Next Article in Journal
Deep Q-Learning Technique for Offloading Offline/Online Computation in Blockchain-Enabled Green IoT-Edge Scenarios
Previous Article in Journal
Seismic Sequence Vulnerability of Low-Rise Special Moment-Resisting Frame Buildings with Brick Infills
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 12
Issue 16
10.3390/app12168233
applsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Predicting Fruit’s Sweetness Using Artificial Intelligence—Case Study: Orange

by
Mustafa Ahmed Jalal Al-Sammarraie
1,
Łukasz Gierz
2,*,
Krzysztof Przybył
3,
Krzysztof Koszela
4,
Marek Szychta
4,5,
Jakub Brzykcy
2 and
Hanna Maria Baranowska
6,*
1
Department of Agricultural Machinery and Equipment, College of Agricultural Engineering Sciences, University of Baghdad, Baghdad 10071, Iraq
2
Institute of Machine Design, Faculty of Mechanical Engineering, Poznań University of Technology, Piotrowo 3, 60-965 Poznan, Poland
3
Department of Dairy and Process Engineering, Faculty of Food Science and Nutrition, Poznań University of Life Sciences, Wojska Polskiego 31, 60-624 Poznan, Poland
4
Department of Biosystems Engineering, Poznań University of Life Sciences, Wojska Polskiego 50, 60-625 Poznan, Poland
5
Lukasiewicz Research Network—Poznań Institute of Technology, Starołecka 31, 60-963 Poznan, Poland
6
Department of Physics and Biophysics, Faculty of Food Science and Nutrition, Poznań University of Life Sciences, 38/42 Wojska Polskiego Street, 60-637 Poznan, Poland
*
Authors to whom correspondence should be addressed.
Appl. Sci. 2022, 12(16), 8233; https://doi.org/10.3390/app12168233
Submission received: 12 July 2022 / Revised: 3 August 2022 / Accepted: 15 August 2022 / Published: 17 August 2022
(This article belongs to the Section Food Science and Technology)
Browse Figures
Versions Notes

Abstract

:
The manual classification of oranges according to their ripeness or flavor takes a long time; furthermore, the classification of ripeness or sweetness by the intensity of the fruit’s color is not uniform between fruit varieties. Sweetness and color are important factors in evaluating the fruits, the fruit’s color may affect the perception of its sweetness. This article aims to study the possibility of predicting the sweetness of orange fruits based on artificial intelligence technology by studying the relationship between the RGB values of orange fruits and the sweetness of those fruits by using the Orange data mining tool. The experiment has applied machine learning algorithms to an orange fruit image dataset and performed a comparative study of the algorithms in order to determine which algorithm has the highest prediction accuracy. The results showed that the value of the red color has a greater effect than the green and blue colors in predicting the sweetness of orange fruits, as there is a direct relationship between the value of the red color and the level of sweetness. In addition, the logistic regression model algorithm gave the highest degree of accuracy in predicting sweetness.

1. Introduction

Oranges are one of the most popular and most commonly cultivated fruits in the world and they predominate in citrus production [1]. According to 2020 statistical data, orange production amounted to 76.29 million tonnes. Brazil is the world’s leader in orange production (17.07 million tonnes). Spain is the leader in Europe (3.23 million tonnes), for which orange production contributes 4% of the worldwide production. Egypt ranks sixth among the largest producers of oranges worldwide and it is the largest exporter of oranges in the world [2]. According to the data of the Central Agency for Public Mobilization and Statistics, the Egyptian orange is one of the most important exported fruits in Egypt’s trade, with its exports amounting to about EGP 493.28 million, which represents 11.3% of the value of all Egyptian agricultural exports. There are specific conditions for the growth of oranges, they grow in a wide variety of soils but deep-rooted clay soils are most ideal for growing oranges and a soil pH of 6.5 to 7.5 is also preferred. Orange fruit harvesting in Egypt usually begins during the months of December and January [3].
Due to the increased preference for a high level of immunity in the body, society is increasingly looking for natural vitamins and minerals. According to research, oranges are characterized by a high content of ascorbic acid (vitamin C), phytochemicals, carotenoids, flavonoids [4], and a little dietary fiber [5]. Consuming oranges or orange juice is recommended due to the varied amounts of bioactive compounds that they contain that contribute to the prevention of cardiovascular diseases and cancer [6,7]. Regular consumption has been shown to have a health-protective effect, for example against stomach cancer [8], and utility in slowing aging [9].
Appearance is the first factor that consumers focus on when they are deciding whether to accept or reject a product and, among this, color is the main appearance characteristic that is assessed. The color of a food product has, historically, provided information about other food qualities or even the safety of a certain food [10,11,12]. Flavor and color are important factors in evaluating fruits. More precisely, the appearance of fruits may cause a change in the perception of their flavor. If the appearance of the fruit is acceptable, then the flavor is the main sensory characteristic that is used to judge the quality of that fruit. Thus, the role of color and flavor in the assessment of the quality of fruits and the controlling factors in maintaining these characteristics must be taken into account [13]. The flavor quality of fruits and vegetables is affected by their genetics as well as pre-harvest, harvest and post-harvest factors. The longer the period between harvest and consumption, the greater the loss of the characteristic flavor (taste and aroma). The post-harvest period that begins to influence the flavor and nutritional quality of a fruit is shorter than that which begins to influence the fruit’s appearance and synthetic quality. Thus, it is necessary to ensure good flavor quality in the future by selecting the best production genotypes, using an integrated crop management system, harvesting at the ripening or ripeness stage that will improve the quality of eating at the time of consumption and using post-harvest procedures that will maintain the optimum flavor and the nutritional quality of fruits and vegetables between their harvest and consumption [14]. After transporting the fruits to warehouses for packaging, different operations need to be performed, such as sorting, grading, washing, packing and storage. The classification of fruits according to their characteristics, including their color, increases their marketing value [15]. There is a connection between the color and flavor, as green fruits are less sweet as compared to more richly colored fruits [10]. Support for this notion was obtained by Maga in 1974 [10], who discovered a decrease in the perceived sweetness of sucrose solutions when they were stained green. The color of the fruit is determined by the wavelength of light that is reflected from its surface. In biological materials, light differs greatly as a function of wavelength. These spectral differences provide a unique key for machine vision and image analysis.
The search for modern food assessment methods aims to allow one to effectively control and evaluate food products. Therefore, in this context, artificial intelligence methods that are supported by computer analysis seem to be the right tool for use to support the various decision-making processes and processes that are related to food processing and preservation [16]. Moreover, it is important to select a noninvasive technology with the lowest possible cost when aiming to obtain high quality food products [17].
At the present time, modern methods are used, including image processing technology [18,19,20]. This technology makes it possible to transmit images from high-resolution cameras to a computer in order for them to be analyzed by special programs.
Image processing technology is used in many areas, for example for the identification, dimensional measurements and quality assessment of kernels, tubers, vegetables and fruits [21,22,23,24,25,26,27], identification of pests [28], evaluation of the health of muscles and joints [28,29] and the evaluation of drying processes [30,31,32], but it is also used in the analysis of fruit color and classification and the measurement of leaf area [33]. Liming and Yanchao [34] worked on a strawberry classification system that classified the fruit according to their shape, size and color, based on image processing technology. Furthermore, Jalal Al-Sammarraie and Özbek used the image processing technique to classify pepper fruits based on their color [35].
Data analysis becomes a difficult task when using commercial software, which hampers the development of the science. The implementation of new algorithms and data analysis workflows is also slow. One solution to these problems is to use free programming environments such as Python or Octave in order to develop data processing algorithms. The Orange data mining tool allows the building of standard data processing workflows for various machine learning problems. Orange is also in the open source domain [36].
Research on the artificial intelligence system has become an interest of many scientists [37,38,39,40,41,42,43,44] as it has a high level of potential for the future. One such innovation is the present attempt to use artificial neural networks for the non-invasive recognition of the flavor of oranges using a color space model. The aim of the present study is to check the relationship between the RGB values of orange fruits and their sweetness and to determine which algorithm has the highest prediction accuracy.

