Ensemble Technique of Deep Learning Model for Identifying Tomato Leaf Diseases Based on Choquet Fuzzy Integral

Abstract

As tomato leaves are often attacked by various microorganisms, pests and bacterial diseases, the yield of tomato is seriously reduced. Accurate and timely identification of tomato leaf diseases is of great significance to reduce farmers’ economic losses. Ensemble learning as a combinatorial optimization method, which can improve the generalization ability and model stability, is widely used in the field of plant leaf disease identification. However, commonly used ensemble methods such as majority voting, weighted averaging, etc. do not consider the interaction between inputs when aggregating the inputs of multiple models, such that they do not produce representative outputs. To solve this problem, this paper adds fuzzy algorithms to the ensemble method, i.e., five pre-trained deep learning models, namely VGG16, VGG19, Xception and InceptionV3 and InceptionResnet V2, are integrated using Choquet integral fuzzy integrals for four classes of tomato disease identification and classification. The experimental results show that the proposed method achieves encouraging results, with the best single deep learning model achieving 98.63% accuracy on the PlantVillage dataset and the proposed fuzzy ensemble method achieving 99.80% accuracy. For the identification of tomato diseases in natural scenarios, the proposed method achieves 97% accuracy. The method can effectively identify tomato diseases.

Keywords:

• Choquet integral
• fuzzy ensemble
• tomato disease identification deep learning

Acknowledgments

We would like to thank the 802 Lab of the College of Arts and Science of Northeast Agricultural University for providing us with the infrastructure support.

Disclosure statement

The authors declare that they have no conflict of interest.

Data availability

The authors are unable or have chosen not to specify which data has been used.

Additional information

Funding

This work was supported by the Postdoctoral Foundation of Heilongjiang Province [Grant LBH-Q21065].