Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
IRJPEE
Vol. 3(2), pp. 056-060, November, 2017. www.premierpublishers.org, ISSN: 3254-1213x
Conference Paper
Many farmers are facing lot of problems in agricultural occupation, every year farmers are
experiencing huge losses due to plant disease. Disease on plants reduces the quality as well as
the quantity of the crops. It is difficult to monitor the plant disease manually, even though farmers
have prior knowledge on the plant disease, it requires tremendous amount of work and excessive
processing time. In this paper we will provide the brief idea to solve the former problem by
developing Agricultural Robot, which will continuously monitor the farm under the guidance of
the operator and we will make use of image processing technique to find the disease on plant
leaf. And if disease is found user can make the robot to spray proper amount of pesticide to the
disease affected region or to the whole plant. This system resulted in detection of plant disease
and spray pesticides in proper amount.
Key Words: Agricultural Robot, Image Processing Algorithm, Robot movement, spraying mechanism.
INTRODUCTION
In this 21st century, the farmers are increasingly under PC. The servo motor, submersible pump, sprayer nozzle
pressure to feed more people as the world population is and floating sensor is used for spraying mechanism.
increasing. Agriculture is the main source of income to the
country. The youths are turning away from the agricultural
profession; thus, availability of human resource is LITERATURE SURVEY
becoming less. The robot technique can make the farming
occupation an exciting Hi-Tech profession. And this may In this section, various method of image processing for
attract the people into agricultural profession and also bring plant disease detection is discussed.
back the people who left this profession. The agricultural The vegetation indices from hyper spectral data have been
robot can perform basic elementary function like picking the shown for indirect monitoring of plant diseases. But they
fruits/ vegetables, harvesting, weeding, pruning, seeding, cannot distinguish different diseases on crop. Wenjiang
grafting, spraying etc. But in this paper, we concentrated Huang et al developed the new spectral indices for
only on disease detection and spraying mechanism. In identifying the winter wheat disease. In this paper three
early days monitoring and analysis of plant disease is done different pests are considered (Powdery mildew, Yellow
manually by expertise persons, which require tremendous rust and Aphids) for their study. To find the disease
amount of work and also excessive processing time. Here wavelengths were extracted using RELIEF-F algorithm (K.
we use the MATLAB software to carry out the image Thangadurai and K. Padmavathi, 2014). Enhanced images
processing technique, and this technique will increase the have got more quality and clarity than original image using
accuracy of the work and also consumes less time. we are RELIEF-F algorithm. Since primary images will be in RGB
using USB webcam mounted on agricultural robot to get form its difficult to process the RGB image because of its
the streaming video on to the connected PC from which the range 0 to 255, hence this primary RGB is converted to
images are captures and are processed using MATLAB grey images. To enhance the image intensities are
software to detect disease. The robot is wirelessly distributed using histogram equalization.
connected to the operators PC and smart phone for two
reasons (a) to receive the instructions regarding the *Corresponding author: Arjun Prakash. R, Department of
mission and (b) to stream live video on to the operators ECE, SIT, Bangalore, India. Email: arjunrp88@gmail.com
To determine the health status of chilli plants, chilli plant Power supply
Arduino Floating sensor 1
leaf image is captured using digital camera. Zulkifli Bin
Submersible pump 1
Husin et al, 2014, uses digital camera to capture the leaf
Tank 1
image and MATLAB software to process the image. GUI for
the MATLAB is build using LABVIEW software. And this
Floating sensor 2
technique ensures that the chemical should be applied to
the diseased plants only. Submersible pump 2
Tank 2
To detect the disease in plant Al-Bashish et al., 2011, uses
Floating sensor 3
image processing. Artificial neural network is used for the
detection of disease in plants. This paper suggests to Submersible pump 3
create to different data bases, one for storing the disease Tank 3
images and other for implementation of query images. In Fig.1. Robot movement and spraying mechanism
this paper mainly 3 feature vectors are considered i.e.,
color, texture and morphology. And it is concluded with USB Webcam
morphological features give better result than other two
features.
