Content area
Agriculture plays a very important role in the provision of food surplus to expanding population, contribution to capital formation, providing raw material to industries, market for industrial products and major contribution in international trade. To enhance the quality and quantity of the agriculture product there is a need to adopt the new technology. Image processing approach is non-invasive technique which provides consistent, reasonably accurate, less time consuming and cost-effective solution for farmers to manage fertilizers and pesticides. The objective of this study is to analyze the nutrition of grape plant using Wavelength algorithmand statistical method regression analysis is used in our work for the nutrient estimation. The relative requirements of nitrogen, phosphorus and potassium in grapes vary with the growth stages of grapes. There is a high-level requirement of N during the vegetative growth stage, P requirement is high during flowering stage and K requirement is more in crop maturity stage. So pruning was carried out in two stages i.e., in April and in October. This study is carried out in grape farms of Theni District of Tamil Nadu for estimating the macro nutrition nitrogen (N), phosphorus (P) and potassium (K) to analyze the yield. The results showed that the overall identification accuracies ofNPK deficiencies were 86.15, 87.69, 90.00 and 89.23% for the two pruning stages.
Keywords: yield estimation, nutrition identification, classification, image processing
Agriculture plays a very important role in the provision of food surplus to expanding population, contribution to capital formation, providing raw material to industries, market for industrial products and major contribution in international trade. To enhance the quality and quantity of the agriculture product there is a need to adopt the new technology. Image processing approach is non-invasive technique which provides consistent, reasonably accurate, less time consuming and cost-effective solution for farmers to manage fertilizers and pesticides. The objective of this study is to analyze the nutrition of grape plant using Wavelength algorithmand statistical method regression analysis is used in our work for the nutrient estimation. The relative requirements of nitrogen, phosphorus and potassium in grapes vary with the growth stages of grapes. There is a high-level requirement of N during the vegetative growth stage, P requirement is high during flowering stage and K requirement is more in crop maturity stage. So pruning was carried out in two stages i.e., in April and in October. This study is carried out in grape farms of Theni District of Tamil Nadu for estimating the macro nutrition nitrogen (N), phosphorus (P) and potassium (K) to analyze the yield. The results showed that the overall identification accuracies ofNPK deficiencies were 86.15, 87.69, 90.00 and 89.23% for the two pruning stages.
Povzetek: Razvili so neinvazivno metodo za oceno pomanjkanja hranil v grozdju s pomocjo analize slik in statisticnih metod, kar omogoca prepoznavanje pomanjkanja dusika, fosforja in kalija ter s tern izboljsanje kakovosti in kolicine pridelka.
1 Introduction
The agriculture sector is the main contributor in the Indian economy and is doing well in white, green and blue revolution. According to APED A, by 2014 exports of Indian agriculture will reach 5% of total production of the world and rank 10th in the ranking [1]. Precision agriculture is a new and developing technology which leads to incorporating the advanced techniques to enhance farm output and also enrich the farm inputs in a profitable and environmentally sensible manner. Farm inputs were important parameters to be controlled and if not will result in adverse effects causing reduction in yield, deteriorating plant health, etc. Irrigation/Water stress, Fertilizers, pesticides and quality of yield were the major factors of concern in agriculture. Most of the time expertise is required to analyze the problems, which may be a time consuming and costly issue in developing countries. Image processing is one of the tools which can be applied to measure the parameters related to agronomy with accuracy and economy. Applications of image processing in agriculture can be broadly classified in two categories: first one depends upon the imaging techniques and the second one based on applications.
Less yield, higher cost of production due to labour scarcity and fertilizer cost are the major challenges before the farmers. To enhance the quality and quantity of the agriculture product there is a need to adopt the new technology. Fertilizer and pesticide management requires early and cost-effective solutions which will lead to higher yield. Image processing approach is a noninvasive technique which provides consistent, reasonably accurate, less time consuming and cost-effective solutions for farmers to manage fertilizers and pesticides. The soft computing techniques are helpful in developing the knowledge-based systems, and may be effectively utilized to develop the expert system. This system will be helpful for farmers to find the solutions to their farming problems in the existing system such as Adaptation to climate change in viticulture, integration with other nutrient assessments and expansion to other cros and environments. The research objective is aimed to identify/develop color models for Fertilizer's estimation viz. Nitrogen, Phosphorous, Potassium and Magnesium for grapes as a horticulture product.
