Next Article in Journal
A Filtering Switch Made by an Improved Coupled Microstrip Line
Previous Article in Journal
An Improved LSTM-Based Failure Classification Model for Financial Companies Using Natural Language Processing
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Applied Sciences
Volume 13
Issue 13
10.3390/app13137885
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

Construction of Color Prediction Model for Damaged Korla Pears during Storage Period

by
Rui Zhang
1,2,
Shiyuan Li
1,2,
Yang Liu
1,2,
Guowei Li
1,2,
Xin Jiang
1,2 and
Xiuwen Fan
1,2,*
1
College of Mechanical Electrification Engineering, Tarim University, Alaer 843300, China
2
Agricultural Engineering Key Laboratory, Ministry of Higher Education of Xinjiang Uygur Autonomous Region, Tarim University, Alar 843300, China
*
Author to whom correspondence should be addressed.
Appl. Sci. 2023, 13(13), 7885; https://doi.org/10.3390/app13137885
Submission received: 17 May 2023 / Revised: 29 June 2023 / Accepted: 3 July 2023 / Published: 5 July 2023
Browse Figures
Versions Notes

Abstract

:
In this work, to scientifically predict the color of damaged Korla fragrant pears during the storage period with lower economic loss and improved added value of the fragrant pears, eight pericarp color prediction models of damaged Korla fragrant pears during the storage period were established. These models had different membership functions, which were based on the adaptive neuro-fuzzy inference system (ANFIS). The optimal model was chosen and verified. Finally, the pericarp color of fragrant pears was accurately predicted through the degree of damage and storage time. According to the acquired test results, the pericarp brightness (L*) decreased, while both the red–green (a*) and yellow–blue (b*) values increased as the storage time prolonged. In addition, the pericarp color of the damaged fragrant pears during the storage period could be well predicted by using the ANFIS model. More specifically, the model with a membership function of trimf showed the optimal prediction effects of L*, a*, and b* (RMSE = 0.1089, R2 = 0.9773; RMSE = 0.5894, R2 = 0.9853; and RMSE = 0.2360, R2 = 0.9772). Our work provides valuable insights for the prediction of the quality of Korla fragrant pears during the storage period.

1. Introduction

The Korla fragrant pear belongs to the white pear system, which is a characteristic fruit and main cultivar in Xinjiang, China, and has a planting history of over 1400 years. The fruit has promising markets in China and overseas due to its dark green color, crisp and sweet taste, rich juice, thick fragrance, and high nutrient values [1,2]. However, Korla fragrant pears can be easily affected by mechanical damages at harvest, transport, classification, and packaging due to the thin pericarp and crisp pulp. During the storage period, fragrant pears begin to rot since the damaged area can quickly expand. Alarmingly, the rotting rate can reach over 15%. To ensure quality, practitioners abandon the damaged fragrant pears directly in storage or before marketing, resulting hence in an annual economic loss of more than CNY 60 million [3]. Moreover, it has been demonstrated in the literature that the tissues of damaged fruits can increase metabolism levels, thus resulting in the quick consumption of nutrients at damaged areas and realizing self-repair functions [4,5]. The fragrant pears with repaired damages still have the possibility to recover their commodity value, and they can be processed into commodities such as pear syrup, pear juice, and pear wine. However, abandoning the damaged fragrant pears directly is not in line with the desired improvement in the market competitiveness of the fragrant pear industry and the added value of the products. Therefore, the storage quality of fragrant pears can be accurately recognized by disclosing and predicting the variation laws of quality of the damaged Korla fragrant pears during the storage period. Along these lines, if the damaged Korla fragrant pears are adequately processed before they lose their commodity values, the economic benefits of the industry can be significantly increased, and practical references for the storage and scientific management of fragrant pears can be provided.
Fruit color is considered one of the primary appearance factors that determine the purchase decision of consumers and is also an important reference for fruit classification [6,7,8,9]. Color is even a core index to evaluate the appearance quality of Korla fragrant pears, and it has an important influence on the commodity value of fresh fruits and processing products. Many works in the literature have evaluated the quality of fruits based on colors. Good and uniform pericarp color means that the fruit is more fresh and mature and has the highest commodity value. The CIELab system is the most complete color model, which was proposed by Commission Internationale de l’Eclairage (CIE) to accurately describe all visible colors [10,11]. It has been extensively applied to the evaluation of fruit colors. The changes in fruit colors have been also examined during the storage period. Particularly, Yifan et al. [12] studied the changes in L*, a*, and b* of blueberry fruits before harvest and 4 weeks after the harvest. The authors found that the color changes in the blueberry surfaces after the harvest had a strong positive correlation with the anthocyanin concentration. Mai et al. [13] collected tomato images by using the RGB (Red, Green, Blue) image acquisition system, which then were transformed into CIELab color spaces for performing the color evaluation. The authors reported that the L* value of tomatoes decreased more quickly under a longer transportation distance, higher storage temperature, and extended storage time. Sierra et al. [14] explored the influencing laws of storage temperature on colors of the avocado and found that with the increase in the storage time and storage temperature, the chlorophyll degradation accelerated, the pigment synthesis volume increased, and both L* and b* values decreased, while the a* value increased. In other words, the brightness of the avocado was negatively affected, accompanied by fading yellow and deepening red. According to the abovementioned works in the literature, the color changes in fruits during the storage period have been successfully evaluated, yielding relatively good results. However, the color changes in the damaged Korla fragrant pears during the storage period have been scarcely reported in the literature.
At present, the damaged fragrant pears during the storage period are only eliminated by conducting simple and rude regular examinations. Nonetheless, this approach not only increases the operation links and labor costs but also cannot accurately recognize the storage quality. Hence, a high-efficiency and low-cost method to predict the colors of fragrant pears is urgently needed. To this end, the adaptive neuro-fuzzy inference system (ANFIS) model has arisen as a promising solution. In 1965, Professor Zadeh of the University of California broke through the classical geometric theory and first proposed the concept of membership function to express the fuzzy nature of things, which laid the foundation for fuzzy theory [15]. More specifically, Takagi and Sugeno established the fuzzy inference model Takagi–Sugeno (TS) in 1985 and 1988, respectively [16]. TS is a fuzzy system with strong adaptability, which can automatically update the parameter values of membership functions [17]. ANFIS is a fuzzy inference system based on the TS model. It can realize the three basic processes of fuzzy control, fuzzy inference, and anti-fuzzification based on neuro networks. The learning mechanism of neural networks is utilized to automatically extract rules from samples, and then the fuzzy inference control rules are adjusted. It has the comparative advantages of self-adaptation, self-organization, and self-learning [18]. On top of that, it has been already applied to predict the quality of fruits and vegetables, achieving promising results. Hao Niu et al. [19] studied the influencing laws of ripeness and storage time on the hardness of Korla fragrant pears and predicted fruit colors based on the ANFIS model. The authors found that the ANFIS model with a membership function of gbellmf (generalized bell) achieved the optimal prediction effect. In another interesting work, Zhengdong Jiang et al. [20] predicted the reserves of ascorbic acid (AA) of fresh-cut pineapples during the storage process based on the ANFIS model and proved that the ANFIS model with a triangular membership function (trimf) achieved the best prediction effect. Although good effects on the quality prediction of fruits and vegetables by using the ANFIS model have been achieved so far, the damage to the Korla pear color during storage based on the application of this method has been scarcely reported in the literature.
Under this perspective, in this work, the variation law of the color of the damaged fragrant pears during the storage period was systematically investigated, and color prediction models were constructed with eight different membership functions in ANFIS. The optimal model was selected and verified. Finally, the colors of the fragrant pears were scientifically and accurately predicted. Moreover, the ANFIS model can provide valuable theoretical references to precisely predict the quality of Korla fragrant pears during storage.

