IMAGE PROCESSING METHOD TO
MEASURE SUGARCANE LEAF AREA
Sanjay B. Patil,
Asst. Prof., Dept of E&TC Engg. Shri Chhatrapati Shivajiraje College of Engineering,
Dhangawadi, Tal-Bhor, Dist-Pune, (M.S.), India [email protected]
Dr. Shrikant K. Bodhe, Principal,
Pandharpur College of Engineering, Pandharpur, Dist-Solhapur,(M.S.),India [email protected]
Abstract:
In order to increase the average sugarcane yield per acres with minimum cost farmers are adapting precision farming technique. This paper includes the area measurement of sugarcane leaf based on image processing method which is useful for plants growth monitoring, to analyze fertilizer deficiency and environmental stress, to measure diseases severity. In image processing method leaf area is calculated through pixel number statistic. Unit pixel in the same digital images represent the same size hence from known reference area and pixel count, unit pixel size can calculate, so that it is easy to calculate leaf area by counting total pixel in leaf area region. The results are compared with the results of graphical area measurement method. The experimentally it is proved that image processing method for measuring sugarcane leaf area is accurate and strong practicability with small relative error.
Key words: Digital Photograph; Image processing; Leaf are\; Matlab; Sugarcane leaf.
1. Introduction
The Sugarcane Industry remains one of the main pillars of the Indian economy, though facing many problems.
The area of land under sugarcane cultivation is still significant in many states of India. In order to increase the average sugarcane yield per acres with minimum cost farmers are adapting precision farming technique.
In overall growth of sugarcane, leaves are treated as photosynthetic engines of the plant, producing the sugar that is stored in the stalks. In sugarcane, leaves attached to the nodes and forms two alternative ranks on either side of the stalk. They consist of two parts, a blade and a sheath, separated by a blade joint. Leaf size, length and number depend on variety and generation of sugarcane.
The leaf area monitoring is an important tool in studying physiological features related to the plant growth, photosynthetic and transpiration process. Also being helpful parameter in evaluating, damage caused by leaf diseases and pastes, to find out micronutrients deficiencies, water and environmental stress, need of fertilization, for effective management and treatment [1].
Leaf area determination can be done by direct methods, which involves the measuring of all the individual leaf areas or indirect methods, which are based on the relation of some plant characteristics with the true leaf area obtained in destructive tests.
In leaf area measurement of coffee plants by using digital image analysis, used two models, one based on the height and width of the canopies and other based on the area of digital image of a tree. Here the images were corrected by frequency histograms and for segmentation thresholding was done by Otsu method the results were compared with real area of the leaves using digital scanner, they found 0.82 and 0.91 correlation. [1]. In other non destructive leaf area measurement , used Hough Transformation to acquire the coordinates of quadrangle corner points in distorted image and thresholding was used for image segmentation. To eliminate the effect of holes in the leaf, contour extraction approach was used where pixel scanning from one side to opposite side was implemented in four directions to extract contour and leaf area was measured by pixel number statistic, they found absolute error 2.88 [2]. In leaf area measurement of cucumber using image processing method they used reference object and picture pixel number statistic to calculate the leaf area ,found coefficient of variation of 3.99 [3]. In two new leaf area determination methods using digital photographs processed in Matlab and computer Aided software they found 99% accuracy [4]. The shape of leaf is also important parameter affecting the leaf area measurement and to evaluate environmental stress [5].
Particularly in sugarcane, leaf width is correlated strongly with stalk diameter and varies from about 20mm to 60mm in commercial varieties. Length is usually between 0.9m to 2m in mature leaves. These leaves have not flat surface, the mid rib height about 3 mm is the major problem to measure the area by above methods. In this method, leaf area is determined by using digital photographs and Matlab tool to process the digital images. The results are compared with the results from grid counting method. To test performance of the system other plants leaf area also calculated and justified by the same method.
I. MATERIALS AND METHODS A.MATERIALS:
The leaf area measuring system consist main four parts: 12 megapixel Digital Camera with 5X zoom , PC, White Paper Sheet for background, One Rupees Coin as a reference Object.