2 Related works
Plants need a certain number of macronutrients (nitrogen, phosphorus, etc.) and micronutrients (Zinc,
Boron, etc.) to grow and stay healthy. These Nutrients alter and regulate the functioning of plants and produce qualitative and quantitative changes in plant yield. Nutrient deficiency may result in weaker plants. Nutrient deficiencies make plants more susceptible to diseases. Fertilizers are the supplements for the plants to grow healthier. In case of grapevines, the deficiencies or overdoses are observed by visual inspections by experts. Availability of experts on time and their consultancy cost are the major issues. Nitrogen application to grapevine in excess form results in excessive growth of shoots at the cost of fruit set. This will also result in a delay in maturity and poor bud formation in the following season [1].
The Proper application of fertilizers and pesticides can save cost of production. In most of the cases, deficiencies are observed by a change in the color of leaves in which Chlorophyll is an important ingredient. Chlorophyll is a molecule in a leaf which is responsible for photosynthesis action. The carbohydrates produced in photosynthesis are used as food for growth of plants and fruit. Various methods are proposed to analyze the nutrients based on chemical analysis such as leaf analysis, petiole analysis [2] Vasifa A. Aglave et al., reported a survey for chlorophyll content, disease severity and leaf area measurement. Leaf area measurement techniques viz. grid counting, paper weighing, leaf area meter, digital image analysis using different techniques were reported. Naked eye observation, image processing techniques like chain code, bounding box, segmentation were reported for measuring disease severity along with different chlorophyll content measurement [3]. M.M. Ali et al., presented different non destructive handheld meter techniques such as leaf color chart, SPAD meter, N-tester, Image analysis to estimate foliar N status of plant. Comparison of these methods was also highlighted on the basis of applicability, accuracy, the effect of environment, etc. [4].
The color image analysis became a popular and cost-effective method for chlorophyll and nutrient estimation. Paula F. Murakami has developed digital imaging software like scion which was used for analysis and quantification of chlorophyll using leaf color. The software calculates the percentage green and red for chlorophyll estimation. RGB and HSV color models were considered in development [5]. Parviz Moghaddam et al., developed an algorithm to estimate chlorophyll using a video camera in which it was shown that the red and blue elements were highly correlated with it. Normalized difference of red and blue was effectively used to estimate chlorophyll under different meteorological conditions. The multilayer perceptron neural network was developed using R, G and B values for chlorophyll estimation. The results show the high coefficient of determination (R2) and low mean square error where the estimated values were compared with chlorophyll SPAD meter [6].
Mario Cupertino da Silva et al., presented the work that showed the correlation between vegetation indices and nitrogen. The Correlation between vegetation indices and nitrogen leaf content and dry matter at different stages for fertilization was calculated with IR camera and digital camera. The study revealed that the high positive correlation decreases as the number of days increases after fertilization and green spectral band is more useful for nitrogen discrimination. Three indices NDVI, GNDVI and SAVI were evaluated and observed that GNDVI was the best [7].
Han Yuzhuet al., proposed a method of nitrogen estimation for pepper in flowering and fruiting using color image processing. Different functions of RGB were considered and correlated with nitrogen. Regression analysis for inorganic nitrogen in soil, total nitrogen, nitrogen concentration and SPAD meter readings were done and all show negative coefficient for the considered function. N applications were given in different treatments [8].
Gloria F. Mata-Donjuan et al., induced five levels of nitrogen deficiencies and proposed improved hue, luminance and saturation color space which is less susceptible to illumination variation. IHLS was used to estimate the nitrogen for tomato seedlings. Image processing methodologies like segmentation, RGB to IHLS conversion and image analysis were implemented. He showed better statistical relation with nitrogen values obtained from chemical analysis. Estimation N was based on components histograms in IHLS and in the fusion of saturation and hue [9].
Table 1 summarizes the literature survey of research methodology, Key focus, research findings, challenges and future research directions.