2. Materials and Methods

2.1. Test Materials

Korla fragrant pear samples were collected from 14-year-old pear trees with generally consistent canopy sizes in Shiertuan Fragrant Garden, Alear City, on 15 September 2021, which is the general harvest period in Xinjiang. Fragrant pears with similar shapes, uniform size, uniform color, no distortion, no insect diseases, and no mechanical damages were chosen as the test samples. The collected samples were weighted at 110 ± 5 g.

2.2. Data Acquisition

The pear damage experiment was carried out on the second day after harvest. The self-made Korla fragrant pear damage test platform based on a negative pressure sucker (Figure 1) was applied in this work. First, the sucker was adjusted to the needed impacting height by using the lead screw, and the fragrant pear was adsorbed by negative pressure. Finally, the negative pressure was relieved to cause the fragrant pear to fall freely onto the testbed surface. After the pear touched the surface, the pear was quickly grasped with a hand wearing a thickened cotton glove to avoid secondary damage caused by the rebound. The contact material on the testbed surface used corrugated board, which is often employed during the packaging of fragrant pears. Fragrant pears with different degrees of damage could be obtained by making them fall from different heights. In the pre-test, the fragrant pears exhibited damages when the impacting height was 30 cm, and the pericarp broke obviously when the impacting height was 130 cm, with a spillover of abundant juices. At this moment, the test was terminated, and the maximum impacting height was set to 130 cm. During the test, the impacting height was set to 30 cm, 50 cm, 70 cm, 90 cm, 110 cm, and 130 cm, while 10 tests were implemented for each group; a total of 560 fragrant pears falling at different heights were obtained, of which 80 were non-destructive pears as controls. The damaged area was measured by the method proposed by Fadiji et al., and it was approximated to be an ellipse [21], as shown in Figure 2. The calculated ellipse area was the damaged area of the fragrant pears:
S = π a b
Based on the natural environment in Xinjiang, the control group of the undamaged fragrant pears and the fragrant pears after the induced damages were stored (in the dark) in the room temperature environment of the Key Laboratory of Modern Agricultural Engineering, Xinjiang Tarim University. The mean temperature and mean relative humidity (RH) were 15 °C and 56%, respectively. The fruits were checked once every 2 days at a total storage period of 40 days.

2.2.1. Tests of L*, a*, and b*

The fruit colors were evaluated by using a high-quality computer colorimeter (Model: NRSC10, 3NH, Hunan, China) from the perspectives of L* (from black to white, 0–100), a* (from green to red, −a–+a), and b* (from blue to yellow, −b–+b). Every 5 days, 20 fragrant pears were randomly selected from the samples falling at the same height, and the color difference of the damaged fragrant pear was tested (Figure 3). A total of 80 points were obtained, and the mean value was used to represent the color data of the falling fragrant pears at this height during the storage period.

