# A Computer Vision System Based on Majority-Voting Ensemble Neural Network for the Automatic Classification of Three Chickpea Varieties

^{1}

^{2}

^{3}

^{4}

^{5}

^{*}

## Abstract

**:**

## 1. Introduction

^{2}to obtain a uniform image. A total of 525 samples were obtained by scanning 75 rapeseed samples from each variety in lab. Three classes of texture properties were extracted from rapeseed image samples: Gray Level Co-ocurrence Matrix (GLCM), defined as a space frequency matrix between adjacent pixels in a digitized image, Gray Level Run Length Matrix (GLRLM), defined as the number of runs with pixels of gray level i and run length j at a given image direction, and local binary patterns, that transform an image into an array of integer labels describing small-scale appearance on an image. Next, three classifiers including support vector machine classifications, k-nearest neighbor, and stochastic gradient descent, were used to classify seven rapeseed varieties. Results showed that the classifiers clustered samples with good accuracy. Other researchers have proposed a new approach based on image analysis techniques to discriminate defective lentil seeds from healthy ones. They extracted several color, shape and texture features. These features were then used as inputs for Support Vector Machine (SVM) classification. Results showed that their method achieved accuracy was 98.9% [13].

## 2. Materials and Methods

#### 2.1. Varieties of Chickpea Used

#### 2.2. Segmentation Operation

#### 2.3. Extraction of Different Properties of Each Chickpea Sample Image

#### 2.4. Feature Selection

#### 2.5. Ensemble Classification of Different Chickpea Varieties: Majority-Voting (MV)

#### 2.5.1. Hybrid ANN-PSO Classifier

#### 2.5.2. Hybrid ANN-ACO Classifier

#### 2.5.3. Hybrid ANN-HS Classifier

#### 2.5.4. Ensemble Final Classification through MV

#### 2.6. Optimal Structures of ANNs Adjusted by Different Algorithms

#### 2.7. Criteria Used to Evaluate the Performance of the Different Classifiers: Confusion Matrices and Receiver Operating Curves (ROC) (Test Set)

- Sensitivity, recall, true positive (TP) rate or probability of detection: measures the proportion of actual positives that are correctly identified as such (2)
- Accuracy or correct classification rate (CCR): total percentage of correct system classifications (3)
- Specificity or true negative (TN) rate: percentage of inaccurate samples that are correctly identified (4)
- Precision or positive predictive value: is the fraction of relevant instances among the retrieved instances (5)
- F1-score: recall and precision harmonic weighted average (6).

## 3. Results

#### 3.1. Effective Discrimiant Property (Feature) Selection

#### 3.2. Classification Using Hybrid ANN-PSO Classifier

#### 3.3. Classification Using Hybrid ANN-ACO Classifier

#### 3.4. Classification Using Hybrid ANN-HS Classifier

## 4. Discussion

- Chickpea bunch imaging acquisition under light controlled conditions, with white LEDs with 425 lux intensity.
- Automatic chickpea image segmentation.
- Automatic extraction of different discriminant features, including: average channels of first, second, and third RGB color space, average first channel of HSI color space, neighborhood Entropy of 90° and 0° in GLCM, and third channel of YCbCr color space, from each input sample image.
- Output chickpea variety classification by a neural network ensemble majority-voting.