2. Materials and Methods

2.1. Preparation of the Samples

The present research used Egyptian oranges that were purchased from Iraqi markets (Iraq, Baghdad). The choice of the research material (the fruits) depended on the fact the prices of Egyptian oranges on the international markets were more competitive than those that were exported by other international suppliers [45].
Before the experiment, the orange samples (50 fruits) were carefully selected, taking into account their different sizes and color levels. The average weight of each orange (sample) that was used in the study was 170.6 ± 0.1 g.

2.2. Measuring the Percentage of Sugar and the Amount of Water

Sugars are the main component of the soluble solids that are found in fruit juice, so the percentage of sugars may be used to estimate the degree of the fruit’s sweetness. A pocket molecular scanner SCiO (Consumer Physics Ltd., Illinois, USA) was used to measure the percentage of sugar (Brix) in oranges with an accuracy of 78%. The SCiO was considered to be the first accurate pocket-size spectrophotometer that is near-infrared (NIR) and smartphone operated and also connected to the cloud. Using the abovementioned device, the amount of water (moisture content) in the orange fruits was measured. The results found that there is an inverse relationship between the amount of water and the sugar content. This has also been previously suggested by other researchers [46].

2.3. Prediction Methodology

Different algorithms were used to determine the accuracy of the prediction of the sweetness of the oranges. The methodology consisted of three steps. The first step was to identify the most effective color in predicting the sweetness. The second step was to determine the most accurate algorithms used and their error scores. The third step was to determine the relationship between the value of the most effective color and the sweetness. Figure 1 shows a diagram of the sweetness prediction methodology.

2.4. Image Acquisition

As part of the preparations for the acquisition of the oranges’ images, a measurement and research platform was designed and constructed (Figure 2). The aim of the project was to obtain graphical data containing information that is encoded in the form of RGB values.
In the first step, the orange fruit was attached to a skewer. Pictures of each fruit were acquired from 6 side views at an angle of 90°. The next stage of the experiment was to adjust the distance of the camera (iPhone 12) in order to obtain the highest image quality, but also the smallest dimensions of the terminal. An iPhone 12 Pro camera was used to take the digital photos, as it features a quad camera system (12 + 12 + 12 MP and TOF 3D LiDAR scanner) with a field of view of 120 degrees. In order to maintain the repeatability of the results of the pictures that were taken, the device had to be calibrated. The iPhone 12 Pro was equipped with the Camera+ application for the advanced control of the image sharpness in manual mode. The following image exposure parameters were used in the research: sensitivity—ISO 320—and exposure time—2/3 s.
The device was mounted on a movable arm that was controlled by the torque of a servo motor. The distance between the camera and the fruit was 20 cm. The servo motor rotated the device through an angle of 90° and took the pictures. As a result of the rotation, a series of four photos was taken horizontally. The fruit was then manually turned to take two pictures vertically. As a result, six photos were obtained for one fruit (Figure 2b).
When studying the outer color of the fruit, a good lighting system is required that should provide uniform radiation, avoiding glare or shadows. Four side lamps were installed at an angle of 45 degrees perpendicular to the fruit. With this apparatus the reflection was reduced, thus avoiding unwanted glare. A similar lighting system was used by Fernandez et al. and Pedreschi et al. [47,48].