3 Materials and methods
3.1 Experimental design
The goal of the experiment was to investigate the grape plant's yield analysis under various levels of NPK nutrition. The experiment was carried out in commercial vineyards in areas around Theni of Tamil Nadu. The vines were planted in a vertical shoot-positioning technique with a north-south row orientation at a distance of 2 m. There were five different grapevine varieties utilised in this experiment. Hydroponics with a nutrient solution formula was applied to cultivate the grape plants. After berry set, the six first basal leaves of the selected plants were carefully plucked.
3.2 Sample/ leaf selection
Changes in nutrients are reflected in plant appearance or chemical composition of plant tissues or petioles. Plant analysis for nutrient contents can be carried out either by chemical analysis or visual diagnosis of deficiencies/ toxicity. Standards are developed for petiole sampling through research in various parts of the world and India (NRC or MRDBS, Pune). These standards are for optimal growth of grapevine. Since the nutritional requirement varies over the season, it is necessary to match the standard requirement of nutrients [3]. While sampling, few factors need to be considered viz. time, site, and plant part of sample.
3.3 Image acquisition
Images are acquired with proper selection of camera and background. In two different seasons-one in May or June and the other in November or December-pictures are taken outside in the sunshine. Days with plenty of sunshine were selected to reduce shadow variations and increase light uniformity. The camera's flash was disabled in order to prevent glare and reflections, which could skew colour and intensity measurements that are essential for precise analysis. Leaf images are captured with a digital camera having a CCD sensor of Nikon Coolpix S570. The model is selected considering the general availability and cost effectiveness. CCD camera is preferred than CMOS as CCD's are less susceptible to noise, to maintain image quality under varying light conditions as compared to CMOS. The camera is kept at a distance of 7-8 inches from the ground level or leaf blade surface to ensure consistency in capturing image details across all samples. At a resolution of 3 mega pixels (2048*1536), the camera was configured in Normal mode. Fine details in leaf texture and colour can be captured at this resolution, which strikes a balance between image quality and file size. In order to keep things simple and make the leaf the main focal point, a neutral background was selected. Segmentation during image processing is aided and colour interference is less likely.
The following image pre-processing steps are needed to ensure high quality image acquisition:
Color calibration: Images are colour calibrated prior to processing in order to correct for lighting-related colour variations. Accurate colour representation is aided by this, and extracting RGB, HSV, and other indices depends on it.
Noise reduction: To improve the accuracy of color-based indices such as Opponent HSV and modified indices, further filtering is carried out to eliminate any remaining noise, even though CCD sensors naturally minimise noise.
Thresholding: The three thresholds (80, 100, and 112) on pixel values are set during pre-processing. This is necessary in order to separate particular colour bands and characteristics associated with estimating nutrients.
Segmentation: To make sure that just the leaf pixels are examined, the leaf region is divided from the background. To extract colour parameters independently of surrounding elements, this segmentation is essential.
3.4 Nutrient analysis for post april pruning
Sampling techniques after April pruning and October pruning are different. The annual cycle growth of grapes involves many processes and events. The timing and duration of development of events vary due to rootstock variety, local climate and seasonal weather [4]. In India because of temporal zone the growth of the plant continues, hence the pruning process was developed for two times. We can classify the development processes into stages after April pruning viz. vegetative growth/ bud differentiation stage and after October pruning viz. flower cluster initiation (Bloom), fruit maturity and harvest. Many of these processes overlap each other during the development process. Vigorous and healthy growth is required during the pre - initiation stage of floral primordial, slow and less growth in fruit bud differentiation is required for higher yield of grapes. The need for N is higher during plant growth, the need for P is higher during flowering and the need for K is higher during crop maturity.
3.5 Nitrogen estimation using wavelength algorithm
The human color perception is scientifically and technically defined by colorimetry. Human visual sensitivity, illumination sources and spectral measurements are the major issues in colorimetry. The tri stimulus values X, Y and Z have been defined by CIE which represents the human color vision mathematically. CIE has defined various illumination sources such as A, B, C and D, but for this experiment we considered daylight D65 source with 1964 color observer. The method of CIE colorimetric specification depends on rules of color matching by additive color mixture. Mathematically, CIEXYZ is represented as equations 1 to 3 where S(A)is object spectrum, E(A)is the spectral power distribution (SPD) of an illuminant, and the color matching functions (CMFs) A(k) ={ x(X), y (X), z (X)} where X is the wavelength.