## 5. Conclusions

## Supplementary Materials

## Author Contributions

## Funding

## Acknowledgments

## Conflicts of Interest

## References

- Sabaghpour, S.H.; Mahmodi, A.A.; Saeed, A.; Kamel, M.; Malhotra, R.S. Study on chickpea drought tolerance lines under dryland condition of Iran. Indian J. Crop Sci.
**2006**, 1, 70–73. [Google Scholar] - Kanouni, H.; Ahari, D.S.; Khoshroo, H.H. Chickpea Research and Production in Iran. In Proceedings of the 7th International Food Legume Research Conference, Marrakech, Morocco, 6–8 May 2018. [Google Scholar]
- Tabatabaeefar, A.; Aghagoolzadeh, A.H.; Mobli, H. Design and Development of an Auxiliary Chickpea Second Sieving and Grading Machine. Agric. Eng. Int. CIGR J. Sci. Res. Dev.
**2003**, 5, FP 03 005. [Google Scholar] - Masoumi, A.A.; Tabil, L. Physical properties of chickpea (c. arietinum) cultivars. In Proceedings of the 2003 ASAE Annual Meeting, Las Vegas, NV, USA, 27–30 July 2003; American Society of Agricultural and Biological Engineers: St. Joseph, MI, USA, 2003; p. 1. [Google Scholar]
- Rybiński, W.; Banda, M.; Bocianowski, J.; Starzycka-Korbas, E.; Starzyck, M.; Nowosad, K. Estimation of the physicochemical variation of chickpea seeds. Agrophys
**2019**, 33, 67–80. [Google Scholar] [CrossRef] - Carl, W.H.; Denny, C.D. Processing Equipment for Agricultural Products, 2nd ed.; The AVI Publishing Inc: Westport, CT, USA, 1979. [Google Scholar]
- Fawzi, N.M. Seed morphology and its implication in classification of some selected species of genus Corchorus, L. (Malvaceae). Middle East J. Agric. Res.
**2018**, 7, 1–11. [Google Scholar] - Sabzi, S.; Abbaspour-Gilandeh, Y.; Hernandez-Hernandez, J.L.; Azadshahraki, F.; Karimzadeh, R. The Use of the Combination of Texture, Color and Intensity Transformation Features for Segmentation in the Outdoors with Emphasis on Video Processing. Agriculture
**2019**, 9, 104. [Google Scholar] [CrossRef][Green Version] - Gino, I.G.C.; Amelia, T. Fault Diagnosis for UAV Blades Using Artificial Neural Network. Robotics
**2019**, 8, 59. [Google Scholar] - LeCun, L.Y.; Bottou, Y.B.; Hafner, P. Gradient-based learning applied to document recognition. Proc. IEEE
**1998**, 86, 2278–2324. [Google Scholar] [CrossRef][Green Version] - Uijlings, J.R.R.; Van De Sande, K.E.A.; Gevers, T.; Smeulders, A.W.M. Selective search for object recognition. Int. J. Comput. Vis.
**2013**, 104, 154–171. [Google Scholar] [CrossRef][Green Version] - Kurtulmus, F.; Unal, H. Discriminating rapeseed varieties using computer vision and machine learning. Expert Syst. Appl.
**2015**, 42, 1880–1891. [Google Scholar] [CrossRef] - HemaChitra, H.S.; Suguna, S. Optimized feature extraction and classification technique for indian pulse seed recognition. Int. J. Comput. Eng. Appl.
**2018**, XII, 421–427. [Google Scholar] - Liu, D.; Ning, X.; Li, Z.; Yang, D.; Li, H.; Gao, L. Discriminating and elimination of damaged soybean seeds based on image characteristics. J. Stored Prod. Res.
**2015**, 60, 67–74. [Google Scholar] [CrossRef] - Golpour, I.; Parian, J.A.; Khazaei, J. Detection of rice varieties, brown and white rice based on image texture and artificial neural network. J. Agric. Mach.
**2015**, 5, 73–81. [Google Scholar] - Pourdarbani, R.; Sabzi, S.; Hernández-Hernández, M.; Hernández-Hernández, J.L.; García-Mateos, G.; Kalantari, D.; Molina-Martínez, J.M. Comparison of Different Classifiers and the Majority Voting Rule for the Detection of Plum Fruits in Garden Conditions. Remote Sens.
**2019**, 11, 2546. [Google Scholar] [CrossRef][Green Version] - Clausi, D.A. An analysis of co-occurrence texture statistics as a function of grey level quantization. Can. J. Remote Sens.
**2002**, 28, 45–62. [Google Scholar] [CrossRef] - Ali, M.Z.; Awad, N.H.; Suganthan, P.N.; Duwairi, R.M.; Reynolds, R.G. A novel hybrid Cultural Algorithms framework with trajectory-based search for global numerical optimization. Inf. Sci.
**2016**, 334, 219–249. [Google Scholar] [CrossRef] - Sivanandam, S.N.; Deepa, S.N. Introduction to Neural Networks Using Matlab 6.0; Tata McGraw-Hill Education: New York, NY, USA, 2006; ISBN1 0070591121. ISBN2 9780070591127. [Google Scholar]
- Grzegorz, Z.; Jarosław, G.; Artur, G.; Ewelina, M.-J.; Thomas, B.; Alexandre, D. Multi-classifier majority voting analyses in provenance studies on iron artefacts. J. Archaeol. Sci.
**2020**, 113, 105055. [Google Scholar] [CrossRef] - Caudill, M. Neural Networks Primer; Miller Freeman Publication: San Francisco, CA, USA, 1989. [Google Scholar]
- Kennedy, J.; Eberhart, R. Particle swarm optimization (PSO). In Proceedings of the IEEE International Conference on Neural Networks, Perth, Australia, 27 November–1 December 1995; pp. 1942–1948. [Google Scholar]
- Sen, T.; Mathur, H.D. A new approach to solve economic dispatch problem using a hybrid ACO–ABC–HS optimization algorithm. Int. J. Electr. Power Energy Syst.
**2016**, 78, 735–744. [Google Scholar] [CrossRef] - Lee, K.S.; Geem, Z.W. A New Meta-Heuristic Algorithm for Continuous Engineering Optimization: Harmony Search Theory and Practice. Comput. Methods Appl. Mech. Eng.
**2005**, 194, 3902–3933. [Google Scholar] [CrossRef] - Wisaeng, K. A comparison of decision tree algorithms for UCI repository classification. Int. J. Eng. Trends Technol.
**2013**, 4, 3393–3397. [Google Scholar] - Guijarro, M.; Riomoros, I.; Pajares, G.; Zitinski, P. Discrete wavelets transform for improving greenness image segmentation in agricultural images. Comput. Electron. Agric.
**2015**, 118, 396–407. [Google Scholar] [CrossRef] - Li, X.; Dai, B.; Sun, H.; Li, W. Corn Classification System based on Computer Vision. Symmetry
**2019**, 11, 591. [Google Scholar] [CrossRef][Green Version] - Men, S.; Yan, L.; Liu, J.; Qian, H.; Luo, Q. A classification method for seed viability assessment with infrared thermography. Sensors
**2017**, 17, 845. [Google Scholar] [CrossRef] [PubMed][Green Version] - Aznan, A.A.; Rukunudin, I.H.; Shakaff, A.Y.M.; Ruslan, R.; Zakaria, A.; Saad, F.S.A. The use of machine vision technique to classify cultivated rice seed variety and weedy rice seed variants for the seed industry. Int. Food Res. J.
**2016**, 23, S31–S35. [Google Scholar] - Kurtulmus, F.; Alibaş, İ.; Kavdır, I. Classification of pepper seeds using machine vision based on neural network. Int. J. Agric. Biol. Eng.
**2016**, 9, 51–62. [Google Scholar]