2.5. Image Processing

During the selection of the samples for the experiment, a difference was observed in the color of the orange fruits. Extracting the color features from each pixel (R, G and B) is mandatory for digital image analysis. Each pixel is made of three separate components that represent the RGB values. RGB values range from 0–255. After taking pictures of the orange fruits, the images from the phone’s camera were saved in the image .jpg format. The RGB values in the different orange fruits’ images were calculated using the ImageJ program that is shown in Figure 3 after selecting the borders of the fruit in the picture. ImageJ is a Java-based image processing application that was inspired by the NIH Image program for Macintosh. The application can view, edit, analyze, process, save and print images. It can also calculate area and pixel value statistics from user-defined selections and measure distances and angles. It can generate frequency distribution plots for intensity and linear profile plots. It supports standard image processing functions such as contrast processing, sharpening, smoothing, edge detection and average filtering. ImageJ was designed with an open architecture that provides extensibility via Java extensions. The following equation was used to measure the brightness of the images.
0.299R + 0.587G + 0.114B

2.6. Data Mining and Structure Algorithms

The Orange data mining tool has been selected for predictions and its ability to check the accuracy of those predictions. Orange is a data mining tool that is easily available online. It is open source and also provides a drag and drop feature. This tool is very useful for machine learning algorithmic applications. It provides an amazing graphic user interface and additional functions as well as Python scripting. This tool is very useful for researchers and much research has already been carried out using it [49].
For the identification of the most effective RGB traits in predicting sweetness, we used a set of scoring methods. These scoring methods are based on different principles, including “Info.gain”: the expected amount of information, “Gain ratio”: a ratio of the information gain and the attribute’s intrinsic information (which reduces the bias towards multivalued features that occurs in information gain), “Gini”: the inequality among the values of a frequency distribution, “ANOVA”: the difference between the average values of the feature in different classes, “X2” (chi-squared): the dependence between the feature and the class as measured by the chi-square statistic, “ReliefF”: the ability of an attribute to distinguish between classes in similar data instances and “FCBF” (Fast Correlation-Based Filter): an entropy-based measure, which also identifies redundancy due to the pairwise correlations between features.
After selecting the inputs (RGB values), we chose the classification algorithms in the Orange data mining tool in order to train the model to get the three outputs (very high sweetness, high sweetness and good sweetness). Figure 4 shows the sweetness prediction structure diagram.
The following classification algorithm was applied.

2.6.1. KNN (K–Nearest Neighbor)

The KNN algorithm is widely applied in pattern recognition and data mining for classification purpose as it is characterized by its simplicity and low error rate. The principle of the algorithm is that if the majority of similar samples of a query point in the feature space belong to a certain class, then a judgment can be made that the query point falls into that class. Similarity can be measured by distance in the feature space [50].

2.6.2. Tree

The decision tree algorithm is represented by a tree-like structure, wherein each inner node represents a test for a particular feature, each branch represents one of the possible test results and each leaf node represents a classification. Depending on the construction algorithms that are being applied, the decision tree models may vary [51].

2.6.3. SVMs (Support Vector Machines)

SVMs are derived from statistical learning theory, in order to classify points by assigning them against one of two separate halved distances. Therefore, an SVM essentially performs binary classification. For linearly separable data, an SVM finds the hyperplane, which increases the distance between the training samples and the class boundary. For cases that are not linearly separable, the samples are mapped to a high-dimensional space where a discrete hyperplane can be found. The allocation is performed by a mechanism that is called a kernel function [52].

2.6.4. Neural Network

A neural network is a set of freely formed algorithms that are based on the principle of the human brain, which is designed to recognize different patterns. The algorithm interprets sensory data through some kind of machine perception, tagging or input method. A neural network works in a similar way to the network of neurons that is found in the human brain. The “neurons” of a neural network are the mathematical applications that collect and index information according to a specific designer’s design. A neural network consists of layers that are connected by nodes [53].

2.6.5. Logistic Regression

Logistic regression is a classification algorithm that is used to generate observations about a set of categories. Some examples of such classification problems are emails being sorted into spam or non-spam and the discernment of fake or non-fake online transactions. Logistic regression changes its output by using a sigmoid logistic function to return the possible values [54].