Often tri stimulus space is represented in 2D space and can be represented by normalized component as equation 4 to 6
Most of the researchers used the chlorophyll meter to estimate or relate the nutrient parameters of the plants. Chlorophyll measurement is a non-destructive method which has been proved to be less costly and effective method. Chlorophyll meter works on the principle of emission of two frequencies of light, one is red at wavelength 660 nm and other is infrared at 940 nm. Leaf chlorophyll absorbs red light but infrared is not absorbed. The difference between the absorptions is calculated as an index of chlorophyll. Since nitrogen is related with the chlorophyll contents, concept of chlorophyll meter can be extended for measurement of nitrogen from images. In this experiment an effort has been made to estimate the wavelength from leaf image and is illustrated as in figure 1. Values of RGB are calculated using XYZ for respective wavelength of CIE 1964 full record. These RGB values are useful in interpolation to estimate the wavelength based on image.
In this experiment an innovative effort has been made to estimate the wavelength of RGB image and used as new index to estimate nitrogen. To achieve the principle of wavelength-based estimation of nutrient parameter following algorithm is developed.
The calculated wavelength is the function of r, g and b values of the image and can be represented as equation 7
April pruning data along with variation is used for analysis purpose. Depending on the values of RGB and chart prepared, wavelength is interpolated for the corresponding image/plot. Using estimated wavelength index Lambda (L) regression analysis using LABfit
3.6 Nutrient analysis for post October pruning
The sample collection after October pruning differs from April pruning, also the nutrient requirement also differs at this stage. After October pruning the leaf which is opposite to the basal bunch collected as a sample. As discussed in a previous chapter the samples are collected from various fields in the Theni district with preferences to collect it from fields of April pruned samples. As time of pruning differs and inputs applied by farmers also vary, the samples collected after October pruning are also with uncontrolled input i.e. information of applied fertilizers is unknown or not considered.
N requirement is high after April pruning to promote vegetative growth and less in the reproductive period. The requirement of P is high during the flowering stage, which promotes fruit bud formation, flower induction and fruit set. Requirement of K is high during the maturity stage of fruit ripening and quality. Mg requirement is required along with K for sugar translocation and fruit ripening. There is an antagonism effect of K and Mg i.e. excess K suppress Mg. October pruning stage is the stage of flowering, fruit setting and maturity stage, hence for this experiment four fertilizer parameters N,P, K and Mg are considered or estimated.
3.7 Statistical approach for nutrient estimation
This section discusses the method of regression analysis for estimation of N, P, K and Mg nutrients for October pruned data using curve fitting methods using two different softwares. The extracted parameters used in the previous chapter are also extended for this analysis.
Same algorithms discussed in the last chapter were extended to extract the individual parameters R, G, B, H, S, V, Ikaw, Ipca, etc. for October pruned data. The parameters are extracted with three different thresholds at pixel value 80, 100 and 112. Three different Igreen indices were obtained with three different thresholds where as other parameters of RGB and HSV color planes along with L, a*, b*, Entropy, Ike and Ipca are calculated at a threshold of 112. These functions and individual parameters are correlated with each nutrient to find a most suitable function for nutrient estimation.
Here in this experiment the two new indices are proposed for nutrient estimation. The first is called as Opponent HSV and second is modified version of equation 8. Opponent HSV is named because HSV values obtained, lies between the range [0 1] and the singular index calculated by subtracting the values of H, S, and V from its maximum value "1" and averaged. As H is represented in angle from 0 to 3600 which is normalized to range 0 to 1, so (1-H) will represent the opposite angle and the same is applicable for saturation and brightness represents the opponent parts. The index OHSV is calculated as shown in equation 8.
Another index, which is a modified version of equation 8 is used to analyze the parameters. Nonlinear mapping of G to R and B is considered with logarithmic function instead of logsig function. It has been observed that newly derived OHSV and Log index gives better correlation with nutrients as compared to other functions if all varieties are considered. To understand the correlation of these parameters with different varieties analysis is carried out and found different results. But the combination of Thompson, Manikchaman and Sonaka (IMS) gave better correlation with these parameters and its correlation compared with all varieties in Table 2.