2.2.2. ANFIS Model

The ANFIS algorithm uses fuzzy treatment based on the traditional neuro network algorithm. Thus the adaptive neuro-fuzzy inference machine is generated. The machine learning system is utilized to perform adjustments and realize supplementation between the neuro network and fuzzy inference [22]. The ANFIS algorithm can also significantly increase prediction accuracy and effectively decrease errors.
In the present study, 70% of data were randomly chosen as the training set, while the remaining 30% were used as the prediction set. The models were constructed using the ANFIS ToolBox in MATLAB. The Matlab software version is R2017a (MathWorks, Natick, MA, USA). Both the damaged area and storage time of fragrant pears were input into the system. The system output the L*, a*, and b* values. The training set data were loaded in Grid Partition, which was chosen to generate the initial fuzzy inference system. Eight membership functions were input, including trimf (triangle), trapmf (trapezoidal), gbellmf (generalized bell), gaussmf (Gaussian), gasuss2mf (two-sided Gaussian), pimf (Pi curve), psigmf (product of two sigmoidal functions), and dsigmf (difference between two sigmoidal functions). The different membership functions could directly determine the output results of models. The optimal membership function could also accurately predict the colors of the damaged fragrant pears during the storage period to the maximum extent. The grid partition method was used to generate the fuzzy inference system, and the “hybrid” optimization method was used in training. The adaptive fuzzy neural network model was established in the MATLAB environment. Eight different membership functions could realize the fuzzy processing of input data. The error-tolerant rate and times of iteration were set to 0 and 100, respectively. The initial values of the parameters are presented in Table 1.

2.2.3. Judgment Standard of Optimal Prediction Model

To screen the optimal prediction model of the storage quality of fragrant pears, the determination coefficient (R2) was applied to evaluate the prediction accuracy of the model for pear peel color. The higher R2 and the lower RMSE indicate the greater prediction precision of the model. The calculation formulas of R2 and RMSE were as follows:
R 2 = 1 − ∑ ( M j − T j ) 2 ∑ M j 2 − ( ∑ M j ) 2 n
R M S E = ∑ i N ( M j − T j ) 2 N
where Mj and Tj are the measured and predicted values of data j, respectively. n refers to the number of measured values, whereas N represents the total number of data, a total of 63.

3. Variation Laws of Colors of the Damaged Korla Fragrant Pears during the Storage Period

3.1. Variation Laws of L* Value

The size, shape, and weight were within a small range and considered similar, and the damaged area of pears falling from the same height was considered similar. Therefore, it was only necessary to use the average value of the damaged area of a batch of pears falling at the same height to represent the damage degree of the pears falling at this height. The mean damaged areas of fragrant pears under six impacting heights were 77.71 mm2, 296.91 mm2, 604.05 mm2, 900.77 mm2, 952.30 mm2, and 993.42 mm2, respectively.
The brightness (L*) expresses the color intensity. The higher L* value indicates a brighter color, and the lower L* value implies a darker color. The variation laws of the L* values of fragrant pears at different damage degrees with storage time are shown in Figure 4. As can be observed, the L* value decreased with the increase in storage time. The difference in the L* value between the control group and the fragrant pears with the maximum degree of damage increased from 0.3573 at 0 d to 1.4584 at 40 d. In other words, the pericarp brightness of fragrant pears gradually declined with the increase in storage time. The L* value of fragrant pears with the minimum degree of damage slowly decreased, while the L* value of fragrant pears with the maximum degree of damage quickly decreased. The degree of damage of fragrant pears was also positively related to the darkening rate of the pericarp. The variation curve of the L* value during the storage period was basically consistent with that of the undamaged fragrant pears when the damaged area was 77.71 mm2. No visible damage on the pericarp except for mild pulp damage after eliminating the pericarp was also detected (Figure 5).
This reveals that the physiological activity of fragrant pears slightly changed under mild damage and no pericarp damage. When the damaged area exceeded 77.71 mm2, the physiological activity of fragrant pears greatly changed with the intensifying damages. This is because fragrant pears are a type of climacteric fruits that can increase respiratory rate and ethylene production after harvest [23]. As a result, the damaged fragrant pears can strengthen their respiration, accelerate their metabolism, and self-repair [24,25]. The pericarp color of fragrant pears also became darker due to respiratory oxidation. Moreover, after the fragrant pears were damaged, the cytomembrane was broken, which destroyed the regional distribution of phenolic substances and polyphenol oxidase (PPO). The contact between the phenolic substances and PPO triggered enzymatic browning, thus deepening the pericarp colors [26,27,28]. Hence, the L* value of fragment pears with a higher degree of damage declined more quickly.