3. Results and Discussion

Based on the results of the RGB values and the use of the rank widget in the different scoring methods, it was found that the trait of the red color had a significant effect on the classification of the sweetness of orange fruits. The rank widget considers class-labeled data sets (classification or regression) and scores the attributes according to their correlation with the class. The following Table 1 shows the classification percentage for the RGB trait.
All of the scoring methods indicated that the red color had a significant effect in predicting the sweetness of orange fruits. The association of the red color with sweetness has also been suggested by other researchers [53].
After selecting the most effective color in the prediction model and evaluating the extracted data, an ANOVA F-test was used at a probability level of p < 0.001. This gave the highest value in the color red in the classification percentage for the RGB traits. Using the resultant box plot, the fruit flavor classification was studied based on the values of the red color (Figure 5).
The orange fruits were categorized into very high sweetness fruits, high sweetness fruits and good sweetness fruits, as is shown in Figure 3. A total of 9 fruits were rated as having good sweetness, 33 fruits as having high sweetness and 8 fruits as having very high sweetness. All of the fruits with good sweetness had a red color value that was less than or equal to 231.38 with a mean of 22.89 (±9.26) and a median of 224.52. The fruits with high sweetness had a red color value that was less than or equal to 249.15 with a mean of 240.06 (±5.51) and a median of 239.9. The fruits that were classified as having very high sweetness had a red color value that was greater than 249.15, the mean of which was 251.66 (±1.27) and the median was 251.52. The relationship was broadly positive with the increasing value of the red color (p < 0.001).
After determining the color affecting the sour flavor classification of orange fruits, we were able to report the results of the predictions, followed by the algorithms that were used. Logistic regression gave us better results for sweetness prediction, because the accuracy of logistic regression is better than that of the KNN, tree, SVM and neural network methods. Conversely, the other algorithms also make good predictions, but none more so than a logistic regression model. Figure 6 shows the prediction values of the different algorithms.
In Figure 6, above, we have compared the results of the different algorithms that we applied to the orange fruit dataset in order to essentially compare the accuracy of the algorithms. We found that logistic regression gave us the best accuracy during prediction. In the KNN model, the classification accuracy was 82%; in the tree model, the classification accuracy was 96%; in the SVM model, the classification accuracy was 93% and in the neural network model, the classification accuracy was 88% (all of which are no more efficient than the logistic regression ratio). In the logistic regression algorithm, the classification accuracy was 97%, which is more efficient than the accuracy of all of the other algorithms. This was demonstrated by Ahmed et al. in predicting pneumonia, in their research they found that logistic regression had the best predictive accuracy [53].
Next, for a better level of understanding, we attached a confusion matrix of positive and negative results. The blue values in the below figure show the positive or accurate results and the red values indicate the negative or false results (Figure 7).
From the confusion matrix it can be seen that a good sweetness fruit was rated as a high sweetness fruit once, a high sweetness fruit was classified as a good sweetness fruit twice and a very high sweetness fruit once. In order to understand the reasons for this error, we used a scatter plot (Figure 8).
Figure 8 shows the scatter plot for the sweetness values based on the red and green color values. These two colors (red and green) gave the best scattered distribution of the data through which errors can be determined. Therein, the spectra belonging to different classes are well separated; for example, the good sweetness fruit group (blue dots) is only just touching the high sweetness fruit group (red dots). Also, the high sweetness fruit group (red dots) is only just touching the very high sweetness fruit group (green dots). The reason for these small overlaps is that the scored results are too close together, which makes it difficult for the algorithm to classify the fruit.
After determining the color that has a significant influence in predicting the sweetness of orange fruits and determining the best algorithms, the relationship between the value of the red color and the sweetness of the orange fruits was determined. Figure 9 shows the relationship between the orange fruits’ red color values and their sweetness.
Figure 9, above, shows that, based on the values of the red color and sweetness of orange fruits, there is a direct relationship between the value of the red color and the sweetness of the orange fruits. This has also been established by previous studies of the sweetness of the fruit [10]. In other words, the reddish oranges are sweeter. With the higher red color of skin, the fruits’ sugar content and sweetness increase [54].
Orange fruits have a color that is close to red, which gives them sweetness. We discovered a more typical flavor in the samples that showed colors that are closer to that of the typical fruit drink in each case. These results are in agreement with those that have been obtained by other authors. Dubose et al. found that the color of fruit-flavored beverages had significant effects on their perceived flavor intensity [55].

4. Conclusions

We conclude that color values can be used to predict fruit’s sweetness. Different algorithms were used, namely logistic regression, KNN, tree, SVM and neural network methods. We applied a logistic regression algorithm to an orange fruit dataset in order to train our model. After the prediction, we concluded that logistic regression gave us a prediction accuracy of 97%, KNN gave us an accuracy of 82%, the tree method gave us an accuracy of 96%, SVM gave us an accuracy of 93% and the neural network gave us an accuracy of 88%. All of the tested algorithms are good for generating predictions; however, in this project, we can say that logistic regression gave us better results for predicting the sweetness of oranges. The results suggest that there is a direct relationship between the value of the red color of orange fruits and their sweetness. Based on the above results, it can be concluded that artificial intelligence technology can be used for predicting the sweetness of orange fruits and, thereby, reduce their sorting time and provide consumers with the best-tasting fruit, especially in convenient and affordable forms—an improvement that is likely to increase their consumption.

Author Contributions

Conceptualization, M.A.J.A.-S. and Ł.G.; methodology, M.A.J.A.-S., Ł.G. and K.P.; software, M.A.J.A.-S. and Ł.G.; validation, M.A.J.A.-S., Ł.G. and K.P.; formal analysis, M.A.J.A.-S., Ł.G. and K.P.; investigation, M.A.J.A.-S., Ł.G. and K.P.; resources, M.A.J.A.-S., Ł.G. and M.S.; data curation, M.A.J.A.-S. and Ł.G.; writing—original draft preparation, M.A.J.A.-S., Ł.G. and K.P.; writing—review and editing, M.A.J.A.-S., Ł.G., K.P. and H.M.B.; visualization, M.A.J.A.-S., Ł.G. and J.B.; supervision, Ł.G., K.K. and H.M.B.; project administration, Ł.G.; funding acquisition, H.M.B., Ł.G. and K.P. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

