14] J.Bishop, H.Mckay, D.Parrott and J.Allan, Review of
international research literature regarding the effectiveness of
The benefit of this system is a clean environment. auditory bird scaring techniques and potential alternatives,
Available:
There is no need chemical substance to chase the pest http://www.defra.gov.uk/environment/quality/noise/research/
birds because the system chases the birds by using birdscaring/birdscaring.pdf,
image processing technique. [15] Albert Froneman and Marius van Rooyen, The Successful
Implementation of a Border Collie Bird Scaring Program at
REFERENCES Durban International Airport, South Africa, Available:
[1] Chomtip Pornpanomchai, Sopit Sahavisit, Paradorn http://www.int-
birdstrike.org/Warsaw_Papers/IBSC26%20WPAE1.pdf,
Kajorndechakul and Tanapol Santhitivatanakul, Ant
Detection and Counting From a Video Frame, The 13th [16] Eyal de Lara and Maria Rbling, New Products, IEEE
Pervasive Computing, pp. 15-17, 2007.

297