3.2. Variation Laws of a* Value

The a* value expresses the red–green degree. It is red if the a* value is positive, and the color intensity is positively related to the a* value. On the contrary, it is green if the a* value is negative, and the color intensity is negatively related to the a* value. The variation laws of a* with the storage time are depicted in Figure 6. The a* value increased with the increase in storage time. More specifically, the mean a* values of fragrant pears with different degrees of damage increased from −9.9456 at 0 d to 2.8608 at 40 d. This meant that the pericarp colors of fragrant pears changed from green to red with the increase in the storage time. The underlying reasons for this effect can be explained as follows. Due to strengthened respiration of the damaged fragrant pears during the storage period, the releasing amount of endogenous ethylene increased to accelerate the degradation of chlorophyll. The chlorophyll content in the pericarp declined with the increase in the storage time [29,30], thus resulting in the fading green. Anthocyanin began to accumulate during the violent degradation of chlorophyll, making the fruits turn red [31,32,33].

3.3. Variation Laws of b* Value

The b* value expresses the yellow–blue degree. It is yellow if the b* value is positive, and the color intensity is positively related to the b* value. On the other hand, it is blue if the b* value is negative, and the color intensity is negatively related to the b* value. The variation laws of b* with the storage time are displayed in Figure 7. The b* value increased with the increase in the storage time. The mean b* value of the fragrant pears with different degrees of damage increased from 39.0444 at 0 d to 43.3331 at 40 d. This meant that the pericarp color of fragrant pears was yellowing with the increase in the storage time. This effect could be explained as follows. As the storage time was prolonged, the chlorophyll was quickly degraded, but the carotenoid content was negatively related to chlorophyll degradation, and it quickly increased [34,35]. With the increase in the ethylene content after harvest, the total carotenoid content accordingly increased [36] to intensify the yellow of the pericarp.

4. Prediction of Storage Quality of the Damaged Pears based on the ANFIS Model

4.1. Predicted L* Value of the ANFIS Model

The number of iterations reached the optimum at the 50th time. The test dataset was input into the trained model to obtain the prediction values.
The RMSE and R2 of the L* values predicted by different membership functions during the training and prediction stages are listed in Table 2 and Figure S1. Clearly, the RMSE and R2 of the predicted L* value by using different membership functions were different. Particularly, they were 0.1089 and 0.9773 when the membership function was trimf, 0.1356 and 0.9649 for trapmf, 0.1310 and 0.9674 for gbellmf, 0.5661 and 0.967 for gaussmf, 0.1347 and 0.9655 for gasuss2mf, 0.1376 and 0.9642 for pimf, 0.1323 and 0.9672 for dsigmf, and 0.1225 and 0.9719 for psigmf. In the prediction stage of the L* value of the fragrant pears, the goodness of the fit between the predicted value and the practical value (R2) was higher than 0.96, showing that the ANFIS model could predict the L* value of the damaged fragrant pears during the storage period effectively. More specifically, trimf achieved the highest R2 and RMSE. This proved that the trimf function was the optimal membership function to predict the L* values of the fragrant pears.

4.2. Predicted a* Value of ANFIS Model

The RMSE and R2 of the a* values predicted by different membership functions during the training and prediction stages are listed in Table 3 and Figure S2. Clearly, the RMSE and R2 of the predicted a* values by using different membership functions were different. They were 0.5894 and 0.9853 when the membership function was trimf, 0.7456 and 0.9750 for trapmf, 0.7927 and 0.9732 for gbellmf, 0.8036 and 0.9729 for gaussmf, 0.8302 and 0.9702 for gasuss2mf, 0.8033 and 0.9713 for pimf, 0.6855 and 0.9790 for dsigmf, and 0.5980 and 0.9831 for psigmf. During the prediction stage, the goodness of fit between the predicted value and the practical value (R2) was higher than 0.97, showing that the ANFIS model could effectively predict the a* value of the damaged fragrant pears during the storage period. More specifically, the R2 values of the trimf and psigmf functions were higher than 0.98, and their RMSE values were lower compared with the remaining six functions. The R2 of the trimf function was the highest, and its RMSE was the lowest, indicating that trimf was the optimal membership function to precisely predict the a* values of the fragrant pears.

4.3. Predicted b* Value of ANFIS Model

The RMSE and R2 of the b* values predicted by different membership functions during the training and prediction stages are listed in Table 4 and Figure S3. Clearly, both the RMSE and R2 of the predicted b* values by using different membership functions were different. They were 0.2360 and 0.9772 when the membership function was trimf, 0.3294 and 0.9528 for trapmf, 0.2597 and 0.9727 for gbellmf, 0.2860 and 0.9672 for gaussmf, 0.2947 and 0.9627 for gasuss2mf, 0.0920 and 0.9551 for pimf, 0.3311 and 0.9526 for dsigmf, and 0.3331 and 0.9520 for psigmf. During the prediction stage, the goodness of fit between the predicted value and the practical value (R2) was higher than 0.95, showing that the ANFIS model could accurately predict the b* value of the damaged fragrant pears during the storage period. More specifically, trimf achieved the highest R2 and RMSE values. This proved that the trimf function was the optimal membership function to predict the b* values of the fragrant pears.