Not applicable.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Zouaoui, B.; Djilali, B.; Ahmed, H.; Mokhtar, B. Effect of thermal pasteurization on phytochemical characteristics and antioxidant capacity of orange juice. Nat. Technol. 2020, 12, 50–53. [Google Scholar]
  2. FAO CF. Processed Statistical Bulletin. 2016. Available online: https://www.fao.org/3/cb6492en/cb6492en.pdf (accessed on 1 June 2022).
  3. Bopitiya, D.; Hearn, M.T.; Zhang, J.; Bennett, L.E. Production of hydrogen peroxide in commercial orange juice products is related to proximate composition, processing conditions and storage time. Food Chem. 2022, 395, 133619. [Google Scholar] [CrossRef]
  4. Galaverna, G.; Dall’Asta, C. Production Processes of Orange Juice and Effects on Antioxidant Components. In Processing and Impact on Antioxidants in Beverages; Preedy, V., Ed.; Academic Press: San Diego, CA, USA, 2014; pp. 203–214. [Google Scholar]
  5. Scheffers, F.R.; Boer, J.M.A.; Verschuren, W.M.M.; Verheus, M.; Van Der Schouw, Y.T.; Sluijs, I.; Smit, H.A.; Wijga, A.H. Pure fruit juice and fruit consumption and the risk of CVD: The European Prospective Investigation into Cancer and Nutrition-Netherlands (EPIC-NL) study. Br. J. Nutr. 2018, 121, 351–359. [Google Scholar] [CrossRef] [PubMed]
  6. Del Rio, D.; Rodriguez-Mateos, A.; Spencer, J.P.E.; Tognolini, M.; Borges, G.; Crozier, A. Dietary (Poly)phenolics in human health: Structures, bioavailability, and evidence of protective effects against chronic diseases. Antioxid. Redox Signal. 2013, 18, 1818.e92. [Google Scholar] [CrossRef] [PubMed]
  7. Bertuccio, P.; Alicandro, G.; Rota, M.; Pelucchi, C.; Bonzi, R.; Galeone, C.; Bravi, F.; Johnson, K.C.; Hu, J.; Palli, D.; et al. Citrus fruit intake and gastric cancer: The stomach cancer pooling (StoP) project consortium. Int. J. Cancer 2019, 144, 2936–2944. [Google Scholar] [CrossRef]
  8. Tan, B.L.; Norhaizan, M.E.; Liew, W.-P.-P.; Rahman, H.S. Antioxidant and Oxidative Stress: A Mutual Interplay in Age-Related Diseases. Front. Pharmacol. 2018, 9, 1162. [Google Scholar] [CrossRef] [PubMed]
  9. Clydesdale, F.M. Color as a factor in food choice. Crit. Rev. Food Sci. Nutr. 1993, 33, 83–101. [Google Scholar] [CrossRef]
  10. Francis, F.J. Quality as influenced by color. Food Qual. Prefer. 1995, 6, 149–155. [Google Scholar] [CrossRef]
  11. Przybył, K.; Gawałek, J.; Gierz, L.; Łukomski, M.; Zaborowicz, M.; Boniecki, P. Recognition of color changes in strawberry juice powders using self-organizing feature map. In Proceedings of the Tenth International Conference on Digital Image Processing (ICDIP 2018), Shanghai, China, 9 August 2018; SPIE: Washington, DC, USA, 2018; Volume 10806, pp. 563–571. [Google Scholar] [CrossRef]
  12. Shewfelt, R.L. Flavor and color of fruits as affected by processing. In Commercial Fruit Processing; Springer: Berlin/Heidelberg, Germany, 1986; pp. 481–529. [Google Scholar]
  13. Kader, A.A. Flavor quality of fruits and vegetables. J. Sci. Food Agric. 2008, 88, 1863–1868. [Google Scholar] [CrossRef]
  14. Örnek, M.N. Havuç Sınıflandırmada Gerçek Zamanlı Görüntü İşleme Makinası Tasarımı ve Bazı Mekanik Sınıflandırma Makinaları ile Boylama Etkinliklerinin Karşılaştırılması. Ph.D. Thesis, Selcuk University, Institute of Science and Technology, Agricultural Machinery, Konya, Turkey, 2014. [Google Scholar]
  15. Przybył, K.; Ryniecki, A.; Niedbała, G.; Mueller, W.; Boniecki, P.; Zaborowicz, M.; Koszela, K.; Kujawa, S.; Kozłowski, R.J. Software supporting definition and extraction of the quality parameters of potatoes by using image analysis. In Proceedings of the Eighth International Conference on Digital Image Processing (ICDIP 2016), Chengu, China, 26 October 2016; SPIE: Washington, DC, USA, 2016; Volume 100332L. [Google Scholar] [CrossRef]
  16. Pawlak, T.; Pilarska, A.A.; Przybył, K.; Stangierski, J.; Ryniecki, A.; Cais-Sokolińska, D.; Pilarski, K.; Peplińska, B. Application of Machine Learning Using Color and Texture Analysis to Recognize Microwave Vacuum Puffed Pork Snacks. Appl. Sci. 2022, 12, 5071. [Google Scholar] [CrossRef]
  17. Przybył, K.; Gawałek, J.; Koszela, K.; Przybył, J.; Rudzińska, M.; Gierz, Ł.; Domian, E. Neural Image Analysis and Electron Microscopy to Detect and Describe Selected Quality Factors of Fruit and Vegetable Spray-Dried Powders—Case Study: Chokeberry Powder. Sensors 2019, 19, 4413. [Google Scholar] [CrossRef]
  18. Gierz, Ł.; Przybył, K.; Koszela, K.; Duda, A.; Ostrowicz, W. The Use of Image Analysis to Detect Seed Contamination—A Case Study of Triticale. Sensors 2021, 21, 151. [Google Scholar] [CrossRef]