TMS shows the improved correlation with N, P, K, and Mg as compared to all four varieties. The parameters G, V and L show significant change in correlation with Nitrogen. Index Igreen Igi, Ig2 and a* has shown a better correlation with P , Index Igi, Ig2 and 2G+B/(R-B) has shown significant changes in correlation for K. Parameters B, H, Ikaw, Lsi, Ls2 and 2G+B/(R-B) shown the changes in the correlation coefficient for Mg. Newly Derived parameters OHSV and Ipki and Ipk2 has shown marginal changes in P and K correlation where as for N and Mg it is considerable. Based on the above formula the Color features extracted for various sample plots with Laboratory values of N, P, K and Mg.
Our research focused on the correlation between nutritional levels and a variety of colour and index-based indicators. To improve accuracy and predictive power, statistical methods such as multiple linear regression and least squares curve fitting can be applied. The association between the dependent variables-nutrient levels-and the independent variables-extracted colour indices-is demonstrated by the regression model. The sum of squared discrepancies between the model's predicted and observed nutrient values is minimised by the use of least squares curve fitting.
4 Discussions
4.1 Estimation of nitrogen
Nitrogen requirement as compared to after April pruning is less after October pruning and is an essential nutrient in development stage. Nitrogen has a role in flowering and have an influence on number of vine bunches and weight i.e. ultimately affects yield. Algorithm discussed in figure 1 is applied for extraction of color feature parameters using three color models - RGB, HSV and La*b*.
Color features/ parameters extracted to derive different functions. The correlation coefficient of these parameters and functions are illustrated in Table 3. Table 3 illustrates the correlation with and without illumination effect. Illumination effect is achieved by multiplying the individual parameter or the function by index Ik. Using derived functions regression analysis is carried out and newly developed Opponent HSV (OHSV) function found most suitable for Nitrogen estimation. The individual name indicates a single plot of a single variety.
The table illustrates the sample points used in regression analysis for both OHSV function i.e. OHSV with and without illumination effect. Estimated Nitrogen for all 35 samples along with percentage error is also shown.
Figure 2(a) shows the regression for Nitrogen estimation using newly derived function OHSV without illumination effect Iix and Labfit software is used for analysis of regression based on the data in Table 3. Figure 2(b) illustrates the graph of regression analysis for the N estimation using OHSV with illumination effect Ik
Figure 3(a) and 3(b) illustrates the error graph between the values of Laboratory and estimated nitrogen respectively.
4.2 Estimation of phosphorus
Though the requirement of phosphorus is less than nitrogen and potassium, it plays an important role in seed and fruit development stimulates flowering and have a favorable effect on bud fertility that promotes the yield of grapes. Table 3 shows the correlation coefficient of various parameters obtained from October pruned images. It shows the correlation for both the cases discussed earlier, i.e. the data without effect of illumination and with illumination.
Figure 4(a) shows the regression curve for the first analysis using Matlab software and figure 4 (b) shows the regression curve obtained using the analysis by LAB fit software. Figure 4(a) and 4(b) shows the error graph between the estimated values of P by these two softwares for Sample data and test data respectively.
4.3 Estimation of potassium (K)
Potassium is needed by grape plants for formation as sugar and starches which is essential for synthesis and cell division. Potassium plays an important role in quality of grapes by controlling its acidity and pH value of the juice. The deficiency will cause the leaves to drop prematurely, resulting in failure of fruit development in terms of color or ripening. Severe deficiency caused unevenly colored small berries and toxicity may induce deficiency of Mg which is known as antagonism effect. A deficiency of potassium leads to susceptibility of grapevine to powdery mildew.
The same algorithm that was applied to previous two nutrients is also used for estimation analysis of potassium. Table 3 shows the correlation of extracted parameters and derived functions which includes newly developed OHSV and Ipki and L?2 with potassium. Also, it shows the correlation of these parameters and functions when illumination index Ik is taken into consideration.