4.4. Model Verification

To verify the general applicability of the optimal prediction model, a storage test of the damaged fragrant pears was carried out on 15 September 2022. The fragrant pears, which fell from the heights of 50 cm, 70 cm, and 90 cm (damaged areas of 306.54 mm2, 628.43 mm2, and 958.62 mm2), were chosen and stored for 0 d, 20 d, and 40 d to test the L*, a*, and b* values of the pericarp. The function trimf was applied for performing the linear fitting of the test and predicted values of the ANFIS model. The extracted results are shown in Figure 8. According to the verification, trimf showed a high prediction accuracy of the L*, a*, and b* values of the pericarp of fragrant pears (RMSE = 0.2303, R2 = 0.9685; RMSE = 0.7593, R2 = 0.9812; and RMSE = 0.5793, R2 = 0.9799). Hence, the L*, a*, and b* values of the pericarp of fragrant pears can be accurately predicted by inputting the degree of damages and storage time into the trained optimal model.
The above research results clearly proved the feasibility of the ANFIS model to predict the colors of fragrant pears based on the degree of damage and storage time. However, the prediction effects among the eight membership functions were different. After performing optimal selection, trimf showed the best prediction effects of the L*, a*, and b* values of the pericarp of fragrant pears. A storage test of the damaged fragrant pears was also carried out in the second year to verify the optimal prediction model. According to the acquired results, the trained optimal prediction model could accurately predict the L*, a*, and b* values of the fragrant pears. Based on the ANFIS model and the CIELab space color system, the variation laws of the colors of the damaged fragrant pears during the storage period were disclosed, and a method for scientific and effective color prediction was put forward. The appearance quality of fragrant pears can be assessed by the color, size, and shape. The shape and size are basic indexes for grading fragrant pears and have explicit execution standards. The color also provides explicit instructions to sensory organs, but there is no relevant execution standard. This work can provide valuable references to formulate color-based grading criteria for Korla fragrant pears and offer theoretical references for carrying out the scientific classification and management of fragrant pears after harvest. Moreover, it can be also used as a novel method for conducting quality prediction of other fruits during the storage period.

5. Conclusions

With the increase in the storage time, the L* value of the pericarp of Korla fragrant pears decreased, but both the a* and b* values increased. The ANFIS model with a membership function of trimf achieved the optimal prediction effect of the pericarp colors of the fragrant pears (RMSE = 0.1089, R2 = 0.9773; RMSE = 0.5894, R2 = 0.9853; and RMSE = 0.2360, R2 = 0.9772). The optimal model was also verified, which proved its best prediction performances in practice (RMSE = 0.2303, R2 = 0.9685; RMSE = 0.7593, R2 = 0.9812; and RMSE = 0.5793, R2 = 0.9799). The peel color of the fragrant pear was accurately predicted and used as an important basis for measuring the storage quality of the fragrant pear. According to the storage quality, the pears with a short shelf life and intolerance to storage can be sold or processed in advance, which can reduce economic losses. The proposed prediction method provides a basis for the storage and scientific management of pears.

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/app13137885/s1: Figure S1: Correlation between measured and predicted L* values of fragrant pears during the training phase and prediction phase by the ANFIS model with different membership functions; Figure S2: Correlation between measured and predicted a* values of fragrant pears during the training phase and prediction phase by the ANFIS model with different membership functions; Figure S3: Correlation between measured and predicted b* values of fragrant pears during the training phase and prediction phase by the ANFIS model with different membership functions.

Author Contributions

Methodology, Y.L. and R.Z.; software, X.J. and G.L.; data curation, Y.L.; writing—original draft preparation, R.Z. and S.L.; writing—review and editing, R.Z. and S.L.; visualization, X.F.; project administration, Y.L. All authors have read and agreed to the published version of the manuscript.

Funding

This research was financially supported by the Innovation and Entrepreneurship Project of the Xinjiang Production and Construction Group Special Commissioner for Science and Technology (Grant No. 2019CB037), the Research Project of Double Employment Academician Work Funds Opening in Tarim University (Grant No. SPYS202002), the University President Fund Project (Grant No. TDZKCQ201902), and the “Strong Youth” Key Talents of Scientific and Technological Innovation (Grant No. 2021CB039), the Innovation Research Team Project of the President’s Fund of Tarim University (TDZKCX202203).

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The data presented in this study are available on request from the corresponding author.