  19. Przybył, K.; Duda, A.; Koszela, K.; Stangierski, J.; Polarczyk, M.; Gierz, Ł. Classification of Dried Strawberry by the Analysis of the Acoustic Sound with Artificial Neural Networks. Sensors 2020, 20, 499. [Google Scholar] [CrossRef]
  20. Nowakowski, K.; Raba, B.; Tomczak, R.J.; Boniecki, P.; Kujawa, S.; Nowak, P.J.; Matz, R. Identification of Physical Parameters of Cereal Grain using Computer image Analysis and Neural Models. In Proceedings of the 5th International Conference on Digital Image Processing (ICDIP 2013), Beijing, China, 21–22 April 2013; SPIE: Washington, DC, USA, 2013; Volume 8878, p. 887823. [Google Scholar] [CrossRef]
  21. Boniecki, P.; Piekarska-Boniecka, H.; Świerczyński, K.; Koszela, K.; Zaborowicz, M.; Przybył, J. Detection of the granary weevil based on x-ray images of damaged wheat kernels. J. Stored Prod. Res. 2014, 56, 38–42. [Google Scholar] [CrossRef]
  22. Boniecki, P.; Piekarska-Boniecka, H.; Koszela, K.; Zaborowicz, M.; Przybył, K.; Wojcieszak, D.; Zbytek, Z.; Ludwiczak, A.; Przybylak, A.; Przybył, J. Neural classifier in the estimation process of maturity of selected varieties of apples. In Proceedings of the 7th International Conference on Digital Image Processing (ICDIP 2015), Los Angeles, CA, USA, 9–10 April 2015; SPIE: Washington, DC, USA, 2015; Volume 9159. [Google Scholar] [CrossRef]
  23. Przybył, K.; Zaborowicz, M.; Koszela, K.; Boniecki, P.; Mueller, W.; Raba, B.; Lewicki, A. Organoleptic damage classification of potatoes with use of image analysis in production process. In Proceedings of the 6th International Conference on Digital Image Processing (ICDIP 2014), Athens, Greece, 16 April 2014; SPIE: Washington, DC, USA, 2014; Volume 9159, p. 91590W. [Google Scholar] [CrossRef]
  24. Przybył, K.; Górna, K.; Wojcieszak, D.; Janczak, D.; Ludwiczak, A.; Przybylak, A.; Boniecki, P.; Koszela, K.; Zaborowicz, M.; Lewicki, A. The recognition of potato varieties using of neural image analysis method. In Proceedings of the Seventh International Conference on Digital Image Processing (ICDIP 2015), Los Angeles, CA, USA, 9–10 April 2015; SPIE: Washington, DC, USA, 2015; Volume 9631, pp. 268–273. [Google Scholar] [CrossRef]
  25. Zaborowicz, M.; Boniecki, P.; Koszela, K.; Przybylak, A.; Przybył, J. Application of neural image analysis in evaluating the quality of greenhouse tomatoes. Sci. Hortic. 2017, 218, 222–229. [Google Scholar] [CrossRef]
  26. Koszela, K.; Raba, B.; Zaborowicz, M.; Przybył, K.; Wojcieszak, D.; Czekała, W.; Ludwiczak, A.; Przybylak, A.; Boniecki, P.; Przybył, J. Computer image analysis in caryopses quality evaluation as exemplified by malting barley. In Proceedings of the Seventh International Conference on Digital Image Processing (ICDIP 2015), Los Angeles, CA, USA, 9–10 April 2015; SPIE: Washington, DC, USA, 2015; Volume 9631, pp. 200–206. [Google Scholar] [CrossRef]
  27. Boniecki, P.; Piekarska-Boniecka, H.; Koszela, K.; Nowakowski, K.; Kujawa, S.; Majewski, A.; Weres, J.; Raba, B. Neural identification of selected apple pests. Comput. Electron. Agric. 2015, 110, 9–16. [Google Scholar] [CrossRef]
  28. Przybylak, A.; Boniecki, P.; Koszela, K.; Ludwiczak, A.; Zaborowicz, M.; Lisiak, D.; Stanisz, M.; Ślósarz, P. Estimation of intramuscular level of marbling among Whiteheaded Mutton Sheep lambs. J. Food Eng. 2016, 168, 199–204. [Google Scholar] [CrossRef]
  29. Zaborowicz, M.; Włodarek, J.; Przybył, K.; Wojcieszak, D.; Czekała, W.; Ludwiczak, A.; Przybylak, A.; Boniecki, P.; Koszela, K.; Przybył, J.; et al. Image Acquisitions, Processing and Analysis in the Process of Obtaining Characteristics of Horse Navicular Bone. In Proceedings of the Seventh International Conference on Digital Image Processing (ICDIP 2015), Los Angeles, CA, USA, 9–10 April 2015; SPIE: Washington, DC, USA, 2015; Volume 9631, pp. 237–241. [Google Scholar] [CrossRef]
  30. Przybył, K.; Samborska, K.; Koszela, K.; Masewicz, L.; Pawlak, T. Artificial neural networks in the evaluation of the influence of the type and content of carrier on selected quality parameters of spray dried raspberry powders. Measurement 2021, 186, 110014. [Google Scholar] [CrossRef]
  31. Koszela, K.; Łukomski, M.; Mueller, W.; Górna, K.; Okoń, P.; Boniecki, P.; Zaborowicz, M.; Wojcieszak, D. Classification of dried vegetables using computer image analysis and artificial neural networks. In Proceedings of the Ninth International Conference on Digital Image Processing (ICDIP 2017), Hong Kong, China, 19–22 May 2017; SPIE: Washington, DC, USA, 2017; Volume 10420, pp. 650–656. [Google Scholar] [CrossRef]