Wirelessly connected
to operator PC
Shantanu Phadikar and Jaya Sil uses pattern recognition Power supply Raspberry pi
- - - - - - - - -
- - - - - - - - - -
- - - - - - - - - -
- - - - - - - - - -
Image segmentation
Spraying mechanism
Fig.3. RC agricultural robot mechanism
a. Robot Movement Image Acquisition: The first stage of any vision system is
For the movement of the robot, DC motors are used which the image acquisition stage. After the image has been
are electronically controlled by Arduino UNO with the help obtained, various methods of processing can be applied to
of L293D. Input signals are received by the HC-05 the image to perform the many different vision tasks
Bluetooth module and sends it to the controller which in turn required today. However, if the image has not been
controller makes the motor to spin. By receiving the signal acquired satisfactorily then the intended tasks may not be
DC motors are turned ON and OFF by enabling particular achievable, even with the aid of some form of image
pin on arduino. 300rpm DC motors are used to provide the enhancement. The captured image will be in the form of
sufficient speed. RGB (red, green and blue). Sometimes we have to convert
image from one form to another. MATLAB contains all the
b. Video Streaming tools require for image processing which covers all color
With the help of USB webcam and Raspberry pi we stream space transformations. Below figure shows loading the
the video to the operator PC. Video streaming can be done image to the MATLAB software.
in many ways i.e., by installing gstreamer software on both
raspberry pi and operators PC or installing VLC player on Image pre-processing: Different pre-processing
both transmitter side and at the receiver side. We prefer to techniques are considered to eliminate the noise like
use VLC player for streaming the video with the https:// cropping, image smoothing and enhancement is carried out
followed by IP address of raspberry pi seems easy and also to increase the contrast. Enhancement is a process of
operator can take the snapshots from the streaming video enhancing the visual quality of images due to non-ideal
for the further disease detection process. acquisition system. Some of the enhancement techniques
are edge enhancement, noise filtering, sharpening. Here,
c. Pesticide Spraying Mechanism the image is processed so that the result is more suitable
Bluetooth module is connected to the digital pin of Arduino than the original image for a specific application.
UNO which receives the signal from the android app
installed on the operators smart phone. Floating sensor The input color of the image are primary colors i.e., red,
and submersible pump is kept inside the pesticide tank. green and blue. It is not possible to implement the
One end of the small diameter pipe is connected to the applications using RGB because of their range 0 to 255. So
submersible pump and the other end is connected to the RGB images are converted to grey images using color
sprayer nozzle. By running the MATLAB algorithm, if the conversion equation as below,
algorithm says plant is affected by some disease, operator F(x) = 0.2989*R + 0.5870*G + 0.114*B (1)
can make use of Android app to spray particular pesticide.
Cumulative distribution is used to distribute intensity values
DISEASE DETECTION ALGORITHM to enhance the plant disease image.
The first phase of algorithms deals with separation of the Image segmentation: Image segmentation is related to
healthy crops from affected one and second phase of partitioning of an image into its constituent parts.
algorithms concentrates on finding the disease on plant Segmentation divides the image into meaningful parts or
leaf. An efficient and speedy response algorithm was regions. Meaning part may be complete object or a part of
developed using MATLAB software. The basic steps for it. Segmentation algorithms use image features to extract
classification using image processing are shown below, regions. Edge detection, thresholding, boundary extraction,
Pesticide Spraying Agricultural Robot
Arjun et al. 059
region growing, splitting and merging are generally used for b. Enhance Contrast:-
segmentation. Here the leaf image is divided into number
of segments, not all the segments consist useful quantity of
information. Hence the patches that contain more than 50%
of useful information is taken for further analysis.
Feature extraction: Feature extraction a type of
dimensionality reduction that efficiently represents
interesting parts of an image as a compact feature vector.
After selecting the segmented part, various features like
color, texture, entropy, mean deviation and etc., are
extracted to describe the infected region.
Detection and classification of disease: After the feature Fig.6. Enhancing the image by increasing the contrast.
extraction is done, the obtained values are compared with
the pre-loaded values of healthy leaves. Based on c. Segment Image:-
difference in the values obtained by comparison,
classification of disease is done.
RESULTS
d. Classification Result:-