Acknowledgments

The authors thank Jiean Liao from Tarim University for thesis supervision. The authors are grateful to the anonymous reviewers for their comments.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Baijunjie, S.; Jia, T.; Zhichao, H.; Qian, W.; Feng, Z.; Yue, W. Effects of Ethephon on Fruit Developmental Process and Quality of Korla Fragrant Pear. Acta Bot. Boreal.-Occident. Sin. 2023, 43, 265–267. [Google Scholar]
  2. An, J.; Luo, X.; Xiong, L.; Tang, X.; Lan, H. Discrimination of Inner Injury of Korla Fragrant PearBased on Multi-Electrical Parameters. Foods 2023, 12, 1805. [Google Scholar] [CrossRef] [PubMed]
  3. Gao, Y.W.; Geng, J.F.; Rao, X.Q. Non-invasive bruise detection in postharvest fruits and vegetables: A review. J. Food Sci. 2017, 38, 277–287. [Google Scholar]
  4. Pan, Y.; Shi, R. Postharvest physiology and biochemistry of fruits anld vegetables responding to mechanical stress. Plant Physiol. 2000, 36, 568–572. (In Chinese) [Google Scholar]
  5. Wei, X.P.; Mao, L.C.; Wei, X.B.; Xia, M.; Xu, C. MYB41, MYB107, and MYC2 promote ABA-mediated primary fatty alcohol accumulation via activation of AchnFAR in wound suberization in kiwifruit. Hortic. Res. 2020, 7, 86. [Google Scholar] [CrossRef]
  6. Palei, S.; Behera, S.K.; Sethy, P.K. A Systematic Review of Citrus Disease Perceptions and Fruit Grading Using Machine Vision. Comput. Sci. 2023, 218, 2504–2519. [Google Scholar] [CrossRef]
  7. Sun, W.; Li, X.; Huang, H.; Wei, J.; Zeng, F.; Huang, Y.; Sun, Q.; Miao, W.; Tian, Y.; Li, Y.; et al. Mutation of CsARC6 affects fruit color and increases fruit nutrition in cucumber. Theor. Appl. Genet. 2023, 136, 111. [Google Scholar] [CrossRef]
  8. Xiong, Y.; He, J.; Li, M.; Du, K.; Lang, H.; Gao, P.; Xie, Y. Integrative Analysis of Metabolome and Transcriptome Reveals the Mechanism of Color Formation in Yellow-Fleshed Kiwifruit. Int. J. Mol. Sci. 2023, 24, 1573. [Google Scholar] [CrossRef]
  9. Liang, J.; Zhang, G.; Song, Y.; He, C.; Zhang, J. Targeted Metabolome and Transcriptome Analyses Reveal the Pigmentation Mechanism of Hippophae (Sea Buckthorn) Fruit. Foods 2022, 11, 3278. [Google Scholar] [CrossRef]
  10. Soraya, M.P. Physicochemical characterization of pomegranate (Punica granatum L.) native to Jordan during different maturity stages: Color evaluation using the CIELab and CIELCh systems. J. Ecol. Eng. 2021, 22, 214–221. [Google Scholar]
  11. Li, S.; Xiao, K.; Li, P. Spectra Reconstruction for Human Facial Color from RGB Images via Clusters in 3D Uniform CIELab* and Its Subordinate Color Space. Sensors 2023, 23, 810. [Google Scholar] [CrossRef] [PubMed]
  12. Yan, Y.; Pico, J.; Gerbrandt, E.; Dossett, M.; Castellarin, S.D. Comprehensive anthocyanin and flavonol profiling and fruit surface color of 20 blueberry genotypes during postharvest storage. Postharvest. Biol. Technol. 2023, 199, 112274. [Google Scholar] [CrossRef]
  13. Aidairi, M.; Pathare, P.B.; Aiyahyai, R. Effect of Postharvest Transport and Storage on Color and Firmness Quality of Tomato. Horticulturae 2021, 7, 163. [Google Scholar]
  14. Neidy, M.S.; Alexandra, L.; José, M.G.; Aníbal, O.H.; Diego, A.C. Evaluation and modeling of changes in shelf life, firmness and color of ‘Hass’ avocado depending on storage temperature. Food Sci.Technol. Int. 2019, 25, 370–384. [Google Scholar]
  15. Zadeh, L.A. Fuzzy sets. Inf. Control 1965, 8, 338–353. [Google Scholar] [CrossRef] [Green Version]
  16. Yong, T.; Jiangbo, L.; Fan, R.; Tianmiao, W.; Shan, J.; He, G.; Yufang, W. Prediction method of robot welding seam shape based on adaptive fuzzy neural network. Comput. Integ. Manuf. Syst. 2022, 28, 3643–3651. [Google Scholar]
  17. Hairu, G.; Zhimin, L.; Jialiang, G. Controller for Automobile’s Rear-end Anti-Collision Based on Adaptive Fuzzy Neural Network. Comput. Simulat. 2012, 29, 344–347. [Google Scholar]
  18. Xiaojuan, Z. Study on the adaptive network-based fuzzy inference system and simulation. Electron. Des. Eng. 2012, 20, 11–13. [Google Scholar]
  19. Hao, N.; Yang, L.; Zhen, T.W.; Hong, Z.; Yong, C.Z.; Hai, P.L. Effects of Harvest Maturity and Storage Time on Storage Quality of Korla Fragrant Pear Based on GRNN and ANFIS Mo dels: Part I Firmness Study. Food Sci. Technol. Res. 2020, 26, 363–372. [Google Scholar]
  20. Zheng, D.; Hong, Z.; Nitin, M.; Zhe, C.; Xiao, D.; Zhuo, N.; Jia, D.; Hong, F.; Zong, S. Prediction of relationship between surface area, temperature, storage time and ascorbic acid retention of fresh-cut pineapple using adaptive neuro-fuzzy inference system (ANFIS). Postharvest. Biol. Technol. 2016, 113, 1–7. [Google Scholar]