  32. Sun, D.W. Computer vision-an objective, rapid and non-contact quality evaluation tool for the food industry. J. Food Eng. 2004, 61, 1–2. [Google Scholar] [CrossRef]
  33. Liming, X.; Yanchao, Z. Automated strawberry grading system based on image processing. Comput. Electron. Agric. 2010, 71, S32–S39. [Google Scholar] [CrossRef]
  34. Jalal Al-Sammarraie, M.A.; Özbek, O. Comparison of the Effect Using Color Sensor and Pixy2 Camera on the Classification of Pepper Crop. J. Mech. Eng. Res. Dev. 2021, 44, 396–403. [Google Scholar]
  35. Toplak, M.; Birarda, G.; Read, S.; Sandt, C.; Rosendahl, S.M.; Vaccari, L.; Demšar, J.; Borondics, F. Infrared orange: Connecting hyperspectral data with machine learning. Synchrotron Radiat. News 2017, 30, 40–45. [Google Scholar] [CrossRef]
  36. Przybył, K.; Gawałek, J.; Koszela, K.; Wawrzyniak, J.; Gierz, L. Artificial Neural Networks and Electron Microscopy to Evaluate the Quality of Fruit and Vegetable Spray-Dried Powders. Case Study: Strawberry Powder. Comput. Electron. Agric. 2018, 155, 314–323. [Google Scholar] [CrossRef]
  37. Mueller, W.; Idziaszek, P.; Gierz, Ł.; Przybył, K.; Wojcieszak, D.; Frankowski, J.; Koszela, K.; Boniecki, P.; Kujawa, S. Mapping and visualization of complex relational structures in the graph form using the Neo4j graph database. In Proceedings of the Eleventh International Conference on Digital Image Processing (ICDIP 2019), Guangzhou, China, 10–13 May 2019; SPIE: Washington, DC, USA, 2019; Volume 11179, pp. 581–587. [Google Scholar] [CrossRef]
  38. Semkło, Ł.; Gierz, Ł. Research on the measurement of spraying time with seed treatment agent using an innovative valve. J. Phys. Conf. Ser. 2021, 1736, 012010. [Google Scholar] [CrossRef]
  39. Gierz, Ł.; Przybył, K.; Koszela, K.; Semkło, Ł.; Kwiecień, S. An Assessment of the Functional and Ecological Aspect of Novel Intermittent Stream Valves for Spraying Seed Potatoes. Agronomy 2020, 10, 541. [Google Scholar] [CrossRef]
  40. Boniecki, P.; Przybył, J.; Kuzimska, T.; Mueller, W.; Raba, B.; Lewicki, A.; Przybył, K.; Zaborowicz, M.; Koszela, K. Neural image analysis in the process of quality assessment: Domestic pig oocytes. In Proceedings of the Sixth International Conference on Digital Image Processing (ICDIP 2014), Athens, Greece, 5 April 2014; SPIE: Washington, DC, USA, 2014; Volume 9159, pp. 138–146. [Google Scholar]
  41. Szychta, M.; Szymczyk, S.; Przybył, K.; Koszela, K.; Duda, A. Comparison of methods of obtaining visual data in the shape of obstacles in wastelands and forest areas for the purpose of automatic control of the stability of self-propelled machines. In Proceedings of the Thirteenth International Conference on Digital Image Processing (ICDIP 2021), Virtual, 20–23 May 2021; SPIE: Washington, DC, USA, 2021; Volume 11878, pp. 463–468. [Google Scholar]
  42. Przybył, K.; Masewicz, Ł.; Koszela, K.; Duda, A.; Szychta, M.; Gierz, Ł. An MLP artificial neural network for detection of the degree of saccharification of Arabic gum used as a carrier agent of raspberry powders. In Proceedings of the Thirteenth International Conference on Digital Image Processing (ICDIP 2021), Virtual, 20–23 May 2021; Jiang, X., Fujita, H., Eds.; SPIE: Washington, DC, USA, 2021; Volume 11878, p. 93. [Google Scholar] [CrossRef]
  43. Przybył, K.; Koszela, K.; Adamski, F.; Samborska, K.; Walkowiak, K.; Polarczyk, M. Deep and Machine Learning Using SEM, FTIR, and Texture Analysis to Detect Polysaccharide in Raspberry Powders. Sensors 2021, 21, 5823. [Google Scholar] [CrossRef]
  44. Hassanain, H.T.; Gabr, R.H. An economic study on Egyptian orange exports and its competitiveness in the international markets. Zagazig J. Agric. Res. 2020, 47, 623–639. [Google Scholar] [CrossRef]
  45. Abobatta, W.F. Development Growth and Productivity of Orange Orchards (Citrus Sinensis L) in Egypt (Delta Region). Adv. Agric. Technol. Plant Sci. 2018, 1, 180003. [Google Scholar]
  46. Rubio-Arraez, S.; Sahuquillo, S.; Capella, J.V.; Ortolá, M.D.; Castelló, M.L. Influence of healthy sweeteners (tagatose and oligofructose) on the physicochemical characteristics of orange marmalade. J. Texture Stud. 2015, 46, 272–280. [Google Scholar] [CrossRef]
  47. Fernandez, L.; Castillero, C.; Aguilera, J.M. An application of image analysis to dehydration of apple discs. J. Food Eng. 2005, 67, 185–193. [Google Scholar] [CrossRef]
  48. Pedreschi, F.; León, J.; Mery, D.; Moyano, P. Development of a computer vision system to measure the color of potato chips. Food Res. Int. 2006, 39, 1092–1098. [Google Scholar] [CrossRef]
  49. Demšar, J.; Zupan, B. Orange: Data mining fruitful and fun. Inf. Družba IS 2012, 6, 1–486. [Google Scholar]
