Disease severity measurement on plant leaf

using image processing

Presented By PIYUSH

CHAUDHARY ID NO.102150011

Guided By Dr. A.N. CHEERAN

CONTENT

AimIntroductionLiterature surveyImage processing techniquesDisease spot detection Disease severity measurementLeaf area measurementConclusion

Aim

Complete analysis of plant leaf for disease severity measurement using image processing.

Introduction

According to UN in 2050:World population will be 9.1 billion. 70% more food production is needed for this

population.In developing countries like India, 60% people are

dependent on agriculture for their livelihood.contributes nearly 19% to India’s G.D.P.Around 51% of the geographical area in India is

already under cultivation as compared to 11% of the world average.

Growth in agriculture has a maximum cascading impact on entire economy and the large segment of population.

Literature survey

Diseases in plants

Losses because of diseasesPlant diseases are important factors, as it

can cause significant reduction in both quality and quantity of crops in agriculture production.

Average loss in agriculture production is 32%

India is losing agricultural production worth Rs. 1.48 lac crore annually

Observation for disease severity measurement

Leaf observation is done by:a) Detection of leaf spot diseaseb) Disease severity measurementc) Leaf area measurement

Disease severity measurement

Disease severity is measured for loss predication and disease control decision.

Two terms are used for taking decision about disease control:Economic threshold levelEconomic injury level

Disease severity measurement process:Calculate disease severity percentage on

plant leaf.Calculate disease intensity for the plant.Calculate disease severity for the whole farm.

Disease severity measurement techniques

Disease spot countRating of sample leaf using disease scale.

Standard area diagrams

Image processing techniques

The discussion of the Image Processing steps used in disease severity measurement should be divided in five major groups:

Disease spot detectionDisease spot countingDisease and leaf area measurement.Disease severity percentage calculationRating.

Disease spot detection techniques

Naked eye observation method.Grid paper method.By image processing technique:

a) Image acquisition.b) Image pre-processing.c) Image segmentation.

Disease spot detection using Image processing

Combination of OTSU threshold and Median filter is used by some researchers.

Main obstacle in this process is noise which is introduced because of –Background noiseCamera flashVeins in plant leaf

Plant Leaf StructureIn “Monocot family” plants, mostly veins are

parallel and less visible.In “Dicot family” plants veins form a netted

pattern, in which larger veins are thicker and straighter. In the process of disease spot detection disturbance mainly occurs because of these thicker veins.

Monocot family plant leaf Dicot family plant leaf

disease spot detection using Image processing

Some researchers used color transform technique before image segmentation to improve results.

Color models used by researchers are:a) HSI color model.b) YCbCr color model.

Brightness and color components are separated from each other.

Threshold is applied on color component.

Image processing steps used in disease severity

measurement

Image Color TransformIn plant, leaf vein is different in intensity

and disease spot is different in color, in comparison to plant leaf.

For minimize the effect of presence of vein, RGB image should be color transformed before segmentation.

Three color models are compared-HSI Color ModelYCbCr Color Model CIELAB Color Model

HSI Color Model‘H’ describe pure color, ‘S’ measures the

colorfulness and ‘I’ shows the amplitude of light.

Threshold is applied on H component.Conversion formula for HSI color model where

YCbCr Color ModelWidely used in digital video.‘Y’ indicates luminance component and ‘Cb’,

‘Cr’ indicates color component.‘Cb’ is the difference between blue

components and ‘Cr’ is the difference between red components.

Conversion Formula-Y = 0.299*R + 0.587*G + 0.114*BCb = -0.168*R – 0.331*G – 0.500*BCr = 0.500*R – 0.418*G – 0.081*B

CIELAB Color Model

Device independent color model, Defined by CIE.

Brightness and color information of LAB color model is independent of each other.

Conversion Formulas for LAB color model are-L = 0.212*R + 0.715*G + 0.0722*BA = 1.4749*(0.221*R – 0.339*G + 0.177*B) + 128B = 0.6245*(0.194*R + 0.605*G – 0.800*B) + 128

CIELAB Color Model

In CIELAB color model, ‘L’ describes color brightness; ‘A’ describes the color ranging from green to red and ‘B’ describes the color ranging from blue to yellow.

CIELAB Color Model

Image Smoothing

During image collection, some noise may be introduced because of camera flash.

To remove unnecessary spot, Image smoothing technique is needed.

Median filter is used for this purpose.

Image Smoothing using Median Filter

Non linear filterThe median of a numerical

collection is such that half the values in collection are less than or equal to median, and half are greater than or equal to median.

Three step algo:-Moving window.Shorting and rearrange .Select middle value.

Image SegmentationObject are separated from each other.

The segmentation process is based on various features found in the image like:Color information that is used to create

histograms.edges or boundariestexture information.

Threshold Based Segmentation

It is important to select a threshold of gray level for extract the disease spot from plant leaf.

If the histogram has sharp and deep valley between two peaks, bottom of the valley can be chosen as threshold.

problem occurs when valley is flat and broad.

OTSU MethodOtsu method is used to automatically select

an optimal threshold.Otsu’s thresholding method is based on

selecting the lowest point between two classes .

Just run through the full range of t values [1,256] and pick the value that :-Minimizes the within class variance.Maximizes the between class variance.