**Figure 1.**System block diagram for the classification of different chickpea (Cicer arietinum L.) varieties.

**Figure 2.**A sample image of each of the three chickpea (Cicer arietinum L.) varieties under consideration: (

**a**) Adel, (

**b**) Arman, and (

**c**) Azad.

**Figure 3.**HSI color space segmentation Equation (1) flowchart for clarification purposes: background and foreground (chickpea) pixel segmentation.

**Figure 4.**An example in the chickpea (Cicer arietinum L.) image segmentation process. (

**a**): Original image, (

**b**): Image converted to HSI color space, (

**c**): Image after application of Equation (1), (

**d**): Binary image, (

**e**): Improved binary after remove objects with pixels less than 100 (this value select with try and error), (

**f**): Segmented image.

**Figure 5.**Boxplots of (

**a**): the correct classification rate (CCR) and (

**b**): area under the ROC curve (AUC) by hybrid ANN-PSO classifier, for Adel (AUC1), Arman (AUC2), and Azad (AUC3) chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test set).

**Figure 6.**Boxplots of (

**a**): the correct classification rate (CCR) and (

**b**): area under the ROC curve (AUC) by hybrid ANN-CO classifier, for Adel (AUC1), Arman (AUC2) and Azad (AUC3) chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test set).

**Figure 7.**Boxplots of (

**a**): the correct classification rate (CCR) and (

**b**): area under the ROC curve (AUC) by hybrid ANN-HS classifier, for Adel (AUC1), Arman (AUC2) and Azad (AUC3) chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test set).

**Figure 8.**Boxplots of (

**a**): the correct classification rate (CCR) and (

**b**): AUC reached by ensemble ANN-PSO/ACO/HS-MV classifier, for Adel (AUC1), Arman (AUC2) and Azad (AUC3) chickpea (Cicer arietinum L.) varieties (1000 uniform random simulations, test set).

**Figure 9.**Mean ROC curves for three chickpea varieties: Adel (blue), Arman (green), Azad (red). (

**a**) ANN-PSO; (

**b**) ANN-ACO; (

**c**) ANN-HS; (

**d**) ANN-PSO/ACO/HS Majority-Voting ensemble (1000 simulations, test set).

**Table 1.**Definition of the various color spaces used in the present work. Equations to obtain these channels from the original RGB values are also provided.