  21. Fadiji, T.; Coetzee, C.; Chen, L.; Chukwu, O.; Opara, U.L. Susceptibility of apples to bruising inside ventilated corrugated paperboard packages during simulated transport damage. Postharvest. Biol. Technol. 2016, 118, 111–119. [Google Scholar] [CrossRef]
  22. Mosavi, M.R.; Ayatollahi, A.; Afrakhteh, S. An efficient method for classifying motor imagery using CPSO-trained ANFIS prediction. Evol. Syst.-Ger. 2019, 12, 1–18. [Google Scholar] [CrossRef]
  23. Jiang, Y.; Wang, Y.; Mao, H.; Yunhao, L.; Chen, G. Delaying the aging process of pears by maintaining cuticular waxes under high humidity storage conditions. Trans. CSAE 2020, 36, 287–295. [Google Scholar]
  24. Ran, Z. Studies on the Quality Changes of HuangHua Pears during Transport and Storage. Ph.D. Thesis, Shanghai Jiao Tong University, Shanghai, China, 2007. [Google Scholar]
  25. Meixia, Z. Study on Respiration Intensity of Different Parts of Several Fruits and Wound Respiration. Master’s Thesis, China Agricultural University, Beijing, China, 2005. [Google Scholar]
  26. Barrett, D.M.; Beaulieu, J.C.; Shewfelt, R. Color, flavor, texture, and nutritional quality of fresh-cut fruits and vegetables: Desirable levels, instrumental and sensory measurement, and the effects of processing. Crit. Rev. Food Sci. 2010, 50, 369–389. [Google Scholar] [CrossRef] [PubMed]
  27. Hu, J.; Bi, J.; Li, X.; Wu, X.; Wang, W.; Yu, Q. Understanding the impact of pectin on browning of polyphenol oxidation system in thermal and storage processing. Carbohyd. Polym. 2023, 307, 120641. [Google Scholar] [CrossRef]
  28. Wang, T.; Yan, T.; Shi, J.; Sun, Y.; Wang, Q.; Li, Q. The stability of cell structure and antioxidant enzymes are essential for fresh-cut potato browning. Food. Res. Int. 2023, 164, 112449. [Google Scholar] [CrossRef] [PubMed]
  29. Liu, Y.; Zhao, J.; Tang, Y.; Jiang, X.; Liao, J. Construction of a Chlorophyll Content Prediction Model for Predicting Chlorophyll Content in the Pericarp of Korla Fragrant Pears during the Storage Period. Agriculture 2020, 12, 1348. [Google Scholar] [CrossRef]
  30. Min, S. Studys on the Postharvest Fruit Quality of ’Starkrimson’ Pear. Master’s Thesis, Northwest Agriculture and Forestry University, Yangling, China, 2019. [Google Scholar]
  31. Modesto, J.E.N.; Martins, M.G.; Pereira, G.A.; Chisté, R.C.; Pena, R.D.S. Stability Kinetics of Anthocyanins of Grumixama Berries (Eugenia brasiliensis Lam.) during Thermal and Light Treatments. Foods 2023, 12, 565. [Google Scholar] [CrossRef]
  32. Qi, Z.; Tianming, H.; Hong, M. Changes of pigment content in the peel of Korla fragrant pear during ripening. J. Tarim Univ. Agr. Reclam. 2002, 14, 16–18. [Google Scholar]
  33. Li, X.; Hou, Y.; Xie, X.; Li, H.; Li, X.; Zhu, Y.; Zhai, L.; Zhang, C.; Bian, S. Blueberry MIR156a/SPL12 module coordinates the accumulation of chlorophylls and anthocyanins during fruit ripening. J. Exp. Bot. 2020, 71, 5976–5989. [Google Scholar] [CrossRef]
  34. Chaonan, K.; Yang, G.; Ming, C.; Chuying, C.; Jinying, C.; Shanjun, L. Effect of Chitosan Coating on the Color and Storage Quality of ‘Cuiguan’ Pear Fruit during ShelfLife at Room Temperature. Food Res. Dev. 2019, 40, 11–16. [Google Scholar]
  35. Zhu, K.; Zheng, X.; Ye, J.; Huang, Y.; Chen, H.; Mei, X.; Xie, Z.; Cao, L.; Zeng, Y.; Larkin, R.; et al. Regulation of carotenoid and chlorophyll pools in hesperidia, anatomically unique fruits found only in Citrus. Plant. Physiol. 2021, 187, 829–845. [Google Scholar] [CrossRef] [PubMed]
  36. Xiao, X.; Shi, L.Y.; Dong, W.Q.; Jin, S.W.; Liu, Q.L.; Chen, W.; Cao, S.F.; Yang, Z.F. Ethylene promotes carotenoid accumulation in peach pulp after harvest. Sci. Hortic. 2022, 304, 111347. [Google Scholar] [CrossRef]
Applsci 13 07885 g001 550
Figure 1. Korla fragrant pear damage test platform based on negative pressure sucker. 1: Air compressing machine; 2: machine frame; 3: vacuum generator; 4: guide rail; 5: cantilever; 6: sucker; 7: guide screw; 8: pneumatic motor.
Figure 1. Korla fragrant pear damage test platform based on negative pressure sucker. 1: Air compressing machine; 2: machine frame; 3: vacuum generator; 4: guide rail; 5: cantilever; 6: sucker; 7: guide screw; 8: pneumatic motor.
Applsci 13 07885 g001
Applsci 13 07885 g002 550
Figure 2. Measurement of the damaged parts of the fragrant pear.
Figure 2. Measurement of the damaged parts of the fragrant pear.
Applsci 13 07885 g002
Applsci 13 07885 g003 550
Figure 3. Measuring points of Korla pear quality index (■).
Figure 3. Measuring points of Korla pear quality index (■).
Applsci 13 07885 g003
Applsci 13 07885 g004 550