B.METHODS:
When a Sugarcane leaf area is to be measure in field condition, it is difficult to keep cameras optical axis vertical with leaf plane due to long curved leaf. So the plane of CCD is not superposition with image plane and the leaf image is distorted because of some geometric distortions. This will affect the measurement precision of leaf area. So the leaf is separated from its stalk and placed horizontally on white background to take the photograph. Photographs of the leaves were arranged in number and immediately packed in wet plastic bag to avoid wrinkling of leaf because evaporation rate of sugarcane leaf is 200 times more than other leaf. To compare results of the proposed method, area of leaf is calculated by using graphical method accurately and results are considered as actual area.
B.1 Graphical Method:
The Leaf whose area is to be measure is placed on the graph paper, with smallest grid size of 1*1 mm. Leaf is outlined with a sharp edge pencil accurately and carefully on the graph paper. The total number of grids covered by outline edge of the leaf is calculated; consideration is, if edge outline occupied more than one half grids of graph paper treated as one otherwise zero.
B.2 Image Processing Method:
B.2.1 Principle: In image processing method leaf area is calculated through pixel number statistic. Unit pixel in the same digital images represent the same size hence from known reference area and pixel count, unit pixel size can calculate, so that it is easy to calculate leaf area by counting total pixel in leaf area region as below.[6] Pc = ∑1 (1)
( x , y) Є Rc
Ac = π (d/2)2 (2) Hence,
Ap = Ac/Pc (3)
And
Pl = ∑1 (4) (x, y) Є Rl
Hence,
Al = Pl*Ap (5) Where,
Ac-Area of coin region, Al-Area of leaf region, Ap-Unit pixel value Rl -Leaf region, Rc-Coin region
B.2.2 Matlab based processing:
Matlab is a high performance language for technical computing where problem and solutions are expressed in familiar mathematical notation. The Matlab processing is a semi automatic method to calculate leaf area for more users easily. The code is written in Matlab version 7.4. This code will work in any higher version of the program.
Algorithm
1. Read the image.
2. Convert RGB to grayscale image. 3. Convert grayscale image to binary image. 4. Remove noise using “ imfill” instruction.
5. Calculate the leaf area using “Regionprops” instruction.
B.2.3Photograph acquisition and processing:
The Sugarcane leaf whose area to be measure is placed on the white background, here white cotton cloth is
plane of the leaf. The photograph distance is neither too close nor too far; it is adjusted such a way that the photograph is covering only background, leaf and known area reference object as shown in fig. 1.
Fig. 1 Sugarcane Leaf with One Rupees Coin as a Reference Object
Photographs captured as shown in fig. 1 is processed by Matlab code in two colors. The function regionprops in Matlab is used to measure area of a selected region of an image in pixel count. Before applying the function regionprops, the actual image converted into a grayscale image as shown in fig. 2
Fig. 2 converting the color image to Gray Image
The formula of graying the collected RGB color image is:
I = 0.3R + 0.59G +0.11B (6)
Some gray value is adopted as threshold to segment the image into two group, object and background. A function bwbounderies, is used to trace the boundary of a selected region in an image, this labeled region will be represented in a two dimensional array of non integers. However due to plant diseases, micronutrient deficiency and insect pests removes some area from the leaf or make the color of this area different from healthy leaf, results in pixel count that will less than the true pixel count of the selected region. These dark patches inside the image are holes that can be easily identified and filled. The Matlab function imfill is used to fill any holes as shown in Fig.3.
Then function regionprops can be used to estimate area enclosed as shown in Fig. 4.The area is the actual number of pixels in the selected region [7, 8].
Fig.4 Number of Pixels and Actual Dimension
Leaf Pixel Count (Pl) = 1530 pixels, Leaf Area (Al) = 36.4578 cm2
B.2.4 Determination of area:
The pixel count of the processed image depends on the distance between the camera and the object when the picture is taken. Smaller the distance, the larger the pixel counts [9, 10]. A reference object is needed to be able to translate the pixel count area. The reference object is an object with known area.