Table 4 shows the estimation of K using regression analysis. Newly developed Ipki and L?2 both are tested for estimation. Estimation of K using Ipki is carried out by MATLAB and LAB fit whereas Ipk2 used to estimate K with LAB fit software. These estimated values are indicated as Kl, K2 and K3 respectively.
From Table 4 it is observed that parameters S, V, functions Ikaw, IpCa, Ipki and Ipk2 show the positive correlation with potassium with correlation of coefficient R ranging from 0.22 to 0.42. Newly obtained function OHSV show negative correlation with the correlation coefficient value of -0.22. If illumination effect Ik taken into consideration S, V, Ikaw and Ipca show the marginal changes in coefficient where as Ipki and Ipk2 shows the significant change in the correlation, i.e. from 0.42 to 0.12. Since all other parameters show the marginal change in correlation coefficient and indices Ipki, IPk2 shows significant change, hence Ipki and Ipk2 are considered for analysis purpose.
Ipki index, which is a newly derived logarithmic index of RGB color parameters used for regression analysis and analyzed with two software MATLAB, LAB fit. Using the MATLAB curve fitting tool, the regression analysis for K, the curve is a quadratic polynomial equation having a coefficient ofdeterminationO.2041 with RMSE of 0. 6204.This quadratic equation is represented as equation 9.
Regression analysis for estimation of K using Ipki with LAB Fit tool have a coefficient of determination, R2 0.1853 and RMSE 0.6084 is represented as equation 10.
The curve shows the characteristics of the second order hyperbola. The curve has reduced chisquare of 0.31001 and error estimated using this curve have 7 points beyond the range of ?30 percent of sample data and test data. Regression analysis using Lm which is a logarithmic function of normalized G gives the best fit curve having a coefficient of determination R2 of 0.2528 and RMSE of 0.601. The curve equation is given by equation 10
The curve is a mixture of hyperbola and exponential curves which gives the reduced chi-square of 0.361336. The estimation errors for equation 11 have 7 points beyond the range of ?30 percent of sample data and 1 in test data.
The error curves plotted for two different regression analysis are compared with each other as shown in figure
5(a) and 5(b). Figure 5 clearly shows that the error is almost same. The error in test samples shows that the last sample is beyond the boundary of ?30 percent. The average errors for two regression equations 8 to 10 including sample and test data are -7.07, -6.84 and -7.27 respectively. Though the average error for K2 is less, but estimation samples within error boundary are less compared to Kl and K3. The temperature effect also affects the uptake of potassium.
The confidence intervals of the traditional models (CNN, SVM and Random Forest) are compared with the proposed OHSV and IPK1/IPK2) with the different parameters such as accuracy, 95% confidence interval, precision, recall, Fl -Score and AUC/ROC are showcased in Table 5. From the table 5, IPK1/IPK2 inferred that the confidence interval lies in the range of 90.8% - 95.2% , which provides the better result among the compared other models. The performance comparison of traditional model with OHSV and IPK1/IPK2 are showcased in Table 6. From it, IPK1/IK2 provides optimal performance when compared with other models.
5 Conclusion and future enhancement
Image processing technique has been proved its importance in Agriculture for detection of weeds, yield estimation, nutrient analysis specially nitrogen and chlorophyll estimation, vegetative growth, fruit sorting and pest detection. Combination of feature extractions like color, size and shape with different classifiers has added accuracy to these applications. This research is carried out to study effectiveness of Image Processing for nutrition analysis of grapes.
Nitrogen is related with the greenness of the leaf. Color image processing can be effectively used to estimate the nitrogen using RGB segmentation which leads to Igreen index Ig. Green index Ig showed better correlation with the nitrogen values obtained by chemical analysis. This index is compared with other indices that are derived by other researchers and it shows the competitive correlation. Regression analysis using statistical technique estimates the nitrogen. As greenness of leaf is related with the nitrogen content the wavelength estimation of green color of leaf is found to be effective method. This method is cost effective and temperature independent wavelength estimation from RGB parameters correlates to nitrogen. The wavelength and green plane ration have shown better correlation than individual parameters. Regression analysis of wavelength and chemical analysis leads to estimation of Nitrogen directly and linear equation is obtained. In the future, this work might be expanded to include other crop varieties besides grapes. Deep learning algorithms for feature selection and classification could be implemented, and remote sensing technologies like drone-mounted cameras could be used for large-scale crop nutrient monitoring. This would automate and streamline the process of analysing nutrient content.