Color Space | Color Channel | Transformation from RGB Color Space |
---|---|---|

$\mathrm{V}$ | $\mathrm{V}=\mathrm{M};\mathrm{with}\mathrm{M}=\mathrm{max}\left\{\mathrm{R},\mathrm{G},\mathrm{B}\right\};\mathrm{m}=\mathrm{min}\left\{\mathrm{R},\mathrm{G},\mathrm{B}\right\};\mathrm{p}=60\mathrm{m}/\mathrm{M}$ | |

HSV | $\mathrm{S}$ | $\mathrm{S}=\left(\mathrm{M}-\mathrm{m}\right)/\mathrm{M}$ |

$\mathrm{H}$ | $\mathrm{H}=\{\mathrm{p}\left(\mathrm{G}-\mathrm{B}\right)\mathrm{if}\mathrm{M}=\mathrm{R};120+\mathrm{p}\left(\mathrm{B}-\mathrm{R}\right)\mathrm{if}\mathrm{M}=\mathrm{G};$ $240+\mathrm{p}\left(\mathrm{R}-\mathrm{G}\right)\mathrm{if}\mathrm{M}=\mathrm{B}\}$ | |

HSI | $\mathrm{I}$ | $\mathrm{I}=\left(\mathrm{R}+\mathrm{G}+\mathrm{B}\right)/3$ |

$\mathrm{S}$ | $\mathrm{S}=255-\mathrm{m}/\mathrm{I}$ | |

YCrCb | $\mathrm{Cr}$ | $\mathrm{Cr}=0.713\left(\mathrm{R}-\mathrm{Y}\right)+128$ |

$\mathrm{Cb}$ | $\mathrm{Cb}=0.564\left(\mathrm{B}-\mathrm{Y}\right)+128$ | |

YIQ | $\mathrm{I}$ | $\mathrm{I}=-0.595716\xb7\mathrm{R}-0.274453\xb7\mathrm{G}-0.321263\xb7\mathrm{B}$ |

$\mathrm{Q}$ | $\mathrm{Q}=0.211456\xb7\mathrm{R}-0.522581\xb7\mathrm{G}-0.311135\xb7\mathrm{B}$ | |

CMY | $\mathrm{C}$ | $\mathrm{C}=255-\mathrm{R}$ |

$\mathrm{M}$ | $\mathrm{M}=255-\mathrm{G}$ | |

$\mathrm{Y}$ | $\mathrm{Y}=255-\mathrm{B}$ |

**Table 2.**Texture features extracted from the gray level co-occurrence matrix (GLCM) [17].

Number | Feature Name | Number | Feature Name |
---|---|---|---|

1 | Contrast | 11 | Inverse difference normalized (INN) |

2 | Sum of squares | 12 | Inverse difference moment normalized |

3 | Second diagonal moment | 13 | Diagonal moment |

4 | Mean | 14 | Sum average |

5 | Sum entropy | 15 | Variance |

6 | Difference variance | 16 | Sum variance |

7 | Difference entropy | 17 | Standard deviation |

8 | Information measure of correlation 1 | 18 | Coefficient of variation |

9 | Information measure of correlation 2 | 19 | Maximum probability |

10 | Inverse difference (INV) is homogeneity | 20 | Correlation |

**Table 3.**Parameter values in the artificial neural network (ANN) that are used in hybrid ANN-CA architecture in order to select the most effective (discriminant) features [19].

ANN Parameter | Value |
---|---|

Number of hidden layers | 2 |

Number of neurons of the hidden layer | 8, 19 |

Transfer function | tribas, tansig |

Backpropagation network training function | trainlm |

Backpropagation weight/bias learning function | learncon |

**Table 4.**Optimal architecture of hidden ANN layers determined by PSO, ACO and HS algorithms. [19]

Classifier | Num. of Layers | Number of Neurons | Transfer Function | Backpropagation Network Training Function | Backpropagation Weight/Bias Learning Function |
---|---|---|---|---|---|

ANN-PSO | 3 | First layer: 16 | First layer: netinv | learnlv1 | traingdx |

Second layer: 9 | Second layer: satlins | ||||

Third layer: 18 | Third layer: compet | ||||

ANN-ACO | 3 | First layer: 12 | First layer: satlin | learnlv2 | traingd |

Second layer: 3 | Second layer: satlin | ||||

Third layer: 13 | Third layer: poslin | ||||

ANN-HS | 3 | First layer: 13 | First layer: tansig | learnp | trainlm |

Second layer: 10 | Second layer: satlin | ||||

Third layer: 17 | Third layer: logsig |

**Table 5.**Confusion matrix including correct classification rate (CCR) for the ANN-PSO classifier: Adel, Arman and Azad chickpea (Cicer arietinum L.) varieties (1000 random iterations, test, train and validation sets).

