Identification of Rice Diseases Based on Batch Normalization and AlexNet Network

Yang Hongyun; Wan Ying; Wang Yinglong; Luo Jianjun

doi:10.3788/LOP202158.0610007

Laser & Optoelectronics Progress, Volume. 58, Issue 6, 610007(2021)

Identification of Rice Diseases Based on Batch Normalization and AlexNet Network

Yang Hongyun¹, Wan Ying², Wang Yinglong^2、*, and Luo Jianjun²

Author Affiliations

¹School of Software, Jiangxi Agricultural University, Nanchang, Jiangxi 330045, China

²School of Computer Information and Engineering, Jiangxi Agricultural University, Nanchang, Jiangxi 330045, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(18)

Fig. 1. Images of rice disease spot. (a)Leaf smut; (b)bacterial leaf blight; (c)bacterial leaf streak; (d)brown spot;(e)aphelenchoides bessyi; (f)red blight; (g)rice blast; (h)rice sheath blight

Download full size

Fig. 2. Structure diagram of AlexNet network

Download full size

Fig. 3. Disease images segmented by FCM algorithm. (a)Leaf smut; (b)bacterial leaf blight; (c)bacterial leaf streak;(d)brown spot; (e)aphelenchoides bessyi; (f)red blight; (g)rice blast; (h)rice sheath blight

Download full size

Fig. 4. Flow chart of the experiment

Download full size

Fig. 5. Loss precision value of training and test for AlexNet_model

Download full size

Fig. 6. Accuracy of training and test for AlexNet_model

Download full size

Fig. 7. Loss precision value of training and test for LeNet _model

Download full size

Fig. 8. Accuracy of training and test for LeNet_model

Download full size

Fig. 9. Loss precision value of training and test forFCM_model

Download full size

Fig. 10. Accuracy of training and test for FCM_model

Download full size

Fig. 11. Flow chart of the convolutional neural network

Download full size

Fig. 12. Loss precision value of training and test for BN_model

Download full size

Fig. 13. Accuracy of training and test for BN_model

Download full size

Fig. 14. ROC of different models. (a)AlexNet_model; (b)LeNet_model; (c)FCM_model; (d)BN_model

Download full size

Table 1. Disease characteristics and image quantity

View table

Table 1. Disease characteristics and image quantity

Disease category of rice	Main characteristic	Number of images
Aphelenchoides bessyi	Yellowish brown and brown, showing "dry tip" shape	754
Bacterial leaf blight	Yellowish brown and finally dry white	1200
Bacterial leaf streak	Yellow-brown streaks, yellow transparent beaded pyogenesis	1140
Brown spot	Dark brown,with yellow halo around disease spot	1200
Leaf smut	Longitudinal scattered short stripe disease spots, gray black	1200
Red blight	Small brown spots or short stripes, irregular rust like spots	800
Rice blast	Fusiform with large middle ends pointed, grayish white in the middle, brown in the edges	1230
Rice sheath blight	Oval, moire shape, gray center, dark brown edges	930
Total	—	8454

Table 2. Structure parameter of Lenet-5 network

View table

Table 2. Structure parameter of Lenet-5 network

Layer name	Kernel size	Stride	Padding	Output shape
Layer1_Conv1	5×5×3×6	1	VALID	28×28×6
Layer2_Maxpooling1	2×2	2	SAME	14×14×6
Layer3_Conv2	5×5×6×16	1	VALID	10×10×16
Layer4_Maxpooling2	2×2	2	SAME	5×5×16
Layer5_Fc1	120	—	—	120
Layer6_Fc2	84	—	—	84
Softmax	—	—	—	8

Table 3. Comparison table of the final results of the models
View table
Table 3. Comparison table of the final results of the models
Model Training loss /10^-3 Test loss Test accuracy /%
AlexNet_model 2.1000 0.0576 98.88
LeNet_model 0.0219 0.1656 94.68
FCM_model 0.3000 0.0577 98.52
BN_model 0.7500 0.0376 99.11

Table 4. Precision rate of disease recognition under each model

View table

Table 4. Precision rate of disease recognition under each model

Precision rate	AlexNet_model	LeNet_model	FCM_model	BN_mode
Aphelenchoidesbessyi	0.99	0.97	0.98	0.99
Bacterial leaf blight	0.99	0.92	0.99	0.99
Bacterial leaf streak	0.97	0.92	0.97	0.98
Brown spot	0.99	0.95	0.99	0.99
Leaf smut	1.00	0.99	1.00	1.00
Red blight	0.99	0.95	1.00	0.99
Rice blast	0.99	0.95	0.99	1.00
Rice sheath blight	1.00	0.94	0.98	1.00

Tools

Get Citation

Copy Citation Text

Yang Hongyun, Wan Ying, Wang Yinglong, Luo Jianjun. Identification of Rice Diseases Based on Batch Normalization and AlexNet Network[J]. Laser & Optoelectronics Progress, 2021, 58(6): 610007

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites