Recognition method of tomato leaf diseases based on convolutional neural network

Journal Title: Journal of Henan Agricultural University - Year 2024, Vol 58, Issue 2

Abstract

[Objective] The convolutional neural network was used to construct a crop disease recognition model to improve the recognition performance and solve the problems of low recognition performance and poor generalization effect of crop diseases. [Method] The data augmentation technology was used to increase the sample diversity. The focal loss was introduced to improve the model learning target, and the sample imbalance problem was solved. The recognition performance of the different convolutional neural network structures was analyzed and compared, and the reliability of the model was measured by class activation map generation technology. The effectiveness of the method was verified on the tomato leaf diseases dataset. [Result] After the application of data augmentation technology, the recognition accuracy of the model on simple background samples was increased by 1.0%, and that on complex background samples was increased by 12.5%. The focus loss improved the accuracy of the model by 0.1%. The recognition accuracy of the model was 99.8%, and the recall rate of various diseases was above 97.3%. The saliency map generated by the class activation map technology could effectively identify the key areas of the model in the recognition process. [Conclusion] This method can effectively solve the problem of unbalanced disease image samples and improve the generalization performance of the disease recognition model, and use the class activation map to analyze the reliability of the model, which can provide reference for the prevention and control of the tomato leaf diseases.

Authors and Affiliations

Junting LIU, Yuncheng ZHOU, Qiong WU, Xiongwei WU, Changyuan WANG

Keywords

Related Articles

Spatial layout of dominant grain crop production in Henan Province

[Objective] This study was conducted to explore the spatial layout of dominant grain crop production in Henan Province, to determine the advantages and disadvantages of grain crop production, and to provide theoretical s...

Screening and optimal field ratios of combined pesticides for the control of Spodoptera frugiperda

[Objective] The aims of this study were to screen the combined pesticides that can significantly enhance the insecticidal efficacy, and explore the optimal formulation which has good control effect on the Spodoptera frug...

Effects of shading on leaf economic traits, hydraulic traits and leaf size traits in Paulownia fortune

[Objective] The purpose of our study is to explore the effect of different shading environments on leaf functional traits for Paulownia fortunei. [Method] Using the Paulownia fortunei as material, three light gradients w...

Research progress in the mechanism of autophagy in coronavirus infection

Autophagy plays a major role in the viral infection and disease pathogenesis. Currently, coronavirus spread around the world, and coronavirus disease 2019 (COVID-19) caused by severe acute respiratory syndrome coronaviru...

Analysis on the microclimate impact mechanism of village streets in the low mountain areas of Zhengzhou

[Objective] This study aims to explore the influencing factors and regulatory mechanisms of micro-climate in village streets, create a more comfortable summer microclimate enviorment, reduce energy consumption and improv...

Download PDF file
  • EP ID EP769339
  • DOI 10.16445/j.cnki.1000-2340.20240018.002
  • Views 63
  • Downloads 0

How To Cite

Junting LIU, Yuncheng ZHOU, Qiong WU, Xiongwei WU, Changyuan WANG (2024). Recognition method of tomato leaf diseases based on convolutional neural network. Journal of Henan Agricultural University, 58(2), -. https://europub.co.uk/articles/-A-769339