Figure 4. Variation laws of the L* value of fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Figure 4. Variation laws of the L* value of fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Applsci 13 07885 g004
Applsci 13 07885 g005 550
Figure 5. Fragrant pear falling at 30 cm height with 77.71 mm2 damaged: (a) No epidermal damage. (b) Minor pulp damage.
Figure 5. Fragrant pear falling at 30 cm height with 77.71 mm2 damaged: (a) No epidermal damage. (b) Minor pulp damage.
Applsci 13 07885 g005
Applsci 13 07885 g006 550
Figure 6. Variation laws of the a* values of fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Figure 6. Variation laws of the a* values of fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Applsci 13 07885 g006
Applsci 13 07885 g007 550
Figure 7. Variation laws of the b* value of the fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Figure 7. Variation laws of the b* value of the fragrant pears at different damage degrees with storage time (the mean values ± standard deviations).
Applsci 13 07885 g007
Applsci 13 07885 g008 550
Figure 8. The ANFIS model of the membership function trimf predicts the scatter plot of the L*, a*, and b* values and the actual values of the fragrant pear.
Figure 8. The ANFIS model of the membership function trimf predicts the scatter plot of the L*, a*, and b* values and the actual values of the fragrant pear.
Applsci 13 07885 g008
Table
Table 1. Initial values of the ANFIS parameters.
Table 1. Initial values of the ANFIS parameters.
Training Optimization MethodMembership Function TypesNumber of Membership FunctionsNumber of EpochsOutput Function Type
HybridANFIS-Tri-mf3, 3100Linear
HybridANFIS-Trap-mf3, 3100Linear
HybridANFIS-Gbell-mf3, 3100Linear
HybridANFIS-Gauss-mf3, 3100Linear
HybridANFIS-Gauss2-mf3, 3100Linear
HybridANFIS-Pi-mf3, 3100Linear
HybridANFIS-Dsig-mf3, 3100Linear
HybridANFIS-Psig-mf3, 3100Linear
Table
Table 2. RMSE and R2 of L* values predicted by different membership functions during the training and prediction stages.
Table 2. RMSE and R2 of L* values predicted by different membership functions during the training and prediction stages.
Membership FunctionR2
(Training Stage)		R2
(Prediction Stage)		RMSE
(Training Stage)		RMSE
(Prediction Stage)
Tri0.99820.97730.04120.1089
Trap0.99830.96490.04000.1356
Gbell0.99840.96740.03920.1310
Gauss0.99830.96790.04040.5661
Gauss20.99830.96550.04000.1347
Pi0.99840.96420.03890.1376
Dsig0.99840.96720.03910.1323
Psig0.99860.97190.03670.1225
R2: determination coefficient; RMSE: root-mean-square error.
Table
Table 3. RMSE and R2 of a* values predicted by different membership functions during the training and prediction stages.
Table 3. RMSE and R2 of a* values predicted by different membership functions during the training and prediction stages.
Membership FunctionR2
(Training Stage)		R2
(Prediction Stage)		RMSE
(Training Stage)		RMSE
(Prediction Stage)
Tri0.99250.98530.38420.5894
Trap0.99480.97500.32130.7456
Gbell0.99510.97320.30960.7927
Gauss0.99510.97290.31130.8036
Gauss20.99490.97020.31550.8302
Pi0.99460.97130.32510.8033
Dsig0.99550.97900.29670.6855
Psig0.99540.98310.30030.5980
R2: determination coefficient; RMSE: root-mean-square error.
Table
Table 4. RMSE and R2 of b* values predicted by different membership functions during the training and prediction stages.
Table 4. RMSE and R2 of b* values predicted by different membership functions during the training and prediction stages.
Membership FunctionR2
(Training Stage)		R2
(Prediction Stage)		RMSE
(Training Stage)		RMSE
(Prediction Stage)
Tri0.99510.97720.10920.2360
Trap0.99640.95280.09290.3294
Gbell0.99660.97270.09130.2597
Gauss0.99610.96720.09700.2860
Gauss20.99660.96270.09130.2947
Pi0.99640.95510.09310.0920
Dsig0.99620.95260.09650.3311
Psig0.99610.95200.09680.3331
R2: determination coefficient; RMSE: root-mean-square error.
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Zhang, R.; Li, S.; Liu, Y.; Li, G.; Jiang, X.; Fan, X. Construction of Color Prediction Model for Damaged Korla Pears during Storage Period. Appl. Sci. 2023, 13, 7885. https://doi.org/10.3390/app13137885

AMA Style

Zhang R, Li S, Liu Y, Li G, Jiang X, Fan X. Construction of Color Prediction Model for Damaged Korla Pears during Storage Period. Applied Sciences. 2023; 13(13):7885. https://doi.org/10.3390/app13137885

Chicago/Turabian Style

Zhang, Rui, Shiyuan Li, Yang Liu, Guowei Li, Xin Jiang, and Xiuwen Fan. 2023. "Construction of Color Prediction Model for Damaged Korla Pears during Storage Period" Applied Sciences 13, no. 13: 7885. https://doi.org/10.3390/app13137885

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