Classifier | Data Set Type | Real/Estimated Class | Adel (1) | Arman (2) | Azad (3) | Total Data | Classification Error Per Class (%) | CCR (%) |
---|---|---|---|---|---|---|---|---|

ANN-PSO | Test | Adel | 57,854 | 1048 | 98 | 59,000 | 1.94 | 98.65 |

Arman | 852 | 57,147 | 1 | 58,000 | 1.47 | |||

Azad | 386 | 0 | 59,614 | 60,000 | 0.643 | |||

Train | Adel | 114,418 | 3582 | 0 | 118,000 | 3.03 | 98.71 | |

Arman | 0 | 115,000 | 0 | 115,000 | 0 | |||

Azad | 936 | 0 | 117,064 | 118,000 | 0.793 | |||

Validation | Adel | 18,721 | 272 | 7 | 19,000 | 1.47 | 98.09 | |

Arman | 68 | 18,932 | 0 | 19,000 | 0.358 | |||

Azad | 741 | 0 | 18,259 | 19,000 | 3.9 |

**Table 6.**Confusion matrix including correct classification rate (CCR) for ANN-ACO classifier: Adel, Arman and Azad chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test train and validation sets).

Classifier | Data Set Type | Real/Estimated Class | Adel (1) | Arman (2) | Azad (3) | Total Data | Classification Error Per Class (%) | CCR (%) |
---|---|---|---|---|---|---|---|---|

ANN-ACO | Test | Adel | 58,059 | 895 | 46 | 59,000 | 1.59 | 98.94 |

Arman | 656 | 57,315 | 29 | 58,000 | 1.18 | |||

Azad | 243 | 0 | 59,757 | 60,000 | 0.405 | |||

Train | Adel | 117,074 | 926 | 0 | 118,000 | 0.785 | 99.52 | |

Arman | 753 | 114,247 | 0 | 115,000 | 0.655 | |||

Azad | 0 | 0 | 118,000 | 118,000 | 0 | |||

Validation | Adel | 18,847 | 153 | 0 | 19,000 | 0.805 | 98.88 | |

Arman | 0 | 18,804 | 196 | 19,000 | 1.03 | |||

Azad | 0 | 289 | 18,711 | 19,000 | 1.52 |

**Table 7.**Confusion matrix including correct classification rate (CCR) for ANN-HS classifier: Adel, Arman and Azad chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test train and validation sets).

Classifier | Data Set Type | Real/Estimated Class | Adel (1) | Arman (2) | Azad (3) | Total Data | Classification Error Per Class (%) | CCR (%) |
---|---|---|---|---|---|---|---|---|

ANN-HS | Test | Adel | 58,059 | 895 | 46 | 59,000 | 1.59 | 98.99 |

Arman | 601 | 57,381 | 18 | 58,000 | 1.07 | |||

Azad | 235 | 0 | 59,765 | 60,000 | 0.392 | |||

Train | Adel | 116,841 | 1159 | 0 | 118,000 | 0.982 | 99.67 | |

Arman | 0 | 115,000 | 0 | 115,000 | 0 | |||

Azad | 0 | 0 | 118,000 | 118,000 | 0 | |||

Validation | Adel | 18,841 | 159 | 0 | 19,000 | 0.837 | 99.56 | |

Arman | 89 | 18,911 | 0 | 19,000 | 0.468 | |||

Azad | 0 | 0 | 19,000 | 19,000 | 0 |

**Table 8.**Confusion matrix including correct classification rate (CCR) for ensemble ANN Majority-Voting (MV): Adel, Arman and Azad chickpea (Cicer arietinum L.) varieties (1000 uniform random iterations, test set).

Ensemble Classifier | Real/Estimated Class | Adel (1) | Arman (2) | Azad (3) | Total Data | Classification Error Per Class (%) | CCR (%) |
---|---|---|---|---|---|---|---|

PSO/ACO/HS ensemble Majority-Voting | Adel | 58,184 | 804 | 12 | 59,000 | 1.38 | 99.10 |

Arman | 508 | 57,490 | 2 | 58,000 | 0.879 | ||

Azad | 266 | 0 | 59,734 | 60,000 | 0.443 |

**Table 9.**Comparison of the performance of the four classifiers (ANN-PSO, ANN-ACO, ANN-HS and ensemble ANN PSO/ACO/HS-MV): recall, specificity, precision, f1-score, AUC and accuracy (1000 iterations, test set).