References
[1] L. Chvyl and C. Williams. Viti-notes. Cooperative Research Centre for Viticulture (CRCV), Australian Wine Research Institute 2006
i- lU-JUUllClil VI 111V IVvJvUlvll lllOllllilV. ?*\J\J\J . [2] Edward W. Hellman. Grapevine structure and function. Oregon viticulture, Oregon State University Press, Corvallis, 5-19, 2003.
[3] VasifaA.Aglave, S. B. Paul and N.B. Sambre. Imaging Technique to Measure Leaf Area, Disease Severity and Chlorophyll Content: A Survey Paper. Journal of Computing Technologies, 1(3),2012.
[4] M.M. Ali, Ahmed Al-Ani, Derek Eamus and Daniel K.Y. Tan. Leaf Nitrogen determination using Handheld Meters. 16th Australian Agronomy conference-precision agriculture,2012.
[5] Paula F. Murakami, Michelle R. Turner, Abby K. van den Berg and Paul G. Schaberg. An Instructional Guide for Leaf Color Analysis using Digital Imaging Software. General Technical Report NE-327,1-37,2005.https://doi.org/10.2737/ NE-GTR-327
[6] Parviz Ahmadi Moghaddam, Mohammadali Haddad Derafshi and Vine Shirzad. Estimation of single leaf chlorophyll content in sugar beet using machine vision. Turk Journal of Agriculture and Forerestry,35(6):563-568,2011. https://doi.org/10.3906/tar-0909-393
[7] Mario Cupertino da Silva Junior, Francisco de Assis de Carvalho Pinto, Daniel Marcalde Queiroz, Enrique Anastacio Alves, Luis Manuel Navas Gracia and Jaime Gomez Gil. Correlation between vegetation indices and nitrogen leaf content and dry matter production in Brachiaria decumbens. Image Analysis for Agricultural Products and Processes, 145-150,2006.
[8] Han Yuzhu, Wang Xiaomei and Song Shuyao, Nitrogen determination in pepper (Capsicum frutescens L.) plants by color image analysis (RGB). African Journal of Biotechnology,10(77): 17737-17741,2011.https://doi.org/10.5897/ajbll.l974
[9] Gloria F. Mata-Donjuan, Adan Mercado-Luna, Enrique Rico-Garcia and Gilberto Herrera-Ruiz. Use of improved hue, luminance and saturation (IHLS) color space in the estimation of Nitrogen on tomato seedlings (Lycopersicon esculentum). Scientific Research and Essays,7(27):2343-2349,2012. https://doi.org/10.5897/srell.966
[10] Fuentes, S., Collins, C. and Rogers, G. Nondestructive estimation of grapevine nutritional status using proximal sensing techniques: A review. Sensors, 19(8), 2019.
[11] Erel, R. and Yermiyahu, U. Remote Sensing-based estimation of grapevine nutrition status under field conditions. Remote Sensing, 7(4):3822-3844,2015.
[12] Lechaudel, M., Tisseyre, B., Darriet, P., Coussement, P. and Bois, B. Comparison of near-infrared spectroscopy and partial least squares regression for the rapid estimation of nitrogen, phosphorus, potassium, and magnesium contents in grapevine leaves (Vitis vinifera cv. Sauvignon). Journal of Agricultural and Food Chemistry, 61(5): 1029-1035,2013.
[13] Diago, M. P., Fernandez-Novales, J. and Tardaguila, J. Assessing the nutritional status of vineyards by proximal sensing: Challenges and opportunities. Journal of the Science of Food and Agriculture,96(7);2277-2292,2016.
[14] Costa J. M., Ortuno M. F. and Chaves M. Deficit irrigation as a strategy to save water: Physiology and potential application to horticulture. Journal of Integrative Plant Biology,49(10),1421-1 434,2007. https://doi.org/10.1111/j. 1672-9072.2007.00556.x
© 2025. This work is published under https://creativecommons.org/licenses/by/3.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.