  50. Kuang, Q.; Zhao, L. A practical GPU based kNN algorithm. In Proceedings of the International Symposium on Computer Science and Computational Technology (ISCSCI 2009), Hyderabad, India, 15–17 January 2009; Academy Publisher: Orlando, FL, USA, 2009; p. 151. [Google Scholar]
  51. Che, D.; Liu, Q.; Rasheed, K.; Tao, X. Decision tree and ensemble learning algorithms with their applications in bioinformatics. Softw. Tools Algorithms Biol. Syst. 2011, 696, 191–199. [Google Scholar]
  52. Alba, E.; Garcia-Nieto, J.; Jourdan, L.; Talbi, E.G. Gene selection in cancer classification using PSO/SVM and GA/SVM hybrid algorithms. In Proceedings of the IEEE Congress on Evolutionary Computation, Singapore, 25–28 September 2007; pp. 284–290. [Google Scholar]
  53. Ahmad, H.M.; Sohail, M.; Ahmad, M.M.; Iqbal, S.; Sarfaraz, A.; Noor, K. Predictions of Pneumonia Disease using Image Analytics in Orange Tool. In Proceedings of the GS International Conference on Computer Science and Engineering 2020 (GSICCSE 20), Beijing China, 22–24 May 2020. [Google Scholar]
  54. Kondo, N.; Ahmad, U.; Monta, M.; Murase, H. Machine vision based quality evaluation of Iyokan orange fruit using neural networks. Comput. Electron. Agric. 2000, 29, 135–147. [Google Scholar] [CrossRef]
  55. DuBose, C.N.; Cardello, A.V.; Maller, O. Effects of colorants and flavorants on identification, perceived flavor intensity, and hedonic quality of fruit-flavored beverages and cake. J. Food Sci. 1980, 45, 1393–1399. [Google Scholar] [CrossRef]
Applsci 12 08233 g001 550
Figure 1. Diagram of the sweetness prediction methodology.
Figure 1. Diagram of the sweetness prediction methodology.
Applsci 12 08233 g001
Applsci 12 08233 g002 550
Figure 2. A diagram of pictures taking platform. (a) Side view; (b) arm rotation angle; (1) the base; (2) servo motor; (3) iPhone mobile; (4) orange fruit; (5) fruit stabilizer.
Figure 2. A diagram of pictures taking platform. (a) Side view; (b) arm rotation angle; (1) the base; (2) servo motor; (3) iPhone mobile; (4) orange fruit; (5) fruit stabilizer.
Applsci 12 08233 g002
Applsci 12 08233 g003 550
Figure 3. Measurement of RGB values using the ImageJ program.
Figure 3. Measurement of RGB values using the ImageJ program.
Applsci 12 08233 g003
Applsci 12 08233 g004 550
Figure 4. The sweetness prediction structure diagram.
Figure 4. The sweetness prediction structure diagram.
Applsci 12 08233 g004
Applsci 12 08233 g005 550
Figure 5. Classification of fruits based on red color values. (N): actual numbers of outlets; (p): probability.
Figure 5. Classification of fruits based on red color values. (N): actual numbers of outlets; (p): probability.
Applsci 12 08233 g005
Applsci 12 08233 g006 550
Figure 6. The prediction values of the different algorithms.
Figure 6. The prediction values of the different algorithms.
Applsci 12 08233 g006
Applsci 12 08233 g007 550
Figure 7. Confusion matrix of positive and negative results.
Figure 7. Confusion matrix of positive and negative results.
Applsci 12 08233 g007
Applsci 12 08233 g008 550
Figure 8. Scatter plot.
Figure 8. Scatter plot.
Applsci 12 08233 g008
Applsci 12 08233 g009 550
Figure 9. The relationship between orange fruits’ red color values and sweetness.
Figure 9. The relationship between orange fruits’ red color values and sweetness.
Applsci 12 08233 g009
Table
Table 1. The classification percentage for RGB traits.
Table 1. The classification percentage for RGB traits.
TraitsInfo.gainGain RatioGiniANOVAX2ReliefFFCBF
R0.8190.410.29351.8525.2760.1421.009
G0.2550.1280.0780.5010.5250.0750.185
B0.2850.1430.1040.4674.7840.0080
(R+G+B)/30.1430.0720.030.6540.4850.0470
Brightness0.2040.1020.0530.7650.6880.060
Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Share and Cite

MDPI and ACS Style

Al-Sammarraie, M.A.J.; Gierz, Ł.; Przybył, K.; Koszela, K.; Szychta, M.; Brzykcy, J.; Baranowska, H.M. Predicting Fruit’s Sweetness Using Artificial Intelligence—Case Study: Orange. Appl. Sci. 2022, 12, 8233. https://doi.org/10.3390/app12168233

AMA Style

Al-Sammarraie MAJ, Gierz Ł, Przybył K, Koszela K, Szychta M, Brzykcy J, Baranowska HM. Predicting Fruit’s Sweetness Using Artificial Intelligence—Case Study: Orange. Applied Sciences. 2022; 12(16):8233. https://doi.org/10.3390/app12168233

Chicago/Turabian Style

Al-Sammarraie, Mustafa Ahmed Jalal, Łukasz Gierz, Krzysztof Przybył, Krzysztof Koszela, Marek Szychta, Jakub Brzykcy, and Hanna Maria Baranowska. 2022. "Predicting Fruit’s Sweetness Using Artificial Intelligence—Case Study: Orange" Applied Sciences 12, no. 16: 8233. https://doi.org/10.3390/app12168233

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop