Machine vision approach for classification of citrus leaves using fused features

Figures & data

Table 1. Local citrus varieties cultivar in Pakistan.^[42]

General Variety	Local Varieties	Region
Grapefruit (Citrus paradisiacal.)	Mash Seedless, Duncan, Foster, and Shamber	Pakistan, India
Mandarin (Citrus reticulata Blanco)	Fuetrells Early and Kinow	Pakistan, India
Sweet Orange (Citrus sinensis (L.)	Moussami, Washington Navel, Succri, Red Blood, Jaffa, Ruby Red and Valencia late.	Pakistan, India, Afghanistan
Bitter Orange (Citrus aurantium L.)	Seville orange, Sour Orange, Marmalade orange or bigarade orange	Pakistan, India
Lime (Citrus aurantifolia Christm.)	Sweet Lime and Kaghazi Lime	Pakistan, India
Lemon (Citrus Limon L.)	Eureka and Lisbon Lemon	Pakistan, India
Rough Lemon (Citrus jambhiri Lush.)	Most common rootstock for the propagation of citrus in the subcontinent.	Pakistan, India
Grapefruit (Citrus paradisiacal.)	It was first developed in California in 1935 and then Punjab Agriculture College and Research Institute Faisalabad (then, Lyallpur) Pakistan, introduced it in the sub-continent in 1940.	Pakistan, India

Figure 1. Digital images of 8 citrus leaf varieties

Figure 2. Gray level images of 8 citrus varieties

Figure 3. Gray level citrus leaf images of four non- overlapping regions of interest (ROIs)

Table 2. Correlation-based Feature Selection (CFS)

S. No	Features
1	Histo. Mean
2	Histo. Standard deviation
3	Texture Energy Average
4	Texture Energy Range
5	Inertia Average
6	Correlation Average
7	Correlation Range
8	Inverse Diff Average
9	Inverse Diff Range
10	Texture Entropy Range
11	B
12	G
13	R
14	NIR
15	SWIR

Table 3. Values of the parameters of MLP used in our experimental work

Parameter	Parameter value
Input Layers	15
Hidden Layers	1
Neurons	10
Learning Rate	0.3
Momentum	0.2
Validation Threshold	20
Epochs	100 (min), 500(max)

Figure 4. Multi-feature data fusion framework for leaf-based citrus varieties classification

Figure 5. Multi-feature data fusion framework for leaf-based citrus varieties classification

Table 4. The overall accuracy of machine learning classifiers on ROIs of citrus leaves images of size (128 × 128)

Classifier	Kaapa value	True Positive	False Positive	Receiver Operating Characteristics ROC	Mean Absolute Error MAE	Root- Mean- Square Error	Time (sec)	Accuracy (%)
MLP	0.851	0.87	0.015	0.935	0.0412	0.1791	9.86	87.03%
RF	0.8090	0.818	0.045	0.962	0.0730	0.2550	6.02	81.75%
J48	0.8013	0.810	0.050	0.890	0.0850	0.2760	12.20	80.34%
NB	0.7360	0.757	0.060	0.941	0.0988	0.3062	16.03	75.70%

Table 5. The confusion matrix showing the classification of citrus leaves images (128 × 128) using MLP

Classes	Grapefruit	Maussami	Malta	Lemon	Kinow	Local Lemon	Fuetrells	Malta Shakri
Grapefruit	695	20	8	10	26	21	10	10
Maussami	20	675	25	0	27	15	8	30
Malta	14	15	630	20	40	10	60	11
Lemon	5	45	15	663	0	30	35	7
Kinnow	10	13	15	28	650	48	20	16
Local Lemon	45	50	0	5	45	605	40	10
Fuetrells	7	30	45	28	0	20	615	55
Malta Shakri	12	37	57	19	48	23	9	595

Table 6. The overall accuracy of machine learning classifiers on ROIs of citrus leaves images of size (256 × 256)

Classifier	Kaapa value	True Positive	False Positive	Receiver Operating Characteristics ROC	Mean Absolute Error MAE	Root- Mean- Square Error	Time (secs)	Accuracy (%)
MLP	0.9787	0.981	0.003	0.986	0.0145	0.0624	1.23	98.14%
RF	0.8090	0.818	0.045	0.962	0.0730	0.2550	6.02	81.75%
J48	0.8013	0.810	0.050	0.890	0.0850	0.2760	12.20	80.34%
NB	0.7360	0.757	0.060	0.941	0.0988	0.3062	16.03	75.70%

Table 7. The confusion matrix showing the classification of citrus leaves images (256 × 256) using MLP

Classes	Grapefruit	Moussami	Malta	Lemon	Kinow	Local Lemon	Fuetrells	Malta Shakri	Accuracy%	Total
Grapefruit	784	0	1	10	5	0	0	0	98.00%	800
Moussami	1	790	0	0	5	0	4	0	98.75%	800
Malta	1	0	794	2	0	0	3	0	99.25%	800
Lemon	1	2	0	780	0	10	0	7	97.50%	800
Kinnow	0	5	5	0	776	0	5	9	97.00%	800
Local Lemon	0	6	7	5	5	767	0	10	95.87%	800
Fuetrells	3	3	0	0	0	10	764	20	95.50%	800
Malta Shakri	0	1	0	0	2	1	0	795	99.37%	800

Figure 6. Accuracy results of leaf-images classification of eight citrus plants (ROI 128 × 128) using MLP

Figure 7. Accuracy results of citrus leaf-images classification of eight citrus plants (ROI 256 × 256) usingMLP

Table 8. Comparison of our proposed approach with existing approaches

Source/Reference	Features	Classifiers	Accuracy
[9]	Digital Morphological Features (DMFs)	Probabilistic Neural Network (PNN)	90.312%.
[10]	Multispectral Features	K-Nearest Neighbor (K-NN)	98%
[11]	Texture + Radiometric Features	Artificial Neural Network (ANN)	96%
[12]	Color + Shape Features	Artificial Neural Network (ANN)	84.94%
[13]	Histogram + Textural Features	Artificial Neural Network (ANN)	97.32%
[14]	Texture, Color and Geometry Features.	Principal Component Analysis (PCA) + Artificial Neural Network (ANN)	86%
[17]	Color Texture Features	Discriminant Analysis (DA)	95.34%
[18]	Morphological Shape Descriptors (MSD), Histogram Oriented Gradients (HOG), Hu invariant moments (Hu) and Zernike Moments (ZM).	Multilayer Perceptron (MLP)	95.25%
[27]	Color, color histogram, and local binary pattern (LBP) fused features	Multi-Support Vector Machine + Genetic Algorithm	98.10%
[28]	Color, texture and geometric based fused features	Multi-Support Vector Machine Genetic Algorithm	92%
[29]	Two deep pre-trained models (VGG16, Caffe Alex Net)	Multi-Support Vector Machine Genetic Algorithm	98%- 98.60%
Proposed Approach	Texture + Histogram + Binary + RST (Rotational, Scalability and Translational) Invariant + Spectral Features.	Multilayer Perceptron (MLP)	98.14%

Figure 8. Comparison analysis of accuracy achieved by MLP in classifying eight citrus leaf images having ROIs 128 × 128 and 256 × 256

Siddique, M. I.; Garnevska, E. Citrus Value Chain (S): A Survey of Pakistan Citrus Industry. Agric. Value Chain. 2018, 37.

 Google Scholar

Hamuda, E.; Glavin, M.; Jones, E. A Survey of Image Processing Techniques for Plant Extraction and Segmentation in the Field. ComputersandElectronicsinAgriculture. 2016, 125, 184–199.

 Google Scholar

Singh, V.; Misra, A. K. Detection of Plant Leaf Diseases Using Image Segmentation and Soft Computing Techniques. Information Process. Agric. 2017, 4, 41–49. DOI: 10.1016/j.inpa.2016.10.005.

 Google Scholar

Wang, S.; He, D.; Li, W.; Wang, Y. Plant Leaf Disease Recognition Based on the Kernel K-means Clustering Algorithm. Nongye Jixie Xuebao = Trans. Chin. Soc. Agric. Mach. 2009, 40, 152–155.

 Google Scholar

Wu, S. G.; Bao, F. S.; Xu, E. Y.; Wang, Y. X.; Chang, Y. F.; Xiang, Q. L. A Leaf Recognition Algorithm for Plant Classification Using A Probabilistic Neural Network. 2007 IEEE international symposium on signal processing and information technology, IEEE, Cairo, Egypt, 2007, pp 11–16.

 Google Scholar

Shifa, M. S.; Naweed, M. S.; Omar, M.; Jhandir, M. Z.; Ahmed, T. Classification of Cotton and Sugarcane Plants on the Basis of Their Spectral Behavior. PakistanJournalofBotany. 2011, 43, 2119–2125.

 Google Scholar

REHMANI, E.; NAWEED, M.; SHAHID, M.; QADRI, S.; ULLAH, M.; GILANI, Z. A Comparative Study of Crop Classification by Using Radiometric and Photographic Data. Sindh Univer. Res. J-SURJ (Science Series). 2015, 47.

 Google Scholar

Szczypin´ski, P. M.; Klepaczko, A.; Zapotoczny, P. Identifying Barley Varieties by Computer Vision. Comput. Electron. Agric. 2015, 110, 1–8. DOI: 10.1016/j.compag.2014.09.016.

 Web of Science ®Google Scholar

Patil, S. B.; Bodhe, S. K. Leaf Disease Severity Measurement Using Image Processing. Int. J. Eng. Technol. 2011, 3, 297–301.

 Google Scholar

Khan, M. A.; Lali, M. I. U.; Sharif, M.; Javed, K.; Aurangzeb, K.; Haider, S. I.; Altamrah, A. S.; Akram, T. An Optimized Method for Segmentation and Classification of Apple Diseases Based on Strong Correlation and Genetic Algorithm Based Feature Selection. IEEE Access. 2019, 7, 46261–46277. DOI: 10.1109/ACCESS.2019.2908040.

 Web of Science ®Google Scholar

Adeel, A.; Khan, M. A.; Sharif, M.; Azam, F.; Umer, T.; Wan, S. Diagnosis and Recognition of Grape Leaf Diseases: An Automated System Based on a Novel Saliency Approach and Canonical Correlation Analysis Based Multiple Features Fusion. Sustainable Comput. 2019, 24.

 Google Scholar

Khan, M. A.; Akram, T.; Sharif, M.; Awais, M.; Javed, K.; Ali, H.; Saba, T. CCDF: Automatic System for Segmentation and Recognition of Fruit Crops Diseases Based on Correlation Coefficient and Deep CNN Features. Comput. Electron. Agric. 2018, 155, 220–236. DOI: 10.1016/j.compag.2018.10.013.

 Web of Science ®Google Scholar