Object Counting Normally, there is more than one spot on plant leaf

image. Therefore counting the disease spot is important to estimate the loss.

A pair of adjoining pixels is part of the same object only if they are both on and are connected along the horizontal, vertical, and diagonal direction.

Disease spot detection

Flow Chart: Disease spot detection

Comparison

Image Collection

Disease Spot Segmentation

Image Smoothing

Image Color Transform•YCbCr Color Model•HSI Color Model •CIELAB Color Model

Experimentation: Disease spot detection

Four methods are used and compare to get best method for disease spot detection.Apply Otsu threshold on RGB image.Apply Otsu threshold on ‘Cr’ component of YCbCr

color space. Apply Otsu threshold on ‘H’ component of HSI color

space. Apply Otsu threshold on ‘A’ component of CIELAB

color space.Research is categorize into three parts-

Experiments with noisy background.Experiment with disturbance because of vein.Experiment with different colored disease spots.

Experiments with noisy background

RGB Image of iris leafMethod:1

Method: 2

Method:3

Method: 4

Experiments with noisy background

RGB Image

Method:1

Method: 2

Method:3

Method: 4

Experiments with noisy backgroundConclusions from these results are:

Using threshold on RGB image disease spot can’t be detected accurately (method 1).

Using threshold on ‘H’ component of HSI color model and ‘Cr’ component of YCBCR color model, disease spots can be detected in some cases but not in all. So results are dependent on type of background (method 2, 3).

Results show that using threshold on ‘A’ component of CIELAB color model in all cases disease spots are detected accurately and results are independent of background (method 4)

Experiment with disturbance because of vein

RGB Image

Method:1

Method: 2

Method:3

Method: 4

Experiment with disturbance because of vein

Conclusions from these results are:Using threshold on RGB image neither we can detect disease

spot nor we can eliminate disturbance because of background and veins (method 1).

Using threshold on ‘Cr’ component of YCbCr color model, some disease spots are detected effectively, but disturbance because of vein is present in results. So results depend on type of leaf and vein (method2).

Using threshold on ‘H’ component of HSI color model only few disease spots can be detected. Disturbance because of vein is also present in some cases. So results depend on type of leaf and background (method 3).

Using threshold on ‘A’ component of CIELAB color model disease spots can be detected accurately in all cases. Experiments show that results are independent of type of plant leaf (method 4).

Experiment with different colored disease spots.

Using CIELAB color transform (Method 4) noise is removed effectively:

This method is also checked with different colored disease spots and results are -

Disease severity measurement

Flow Chart: Disease severity measurement

Disease and leaf Region Segmentation

Threshold is applied on ‘A’ component of CIELAB color transformed image.

Binary image is inverted in color and region filling technique is used for leaf region segmentation

Disease severity measurement

Objects are counted using 8-connected neighborhood technique.

Formulas for area calculation and disease severity measurement are:

Disease scale is used for assessment of disease severity on plant leaf:

Experimentation: Disease severity measurementSample leaves Rating of sample leaves

according disease severity percentage

Accuracy test: Disease severity measurement

Error in disease spot counting

Error in closer spot counting

Error because of vain color difference

Leaf area measurement

Review of Leaf area measurment techniques

Using regression equation: A = b*l*wGrid count methodGravimetric methodImage processing techniques:

a) Threshold based techniqueb) Edge detection based technique

Flowchart: Leaf area measurement

Leaf area measurement

Image acquisition Leaf region segmentation

Leaf area calculation

Experimentation: Leaf area measurement

Sample leaves Results

ConclusionApplying threshold on color component one can

get good segmentation of disease spot.In this research HSI, YCbCr and CIELAB color

models are compared and finally ‘A’ component of CIELAB color model is used.

Following this method different disease spots are detected accurately and results are not affected by background, type of leaf, type of disease spot and camera.

Average accuracy of disease severity measurement algorithm is above 99% which is confirmed by calculating area of known area objects.

Cont..Leaf area is also an important part of

plant to analyze the growth and predict the yield.

Average accuracy of this algorithm is above 99% which is confirmed by comparing the results with measurements of grid count method.

Future Work

Further to this it is needed to compute disease severity in other parts of plant like fruits, stem and root.

PublicationsPiyush Chaudhary, Dr. A.N. Cheeran, Anand

Chaudhari and Sharda Godara “Color transform based approach for disease spot detection on plant leaf”, International Journal of Computer Science and Telecommunications (IJCST), Vol.3, Issue.6, June 2012, pp. 65-70.

Piyush Chaudhary, Sharda Godara, Dr. A.N. Cheeran and Anand Chaudhari, “Fast and accurate method for leaf area measurement” International Journal of Computer Applications(IJCA), July 2012 edition (will published on 28-JUL-2012)

PublicationsPiyush Chaudhary and Dr. A.N. Cheeran, “Fast

and reliable method for disease severity measurement”, 2012Third International Conference on Emerging Applications of Information Technology (EAIT), IEEE, 29 Nov-1 Dec 2012, (Decision Pending)

Anand Chaudhari, Piyush Chaudhary, Dr. A.N. Cheeran and Dr. Yashant Aswani, “Improving signal to noise ratio of low does CT image using wavelet transform”, International Journal on Computer Science and Engineering (IJCSE), Vol.4 No.5, May 2012, pp. 779-787.