Classifier | Class | Recall (%) | Specificity (%) | Precision (%) | F1-Score (%) | AUC (Mean ± Std. Dev.) | Accuracy % (Mean ± Std. Dev.) |
---|---|---|---|---|---|---|---|

ANN-PSO | Adel | 97.91 | 99.03 | 98.06 | 97.98 | 0.9963 ± 0.0097 | 98.65 ± 1.31 |

Arman | 98.19 | 99.28 | 98.53 | 98.36 | 0.9988 ± 0.0026 | ||

Azad | 99.83 | 99.66 | 99.36 | 99.59 | 0.9999 ± 0.0011 | ||

ANN-ACO | Adel | 98.47 | 99.2 | 98.4 | 98.44 | 0.9978 ± 0.0047 | 98.94 ± 0.89 |

Arman | 98.46 | 99.42 | 98.82 | 98.64 | 0.9888 ± 0.0029 | ||

Azad | 99.87 | 99.79 | 99.59 | 99.73 | 0.9999 ± 0.0006 | ||

ANN-HS | Adel | 98.58 | 99.2 | 98.4 | 98.49 | 0.9975 ± 0.0051 | 98.99 ± 0.87 |

Arman | 98.46 | 99.48 | 98.93 | 98.69 | 0.9984 ± 0.0035 | ||

Azad | 99.89 | 99.79 | 99.61 | 99.75 | 1.0000 ± 0.0004 | ||

ensemble ANN PSO/ACO/HS Majority-Voting | Adel | 98.69 | 99.31 | 98.62 | 98.65 | 0.9898 ± 0.0088 | 99.10 ± 0.75 |

Arman | 98.62 | 99.57 | 99.12 | 98.87 | 0.9822 ± 0.0083 | ||

Azad | 99.97 | 99.77 | 99.56 | 99.77 | 0.9977 ± 0.0037 |

**Table 10.**Mean ± std. AUC values for the precision-recall curves(pr-AUC): ANN-PSO, ANN-ACO, ANN-HS and ensemble ANN PSO/ACO/HS-MV classifiers (1000 iterations, test set).

Chickpea Variety/Classifier | Adel | Arman | Azad |
---|---|---|---|

ANN-PSO | 0.9763 ± 0.0206 | 0.9794 ± 0.0098 | 0.9822 ± 0.0295 |

ANN-ACO | 0.9799 ± 0.0083 | 0.9796 ± 0.0085 | 0.9831 ± 0.0031 |

ANN-HS | 0.9795 ± 0.0081 | 0.9775 ± 0.0132 | 0.9832 ± 0.0024 |

ensemble ANN PSO/ACO/HS Majority-Voting | 0.9766 ± 0.1001 | 0.9756 ± 0.0122 | 0.9818 ± 0.0029 |

**Table 11.**Accuracy of the classifiers reported in different food seeds. Please note no direct comparison possible given the various works do not share the same image database and are for different food seeds.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## Share and Cite

**MDPI and ACS Style**

Pourdarbani, R.; Sabzi, S.; Kalantari, D.; Hernández-Hernández, J.L.; Arribas, J.I. A Computer Vision System Based on Majority-Voting Ensemble Neural Network for the Automatic Classification of Three Chickpea Varieties. *Foods* **2020**, *9*, 113.
https://doi.org/10.3390/foods9020113

**AMA Style**

Pourdarbani R, Sabzi S, Kalantari D, Hernández-Hernández JL, Arribas JI. A Computer Vision System Based on Majority-Voting Ensemble Neural Network for the Automatic Classification of Three Chickpea Varieties. *Foods*. 2020; 9(2):113.
https://doi.org/10.3390/foods9020113

**Chicago/Turabian Style**

Pourdarbani, Razieh, Sajad Sabzi, Davood Kalantari, José Luis Hernández-Hernández, and Juan Ignacio Arribas. 2020. "A Computer Vision System Based on Majority-Voting Ensemble Neural Network for the Automatic Classification of Three Chickpea Varieties" *Foods* 9, no. 2: 113.
https://doi.org/10.3390/foods9020113