In this study, One Rupees coin is chosen as reference object whose area is Area of coin (Ac) = π(d/2)2
Where “d” is diameter of the coin
=
2
2.5
2
cm
π
= 4.9063 cm2
The pixel count of the coin from the image is 206 Hence,
1 pixel value (Ap) = Area of coin (Ac)/ pixel count (Pc) = 4.9063/206
= 0.0238 cm2
For Leaf, the pixel count of the leaf region (Pl) is 1530 pixels. Hence,
Area of Leaf (Al) = pixel count (Pc)X 1 pixel value = 1530 *0.0238
= 36.45 cm2
B.2.5 Determination of Relative Error:
The Relative error can compare values of true value and measured value to describe the accuracy of measure values as below
Leaf area of grid counting is assumed to be standard area and measure true value.
III. RESULTS AND DISCUSSIONS
To test the performance of the of new measurement system, leaves are selected from different plots and different varieties of sugarcane. Leaf area of other plants like Maize, betel is also tested by this method. Area of leaf calculated by square grid method considered as standard area. Table-1 lists the measurement results by image processing method and grid counting method with relative error. It indicates that the relative error value is smaller, the accuracy is higher.
TABLE 1 Comparison of leaf area values of grid counting and proposed method
Leaf No.
Actual Area (cm2)
Measured true value Mt
Area by New Method (cm2) Measured value M
Error (Er)
1 26.06 25.74 0.012
2 36.86 36.45 0.011
3 55.13 55.49 0.006
4 55.52 54.89 0.011
5 55,91 55.17 0.013
With the implemented algorithms the error between area measurements by the proposed system and the actual values is typically 0.0106. . The Relative error can compare values of true value and measured value to describe the accuracy of measure values. It indicates that the relative error value is smaller, the accuracy is higher. Relative error will increase if leaf has cuts due to insect bites, color patches due to diseases or different stresses. IV.CONCLUSION
This work used digital image processing techniques to calculate sugarcane leaf area, the implemented algorithm proved to adequate to compute the leaf areas.
The grid counting methods have very high accuracy, but take more time to measure leaf area if leaf length and width is more. The image processing method also has high precision and accuracy whether leaf with maximum length and width, it takes less processing time. Image can be processed at any time if only the images of the leaves have stored. This method is also feasible to measure area of any type of leaf with same accuracy.
At the present other experiments are being conducted to detect the presence of holes and changes in leaf color in order to infer about the presence of diseases. This component will be valuable to achive a better estimate of the leaf area and diseases severity for application of pesticide and fertilizers.
ACKNOWLEDGMENT:
This work is supported by the Vasantadada Sugarcane Research Institute, Manjari Dist.-Pune. (M.S.), India. REFERENCES:
[1] Morlon Marcon, Kleber Mariano ( et.al),”Estimation of total leaf area in perennial plants using image analysis”, R.Bras.Eng.Ambiental,2011,v.15,96-101.
[3] Tian You-wen,Wang Xiao-juan,”Analysis of leaf parameters measurement of cucumber based on image processing,”World congress on software engineering,2009,34-37.
[4] Enrique Rico-Garcia,Fabiola Hernadez—Hernadez ,” Two new methods for estimation of leaf area using digital photography,”International journal of agriculture and biology,”2009,397-400.
[5] Hiroya Kondou,Hatuyoshi Kitamura,”Shape evaluation by digital camera for grape leaf.”Science and Technology promotion center, 2002, 586-590.
[6] Shen Weizheng, Wu Yachun,[et.al],” Grading method of leaf spot disease based on image processing”,IEEE, 2008,pg.491-494. [7] Nur Badarish Ahmad Mustafa [et.al],”Image processing of an agriculture produce: Determination of size and ripeness of a
banana,”IEEE,2008,pg 464-469.
[8] Sanjay B. Patil, Dr. S.K.Bodhe,”Betel leaf area measurement using image processing,”IJCSE,2011,pg.2856-2660.
